Home

Eventus User's Guide

1. 80 Figure 6 6 Remaining results for the event parameter approach demonstration Event Parameter Approach Demonstration 4 Mean Median Abnormal Abnormal Day Return Return 5 1 01 1 32 4 0 29 0 21 3 1 34 1 17 2 1 04 0 83 1 0 21 0 43 0 4 10 3 52 1 1 34 0 83 2 0 92 0 23 3 0 26 0 37 4 2 46 2 57 5 0 43 0 49 81 82 Chapter 7 Obtaining Returns Prices Volume Number of Trades and Shares Outstanding from the CRSP Database Eventus provides statements to read selected stock and index returns stock prices or trading volume data from the CRSP database for future analysis by non Eventus software The selected returns can go in to any type of file that SAS can handle including sas data sets and external files for example ASCII character files Explanations of the statements and their options appear below 7 1 The EVENTUS statement The option GETDATA must appear on the EVENTUS statement preceding any RETURNS PRICES or VOLUME statement The default data frequency is daily To use an installed monthly CRSP database specify MONTHLY on the EVENTUS statement For example EVENTUS MONTHLY To specify landscape orientation of the printed output use the PAGE WIDE option on the EVENTUS statement the default is portrait orientation 83 Figure 7 1 Eventus statements to read stock returns from a CRSP database filename request G Some Folder Filename extension EVENT
2. 2 22 004 105 A 3 Different estimation and event return intervals 115 A 4 Variable Names In Eventus Output SAS Data Sets 116 Ady Missing Returns 2 24 64 44 4 4 4 64 Ha OS Po Oe eS 118 Reference Guide to Eventus Statements 121 B 1 CUSIPERM Statement 0 000000084 121 B 2 DATECONV Statement 64 65 pak e e o a i aa 122 Bs EVENTUS statement oa aaien ee heh Re 126 BA EVISTUDY Statement s o iia oa aa i ie oedi e i i o a e a 130 B 5 EXTRACT Statement 2 8 4 poa 48444498444 254 136 B6 OLDSTUDY Statement 6 ee eee wee 138 By PRICES Statement o c ca sa 464 448 Oe be Bee we ee 141 B S REQUEST Statement sao sacca desiu we RE ee 145 B 9 RETURNS Statement 0 0 0 0 00 00 0084 152 B 10 TITLE and TITLE2 Statements 004 154 B 11 UPCUSIP Statement 0 0 0 0 0 0 0 0 000004 154 B 12 VOLUME Statement 0 0 0 0 0 0 0 0000000000848 155 B 13 WINDOWS Statement 0 0 0 0 0 0 0 000 0004 158 How Eventus Finds the CRSP Stock Database 159 Cel ORSPACCESB nb eb wa ee oe RGR SEDBERGH 159 C2 SFA POrmmat sia beh eb eh ee Ee hh EEG eS 160 il C 3 Size index files used with both cRSPAccess and SFA format databases 2 00 0222s C 4 CRSP data permanently stored in SAS data sets ill iv List of Tables A l A 2 A 3 B 1 Adjustments factors for different estimation and event return MAWE VAl ee oe eie e oa ke e g dod e ER Ew eS DG 116 Variable names in
3. title2 Replication of Table III C Insolvent Bank Sample request est 6 estlen 50 id Swarynum idfmt 2 0 evtstudy nomar nostd value post 5 pre 5 tail 2 Swary s 1986 list so we call the identifying variable Swarynum The spec ification IDFMT 2 0 means that the identifying values are integers of up to two digits O Hara and Shaw 1990 report only market model abnormal returns the option NOMAR on the EVTSTUDY statement suppresses the computation and reporting of market adjusted returns Because the firms are all in the same industry and have a common event date the authors do not use a standardized residual test but base their test statistic on a standard devi ation estimated for the portfolio of sample firms from residual returns in the estimation period See Chandra and Balachandran 1990 and Chandra Moriarity and Willinger 1990 for further discussion of cross sectional de pendencies in event studies We suppress standardized tests with the NOSTD option The NOMAR and NOSTD options are not necessary they only elimi nate extra output that is not of interest in this study The VALUE option is necessary because O Hara and Shaw use the CRSP value weighted index in stead of the Eventus default equally weighted index The POST 5 and PRE 5 specifications limit the event period to days 5 through 5 Finally TAIL 2 means that significance level symbols will reflect two tailed tests Figure 3 2 shows the s
4. 0 5 Ej 6 E 1 s The jackknife estimator for stock 7 on day t is the mean of the pseudo values To gain efficiency the estimates are averaged across the sample of stocks es ae O Dei Nz Finally the jackknife test statistic for the sample of stocks on day t is O t Jackknife AA a S Jackknifet VN l where i lt oe S Jackknife t NI 5 0 A 5 i 1 The distribution of t yackknife under the null hypothesis is approximately nor mal with mean zero and unit variance To test the significance of the cumulative average abnormal return over the window from date T through date Tp define Lan Ajt nF A 6 OT T Sequentially delete one abnormal return A r from equation A 3 and re compute o 4 using the new value in turn to re compute 0 using equation jt A 6 Call the latter value 6s Ti Ta Form pseudo values O s 7 E Onn E 1 Pann 114 The jackknife estimator for stock j in window T T2 is the mean of the pseudo values 1 E OiT T E As J t Ey The estimates are averaged across the sample of stocks 1 2 Onn VF gt GTT j l The jackknife test statistic for the sample of stocks in window 7 T2 is Or tJackkni e L A T i S Jackknife T Ta VN where 1 Ny _ 2 2 S Jackknife T T wo 5 Oirr Orra i A 8 i 1 A 3 Different estimation and event return in tervals The ESTINTER option on the EVENTUS
5. 0 73 1 39 0 38 0 60 0 52 0 08 0 02 0 16 1 39 0 40 6 76 0 92 0 38 0 00 0 96 0 16 2 90 0 70 1 22 0 14 2 08 0 04 4 65 1 04 5 16 08 18 86 15 08 0 19 0 26 2 13 0 29 2 01 0 33 2 19 0 26 1 46 0 40 2 46 0 18 0 63 0 33 0 10 0 06 1 63 0 66 0 67 0 19 5 36 0 02 1 27 0 18 1 33 0 10 1 01 0 12 Cumulative Average Abnormal Return Equally Precision Weighted Weighted 18 78 16 18 24 02 18 81 11 41 8 39 significant at 3 58 x 0 25 1 42 0 80 1 32 0 25 2 63 4 70 19 82 2 43 0 79 1 15 1 44 1 41 1 50 0 05 0 94 1 02 0 57 2 15 0 56 0 19 0 37 Median Cumulative Abnormal Return 11 99 17 12 4 20 Generalized N Positive Sign Negative Z 16 10 6 1 23 16 6 10 0 78 16 7 9 0 28 16 8 8 0 22 16 7 9 0 28 16 9 7 0 72 16 11 5 1 73 gt 16 8 8 0 22 16 9 7 0 72 16 5 11 1 28 16 11 5 1 73 gt 16 8 8 0 22 16 9 7 0 72 16 12 4 2 23 gt 16 12 4 2 23 gt 16 8 8 0 22 16 6 10 0 78 16 4 12 1 78 lt 16 7 9 0 28 16 10 6 1 23 16 7 9 0 28 16 9 7 0 72 16 8 8 0 22 16 5 11 1 28 16 8 8 0 22 16 8 8 0 22 16 6 10 0 78 16 10 6 1 23 16 11 5 1 73 gt General Positive ized Z Negative Sign Z 3 92 11 5 1 73 gt 17 34 11 5 1 73 gt 1 94 12 4 2 23 gt 10 lt gt significant at 05 lt lt lt gt gt gt significant at 001 11 12 Chapter 3 Event Stud
6. 1 0 4 1 8 I 1 Bn 1 j l t T t T Scholes Williams Beta Estimation When the SW option appears on the EVTSTUDY statement Eventus reports market model results using betas estimated by both ordinary least squares and the method of Scholes and Williams 1977 The Scholes Williams beta estimator is je where oe is the OLS slope estimate from the simple linear regression of Rj on Rmt 1 a is the OLS estimate from the regression of Rj on Rmi 1 and Pm is the estimated first order autocorrelation of Rm As in OLS the intercept estimator forces the estimated regression line through the sample mean a yj BR R is the mean return of stock j over the estimation period and Rpm is the mean market return over the estimation period Eventus uses the simplifying pose pion that the use of Scholes Williams estimates does not affect the formula for s a below Analytically this assumption is not strictly correct but simulation results obtained by the author show that tests using the assumption are swell specified 104 Market Adjusted Returns Market adjusted returns are computed by subtracting the observed return on the market index for day t Rm from the rate of return of the common stock of the j firm on day t Ast Rj a Rmt The definitions of the average abnormal return and cumulative average ab normal return follow those for market model abnormal returns above Comparison Perio
7. 18 same group receive weights of 08333 08333 25 25 and 25 respectively Other observations are part of other groups Notice that the groups need not all have the same number of members Had the GRWEIGHT option not appeared on the REQUEST statement the weight information would have been ignored Each observation in a group portfolio would have received equal weight Thus the researcher could check the sensitivity of the results to the group weighting by changing only one option Creating arbitrage portfolios by specifying stocks to sell short Each stock normally has a positive weight in the event time portfolio To indicate that some stocks should enter the portfolio as if they are sold short specify SHORT on the REQUEST statement In the request file place either an S or an L as the last item of each line to indicate short or long When an S appears Eventus reverses the sign of each estimation period and event period stock return for that observation The sign reversal occurs just after the re turn is read from the CRSP database and verified as non missing Thereafter Eventus treats the returns the same as the returns for any other observation The same stock can be held short in one line of the request file and long in another Eventus re reads the stock returns from the CRSP database each time even if the dates are the same Note that Eventus does not make any explicit adjustment to the portfolio weights the calculations s
8. TITLE text TITLE2 text 3 WINDOWS begin end begin end OLDSTUDY INSAS libref membername INSAS2 libref membername INSAS3 libref membername ID ID_variable name IDFMT format DETAIL DETAIL FULL NOSTD STDONLY RANKTEST SERIAL TAIL 1 2 3 4 Reprinting or Merging Saved Event Stud ies with the OLDSTUDY Statement The EVTSTUDY statement allows you to save abnormal returns in a SAS data set with the OUTSAS option The OLDSTUDY statement lets you reprint a single event study possibly with a different set of windows or merge two or three event studies from a saved EVTSTUDY SAS data set Figure 3 9 lists the statements and options used in an OLDSTUDY program Most of the options can be used in EVTSTUDY programs and are described in Section 3 1 above The options have a similar meaning here but placement and usage differ in certain instances For example the ID option goes on the OLDSTUDY statement and specifies the name of the original identification variable The TWIN MONTHLY WEEKLY or EXCESS option is needed on the EVENTUS statement when used in the program that created the saved event study data only for proper labeling of OLDSTUDY output No other option is needed on the EVENTUS statement The windows specified on the WINDOWS statement need not be those used with the original EVTSTUDY program s 40 The INSAS specification and the INSAS2 and INSAS3 options on the OLD STUDY statement tell Ev
9. Table A 2 Variable names in a saved event study data set Data Variable Name Estimation Event Period Period Daily abnormal return AR1 AR255 AR256 AR9999 Daily abnormal return ARX1 AR744 Daily raw return Daily raw return Daily market index return Equally weighted Equally weighted Value weighted Value weighted SA Dummy variable 1 if abnormal return gt 0 Parameters j bj mean return Rm Lezio Bnk Rm return variance SA Fraction of returns gt 0 Dj Name of market index Type of abnormal return RETN1 RETN255 RETNX1 RETNX744 MKT1 MKT255 MKTX1 MKTX744 VWMK1 VWMK255 VWMKX1 VWMKX744 ALPHA BETA CPR RMBAR SSRM OWNVAR _RMSE_ NPOS_EST Ti RETN256 RETN9999 MKT256 MKT9999 VWMK256 VWMK9999 WDEN256 WDEN9999 NPOS256 NPOS9999 _WEIGHT_ RESTYPE 117 a Additional variable names used only when the estimation period exceeds 255 trading days or other intervals X variable subscripts start with 1 and go up to T 255 where T is the length of the estimation period The X variables contain the first T days of the estimation period then subscript 1 of the regular variables continues with the T 1st day For example if ESTLEN 300 the first 55 days of event period abnormal stock returns are in ARX1 through ARX45 and the 56th through 300th days are in AR1 through AR255 stored observations per stock event date combination per abnormal return method one
10. The Eventus statements in figure 3 8 run a TWIN event study For the most part the options also are options for single event date event studies and function similarly in the two contexts The only differences are the word TWIN If O Hara and Shaw used the same untruncated average abnormal return that Eventus used their result would differ from ours by a factor of 49 48 1 010363 and 1 010363 x 0 77 0 78 rounded 37 Figure 3 7 Event study results O Hara and Shaw JF Dec 1990 pp 1587 1600 4 Replication of Table III C Insolvent Bank Sample Market Model VW Index Average Median Generalized Day Abnormal Abnormal t N Positive Sign Return Return Negative Z 5 1 27 1 51 2 29 22 16 6 2 52 gt 4 1 21 1 08 2 18 22 17 5 2 94 gt gt 3 0 27 0 14 0 49 22 12 10 0 80 2 0 18 0 32 0 33 22 15 7 2 09 gt si 0 21 0 12 0 38 22 10 12 0 05 0 0 43 0 24 0 77 22 10 12 0 05 1 0 30 0 29 0 54 22 15 7 2 09 gt 2 0 55 0 84 1 00 22 8 14 0 91 3 0 38 0 41 0 68 22 9 13 0 48 4 0 61 0 46 1 10 22 8 14 0 91 5 0 33 0 12 0 59 22 10 12 0 05 Cumulative Average Median Cumulative t Positive Gen Sign Days Abnormal Return Abnormal Return Negative Z 5 2 2 92 3 40 2 64 19 3 3 81 gt gt gt 1 0 0 63 0 44 0 81 12 10 0 81 1 5 0 30 1 14 0 24 7 15 1 33 significant at 10 lt gt significant at 05 lt lt gt gt significant at 01 x lt lt lt gt gt gt sign
11. as A SAjt Under the null hypothesis each SAR follows a Student s t distribution with D 2 degrees of freedom Summing the SAR across the sample we obtain N TSAR X SAR je j l The expected value of TSAR is zero The variance of TSAR is N D3 Q p 2E j l The test statistic for the null hypothesis that CAARy 7 0 is ee Zn gt Zh n 1 42 VN c 1 12 where j 3 SA Zr Ty SS Rit Qn 1m t T and oe Orn h Tit Ya e Under cross sectional independence of the Zh ty and other conditions see Patell 1976 Zr m follows the standard normal distribution under the null hypothesis 106 Eventus reports a precision weighted cumulative average abnormal return with the standardized abnormal return method The precision weighted av erage is constructed using the relative weights implied by the definition of ZT m Thus the precision weighted average will always have the same sign as the corresponding Zr m The formula for the precision weighted average is N To PWCASARr 7 X Se wjAjt j l t T where 1 2 T 2 Sin Shy N To 32 Aiai an s2 T 2 TWIN Event Studies Two Firm Specific Event Dates The major difference between TWIN and single date event studies is that TWIN cumulates returns over intervals of security specific length Instead of defining a window for return cumulation with reference to a single event date the window is defined as the period between two e
12. 0 43 2 29 0 55 2 39 0 49 1 56 0 22 2 29 0 03 0 70 0 00 0 27 0 30 1 84 0 56 0 84 0 10 5 33 0 08 1 37 0 35 1 55 0 03 1 14 0 01 46 DKK 51 74 57 83 O7 54k OT KKK 30 06 76 92 00 31 92 59 22 54 70 4T 14 30 95 16 Cumulative Average Median Cumulative Abnormal Return 11 56 22 98 5 82 significant at 10 lt lt gt gt significant at 01 85 47 89 Abnormal Return 11 14 0 1 ON OTN WAAWAAN AN ON O NFO OOO O ONG OU ON eaP POCO SO OU oat te OO ee aa weal ioe Megs sal BE Jane fed oo 51 fale nies Zea aoe NG pk ee OFRFROArRWOMAANFRH OD e e O N e N e o oe t 1 80 3 63 0 89 57 gt gt 55 04 1 47 0 97 1 05 0 55 0 55 0 04 1 47 0 55 0 46 0 97 0 04 0 55 1 0 0 2 06 gt 2 0 0 Positive Negative 11 5 11 5 10 6 lt gt significant at 05 lt lt lt gt gt gt significant at 001 RK Gen Sign 2 06 gt 2 06 gt 1 56 9 Figure 2 7 Sample Eventus output Fifth page Eventus R Software from Cowan Research L C Market Model EW Index Positive Negative Generalized Sign Average Median Day Abnormal Abnormal Return Return 14 1 13 0 30 0 13 0 73 1 39 0 12 0 38 0 60 0 14 0 52 0 08 0 10 0 02 0 16 0 9 1 39 0 40 1 8 6 76 0 92 5 f 0 38 0 0
13. Cowan and Nayar 1991 and Chen Hu and Shieh 1991 Chen Hu and Shieh refer to the test as a binomial sign test For a more detailed explanation of the generalized sign test see Sprent 1989 and Cowan 1992 Rank Test Corrado 1989 describes the rank test for a one day event window The ranks of the abnormal returns of different days are dependent by construction However the effect of ignoring the dependence should be negligible for short event windows Eventus extends the rank test to multiple day windows by assuming that the daily return ranks within the window are independent The rank test procedure treats the combined estimation period and event period as a single set of returns and assigns a rank to each daily or monthly etc return for each firm Let Ay represent the rank of abnormal return Aj in the sample of D Ej abnormal returns of stock j E is the number of non missing returns of stock 7 in the event period if there are no missing returns E E POST PRE 1 and D D ESTLEN Rank one signifies the smallest abnormal return The mean and median rank across Corrado and Zivney 1992 analyze a similar sign test 3Eventus does not require that the estimation and event periods be contiguous The estimation and event period returns used for the other computations in the current event study are used for the rank test also 112 the combined estimation and event period is D E 1 K The r
14. For single event date event studies only either EVENTDAT SAS date variable containing day 0 or CRSPDAY integer representing trading date 0 as a sequence number corresponding to the CRSP calendar The integer may represent the day week month quar ter or year from the beginning date of the index file For TWIN event studies and for data retrieval using RETURNS or PRICES the two SAS date variables EVENTDA1 and EVENTDA2 must be present or if DATEFMT CRSP is in effect then the variables CRSPDAY1 and CRSPDAY2 must be present However if the NDAYS option is specified then omit EVENTDA2 or CRSPDAY2 If you use EST SPECIFIC see below either ESTEND SAS date variable or CRSPEST CRSP day week or month number must also be present on the data set If you specify ID variable see below the variable you name If GROUP group_variable is specified then the name given in place of group_variable must be present If GRWEIGHT is specified then a numeric variable named GRWEIGHT containing the weight of the firm event within the group must be present If the NAME option appears a character variable of length 33 NAME must be present If the SHORT option is specified a character variable of length one named SL must have a value of either S or L If DATEFMT CRSP see above is in effect the event day or month number CRSPDAY or for a TWIN event study the beginning and ending day or month numbers CRSPDAY1 a
15. Using Daily Stock Returns The Case Of Event Studies Journal of Financial Economics 1985 14 1 3 31 163 Cantrell Steve Michael T Maloney and Mark Mitchell On Estimating the Variance of Abnormal Stock Market Performance working paper 1989 Clemson University Chandra Ramesh and Bala V Balachandran A Synthesis of Alterna tive Testing Procedures for Event Studies Contemporary Accounting Research 1990 6 2 611 640 Chandra Ramesh Shane Moriarity and G Lee Willinger A Reexami nation of the Power of Alternative Return Generating Models and the Effect of Accounting for Cross Sectional Dependencies in Event Stud ies Journal of Accounting Research 1990 28 2 398 408 Chen Haiyang Michael Y Hu and Joseph C P Shieh The Wealth Effect of International Joint Ventures The Case of U S Investment in China Financial Management 1991 20 4 31 41 Corhay A and A Tourani Rad Conditional Heteroscedasticity Adjusted Market Model and an Event Study Quarterly Review of Economics and Finance 1996 36 4 529 538 Collins Daniel W and Warren T Dent A Comparison of Alternative Methodologies used in Capital Market Research Journal of Accounting Research 1984 22 1 48 84 Corrado Charles J A Nonparametric Test for Abnormal Security Price Performance in Event Studies Journal of Financial Economics 1989 23 2 385 396 Corrado Charles J and Terry L Zivn
16. matched see the SPORT option decile portfolio return is to be used as the market index The results labeled market adjusted returns correspond to what were traditionally called CRSP Excess Returns However the risk matched decile portfolio returns are used with all event study methods For example with the EXCESS option market model event study results use risk matched decile portfolio returns in place of a broad market index and the returns statement with the index option reports raw stock returns with the risk matched decile portfolio return labeled as the index return ESTINTER The ESTINTER option indicates that the event study will use two different return intervals one for the estimation of benchmark model pa rameters and one for the event period For example to conduct a daily return event study using market model parameters estimated from monthly returns specify ESTINTER MONTH No option is needed to specify the return interval for the event period in this illustration because daily is the default return interval To use the ESTINTER option with CRSP data you must subscribe to a CRSP stock database with the desired estimation period and event period return computation intervals To run a daily event study with monthly parameter estimation you need both daily and monthly CRSP stock databases When working with a CRSP SFA legacy format database a filename crspes statement should point to the main stock data fi
17. see page 51 below is available with either wide or tall output to force Eventus to use additional pages to avoid the limitation on the number of days 45 For example a line of the request file could look like this 80234H10 950920 Spacing does not matter as long as the CUSIP comes first then the event date with at least one space between The rest of this section describes REQUEST statement options for the pro cessing of dates and market and stock returns and for the construction of the estimation period A SAS filename statement or the operating system must associate the fileref REQUEST with the request file Including an identification variable Each line of your request file may include an identification variable following the event date Eventus will read it and print it on subsequent output if you name it with ID ID ID works or you can choose another name Use IDFMT to tell Eventus what format to use for reading and printing the identifying variable IDFMT 4 means a 1 4 digit integer while IDFMT 4 means a four character string Other lengths and other SAS formats are permitted also With IDFMT 4 specified a line of your request file might look like this 16372210 19840920 AY4E Using date formats other than YYMMDD The REQUEST statement allows you to specify the format of dates in your request file You may list calendar dates in nearly any conventional format If you use either six or eight digit YyYMMDD the default
18. you need not specify DATEFMT Besides MMDDYY and DDMMYY you can use the SAS date format DATE An example of a date in DATE format is 19OCT1987 Options for constructing the estimation period The estimation period is the time period used for running market model regressions computing comparison period mean returns and so on EST and POOL By default Eventus determines the estimation period for each firm by subtracting 46 trading days from the event date in your request file The resulting date becomes the last day of the estimation period If you want the estimation period moved back say to 90 days before the event date specify EST 90 If you need an estimation period following the event 46 date specify a positive number EST 61 the plus sign is optional gets you an estimation period that begins on day 61 The default estimation period for weekly event studies ends with week 10 for monthly event studies the default is to end the estimation period with month 13 EST has no effect on the length of the estimation period The ESTLEN option described next changes the estimation period length ESTLEN By default the estimation period is 255 days long when using daily data and 52 weeks or 60 months long when using weekly or monthly data You can change the estimation period length with ESTLEN The largest number you can use is 999 and the smallest is 3 Eventus assumes that you are specifying the number of months for a m
19. your system s rules TRADES Specifies that the number of trades on each date should be read from the Supplemental Nasdaq Data Arrays and included in the output file VSAS Creates a vertical format SAS data set A SAS data set cannot be read by non SAS software The following variables are included in the data set PERMNO CUSIP if the CUSIPERM option is in effect CRSPDAY1 DATE VOLUM SHARES when applicable and any identifying variable you list on the REQUEST statement CRSPDAY1 is the CRSP day number for the beginning date you request DATE is the actual date of the volume figure recorded as a SAS date variable VOLUM is the trading volume in shares and SHARES is the number of shares outstanding in thousands There is one observation in the SAS data set for each trading day in the interval you request Use the VSAS option if you plan to process the data with a SAS program that uses BY variable techniques You must specify the SAS data set name using the OUTSAS option 157 B 13 WINDOWS Statement A single WINDOWS statement may precede the EVTSTUDY and OLDSTUDY state ments and must precede the EXTRACT statement For a single event date event study use WINDOWS to list up to 99 event windows for which cumulative or compounded abnormal returns and test statistics are to be reported on the output The earliest and latest possible dates are determined by the value of the PRE and POST options respectively If the WINDOWS s
20. 001179 840925 56480910 Manufactur 0 028455 0 003405 840926 56480910 Manufactur 0 005217 0 003868 840927 57 Figure 4 5 The Eventus sample listing and input report O Hara and Shaw JF Dec 1990 pp 1587 1600 Replication of Table III C Results of Daily Stock Returns Input Bankname Bankameric BankersTru BankofBost ChaseManha ChemicalNY Citicorp CrockerNat Equimark FirstChica FirstInter FirstPenn FirstWisco Interfirst Irving JPMorgan Manufactur MarineMidl MellonNati Republicba RepublicNY SecurityPa WellsFargo CUSIP 06605010 06636510 06071610 16161010 16372210 17303410 22682210 29443250 31945510 32054810 33607210 33761010 45891610 46371210 61688010 56480910 56828710 58550910 33616010 76071910 81482310 94974010 Event Date 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 09 20 84 Esti mation Period Returns lt 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 50 Insolvent Bank Sample Event Period Returns lt 11 11 11 11 t4 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 14 11 58 Figure 4 6 The Eventus parameter estimate listing O Hara and Shaw JF Dec 1990 pp 1587 1600 3 Replication of Table III C Insolvent Bank Sample Parameter Es
21. 19 FACTORS option 76 only allowed in NONCRSP mode 76 firm by firm results exporting 63 for further analysis 63 printing 24 storing 31 GARCH errors 26 Garven James 26 50 generalized sign test 29 37 53 112 Giaccotto Carmelo 29 53 group weight variable 147 grouping variable 17 123 147 Haw In Mu 105 holidays 20 96 Holthausen Robert W 111 Hu Michael Y 112 industry codes 24 88 interevent CAAR 37 iterated JGLS 76 iterated SUR 76 132 and PACKAGE 31 54 jackknife test 29 53 132 example 67 joint generalized least squares 76 132 options 135 REQUEST required before 145 EXTRACT statement BUYHOLD 67 137 138 CDCSI 137 EXTEND 66 137 EXTFILE 137 ID 65 137 IDFMT 137 INSAS 64 OUTSAS 137 SERIAL 67 137 TYPE 66 138 VALUE 138 VPREFIX 65 138 WPREFIX 65 Karafiath Imre 29 52 71 108 Kramer Lisa 30 Kroner Kenneth 26 51 Lee D Scott 108 Leftwich Richard W 111 Lilien Steven B 105 Linn Scott C 105 log transformed returns 22 multiperiod 136 MacKinlay A Craig 1 Mais Eric 108 Malatesta Paul H 71 Maloney Michael T 108 172 Mann Steven V 108 market adjusted returns 25 49 66 110 market closed 20 market index excluding dividends 20 149 Nasdaq Composite 20 151 Standard amp Poor s 20 market maker count 91 156 market model abnormal return definition 103 estimation 5 GARCH 26 Scholes Williams 26 multi factor 76 standardized abnormal return 106 Marr M Wayne 26 MAXMISS op
22. 76 Poulsen Annette B 28 52 109 111 135 140 Prabhala N R 1 PRE option 27 51 76 PRE95 option 129 precision weighted returns 107 prices bid low and ask high 89 Nasdaq bid and ask 89 split adjusted 143 PRICES statement 141 default output 144 options BIDASK 89 141 BINARY 87 141 DISTRIB 90 141 EXTFILE 92 141 HSAS 88 141 NMS 89 142 NOCLOSE 90 143 OUTSAS 92 143 SHARES 90 143 SHRCODE 89 143 SIC 143 SPLITADJ 91 143 TRADES 91 144 VSAS 88 144 REQUEST required before 145 printed output orientation 16 45 129 suppressing 24 48 49 75 132 printing individual firm results 24 PROC REG 65 69 Rad A Tourani 26 50 rank test 29 53 134 RAW option 25 50 raw returns 25 50 66 83 110 REG procedure SAS 65 69 regression cross sectional 65 example 69 REQFILE option 121 155 request file 3 16 19 39 45 46 61 73 85 86 96 124 146 150 155 event parameter approach 72 example 4 33 for CUSIPERM 100 for DATECONV 94 125 for PRICES 84 for RETURNS 84 for VOLUME 84 how to format 150 more than two 129 SAS data set as 147 second 149 150 sorting 4 with CUSIPERM 121 with NDAYS 149 REQUEST statement 174 options AUTODATE 16 20 84 86 145 AUTODATE BACK 145 COMPOSIT 20 87 151 CUSIP 145 CUSIPERM 17 145 DATEFMT 20 46 73 86 145 EST 22 46 74 146 ESTLEN 22 47 74 GROUP 17 147 GRWEIGHT 147 ID 17 46 73 85 IDFMT 85 INSAS2 149 IX2Y 149 LOG 22 136 149 MINESTN 23 75 149
23. DATECONV 93 paired 37 VAXFMT 130 shifting 150 WEEKLY 93 130 event parameter approach 71 133 EVTSTUDY statement 5 23 25 41 48 EVENTDA1 148 49 50 75 130 139 EVENTDA2 148 and EXTRACT 65 EVENTDAT 148 and OLDSTUDY 40 events options CAARs between two 37 ALLDAYS 28 51 130 EventStream software 43 BOOT 30 130 EVENTUS statement BOTH 25 50 136 SASCRSP 129 options BTAIL 130 139 ACCESS97 126 BUYHOLD 30 53 130 ANNUAL 126 CDCSI 29 52 130 CHAR 126 CP 25 50 131 CHAR4 126 CSECTERR 28 51 131 DBFNSTMT 126 DETAIL 25 49 131 DUAL 126 EGARCH 131 ELIBNAME 16 127 ESTINTER 15 115 127 EXCESS 15 40 84 127 FBIN 127 GETDATA 83 127 HOSTBIN 127 IDXLEAD 127 LIBNAME 128 MONTHLY 13 83 93 128 NASDAQ 128 NLIBNAME 90 128 NONCRSP 45 128 PAGE 16 45 129 PRE95 129 QUARTER 129 REQFILES 129 EGLS 29 52 131 FACTORS 76 131 GARCH 131 INSAS 131 ITSUR 76 132 JACKNIFE 29 53 132 MAXMISS 28 132 MEDIAN 30 53 132 NOMAR 25 50 NOMM 25 50 NONAMES 24 132 NOPLIST 24 49 132 NOPRINT 132 NOSTD 25 50 NUMFM 133 OLSPARAM 76 133 OUTSAS 31 54 133 OVERLAP 28 51 133 PACKAGE 31 54 133 POST 27 51 76 134 PRE 27 51 76 134 RANKTEST 29 53 134 RAW 25 50 SERIAL 29 52 108 134 SHRCODE 135 SIC 24 135 SKIP 135 SPORT 26 135 STDALL 25 50 135 STDCSECT 28 52 135 STDONLY 25 50 SUR 76 135 SW 26 50 104 135 TAIL 31 54 135 TIMEUNIT 26 136 VALUE 25 50 76 136 Eyssell Thomas H
24. Data Arrays of the CRSP database which report closing bid and ask quotes and number of trades When you specify both the BIDASK and NMS options Eventus attempts to supply as many true bid and To use the NMS option with SFA format CRSP data files a SAS filename statement or the operating system must associate the fileref CRSPNMS with the Supplemental Nasdaq File If your system permanently stores the CRSP stock databases in SAS data sets do not use the filename statement Instead use a libname statement to point to the location SAS data library of the Supplemental Nasdaq File and give the libref using the NLIBNAME option on the EVENTUS statement If it is the same libref as in the LIBNAME specification you still must repeat it in the NLIBNAME option 89 ask quotations instead of intraday high and low transaction prices as possi ble Eventus reports bid and ask prices from the Supplemental Nasdaq Data Arrays when they exist and secondary price data from the main time series arrays otherwise Nasdaq stocks not represented in the Supplemental Nas daq Data Arrays generally have closing bid and ask quotations in the main time series arrays If the sample contains both Nasdaq and exchange listed stocks the BIDASK NMS option combination is likely to result in a mixture of bid ask quotations and intraday high low transaction prices To prevent Eventus from reading the closing transaction primary prices in addition to the secondary pric
25. GARCH or EGARCH 1 1 specify the EGARCH option With either the GARCH or EGARCH option maximum likelihood estimates using the quasi Newton algorithm are produced Eventus reports the alpha and beta estimates as it does in the case of ordinary least squares but does not report the estimated parameters of the conditional error variance model No more than 40 iterations will be performed for each stock In general convergence will be better the longer the estimation period ESTLEN and better with the EGARCH 1 1 than the GARCH 1 1 model Cowan and Sergeant 1996 report that event study test specification and power are insensitive to the use of Scholes Williams versus OLS estimation Corhay and Rad 1996 and Brockett Chen and Garven 1999 discuss the potential benefits of GARCH estimation in event studies Bollerslev Chou and 50 Kroner 1992 provide an overview of GARCH EGARCH and related models in finance Specifying the number of days or months in the event period EVTSTUDY computes and reports abnormal returns for the event period which is defined by default as days or weeks 30 through 30 or months 12 through 12 To change from these defaults specify the number of days before and after on the EVTSTUDY statement using PRE and POST These options may be used singly or in combination For example PRE 60 means that the abnormal returns are to start with day 60 unless POST is also specified the latter retains its defa
26. Prior to Mergers The Effect of Capital Structure Accounting Review 1990 65 2 432 451 Karafiath Imre Using Dummy Variables In The Event Methodology F i nancial Review 1988 23 3 351 358 Karafiath Imre and David E Spencer Statistical Inference in Multiperiod Event Studies Review of Quantitative Finance and Accounting 1991 1 4 353 371 Kramer Lisa Alternative Methods for Robust Analysis in Event Study Applications working paper Simon Fraser University 2000 Available from the World Wide Web http www sfu ca kramer Papers event pdf Lee D Scott Management Buyout Proposals and Inside Information Journal of Finance 1992 47 3 1061 1079 Linn Scott C and John J McConnell An Empirical Investigation of the Impact of Antitakeover Amendments on Common Stock Prices Journal of Financial Economics 1983 11 1 361 399 MacKinlay A Craig Event Studies in Economics and Finance Journal of Economic Literature 1997 35 1 13 39 Mais Eric L William T Moore and Ronald C Rogers A Re Examination of Shareholder Wealth Effects of Calls of Convertible Pre ferred Stock Journal of Finance 1989 44 5 1401 1410 165 Malatesta Paul H Measuring Abnormal Performance The Event Pa rameter Approach Using Joint Generalized Least Squares Journal of Financial and Quantitative Analysis 1986 21 1 27 38 Mann Steven V and Neil W Sicherma
27. RETN1 through RETNnnn MKT1 through MKTnnn if you specify INDEX on the RETURNS statement PERMNO EVENTDA1 EVENTDA2 and any identifying variable that you list on the REQUEST statement If you specify the BOTH option value weighted in dex returns will be stored as VWMK1 through VWMKnnn You must specify the SAS data set name using the OUTSAS option EVENTDA1 and EVENTDA2 are SAS date variables representing the begin ning and ending trading days of the interval you requested The one to three digit number nnn is the maximum number of trading days in any of the intervals you requested For example if the number of trading days you ask for ranges from 2 for some firms to 90 for others nnn is 90 For a firm in this example with only two trading days RETN3 through RETN90 contain missing values Because the SAS system stores a large amount of information on each variable for its own use a horizontal 152 format data set occupies considerably more disk storage space than a vertical format data set INDEX Includes market index returns in the file The default is not to include market index returns OUTSAS Specifies a two level SAS name libref membername under which to create the SAS data set containing the returns This is valid only with the HSAS or VSAS option SHRCODE Causes Eventus to store the two digit share code from the CRSP database s name history array as the variable SHRCODE in any OUTSAS data set SIC Causes Eventus to
28. SHORT WINDOWS begin end begin end EVTSTUDY NONAMES NOPLIST SIC DETAIL DETAIL FULL RAW CP NOMAR NOMM NOSTD STDONLY STDALL VALUE BOTH SPORT SW GARCH EGARCH TIMEUNIT periods PRE periods POST periods OVERLAP MAXMISS n CSECTERR STDCSECT EGLS CDCSI j SERIAL RANKTEST JACKNIFE BOOT MEDIAN BUYHOLD TAIL 1 2 BTAIL 1 2 OUTSAS libref membername PACKAGE specifier 14 EXCESS This option requires a separate add on subscription to the CRSP Ex cess Returns or CRSP Index File Portfolio Assignments File With CRSPAc cess format and the add on subscription installed the database includes risk class decile numbers for individual stocks that Eventus uses to select appropri ate size decile portfolio returns To do this the plain text ASCII or EBCDIC character versions of the Indices subscription files must be accessible Specif ically Eventus needs the files dsbo dat and dsbc dat to be associated with the SAS filerefs statidx1 and statidx2 This is done by submitting from the Editor window or including in a batch program file statements similar to the following where crsp indices should be changed to the full Unix Windows or OpenvMS specification for the directory or folder containing the character indices files filename statidx1 crsp indices dsbc dat filename statidx2 crsp indices dsbo dat The EXCESS option indicates that the beta matched or standard deviation
29. Significance symbols for the default generalized sign test or optional rank test still are reported Computing buy and hold compounded window returns To obtain buy and hold abnormal returns for multiperiod windows spec ify BUYHOLD on the EVTSTUDY statement The option applies only to non standardized abnormal return methods By default Eventus reports non standardized results for each abnormal return method used in an event study and standardized results in addition for the market model abnormal returns Eventus computes buy and hold abnormal returns by compounding successive daily or other period raw returns and market index returns then adjusting the raw returns according to the abnormal return method used Comparison period mean returns and market model alphas are adjusted for the window length The Eventus output labels the mean buy and hold window returns Average Compounded Abnormal Return while the default additive win dow abnormal returns appear as Cumulative Average Abnormal Return 53 Reporting one instead of two tailed tests Eventus reports all significance levels for one tailed tests by default TAIL 2 changes to two tailed tests Saving results to a file Normally none of the event study results are saved in a disk file To store firm by firm results specify the SAS file to store them in using the OUTSAS option The libref parameter specifies the libref of the SAS data library in which to store t
30. Users who want to write their own programs to process event study data saved by Eventus need to know the variable names used Data sets created with the OUTSAS option on the EVTSTUDY statement rep resent each day by one or more variables Any variable ending in the number 255 represents the last day of the estimation period The first estimation period variable in a series depends on the length of the estimation period If the estimation period is 100 trading days long for example the first day is represented by a variable ending in the number 156 The sequence of vari able numbers is in ascending time Thus variable number 255 contains a data item for day 46 if EST 46 is specified on the REQUEST statement If EST 91 ESTLEN 100 were specified variable number 255 would represent day 190 Variable names ending in the numeral 256 represent the first day of the event period regardless of estimation period length For example if PRE 20 is specified then variable number 256 corresponds to day 20 Table A 2 presents the possible range of variable names in an output event study data set OUTSAS data set Only those variables needed to represent days as described above are included in an actual data set _WEIGHT_ is a five byte character variable It will have a value of Equal Value SP500 NASDQ depending on the market index used When both the equally weighted index and another index are used there are two 116
31. a base event date day or month 0 as in the conventional approach then creating a dummy variable for each window listed on the WINDOWS statement Since a window can consist of one or more dates relative to day 0 the researcher has considerable flexibility in defining the dummy variables 6 1 Statements for the Event Parameter Ap proach Figure 6 1 displays the Eventus statements to run a single event date event study The options that are specific to the event parameter approach are 71 Figure 6 1 Eventus statements for an event parameter approach event study EVENTUS NONCRSP MONTHLY PAGE WIDE TITLE text TITLE2 text 3 REQUEST CUSIP CUSIPERM ID variable IDFMT format DATEFMT MMDDYY YYMMDD DDMMYY DATE EST value value POOL ESTLEN n WINDOWS begin end begin end EVTSTUDY NONAMES OLSPARAM SUR ITSUR FACTORS n VALUE PRE periods POST periods described below The EVENTUS statement After any needed filename or libname statements an Eventus program starts with an EVENTUS statement The event parameter approach can be used either with direct CRSP database retrieval or with the NONCRSP option For more details on direct CRSP access and the options that can be used with it please see Chapter 3 For more information on event studies with non CRSP data please see Chapter 4 The REQUEST statement The REQUEST statement instructs Eventus to read the request file and
32. a single equally weighted portfolio The value of the grouping variable for each observation is listed on the appropriate line of the request file after the dates and ID variable if any The grouping variable must be an integer between 0 and 9999 inclusive leading zeros in the request file are optional and ignored The grouping variable is included in the converted request file GRWEIGHT Valid only if the GROUP option is specified Denotes that the request file contains a group weight variable This variable expressed as a decimal specifies the weight to be given the individual stock within its group portfolio All the weights for a single group should sum to 1 This option is included on the DATECONV statement for compatibility with the REQUEST statement weights are copied to the output file ID variable Optionally names a variable to be used as an observation iden tifier The identifying variable may be of any data type If INSAS is specified the identifying variable must exist on the permanent SAS data set specified IDFMT format Gives the format of the identifying variable in the external request file For example if the identifying variable is an integer that varies from one to four digits specify IDFMT 4 0 INSAS libref membername Used in place of REQFILE when the input re quest file is a SAS data set The input SAS data set must contain these variables e CUSIP 8 character string Required if and only if CUSIP or CUSIPERM
33. and Partch 1988 include Mais Moore and Rogers 1989 Cowan Nayar and Singh 1990 Mann and Sicherman 1991 and Lee 1992 Simulation evidence of the properties of the corrected and un corrected Patell tests appears in Karafiath and Spencer 1991 using Monte Carlo experiments and Cowan 1993 using sampling experiments with CRSP data Both papers report that the bias in the uncorrected test is small in event windows shorter than 60 days but serious in event windows longer than 100 days Mikkelson and Partch 1988 acknowledge Craig Ansley for the original derivation of the corrected test statistic in an event study context For other derivations and discussion see Cantrell Maloney and Mitchell 1989 and Sweeney 1991 108 When the SERIAL and STDALL options both appear on the EVTSTUDY statement Eventus uses the following definitions for the standardized method tests with non market model abnormal returns For comparison period mean adjusted returns 2 eee L Li SCART T2 SA J D For raw returns and market adjusted returns there is no estimation of the mean Instead the mean is assumed to be equal to a known constant with probability one The constant is zero in the case of raw returns and the realized market index return in the case of market adjusted returns Thus 2 2 SCART 5 7 SA Lj s Standardized Cross Sectional Method Eventus uses the standardized cross sectional method for market model ab normal
34. are to be used Eventus does not try to search a CRSP database when NONCRSP is in effect PAGE TALL WIDE The default is PAGE TALL which causes printed out put to be formatted with the SAS system option LS 79 for portrait orientation PAGE WIDE causes output to be formatted with the SAS 128 system option LS 132 for landscape orientation this requires a small font or a wide carriage printer PRE95 Specifies that the ending date of the input CRSP database is before 1995 This is not needed and has no effect with SFA character format CRSP database QUARTER This tells Eventus that it is reading quarterly returns from a USERSTOK file also see the NONCRSP option or from a quarterly returns database formatted like CRSP stock databases REQFILES n An extension of the DUAL option for event studies this option indicates that there is more than one CRSP stock file with a request file corresponding to each The filerefs of the files must be CRSPSTOK CRSPST2 CRSPSTn The user must supply corresponding request files with filerefs REQUEST REQUEST2 REQUESTn If n 2 the request files can be SAS data sets in this case the filerefs would not be used The value of n may be from 2 to 9 If neither DUAL nor REQFILES is specified Eventus looks for only one request file and one CRSP stock file SASCRSP 0 Indicates that the CRSP data are in the form of SAS data sets Specifying 0 turns the option off Normally set at Eventus inst
35. by PERMNO the UPCUSIP statement is becoming obsolete It still is available for users who may be working with 1993 or earlier releases of the CRSP stock files The statement reads the user request file and performs the updating operation The user may specify these options COLUMN n Use when the CUSIP is not the first item on each line of the request file Substitute the starting column for the CUSIP for n This option is not needed if only blanks precede the CUSIP on each line 154 EXTFILE Gives the fileref of the external file in which Eventus is to store the updated copy of the request file The default is userdata On mainframe systems where applicable the file should have a disposition of NEW or OLD and a logical record length of 80 REQFILE fileref The request file for UPCUSIP must be an external non SAS formatted file such as a card image format file Replace fileref with the fileref associated with your file The REQFILE specification may be omitted if the fileref of the request file is REQUEST The file need not be sorted by CUSIP B 12 VOLUME Statement Reads trading volume from the CRSP stock files Eventus stores the volume data in your choice of disk file Choose from the following options BINARY This is identical to the TEXT format except that the volume data are written in real binary FLOAT4 RB4 or S370RB4 depending on the system format This uses only one third as much disk space for storing volume data as t
36. dates in nearly any conventional format If you use either six or eight digit YYMMDD the default you don t need to specify DATEFMT Besides MMDDYY and DDMMYY you can use the SAS date format DATE CRSP Trading Day Numbers To convert CRSP day numbers back to calendar dates specify DATEFMT CRSP on the DATECONV statement AUTODATE If some of the calendar dates in your request file may be non trading days specify AUTODATE on the DATECONV statement AUTODATE tells Eventus to convert automatically all calendar dates to trading days Non trading days are converted to the following trading day For example a Saturday would be changed to the following Monday or Tuesday if Monday were a holiday To convert non trading dates to the previous trading date instead of the next specify AUTODATE BACK The AUTODATE option has no effect when converting between calendar dates and CRSP month numbers because all months within the range of the CRSP calendar are trading months An example of a date in DATE format is 190CT1987 96 NDAYS Specify NDAYS n if you want create a second date to be the same fixed number of trading days from the first date for every firm in your request file Specify TWIN on the EVENTUS statement but omit the second date from your request file if you use NDAYS SHIFT1 and SHIFT2 SHIFT1 and SHIFT2 shift the dates from the request file by a specified number of days When the request file contains calendar d
37. does not make the log transformation Options for constructing the estimation period The estimation period is the time period used for running market model regressions computing comparison period mean returns and so on EST and POOL By default Eventus determines the estimation period for each firm by subtracting 46 trading days from the event date in your request file The resulting date becomes the last day of the estimation period If you want the estimation period moved back say to 90 days before the event date specify EST 90 If you need an estimation period following the event date specify a positive number EST 61 the plus sign is optional gets you an estimation period that begins on day 61 The default estimation period for weekly event studies ends with week 10 for monthly event studies the default is to end the estimation period with month 13 If you need to split your estimation period between pre and post event dates specify POOL on the REQUEST statement Then Eventus will chop your estimation period into two equal halves For example REQUEST EST 50 POOL gets you an estimation period of which the first half ends with day 50 and the second half begins on day 50 EST has no effect on the length of the estimation period The ESTLEN option described next changes the estimation period length ESTLEN By default the estimation period is 255 days long when using daily data and 52 weeks or 60 mo
38. event study normally Eventus will tell you which observations have non trading dates You can then make the necessary corrections and re run the event study to have the observations included Your second choice is to specify AUTODATE on the REQUEST statement AUTODATE tells Eventus to convert automatically all calendar dates to trading days Non trading days are converted to the following trading day For example a Saturday would be changed to the following Monday or Tuesday if Monday were a holiday To convert non trading dates to the previous trading date instead of the next specify AUTODATE BACK Market index options NODIVIDX Eventus normally uses the returns including dividends of the basic equally weighted and value weighted indexes in the CRSP database Specify NODIVIDX to instruct Eventus to use the index returns excluding dividends SP500 and COMPOSIT The CRSP NYSE AMEX Nasdaq database reports the Standard and Poor s 500 Composite Index in addition to the equally weighted and value weighted indexes of all stocks Specify SP500 to tell Eventus to read the Standard and Poor s index return instead of the value weighted index A CRSP SFA database containing only Nasdaq data may report the Nasdaq Composite Index return instead of the Standard and Poor s index To use 3An example of a date in DATE format is 190CT1987 20 the Nasdaq Composite Index instead of the value weighted index specify COMPOSIT Eventus does not det
39. filename userdata statement before the EVENTUS state ment to give the path and name of the file to be created To use a different fileref besides userdata specify the fileref with the EXTFILE option For maximum safety do not specify the fileref request or the input request file may be overwritten You may also have Eventus store the converted dates in a SAS data set Specify the two part SAS data set name libref membername using the OUTSAS option 98 Chapter 9 Converting CUSIP Identifiers Using CUSIPERM CUSIPERM converts 8 character common stock CUSIP values to the correspond ing CRSP permanent identification number or PERMNO Before 1995 when the 1994 stock files appeared CRSP sorted its files by the CUSIP at the end ing date of the file or if the stock went off earlier for example due to the company being taken over the last CUSIP it had This resulted in CUSIPs that could change from one annual edition of the files to the next The CRSP stock databases currently are sorted in PERMNO order The purpose of the CUSIPERM statement is to convert CUSIPs from any previous edition of the CRSP stock files to PERMNOs Whoever takes care of Eventus at your site should run the PermnoUp program supplied with Eventus to generate databases for CUSIPERM after installing Eventus and after each annual or quarterly update of the CRSP stock database This chapter describes the stand alone version of CUSIPERM which re quires its own E
40. included in the output file This option is ignored unless either NDAYS 1 is specified on the REQUEST statement or VSAS is specified on the PRICES statement The shares outstanding data may not be as timely as the price data refer to CRSP documentation for more information SHRCODE Causes Eventus to store the two digit share code from the CRSP name history array as the variable SHRCODE in any OUTSAS data set SIC Causes Eventus to read the SIC code from the CRSP database and to add the variable SICCODE to any output SAS data set SPLITADJ Causes Eventus to adjust for stock splits reverse splits and stock dividends using distribution data from the CRSP database The adjust ment takes place only within the range of data being extracted for each stock For splits and stock dividends occurring after the first date extracted Eventus multiplies prices and cash distributions by a split factor and divides shares outstanding by the same factor The split factor is equal to 1 0 on the first date being extracted and is cumulative within the range of data being extracted The split factor increases by a factor of 1 the CRSP factor to adjust price each time there is a 143 split or stock dividend The split factor variable SPLFAC is added to any output SAS data set The SPLITADJ option is not supported with the HSAS output option TEXT This is the default Eventus writes the prices on a file that people not just computers can read The
41. specified on the EVENTUS and REQUEST statements The file need not be sorted by PERMNO If the CUSIPERM option or the CUSIP option is specified the first variable on the line should be CUSIP instead of PERMNO REQFILE2 fileref Used to specify the fileref of the second request file when DUAL is specified on the EVENTUS statement the default is REQFILE2 REQUEST2 SHIFT1 n SHIFT2 n The SHIFTn options are intended primarily for the DATECONV statement and for the REQUEST statement in a RETURNS or PRICES program but may also be used on the REQUEST statement in a TWIN EVTSTUDY program The first date in the request file is shifted by n periods and the second date is shifted by ng periods For the monthly file the periods are months For the daily file the periods are trading days if DATEFMT CRSP and calendar days otherwise Both n and ny may be specified as any integer value For example SHIFT1 1 shifts June 1 2000 back to May 31 2000 SHIFT1 and SHIFT2 may be specified singly as well as together Using these options with calendar dates may result in invalid date messages 150 unless AUTODATE is also specified If the researcher needs to shift by trading days rather than calendar days it may be necessary to convert to CRSP trading day numbers using DATECONV first SHORT Specifies that an S for short position or L for long position code appears at the end of each line in the request file When S appears all raw stock return
42. splits and stock dividends using distribution data from the CRSP database The adjust ment takes place only within the range of data being extracted for each stock For splits and stock dividends occurring after the first date ex tracted Eventus multiplies share volume and cash distributions by a 156 split factor and divides shares outstanding by the same factor The split factor is equal to 1 0 on the first date being extracted and is cu mulative within the range of data being extracted The split factor increases by a factor of 1 the CRSP factor to adjust price each time there is a split or stock dividend The split factor variable SPLFAC is added to any output SAS data set The SPLITADJ option is not supported with the HSAS output option TEXT This is the default Eventus writes the prices on a file that people can read The data are arranged in columns with each row reporting the PERMNO identification variable if applicable crsp date and volume All the data for the first stock are listed with each trading day on its own row followed by all the data for the second stock and so on Specify the fileref a SAS file shortcut or on mainframe systems a DDname to which to write the prices with the EXTFILE option on the VOLUME statement On a mainframe operating system that defines record lengths and blocks the file should have a logical record length of 80 or longer a fixed block format and a block size that conforms to
43. standardized Unlike the standardized abnormal return method the time series standard deviation method calculates a single variance estimate for the entire portfolio Therefore the time series standard deviation method does not take account of unequal return variances across securities On the other hand it does avoid the potential problem of cross sectional correlation of security returns The estimated variance of AAR is Tp E gt AAR AAR 22 t T O AAR D 2 t where the market model parameters have been estimated over the estimation period of D Tp Tp 1 days and The portfolio test statistic for day t in event time is t AAR Gar 110 Assuming time series independence the test statistic for CAAR7 7 is CAAR aT oak t Many studies use the time series standard deviation method see for ex ample Dopuch Holthausen and Leftwich 1986 and Brickley Dark and Weis bach 1991 Cross Sectional Standard Deviation Method When the CSECTERR option appears on the EVTSTUDY statement Eventus substitutes a daily cross sectional standard deviation for the portfolio time series standard deviation in the non standardized tests The portfolio test statistic for day t in event time is po AAR sar VN where T wee T 3 5 _ Ay A O AAR Noi t np jt The estimated variance of CAARz 7 is 1 x 1X 2 TCAART To NI E D CARIT To N a CARIT T The test statistic for
44. the potential benefits of GARCH estimation in event studies Bollerslev Chou and Kroner 1992 provide an overview of GARCH EGARCH and related models in finance Using multiperiod returns as the basic unit of analysis To combine each consecutive n days weeks or months into a single time unit for purposes of estimation and testing specify TIMEUNIT n For example in an event study using the CRSP daily stock database TIMEUNIT 2 specifies that two day returns are to be computed and used as if they were daily returns For an illustration of the technique see the lower panel of Table VI in Bhagat Marr and Thompson 1985 26 When n is even period 0 in event time contains day 0 the date in the request file 4 1 days following day 0 and preceding day 0 Additional periods are formed on either side of period zero For example TIMEUNIT 2 results in period zero containing days 1 and 0 period 1 contains days 1 and 2 and period 1 contains days 3 and 2 If n is odd period 0 is centered on day zero TIMEUNIT 3 produces a period 0 containing days 1 0 and 1 a period 1 containing days 2 3 and 4 and so on Eventus combines daily weekly or monthly returns into period returns by addition If the LOG option is specified on the REQUEST statement individual returns are converted to logarithmic form before adding When TIMEUNIT is specified Eventus interprets the WINDOWS statement ar guments and the PRE POST and
45. the CRSP database is sorted by PERMNO and the request file contains PERMNOs CUSIPERM Specifies that the request file contains CUSIPs but does not change the default assumption that the CRSP database is sorted by PERMNO Eventus will attempt to convert CUSIPs to PERMNOs during execution This requires conversion files that normally are created dur ing installation or upgrading If these files have not been created please see the Eventus installation instructions for the required procedure DATEFMT format Specifies the format of the dates in the request file The specification must be either a valid SAS date informat or the word CRSP The word CRSP tells Eventus to look for a 1 to 4 digit integer representing a CRSP trading day or month number 1 July 2 1962 for all the daily files except 1 December 14 1972 for the old format 1985 and 1987 of the Nasdaq files Leading zeroes need not but 145 may be included in the CRSP day number Any format other than CRSP must be a valid SAS date format although the period at the end is optional The default is DATEFMT YYMMDD8 EST periods and POOL The EST option lets you choose the estimation pe riod Eventus uses to estimate the benchmark return parameters for the event study If you specify a negative value Eventus constructs an es timation period ending that number of trading days months before the event date If you give a positive number with or without the plus sign the esti
46. the estimation period for each firm by subtracting 46 trading days from the base date in your request file The resulting date becomes the last day of the estimation period If you want the estimation period moved back say to 90 days before the base date specify EST 90 For an estimation period following the base date specify a positive number EST 61 the plus sign is optional produces an estimation period that begins on day 61 The default estimation period for weekly event studies ends with week 10 for monthly event studies the default is to end the estimation period with month 13 To split the estimation period between pre and post base dates specify POOL on the REQUEST statement Then Eventus will chop your estimation period into two equal halves For example REQUEST EST 50 POOL defines an estimation period of which the first half ends with day 50 and the second half begins on day 50 The POOL option has no particular meaning in a non CRSP job since the researcher must have assembled the estimation period data previously EST has no effect on the length of the estimation period The ESTLEN option described next changes the estimation period length ESTLEN By default the estimation period is 255 days long when using daily data and 52 weeks or 60 months long when using weekly or monthly data You can change the estimation period length with ESTLEN The largest number you can use is 999 and the smallest
47. the window z tests should be corrected for the serial correlation of abnormal returns that is present by construction This option is implied by the STDCSECT option 134 SHRCODE Causes Eventus to store the two digit share code from the CRSP name history array as the variable SHRCODE in any OUTSAS data set SIC Causes Eventus to report the Standard Industrial Classification code from the CRSP name history array and to add the variable SICCODE to any output data set The code is reported on the estimation period statistics listing SKIP n Gives the number of header lines in the USERSTOK external data file Eventus ignores the header lines SPORT When the EXCESS option appears on the EVENTUS statement this option switches from beta based to standard deviation based excess returns or decile index portfolios STDALL Requests that the standardized test statistic be computed for the non market model benchmark s as well as for the market model The STDALL option is ignored in TWIN event studies Any applicable stan dardized test options then affect all benchmarks in use For example the SERIAL STDCSECT and BOOT options affect standardized tests for market adjusted returns as well as market model returns when STDALL is in effect STDCSECT Specifies that the standardized cross sectional test Boehmer Musumeci and Poulsen 1991 be substituted for the Patell z test in the standardized method This option implies SERIAL option STDONLY Ple
48. variable Bankname The specification IDFMT 10 means that the identifying values are up to ten characters long No ID variable is required the CUSIP alone is enough If there is no separate identifying variable ID and IDFMT are omitted 54 O Hara and Shaw 1990 report only market model abnormal returns the option NOMAR on the EVTSTUDY statement suppresses the computation and reporting of market adjusted returns Because the firms are all in the same industry and have a common event date the authors do not use a standardized residual test but base their test statistic on a standard devi ation estimated for the portfolio of sample firms from residual returns in the estimation period See Chandra and Balachandran 1990 and Chandra Moriarity and Willinger 1990 for further discussion of cross sectional de pendencies in event studies We suppress standardized tests with the NOSTD option The NOMAR and NOSTD options are not necessary they only eliminate extra output that is not of interest in this study Because no WINDOWS statement is specified Eventus automatically gener ates multiday windows around the event date to supplement the day by day results The BUYHOLD option on the EVTSTUDY statement produces buy and hold results for the windows around the event date The default when the user does not specify BUYHOLD is to cumulate additively abnormal returns across windows The VALUE option serves only to label the Eventus r
49. version of the CRSP stock file When a change is made only the updated CUSIPs appear in the converted request file The option requires a conversion table a component of Eventus created by the PermnoUp program at Eventus installation and after each update of the CRSP database This option is nearly obsolete now that CRSP stock files are sorted by PERMNO B 3 EVENTUS statement The EVENTUS statement is always required It has the following options ACCESS97 0 Indicates that the input CRSP database is CRSPAccess for mat Specifying O turns the option off Normally set at Eventus in stallation but may be specified at run time ANNUAL Tells Eventus that it is reading annual returns from a USERSTOK file also see the NONCRSP option or from an annual returns database formatted like CRSP stock databases CHAR 0 CHAR4 0 CHAR indicates that the input CRSP database is SFA character format CHAR4 further specifies that the SFA character format database contains four digit years Specifying 0 turns the option off Normally set at Eventus installation but may be specified at run time DBFNSTMT 0 Indicates that input CRSPAccess format database is iden tified by a SAS filename crspdb statement instead of by environment variables or logicals Specifying 0 turns the option off Normally set at Eventus installation but may be specified at run time DUAL Available for event studies only the DUAL option indicates that there are two CRSP SFA
50. 0 0 6 0 96 0 16 0 5 2 90 0 70 2 4 1 22 0 14 1 3 2 08 0 04 1 2 4 65 1 04 3 1 5 16 08 4 0 18 86 15 08 15 1 0 19 0 26 0 2 2 13 0 29 1 3 2 01 0 33 1 4 2 19 0 26 1 5 1 46 0 40 1 6 2 46 0 18 2 7 0 63 0 33 0 8 0 10 0 06 0 9 1 63 0 66 1 10 0 67 0 19 0 11 5 36 0 02 4 12 1 27 0 18 1 13 1 33 0 10 1 14 1 01 0 12 0 Cumulative Average Median Cumulative Days Abnormal Return Abnormal Return 30 2 18 78 11 99 1 0 24 02 17 12 1 30 11 41 4 20 significant at 10 lt lt gt gt significant at 01 6T7 x 32 81 43 03 74 90 x 33 x 83 x 16 79 69 84 23 O7 53 09 37 56 49 x 06 lt 12 35 lt gt significant at 05 RK 16 1 10 6 6 10 t 2 93 4 25 1 75 ONNDWORRFROOR O 1 l e O 73 gt 22 72 28 73 gt 22 72 23 gt 23 gt 22 78 78 lt 0 23 28 72 28 0 22 28 0 22 22 78 23 73 gt Positive Negative 11 5 11 5 12 4 Gen Sign Z 1 73 gt 1 73 gt 2 23 gt lt lt lt gt gt gt significant at 001 10 Figure 2 8 Sample Eventus output Final page Market Model Standardized Residual Method EW Index 11 Days 30 2 1 0 1 30 lt lt gt gt Significant at 01 Average Median Abnormal Abnormal Return Return 1 13 0 30
51. 1 4 0 eventdat yymmdd name 10 run eventus getdata title Returns for Event Parameter Example request insas project request cusiperm datefmt crsp shift1 105 ndays 111 returns index vsas outsas work stockret proc sort data stockret by date run data t_rates read in separate file containing T note yields infile trates input date mmddyy8 treaslyr 5 2 factor2 dif1 treaslyr 100 change in 1 yr T note rate drop treaslyr run data stockret merge stockret in needed t_rates by date if needed keep only matched t note dates keep date permno cusip return market factor2 run proc sort data stockret out project stockret by cusip date run TT Figure 6 3 Request file for the event parameter approach demonstration 00620310 5432 19840216 Adams 04557310 5513 19840613 Associated 14348310 5538 19840719 Carnation 17444010 5630 19841128 Citizens 6 2 An Event Parameter Approach Example The example includes a two factor return generating model The FACTORS option is available only in NONCRSP mode so we first retrieve the needed data from the CRSP database Figure 6 2 shows the Eventus program to extract the returns with additional SAS code to process the second return generating factor The EVENTUS statement needs only the NONCRSP option The program sets up a permanent SAS data library with the libref project The specific methods for establishing a permanent SAS data library vary by system
52. 28 29 51 52 standardized abnormal return 52 106 standardized cross sectional test 28 D2 EVTSTUDY 135 OLDSTUDY 140 STDALL option 25 50 STDONLY option 25 50 stock splits 91 Supplemental Nasdaq File 89 91 142 157 SUR 76 135 SW option 26 50 Swary Itzhak 31 Sweeney Richard J 108 TAIL option 31 54 135 test statistic Boehmer Musumeci Poulsen 28 52 bootstrap 30 Collins Dent 29 52 cross sectional method 111 EGLS 29 52 generalized sign 29 53 112 jackknife 29 53 113 Patell 28 52 109 portfolio 110 rank 29 53 112 serial dependence adjustment 29 92 standardized abnormal return 28 52 109 standardized cross sectional 28 52 Thompson G Rodney 26 time series standard deviation 110 TIMEUNIT option 26 versus WINDOWS 27 trademarks 2 TRADES option 91 trading status 91 156 trading volume 155 TWIN option 37 59 UPCUSIP option 151 UPCUSIP statement 154 options COLUMN 154 EXTFILE 155 REQFILE 155 USERSTOK file 43 using SAS data set 48 131 example 78 VALUE option 25 50 value weighted index 25 50 87 88 136 variable names 116 _WEIGHT_ 116 CRSPDAY 148 CRSPDAY1 148 CRSPDAY2 148 event study OUTSAS data set 117 EVENTDA1 148 EVENTDA2 148 EVENTDAT 148 for group weight 148 in INSAS for non CRSP 131 in INSAS for request file 147 in PRICES OUTSAS data set 144 in RETURNS OUTSAS data set 154 176 in VOLUME OUTSAS data set 157 RESTYPE 118 with DATEFMT CRSP 149 variance increase 28 52 111 VAXFMT optio
53. 6 48071 48354 50024 52863 19043 53858 53903 60839 51772 59379 53209 59109 53938 56805 Name on Event Date MANUFACTURERS HAN CHASE MANHATTAN C CITICORP BANKAMERICA CORP CHEMICAL NEW YORK IRVING BANK CORP CROCKER NATIONAL MORGAN J P amp CO I BANKERS TRUST NY WELLS FARGO amp CO EQUIMARK CORP MARINE MIDLAND BK FIRST CHICAGO COR FIRST PENNSYLVANI SECURITY PACIFIC BANK OF BOSTON CO MELLON NATIONAL C FIRST WIS CORP REPUBLICBANK CORP REPUBLIC NEW YORK INTERFIRST CORP Figure 3 5 The Eventus sample listing and input report O Hara and Shaw JF Dec 1990 pp 1587 1600 Esti mation Period Event Returns Date lt 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 09 20 84 50 Insolvent Bank Sample Event Period Returns lt 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 Reason if no usable returns 35 Figure 3 6 The Eventus parameter estimate listing O Hara and Shaw JF Dec 1990 pp 1587 1600 3 Replication of Table III C Insolvent Bank Sample Parameter Estimates and Estimation Period Statistics Market Residual Mean Model Res Standard SWARYNUM PERMNO Alpha Beta Return iduals gt 0 Deviation 1 48223 0 00102 1 120 0 00285 50 00 0 02192 2 41
54. 718 0 00081 1 095 0 00260 48 00 0 01216 4 47079 0 00019 1 703 0 00297 46 00 0 01388 5 58827 0 00042 1 394 0 00185 52 00 0 01748 6 47896 0 00190 1 433 0 00424 42 00 0 01909 7 46842 0 00103 1 052 0 00275 46 00 0 02000 8 51676 0 00716 0 309 0 00766 44 00 0 03187 9 48071 0 00089 1 042 0 00259 42 00 0 01129 10 48354 0 00273 1 089 0 00451 50 00 0 01107 11 50024 0 00395 0 952 0 00550 52 00 0 01283 12 52863 0 00140 1 132 0 00045 50 00 0 02894 13 19043 0 00321 0 464 0 00396 48 00 0 01535 14 53858 0 00171 1 519 0 00419 44 00 0 01643 15 53903 0 00258 0 833 0 00394 50 00 0 02685 16 60839 0 00196 0 822 0 00330 44 00 0 01171 17 51772 0 00134 1 007 0 00299 48 00 0 01261 18 59379 0 00069 1 125 0 00114 44 00 0 01719 19 53209 0 00210 0 494 0 00291 48 00 0 01944 20 59109 0 00073 0 858 0 00213 44 00 0 01932 21 53938 0 00070 0 830 0 00205 44 00 0 01494 22 56805 0 00478 0 261 0 00436 32 00 0 02361 69 26550 0 00254 0 652 0 00360 44 00 0 01539 MEAN 0 00176 0 939 0 00330 46 00 0 01788 MEDIAN 0 00153 1 025 0 00298 46 00 0 01681 36 residual standard deviation would tend to make the statistic that Eventus reports smaller and so may explain the difference Eventus reports that 10 of the 22 abnormal returns on day 0 are positive which is the same 45 5 fraction that the original study reports The authors provide a significance level for the median abnormal return of 0 516 The article does not seem to define the statistic explicitly Ev
55. 840809 56480910 Manufactur 0 008811 0 001881 840810 56480910 Manufactur 0 008734 0 000570 840813 56480910 Manufactur 0 079295 0 004966 840814 56480910 Manufactur 0 000000 0 008240 840815 56480910 Manufactur 0 023923 0 005825 840816 56480910 Manufactur 0 000000 0 001539 840817 56480910 Manufactur 0 004673 0 003809 840820 56480910 Manufactur 0 014085 0 016093 840821 56480910 Manufactur 0 009259 0 003148 840822 56480910 Manufactur 0 009174 0 001218 840823 56480910 Manufactur 0 000000 0 002024 840824 56480910 Manufactur 0 013889 0 005190 840827 56480910 Manufactur 0 000000 0 005274 840828 56480910 Manufactur 0 018779 0 000454 840829 56480910 Manufactur 0 004608 0 002779 840830 56480910 Manufactur 0 004587 0 001160 840831 56480910 Manufactur 0 013825 0 008776 840904 56480910 Manufactur 0 037383 0 003732 840905 56480910 Manufactur 56480910 Manufactur 0 019417 0 007350 840906 0 014286 0 006375 840907 56480910 Manufactur 0 009390 0 000346 840910 56480910 Manufactur 0 027907 0 001752 840911 56480910 Manufactur 0 013575 0 000353 840912 56480910 Manufactur 0 080357 0 017459 840913 56480910 Manufactur 0 024793 0 006062 840914 56480910 Manufactur 0 004032 0 000738 840917 56480910 Manufactur 0 004016 0 005255 840918 56480910 Manufactur 0 016129 0 003625 840919 56480910 Manufactur 0 028689 0 003045 840920 56480910 Manufactur 0 007968 0 008275 840921 56480910 Manufactur 0 016064 0 002395 840924 56480910 Manufactur 0 004082 0
56. AR WLS weight pair wincar1 weight1 corresponds to the first window listed on the last WINDOWS state ment 30 2 in the example and the second pair corresponds to the sec ond window To conform to the requirements of the weight statement in the SAS regression procedure proc reg the weights are reciprocals of variance not portfolio weights Assume that the researcher creates a SAS data set explanatory with one observation for each firm in the sample of various explanatory variables identified by PERMNO The researcher can then merge the two data sets and 68 Figure 5 3 Contents of SAS data set abnormalreturns produced by Figure 5 2 code US Targets of Canadian Acquirers 1997 1998 7 PERMNO _weight_ wincar1 weight1i wincar2 weight2 10506 Equal 0 56738 5 777 0 03257 83 88 10914 Equal 0 13304 55 647 0 41793 807 09 36150 Equal 0 20945 91 579 0 19949 1331 05 67652 Equal 0 29229 46 506 0 44130 675 01 72100 Equal 0 08272 21 992 0 81411 319 29 75111 Equal 0 52603 37 952 0 10444 552 85 75241 Equal 0 04506 88 500 0 04199 1241 09 76263 Equal 0 03851 209 194 0 02503 3045 64 76369 Equal 0 01549 2 847 1 34032 41 35 76754 Equal 0 49761 30 693 0 14292 438 04 77142 Equal 0 10692 41 911 0 06721 600 09 77170 Equal 0 54004 6 870 0 35172 100 62 77446 Equal 0 08327 19 854 0 22735 290 88 77833 Equal 0 07545 13 418 0 22582 191 68 79739 Equal 0 10539 44 350 0 39443 644 39 83447 Equal 0 29044 27 797 0 08052 396 70 estimate a cross s
57. ATE option 20 86 96 145 with weekly returns 16 Balachandran Bala V 32 55 beta GARCH 26 Scholes Williams 26 104 Bhagat Sanjai 26 bid and ask prices 90 binary ASCII output files 87 141 152 155 Binder John J 1 binomial test 112 Boehmer Ekkehart 28 52 109 111 135 140 Bollerslev Tim 26 51 BOOT option 30 130 bootstrapped tests 130 139 1 or 2 tailed 130 139 bootstrapping 30 BOTH option 25 50 Brickley James A 111 Brockett Patrick L 26 50 Brown Stephen J 111 buy and hold returns 30 53 104 with EXTRACT statement 67 BUYHOLD option 30 53 Cantrell Steve 108 cash distributions 90 141 Chandra Ramesh 32 55 Chen Haiyang 112 Chen Hwei Mei 26 50 Chou Ray 26 51 Collins Daniel W 29 52 Collins Dent test 29 52 comparison period returns 25 50 66 105 compounded returns 30 53 104 continuously compounded returns 22 copyright notice 2 Corhay Albert 26 50 Corrado Charles J 29 53 112 Cowan Arnold R 26 29 50 52 53 108 111 112 CP option 25 50 cross sectional analysis 63 cross sectional regression 65 example 69 cross sectional standard deviation 111 cross sectional standard error 111 cross sectional test 28 51 formula 111 CRSP database cRSPAccess 159 environment variables 159 excess returns 15 26 84 88 127 how Eventus finds 159 169 Indices 21 monthly 128 Nasdaq only and NYSE AMEX only together 126 old Nasdaq 128 156 pre 1995 129 reading multiple databases in o
58. AYS 1 VOLUME BINARY VSAS HSAS SIC SHRCODE SHARES TRADES SPLITADJ EXTFILE fileref OUTSAS libref SAS_dataset need to specify DATEFMT Besides MMDDYY and DDMMYY you can use the SAS date format DATE CRSP Trading Day Numbers Eventus never requires you to manu ally convert calendar dates to CRSP day numbers If you happen to have CRSP day numbers you can use them in your request file Specify DATEFMT CRSP on the REQUEST statement AUTODATE If some of the calendar dates in your request file may not be trading days you can specify AUTODATE on the REQUEST statement AUTODATE tells Eventus to convert automatically all calendar dates to trading days Non trading days are converted to the following trading day For example Eventus changes a Sunday to the following Monday or Tuesday if Monday is a holiday The AUTODATE option is ignored when reading a monthly file and when you specify DATEFMT CRSP NDAYS Specify NDAYS n when you want the same fixed number of trading days weeks or months of returns prices or volume data for every firm in the request file Omit the ending date from your request file if you specify this option The valid range of values for n is from 1 to 9999 but some computers will be unable to handle values above 1000 or so The main constraint on 1 An example of a date in DATE format is 190CT1987 86 the number of trading days weeks or months of returns is the amount of space in the default WO
59. And Insider Trad ing New Evidence Journal of Business 1988 61 1 25 44 Sanders Ralph W Jr and Russell P Robins Discriminating between Wealth and Information Effects in Event Studies in Accounting and Fi nance Research Review of Quantitative Finance and Accounting 1991 1 3 307 329 166 Sanger Gary C and James D Peterson An Empirical Analysis Of Com mon Stock Delistings Journal of Financial and Quantitative Analysis 1990 25 2 261 272 Schipper Katherine and Abbie Smith Effects of Recontracting on Share holder Wealth The Case of Voluntary Spin Offs Journal of Financial Economics 1983 12 4 437 468 Scholes Myron and Joseph T Williams Estimating Betas from Nonsyn chronous Data Journal of Financial Economics 1977 5 3 309 327 Singh Ajai K Arnold R Cowan and Nandkumar Nayar Underwrit ten Calls of Convertible Bonds Journal of Financial Economics 1991 29 1 173 196 Sprent Peter Applied Nonparametric Statistical Methods Chapman and Hall London 1989 Swary Itzhak Stock Market Reaction to Regulatory Action in the Conti nental Illinois Crisis Journal of Business 1986 59 3 451 474 Sweeney Richard J Levels of Significance in Event Studies Review of Quantitative Finance and Accounting 1991 1 4 373 382 167 168 Index Ansley Craig 108 arbitrage portfolio 19 151 Arshadi Nasser 19 ask and bid prices 90 AUTOD
60. BOOT option on the EVTSTUDY statement produces bootstrapped versions of parametric tests Eventus performs bootstrap tests only for the windows not each individual day or month However you can obtain bootstrap tests for an individual day or month by specifying a window on the WINDOWS statement with the same beginning and ending date The bootstrap tests do not replace other results in the output but appear on a separate page after the regular parametric and nonparametric tests The BOOT option produces true bootstrap or resampling tests using the approach described by Kramer 2000 Eventus does not compute out of sample bootstrap more precisely called simulation tests Eventus restricts the parametric test methods that it bootstraps to cross sectional tests on buy and hold abnormal returns and the standardized cross sectional test on cumulative abnormal returns Reporting median abnormal returns This option is needed only when the PAGE WIDE option appears on the EVENTUS statement With the default output format PAGE TALL Eventus always reports median and mean abnormal returns With the optional wide output format Eventus reports median abnormal returns in place of the num ber positive and negative if you specify MEDIAN on the EVTSTUDY statement Significance symbols for the default generalized sign test or optional rank test still are reported Computing buy and hold compounded window returns To obtain buy and hold abnormal
61. CAARr 7 is CAARr 1 cure n VN taR Brown and Warner 1985 report that the cross sectional test is well specified for event date variance increases but not very powerful Boehmer Musumeci and Poulsen 1991 report that the standardized cross sectional test see above is more powerful and equally well specified Cowan 1992 reports that the generalized sign test see below also is well specified for event date variance increases and more powerful than the cross sectional test 111 Generalized Sign Test For each trading day or month in the event period and for each window Eventus reports the number of securities with positive and negative average abnormal returns cumulative in the case of windows Also reported is a test statistic in the default output format and significance level symbols for the generalized sign test The null hypothesis for the generalized sign test is that the fraction of positive returns is the same as in the estimation period For example if 46 of market adjusted returns are positive in the estimation period while 60 of firms have positive market adjusted returns on event day 1 Eventus reports whether the difference between 60 and 46 is significant at the five percent one percent or one tenth of one percent level The actual test uses the normal approximation to the binomial distribution For examples of the generalized sign test in recent research see Sanger and Peterson 1990 Singh
62. CRSP data and non CRSP data The filename request statement points to the request file prepared by the researcher The request file is a separate file that the researcher creates to define the sample for the study Each line of the request file should contain a five digit PERMNO identifier and a date in the form CCYYMMDD Later filename is a base SAS statement this one defines the fileref request A SAS fileref is a simple reference label associated with an external file An external file means a file not in a format exclusive to SAS Depending on the operating system filerefs may be strictly internal to SAS as when they are defined by filename statements or they can be defined by operating system DDnames logicals or environment variables Figure 2 1 The simplest Eventus event study program filename request F Any Folder Filename extension eventus request evtstudy Figure 2 2 Request file PERMNOs and dates illustrating tolerated irregularities 72100 19970626 75111 19980209 77142 981016 36150 19970616 77170 19981118 75241 19980805 76263 19980210 76369 19970814 77446 19981117 83447 19980615 10506 19971013 67652 19980330 91732 19980708 79739 970617 76754 19980615 77833 19970331 10914 19970306 chapters describe how to use CUSIP and similar identifiers instead of PERMNO The date will be used as day 0 Spacing is unimportant as long as at least one blank separates the PERMNO and da
63. ERR option specifies that the standard error for each date in event time should be computed across securities instead For an example of the application of the cross sectional method see Pilotte 1992 This option has no effect on the standardized abnormal return method The standardized cross sectional test The z statistic that Eventus normally computes for the standardized abnor mal return method is the widely used statistic described by Patell 1976 The STDCSECT option substitutes the standardized cross sectional test This is an extension of the Patell test introduced by Boehmer Musumeci and Poulsen 1991 The standardized cross sectional test compensates for a pos sible variance increase on an event date by performing a simple cross sectional variance adjustment The SERIAL option explained below is automatically invoked by the STDCSECT option 28 EGLS and Collins Dent tests The EGLS and CDCSI options replace the default standardized test with the estimated generalized least squares test and Collins Dent test assuming cross sectional independence respectively These tests are discussed in detail by Sanders and Robins 1991 The EGLS and CDCSI options automatically invoke the SERIAL option Adjusting window z tests for serial dependence The SERIAL option applies to standardized abnormal returns only Normally the test statistics for abnormal returns cumulated over intervals you define in the WINDOWS statement are not adjus
64. Eventusa Version 6 3C Software for Event Studies and CRSP Data Retrieval http www eventstudy com User s Guide June 2000 Edition Revised Printing October 2000 copyright 1989 2000 Cowan Research L C Notice Concerning Use of Eventus Software Eventus is a proprietary software product 1989 2000 by Cowan Re search L C Eventus software is licensed and not sold Eventus software is licensed to an organization or an individual to be used only by the licensee or in the case of an organization its employees faculty and students as ap plicable The software is not to be copied except in such manner as may be expressly permitted by the license sold and or given to anyone Eventus is a registered trademark for software sold by Cowan Research L C Registered refers to U S trademark registration Australian trademark registration ap plied for EventStream request file and USERSTOK are trademarks of Cowan Research L C Eventus licensees may reproduce this manual for internal use only pro vided that each copy contains this copyright notice page ISBN 1 893112 09 8 Contents 1 Introduction 2 Event Studies The Essentials 3 Event Studies The Options 3 1 Event Studies Centered on a Single Event Date 3 2 An Event Study Example 04 3 3 Abnormal Returns between Paired Events The TWIN Option 3 4 Reprinting or Merging Saved Event Studies with the OLDSTUDY Gtatemient wc a we eR a Oe E
65. MAXMISS options or their defaults in terms of multiday multiweek or multimonth periods All results are reported in terms of multiday multiweek or multimonth periods as well However Eventus in terprets the EST ESTLEN MINESTN and other REQUEST statement options in terms of the original days weeks or months For example in a daily event study suppose the user specifies ESTLEN 100 and TIMEUNIT 2 The market model and other estimates are computed on 50 two day returns a total of 100 single days in the estimation period The use of TIMEUNIT precludes the examination of the returns of single actual days weeks or months Eventus converts each sequence of n returns to a multiday multiweek or multimonth return as soon as it is read from the CRSP file Another method of evaluating individual dates is to use the WINDOWS statement without TIMEUNIT WINDOWS lets the user aggregate selected date ranges in event time for reporting and testing Specifying the number of days or months in the event period EVTSTUDY computes and reports abnormal returns for the event period which is defined by default as days or weeks 30 through 30 or months 12 through 12 To change from these defaults specify the number of days before and after on the EVTSTUDY statement using PRE and POST These options may be used singly or in combination For example PRE 60 means that the abnormal returns are to start with day 60 unless POST is also specified the la
66. NAME 149 NDAYS 86 149 149 NODIVIDX 20 87 149 POOL 22 74 REQFILE 150 REQFILE2 150 SHIFTn 150 SHORT 19 151 SIZEINDX 21 151 SP500 20 87 151 UPCUSIP 151 required before EVTSTUDY 145 required before PRICES 145 required before RETURNS 145 required before VOLUME 145 results output for cross sectional anal ysis 63 RETURNS statement 152 default output 153 options BINARY 87 152 BOTH 88 153 EXTFILE 92 152 HSAS 88 INDEX 88 153 OUTSAS 92 153 SHRCODE 89 153 SIC 153 SPORT 88 VALUE 88 153 VSAS 88 152 154 REQUEST required before 145 Robins Russell P 29 52 109 Rogers Ronald C 108 Rozeff Michael S 19 Sanders Ralph W Jr 29 52 109 Sanger Gary C 112 SAS data sets storage of CRSP data 128 129 161 variable names 116 Schipper Katherine 105 Scholes Myron M 26 50 104 Scholes Williams beta 26 50 66 104 secondary prices 141 seemingly unrelated regressions 76 132 135 selling short 19 151 Sergeant Anne M A 26 50 serial correlation correction for 108 Sfiridis James M 29 53 share type code 89 135 143 153 156 shares outstanding 90 143 156 Shaw Wayne 31 32 34 54 55 56 Shieh Joseph C P 112 short position 19 SIC codes 24 88 135 SIC option 24 Sicherman Neil W 108 175 sign test 29 53 112 significance level 31 54 135 Singh Ajai K 108 112 size adjusted returns 21 Smith Abbie 105 Spencer David E 29 52 108 split adjustment 156 Sprent Peter 112 standard errors
67. NO identifiers and converted dates if you name it with ID ID ID works or you can choose another name Use IDFMT to tell Eventus the format to use for reading and printing the identifying variable IDFMT 4 means a 1 4 digit integer while IDFMT 4 means a four letter word Other lengths and other sas formats also are valid Using grouping variables group weights and short long indicators If the input request file contains a grouping variable see page 17 spec ify the GROUP option on the DATECONV to have Eventus include the grouping variable in the updated file Add the GRWEIGHT option if the request file also contains a within group weight for each observation If the input request file 95 contains an S or an L to indicate short or long see page 19 specify SHORT on the DATECONV statement The S or L must follow the PERMNO date s and any identifying variable grouping variable and group weight These options have no effect on the operation of DATECONV except to copy the grouping variable weights or short long indicator to the output file DATECONV state ment processing makes no use of the grouping variable group weight or short long indicator except to copy them from the input request file to the correct position in the updated request file Options for processing dates The DATECONV statement allows you to specify how Eventus should handle the dates in your request file DATEFMT Calendar dates You can list calendar
68. OL3 through VOL90 contain missing values Because the SAS system stores a large amount of information on each variable for its own use a horizontal format data set occupies considerably more disk storage space than a vertical format data set NASDINFO Retrieves trading status trait code NMS indicator and market maker count from the Nasdaq information arrays of the CRSP database Valid only when SHARES see below is a valid option OUTSAS Gives the two level SAS name libref membername in which to store the volume data A SAS libname statement or the operating system must associate the libref with the aggregate storage location used for the SAS data library The data set is created if it does not exist already or replaced if it does exist SHARES Specifies that the number of shares outstanding on the event date in thousands should be included in the output file This option is ignored unless either NDAYS 1 is specified on the REQUEST statement or VSAS is specified on the VOLUME statement Note that the shares outstanding data may not be as timely as the volume data Refer to CRSP publications for more information SHRCODE Causes Eventus to store the two digit share type code from the CRSP name history array as the variable SHRCODE in any OUTSAS data set SIC Causes Eventus to read the SIC code from the CRSP database and to add the variable SICCODE to any output SAS data set SPLITADJ Causes Eventus to adjust for stock splits reverse
69. RK SAS data library See the SAS Companion for your environment or ask a consultant at your site if you need to enlarge the WORK library Two additional date processing options SHIFT1 and SHIFT2 are avail able These options shift the dates from the request file by a specified number of days See page 150 in Appendix B Market index options NODIVIDX Eventus normally supplies the returns including dividends on the equally weighted and value weighted if the VALUE option is specified indexes from the CRSP index file Specify NODIVIDX to instruct Eventus to supply the index returns excluding dividends SP500 and COMPOSIT The CRSP NYSE AMEX and NYSE AMEX Nasdaq in dex files contain the Standard and Poor s 500 Composite Index in addition to the equally weighted and value weighted indexes Specify SP500 to tell Eventus to read the Standard and Poor s index return instead of the value weighted index The CRSP Nasdaq index file includes the Nasdaq Composite Index return instead of the Standard and Poor s index You instruct Eventus to read the Nasdaq Composite Index instead of the value weighted index when you specify COMPOSIT Eventus does not determine whether the index file is actually the NYSE AMEX Nasdaq or the Nasdaq only file 7 3 The RETURNS PRICES and VOLUME statements Use only one of the RETURNS PRICES or VOLUME statements in a single pro gram The following options are available Selecting an output file format Fi
70. RSP Daily eventus etc indicates that the SAS data sets containing the CRSP data for the current run are in the folder SASCRSP Daily on drive H of a Windows computer or network The libref that Eventus looks for can be changed using LIBNAME option on the EVENTUS statement To use a different data frequency in the estimation period of an event study from that used in the event period in conjunction with the EVENTUS statement option ESTINTER the libref specified in the EVENTUS statement option ELIBNAME must point to the SAS data library containing the CRSP data to be used for the estimation period For example libname crsp home crsp sasdata daily libname estcrsp home crsp sasdata monthly eventus estinter month elibname estcrsp etc could be used to run a daily data event study with parameters estimated from monthly data on a Unix system where the daily and monthly CRSP data are stored in SAS data sets in the directories home crsp sasdata daily and home crsp sasdata monthly respectively To use a size index file in conjunction with the REQUEST statement option SIZEINDX the EVENTUS statement option SIZEIND must point to the SAS data set containing the size index data For example libname crsp home crsp sasdata daily eventus sizeind crsp sizeindx request sizeindx etc 161 indicates that the size index data set is in the same CRSP data library direc tory home crsp sasdata daily in this cas
71. RSP files The default is PACKAGE 1 133 Table B 1 PACKAGE specifiers for the EVTSTUDY statement Data Package Specifier Remarks Estim ation Event Period Period Cumulative abnormal return 0 TWIN only Data required for OLDSTUDY 1 Same as OADGHP Daily abnormal returns A D Daily raw returns B E Even if RAW not specified Daily market index returns C F Daily standard deviation G Ignored if NOSTD specified Dummy variable 1 if abnormal return gt 0 H Mainly for OLDSTUDY use Parameters P Mean returns a 8 etc Table A 2 on page 117 lists SAS variable names You do not need to know variable names to use the EXTRACT or OLDSTUDY statements The EXTRACT statement will not extract raw returns unless the EVTSTUDY statement that produced the SAS data set included RAW and PACKAGE D POST periods Specifies the number of trading days or months immediately following the event date for which to compute abnormal returns The default is POST 30 for daily and weekly and POST 12 for monthly event studies PRE periods Specifies the number of trading days or months immediately preceding the event date for which to compute abnormal returns The default is PRE 30 for daily and weekly event studies and PRE 12 for monthly event studies RANKTEST Designates the rank test instead of the generalized sign test as the nonparametric statistic to appear with the non standardized parametric tests SERIAL Specifies that
72. S data sets to input the USERSTOK data This chapter does not provide the details of the approach but reference documentation for the INSAS option used with the sas data set approach appears on page 131 Chapter 6 presents an example of SAS data set input that the user may adapt as needed Additional examples appear on our web site The WINDOWS statement The WINDOWS statement is optional Use it to specify ranges of dates windows in event time over which Eventus should report cumulative or compounded abnormal returns and associated test statistics Each window specification requires a begin and end which can be the same if desired For example 3 0 specifies a four day or four month window ending with the event date 2 2 specifies a one day or one month window containing the second day or month after the event date When begin and end are equal no cumulation or compounding occurs the only purpose for such a specification would be to cause a single date to appear in the window output section for ease of reference If the WINDOWS statement is omitted Eventus reports three windows PRE 2 1 0 and 1 POST The results for the windows appear at the bottom of the same page on which the results for individual dates are printed or on a separate page if there is not enough space The EVTSTUDY statement The EVTSTUDY statement tells Eventus to run an event study The options permit you to change the default event p
73. S option also specify the two part SAS data set name libref membername using the OUTSAS option For a temporary data set which will no longer exist after the user closes SAS or after the batch run completes use the libref work For a permanent data set use a libname libref statement before the EVENTUS statement to point to the aggregate storage location replacing libref with the desired libref 92 Chapter 8 Converting Calendar Dates to CRSP Trading Day or Month Numbers Using DATECONV DATECONV is an Eventus statement that converts calendar dates to CRSP style trading day week or month numbers It is not necessary to use DATECONV before running most Eventus applications Eventus automatically con verts calendar dates to CRSP trading day numbers without explicit instruc tions Only researchers who have special requirements for CRSP day numbers perhaps for use with non Eventus software need DATECONV Figure 8 1 lists the Eventus statements needed to use DATECONV The op tions are described below 8 1 The EVENTUS statement If you are converting pairs of dates for each PERMNO specify TWIN If you are converting dates for use with a monthly or weekly database specify MONTHLY or WEEKLY on the EVENTUS statement Eventus uses the CRSP database s calendar to determine the association between calendar dates and CRSP day week or month numbers 93 Figure 8 1 Eventus statements for converting calendar dates to CRSP
74. SIP or converting CUSIPs to PERMNOs 1993 and earlier editions of the CRSP stock files are sorted by CUSIP identifiers rather than permanent identification numbers If you are using one of these files provide CUSIPs instead of PERMNOs in the request files You also need to specify the CUSIP option on the REQUEST statement Follow the same request file format as described elsewhere in this chapter but supply an eight character CUSIP instead of a numeric PERMNO for each stock Sites where the CRSP data exist as SAS data sets often index the data sets so that programs can search them by either CUSIP or PERMNO without sorting The default in Eventus is to search by PERMNO To search by CUSIP provide CUSIPs instead of PERMNOs in the request files and specify the CUSIP option on the REQUEST statement If you have CuSIPs in the request file you can specify the CUSIPERM option to convert them to CRSP PERMNOs Eventus uses the PERMNOs to search the CRSP database and label the results but does not change your request file Only the PERMNOs not the CUSIPs appear in the output file The CUSIPERM option works properly only if the PermnoUp program de scribed in the Eventus installation instructions is run once after each annual or quarterly update of the CRSP stock database Including an identification variable Each line of your request file may include an identification variable following the event date Eventus will read it and store it with the PERM
75. TLE text 3 TITLE2 text 3 REQUEST ID variable IDFMT format DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP EST value value POOL ESTLEN n NAME WINDOWS EVENT1 descriptor EVENT2 descriptor2 EVTSTUDY NOPLIST DETAIL NOSTD VALUE BOTH SW GARCH EGARCH iCSECTERR STDCSECT SERIAL MEDIAN BUYHOLD TAIL 1 2 OUTSAS libref membername on the EVENTUS statement and the specification of the WINDOWS statement The descriptors in the WINDOWS statement are 1 11 character names by which you want Eventus to designate the two event dates on the output Blank spaces are not allowed within either descriptor See Section 4 1 for descriptions of the remaining options Each line of the request file must contain a pair of event dates separated by one or more blank spaces in the location where the single event date goes in a normal event study request file 61 62 Chapter 5 Extracting Event Study Results for Further Analysis This chapter describes how to use the EXTRACT statement The EXTRACT statement selects and organizes daily or monthly firm by firm cumulative or buy and hold compounded abnormal returns from results saved by the EVTSTUDY statement option OUTSAS The most common use of EXTRACT is to create a data file for cross sectional analysis of abnormal returns Be careful to distinguish EXTRACT from RETURNS RETURNS covered in Chapter 7 selects returns directly from the CRSP database and s
76. TUS statement By default when the EXCESS option is in effect and the appro priate add on CRSP module is available Eventus based on decile portfolios of stocks ranked by beta Optionally Eventus can use excess returns based on a ranking by standard deviation of return To use standard deviation excess returns instead of beta excess returns specify SPORT on the EVTSTUDY statement Scholes Williams and GARCH market model estimation Eventus reports market model results using both ordinary least squares and Scholes Williams 1977 beta estimation when you specify SW When you specify the GARCH option Eventus estimates the market model assuming a GARCH 1 1 error structure For exponential GARCH or EGARCH 1 1 specify the EGARCH option With either the GARCH or EGARCH option maximum likelihood estimates using the dual quasi Newton algo rithm are produced Eventus reports the alpha and beta estimates as it does in the case of ordinary least squares but does not report the estimated pa rameters of the conditional error variance model No more than 40 iterations will be performed for each stock In general convergence will be better the longer the estimation period ESTLEN and better with the EGARCH 1 1 than the GARCH 1 1 model Cowan and Sergeant 1996 report that event study test specification and power are insensitive to the use of Scholes Williams versus OLS estimation Corhay and Rad 1996 and Brockett Chen and Garven 1999 discuss
77. The program in figure 6 2 includes a preliminary data step to read the request file into the permanent SAS data set project request The request file includes the event date in both CRSP format crspday1 and YYMMDD format eventdat This allows the use of trading days in the REQUEST state ment where the SHIFT1 option specifies that the extracted returns should start 105 trading days before day 0 and extend a total of 111 days implying that the return series ends on day 5 The request file contains CUSIPs so the REQUEST statement includes the CUSIPERM option Figure 6 3 displays the request file The program merges the stock and market returns extracted by Eventus with the data set contain ing the second factor creating the permanent data set project stockret The permanent data set will become input to the second Eventus program that actually runs the event study The name of the first return generating factor must be market which also is the name automatically constructed by the RETURNS statement The second factor name must be factor2 additional factor names follow the obvious pattern Figure 6 4 shows the Eventus program to complete the event study Note that the REQUEST statement includes EST and ESTLEN parameters consistent with the length of return series extracted in the first program The first 100 trading days of the 111 day return series for each stock are arbitrarily called the estimation period but the model estimation uses the e
78. US GETDATA PAGE WIDE MONTHLY EXCESS REQUEST ID variable IDFMT format DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP AUTODATE NDAYS n NODIVIDX SP500 COMPOSIT RETURNS BINARY VSAS HSAS INDEX VALUE BOTH SIC SHRCODE EXTFILE fileref OUTSAS libref SAS_dataset You may also select excess returns based upon beta decile portfolios or standard deviation decile portfolios if the appropriate CRSP add on module is installed To do so specify EXCESS on the EVENTUS statement 7 2 The REQUEST statement The REQUEST statement instructs Eventus to read your request file which is an external file containing PERMNO identifiers dates and sometimes other information This section describes several REQUEST statement options for processing dates and returns If you are familiar with permanent SAS data sets you might want to take advantage of the INSAS option see page 147 Otherwise a SAS filename statement or the operating system must associate the fileref REQUEST with the request file Each line of the request file for a RETURNS PRICES or VOLUME program should contain the following items PERMNO starting date of data to retrieve ending date of data to retrieve and identifying variable value if any Omit the ending date if you specify NDAYS explained below on the REQUEST statement When reading a monthly database the dates in the request file may be any day of the month 84 Figure 7 2 Eventus statements to read st
79. a saved event study data set 117 Eventus special SAS missing values for missing returns from the CRSP databas o o c c ek ee niea we Re we we eS 119 PACKAGE specifiers for the EVTSTUDY statement 134 vi List of Figures zal a 2 3 2 4 2 5 2 6 2 2 8 3 1 3 2 3 3 3 4 3 9 3 6 3 7 3 8 3 9 3 10 4 1 The simplest Eventus event study program 4 Request file PERMNOs and dates illustrating tolerated irreg HOMES 6 e o e ea AS ES aa ae ad a e Ae A 4 Sample Eventus output First page a ooo 6 Sample Eventus output Second page ooo aaa 7 Sample Eventus output Third page ooa aaa 8 Sample Eventus output Fourth page 9 Sample Eventus output Fifth page 10 Sample Eventus output Final page oaoa aaa aaa 11 Eventus statements for an event study centered around a single UAE a oe Poe eas SA ear p a a Bede R 14 Eventus program for an event study of the Too Big to Fail Oi ee Re ee ee eee Se ek Oe Bath a 32 Request file for the Too Big to Fail study 33 The first page of Eventus output aooaa a a 34 The Eventus sample listing and input report 35 The Eventus parameter estimate listing o 0a aa aaa 36 Eyent study Teslis s oc oca ROAR KERR OR erk o peth 38 Eventus statements for event studies cumulating returns be tween two firm specific event dates o oo ooo a 39 Eventus statements for reprinting an
80. ained the normal number of returns for the window or until it has exhausted the n days following the window Specifying EXTEND without n is equivalent to specifying EXTEND 1 You may want to consider the use of weighted least squares regression with abnormal returns generated using this option Adjusting the weights for serial correlation If you ran the original event study with the SERIAL option described on page 29 or the STDCSECT option which implies SERIAL Eventus adjusted the window test statistics to reflect the serial correlation that is inherent 66 in abnormal returns To make such an adjustment to the weights for cross sectional analysis specify the SERIAL option on the EXTRACT statement This option only is relevant when you use the WPREFIX option The default is not to make any adjustment to the weights for serial correlation even though the test statistics for the original event study incorporated the adjustment Computing buy and hold compounded returns To obtain buy and hold abnormal returns for windows specify BUYHOLD on the EXTRACT statement This feature requires that the BUYHOLD option also appeared on the EVTSTUDY statement The default is to generate cumula tive abnormal returns even when the original event study used buy and hold returns 5 4 Usage example Figure 5 2 displays the Eventus statements to perform an event study using the request file from Chapter 2 with daily CRSP data In this example
81. alid only when CUSIP also is specified makes Even tus attempt to update CUSIP identifiers in the request file to match the latest version of the CRSP stock file before searching CRSP Requires a conversion file a component of Eventus that normally should have been created when the software was installed or last upgraded 151 B 9 RETURNS Statement Reads returns from a CRSP stock database No processing occurs except that missing returns are converted from CRSP to SAS missing value codes Eventus stores the returns in your choice of file Choose from the following options BINARY This is identical to the TEXT format except that the returns them selves are written in real binary FLOAT4 RB4 or S370RB4 depend ing on the system format This uses only one third as much disk space for storing returns as the TEXT format This type of file like the TEXT file can be read by any major computer language including FORTRAN SAS and C BOTH See VALUE EXTFILE Gives the fileref of the external file in which Eventus is to store the returns The default is userdata On mainframe systems where applicable the file should have a disposition of NEW or OLD and a logical record length of 80 This option is only valid with the default TEXT or optional BINARY format HSAS Creates a horizontal format SAS data set A SAS data set cannot be read by non SAS software The data set will contain as many of the fol lowing variables as applicable
82. all other dates in the combined estimation and event period The EVTSTUDY statement The EVTSTUDY statement tells Eventus to run the event study The options permit you to change the default event period vary the amount of printed output and select a market index and estimation method Selecting the amount of printed output EVTSTUDY produces a report of the results of reading your data The report lists the PERMNO or CUSIP identifying variable and event date from your request file and the name of the issuer if direct CRSP access is used or you specified the NAME option on the REQUEST statement The report also tells for each firm how many returns Eventus found in the estimation period and in the event period If you prefer not to get this report specify the NONAMES option 75 Selecting the value weighted market index With direct CRSP access the default is to use the equal weighted index with dividends from the CRSP index file To use the value weighted index specify VALUE on the EVTSTUDY statement Specifying additional return generation factors To customize the return generating model specify the FACTORS option on the EVTSTUDY statement For example FACTORS 2 specifies a two factor model The option is only available in NONCRSP mode and requires that the return data be supplied in the form of SAS data sets The use of a customized return generating model is illustrated in the example below Specifying the number of days
83. allation but can be specified at run time SIZEIND libref membername Specifies the two level name of the sas data set containing the size decile return data This option is for systems that store the CRSP data in the form of SAS data sets and is needed only in jobs where the REQUEST statement specifies the SIZEINDX option Eventus expects to find variables named caldt and decret1 through decret10 in the SIZEIND data set TWIN Specify TWIN in event study programs when you want to estimate variable length firm specific event windows instead of the conven tional abnormal returns and windows around a single firm specific event date When TWIN is in effect you provide two event dates for each ob servation instead of just one See the REQUEST statement description below for an explanation of how to indicate the event dates TWIN is also used this way in DATECONV programs to convert a pair of dates instead of just one 129 VAXFMT Causes Eventus to use the VAXRB4 SAS informat instead of the default FLOAT4 SAS informat to read an input CRSP database that is SFA binary or CRSPAccess format This option is intended primarily for OpenvMS Alpha systems It has no effect when the input CRSP database is SFA character format WEEKLY Tells Eventus that it is reading weekly returns from a USERSTOK file also see the NONCRSP option or from a weekly returns database formatted like CRSP stock databases B 4 EVTSTUDY Statement EVTSTUDY is required
84. also are options for single event date event studies and function similarly in the two contexts The only differences are the word TWIN 59 Figure 4 7 Event study results O Hara and Shaw JF Dec 1990 pp 1587 1600 4 Replication of Table III C Insolvent Bank Sample Market Model VW Index Average Median Generalized Day Abnormal Abnormal t N Positive Sign Return Return Negative Z 5 1 27 1 51 2 29 22 16 6 2 52 gt 4 1 21 1 08 2 18 22 17 5 2 94 gt gt 3 0 27 0 14 0 49 22 12 10 0 80 2 0 18 0 32 0 33 22 15 7 2 09 gt Sd 0 21 0 12 0 38 22 10 12 0 05 0 0 43 0 24 0 77 22 10 12 0 05 1 0 30 0 29 0 54 22 15 7 2 09 gt 2 0 55 0 84 1 00 22 8 14 0 91 3 0 38 0 41 0 68 22 9 13 0 48 4 0 61 0 46 1 10 22 8 14 0 91 5 0 33 0 12 0 59 22 10 12 0 05 Average Compounded Median Compounded t Positive Gen Sign Days Abnormal Return Abnormal Return Negative Z 5 2 2 96 3 48 2 68 19 3 3 80 gt gt gt 1 0 0 62 0 43 0 80 12 10 0 80 1 5 0 50 1 15 0 41 7 15 1 33 significant at 10 lt gt significant at 05 lt lt gt gt significant at 01 lt lt lt gt gt gt significant at 001 60 Figure 4 8 Eventus statements for event studies cumulating returns between paired event dates filename request G Some Folder Filename extension filename userstok G Some Folder Filename extension EVENTUS TWIN NONCRSP MONTHLY WEEKLY PAGE WIDE TI
85. ank test statistic for the event window composed of days T through 75 is Kam K z h T 1 T 3 D E E Ri RY D where a 1 K Thb T T 1 A iEn is the average rank across the n stocks and T 7 1 days of the event window and Ay 1 n X7 K is the average rank across n stocks on day t of the D E day combined estimation and event period The expected rank still is K for event windows shorter than E days because the full D E day set of returns is used for the assignment of ranks Jackknife Test The discussion in this subsection is adapted from Giaccotto and Sfiridis 1996 The jackknife test incorporates the standardized abnormal return for each stock 7 computed using the event period sample standard devia tion The standardized abnormal return for day t is oe O Aji where i Te A A Gy a A 3 t T j and A is the mean abnormal return of stock j during the event period of E Te Tp 1 days If there is an event induced transient variance change on day t then 4 is a biased estimator of o4 and is a biased statistic Giaccotto and Sfiridis propose reducing the bias by jackknifing the 6 values 113 The first step of the jackknife is to sequentially delete one abnormal return Ajr from equation A 3 and re compute o 4 using the new value in turn to re compute 6 using equation A 2 Call the latter value i s The next step is to form pseudo values
86. appears on the DATECONV statement e PERMNO 5 digit integer Required unless CUSIP or CUSIPERM ap pears on the DATECONV statement 123 e The dates to be converted For a single date per PERMNO either EVENTDAT SAS date variable or CRSPDAY integer representing trading date 0 as a sequence number corresponding to the CRSP index calendar The integer may represent the day week month quarter or year from the beginning date of the index file For programs that convert two dates per PERMNO specify TWIN on the EVENTUS statement two names must appear EVENTDA1 and EVENTDA2 or CRSPDAY1 and CRSPDAY2 If you want Eventus to look for CRSPDAY and CRSPEST specify DATEFMT CRSP see below e If you specify ID variable see above the variable named there INSAS2 libref membername2 This option is used only with an SFA format CRSP database Specifies a second request file in the form of a SAS data set A second CRSP stock file which must be associated with the CRSPST2 libref is searched for the PERMNOs in the second request file The format required for the second request file is the same as for the first IX2Y Included for compatibility with the REQUEST statement only NAME Indicates that the input request file includes a firm name Rarely used NDAYS n Specifies that you want to create a second date by adding n trading days months to the first date for each PERMNO on each line of the request file If you specify NDAYS omit the
87. ase see NOSTD SUR Causes Eventus to use the seemingly unrelated regressions approach where the market model augmented by a dummy variable for each window specified on the WINDOWS statement is estimated by joint gen eralized least squares over the combined estimation period and event period return series SW Produces Scholes Williams 1977 market model results in addition to the OLS results The EGARCH GARCH and SW options are mutually exclusive TAIL 1 2 Specifies the significance levels of the reported test statistics are to be based on one or two tailed tests The default is TAIL 1 The 135 TAIL option does not set the tails for bootstrapped tests please see the BTAIL option TIMEUNIT n Specifies that Eventus should aggregate returns in the es timation and event periods into n day when the original returns are daily returns before performing the analysis When n is even period 0 in event time contains day 0 the date in the request file 3 1 days following day 0 and 5 preceding day 0 Additional periods are formed on either side of period zero Eventus combines daily weekly monthly or quarterly returns into pe riod returns by addition If the LOG option is specified on the REQUEST statement individual daily or monthly returns are converted to loga rithmic form before adding Eventus interprets the WINDOWS statement arguments and the PRE POST and MAXMISS options or their defaults in terms of multiday multiwee
88. atement if any that preceded the EVTSTUDY statement When the saved data are from a TWIN event study omit the WINDOWS statement Otherwise your program must include exactly one WINDOWS statement before the EXTRACT statement 5 3 The EXTRACT Statement Identifying the saved event study file Use INSAS to tell Eventus where to find the saved event study data The libref and membername should match the OUTSAS specification of the original EVTSTUDY statement See page 31 in Chapter 3 64 Selecting a stored identification variable If the EVTSTUDY program that saved the data included ID and IDFMT on the REQUEST statement then you can repeat them on the EXTRACT statement If you do the identifying variable values will be added to the OUTSAS data set or the EXTFILE file created by EXTRACT Naming the window variables and selecting weights The VPREFIX option selects a prefix for the variable names under which the window cumulative abnormal returns are stored in the OUTSAS data set The prefix may be up to 6 characters long of which the first must be a letter or underscore Eventus completes the variable name by appending an integer from 1 to 20 indicating the position of the window on the WINDOWS statement For example if you specify VPREFIX DEP and WINDOWS 1 0 2 2 the OUTSAS data set will include a variable named DEP1 containing the days 1 0 cumulative abnormal return for each firm The cumulative abnormal return f
89. ates these options determine the number of calendar days by which to shift When the request file contains CRSP trading day numbers these options determine the number of trading days by which to shift A positive shift adds days a negative shift subtracts days OUTDTFMT Many DATECONV programs do not need this option Use it only when you are converting from CRSP style day week or month numbers back to calendar dates In this case DATEFMT CRSP also should appear on the DATECONV statement This option specifies the calendar date format into which to convert CRSP trading day numbers You do not need this option when your request file contains calendar dates because in that case Eventus always converts to CRSP day numbers You can specify any SAS date format The default is YYMMDD Sorting the updated request file The observations in the updated request file will appear in PERMNO order by default To have the observations sorted by the identifying variable include the SORTBYID option on the DATECONV statement For this to work the input request file must contain an identifying variable value on each line and you also must specify the ID option Selecting an output file location The converted request file is directed by default to the fileref USERDATA If the fileref is undefined a file named userdata dat is usually created in the 97 current working directory To select a specific disk file for the converted request file use a
90. ation date must be a variable named ESTEND CRSPEST if DATEFMT CRSP is in effect Do not specify POOL with EST SPECIFIC e The default is EST 46 for daily event studies EST 10 for weekly event studies and EST 13 for monthly event studies 146 ESTLEN n Specifies the length of the estimation period in trading days weeks months quarters or years depending on the return interval being used for estimation in the current run The maximum permitted is 999 The default is 255 days about one year for daily returns 60 months for monthly returns 52 weeks for weekly returns 20 quarters for quarterly returns or 10 years for annual returns Odd values of ESTLEN will be reduced by one day when POOL is specified The number of usable returns in the estimation period may be lower than ESTLEN in individual cases if there are missing returns on the CRSP database GROUP group_variable Names a grouping variable to be used in an event study to combine multiple observations into a single equally weighted portfolio The value of the grouping variable for each observation is listed on the appropriate line of the request file after the dates and ID variable if any The grouping variable must be an integer between 0 and 9999 inclusive leading zeros in the request file are optional and ignored Two or more observations with identical grouping variable values are combined into a portfolio treated as a single observation The default is to use no group
91. ave the event study results in a permanent SAS data set see the EVTSTUDY statement options below only the PERMNOs will be included in the file Including an identification variable Each line of your request file may include an identification variable following the event date Eventus will read it and print it on subsequent output if you name it with ID ID ID works or you can choose another name Use IDFMT to tell Eventus what format to use for reading and printing the identifying variable IDFMT 4 means a 1 4 digit integer while IDFMT 4 means a four letter word Other lengths and other SAs formats are permitted also Grouping variables and group weights When you specify GROUP variable name Eventus looks for a grouping variable in the request file Replace variable name with a valid SAS name for example GROUPID The grouping variable should follow the event date and the ID variable if there is one on each line of the request file The grouping variable must be an integer between 0 and 9999 inclusive Leading zeros are optional If any two or more observations have the same grouping variable value Eventus combines the observations into an event time portfolio and treats the portfolio as a single observation in the event study An observation with a unique value of the grouping variable is treated normally To use a SAS data set as the request file please see the INSAS option on page 147 17 By default the observations
92. based on the CRSP file name structures The names are truncated in some cases but if there is a letter distinguishing a class of common stock Eventus always includes it at the end of the name None of the 22 sample firms has classified stock Eventus prints the untruncated name in the wide output format PAGE WIDE on the EVENTUS statement If Eventus drops an observation from the sample an explanatory message appears under Reason if no usable returns Figure 3 6 presents the Eventus parameter estimate listing The user can suppress this listing by specifying NOPLIST on the EVENTUS statement Eventus prints more statistics including the first order autocorrelation of the market model residuals if the PAGE WIDE option appears on the EVENTUS 33 Figure 3 4 The first page of Eventus output O Hara and Shaw JF Dec 1990 pp 1587 1600 Replication of Table III C Insolvent Bank Sample ESTIMATION PERIOD Ends 6 days before the event date 50 days in length TOTAL NUMBER OF EVENTS 22 EVENTS WITH USEABLE RETURNS 22 EVENTS DROPPED 0 STATISTICAL SIGNIFICANCE LEVELS 2 tailed NONPARAMETRIC TEST Generalized sign test in all event studies For nonparametric tests significance levels of 10 05 01 and 001 are denoted by lt lt lt lt lt lt or gt gt gt gt gt gt respectively Left brackets lt appear when the ratio of positive to negative is less than in the parameter estimation period R
93. ble to man ually tell Eventus where the database is located To do so use a filename crspdb statement to point to the aggregate storage location folder or di rectory on most systems containing the data and also specify the option DBFNSTMT on the EVENTUS statement for example filename crspdb c dai99912 eventus dbfnstmt etc To use a different data frequency in the estimation period of an event study from that used in the event period in conjunction with the EVENTUS statement option ESTINTER when the DBFNSTMT option is in effect the fileref mcrspdb must point to the aggregate storage location containing the CRSP Ac cess database from which Eventus is to read the estimation period returns 159 If the EVENTUS statement option EXCESS is specified the SAS filerefs statidx1 and statidx2 must point to the daily files dsbo dat and dsbc dat respectively or to the corresponding monthly files msbo dat and msbc dat The files are available from the ascii folder of the Indices CD ROM C 2 SFA Format To work with the CRSP SFA database format the SAS filerefs crspstok should point to the standard stock data file and the fileref crspindx should point to the stock associated calendar indices file If the Supplemental Nasdaq stock data file is to be read by the Eventus PRICES or VOLUME statements the fileref crspnms should point to it The same filerefs are used for daily and monthly data Therefore to change from daily to m
94. c compu tations DETAIL Causes Eventus to report the windows for each firm individually under the standardized abnormal return method in addition to the usual portfolio statistics DETAIL FULL This option gives you the output described under DETAIL plus the individual daily returns for each observation Warning this option produces at least one page and over 50 lines of printed output for each firm EGARCH Causes the market model to be estimated assuming an exponen tial GARCH 1 1 error process The EGARCH GARCH and SW options are mutually exclusive EGLS Requests the estimated generalized least squares test Sanders and Robins 1991 instead of the default standardized test FACTORS n Specifies how many exogenous factors are to be used for the OLSPARAM SUR or ITSUR option The default is FACTORS 1 indicating that only one factor typically the market index is to be used Non default values of this option are allowed only when the NONCRSP option appears on the EVENTUS statement GARCH Causes the market model to be estimated assuming a GARCH 1 1 error process The EGARCH GARCH and SW options are mutually exclu sive INSAS libref membername Points to the SAS data set from which to take the user supplied return information when the NONCRSP option appears on the EVENTUS statement The file must include the following variable names CUSIP RETURN MARKET and the variable specified in the ID option if any While the variable na
95. cify NDAYS the first day or month is the date in the request file omit the ending date which must be specified in the above situa tions when NDAYS is not used from the request file NODIVIDX Causes Eventus to market indices excluding dividends instead of the default indices including dividends POOL See EST 149 PORTYPE 1 2 Specifies the portfolio type position within a CRSPAc cess portfolio assignment structure to which Eventus reads when ex tracting decile ranks for the SIZEINDX option Users should not specify this option except as explicitly directed by Eventus documentation REQFILE fileref The most often used method of storing the request infor mation PERMNO identifiers dates etc is to use an external non SAS formatted file typically an ASCII text file Replace fileref with the fileref associated with your file The REQFILE specification may be omitted if the fileref of the request file is REQUEST Each line of the request file should contain the following variables in order PERMNO event event specific ID number grouping group S or date date 2 estimation date or string variable weight L Each value must be separated by at least one blank but the exact position of the values is unimportant as long as they appear in the order shown The square brackets simply indicate items that need not always appear do not include them in the file Whether an optional item should appear is determined by the options
96. cond firm then start on a new line and so on When extracting data from a TWIN event study omit VPREFIX even if you are creating a SAS data set EXTRACT assigns the name CAAR to the cumulative abnormal return from a TWIN event study Selecting the type of abnormal return to extract By default Eventus extracts market model returns unless the event study program used the CRSP excess return file Alternatively you may specify CP RAW MAR or SW to get comparison period raw market adjusted or Scholes Williams returns provided that the original event study included the type of abnormal return you specify Note that EXTRACT considers RAW to be a type of abnormal return In order to extract raw returns the EVTSTUDY statement in the event study program must include the RAW option and the package specifier D Extending a window to make up for missing days If you specify the EXTEND option Eventus will attempt to make up for missing returns within the window For example suppose you want to output a two day window for a sample of takeover targets some of which experienced trading halts on day 1 or 0 or both You could specify EXTEND 3 to attempt to obtain 2 days worth of abnormal returns for each firm In this case if one of days 1 and 0 were missing Eventus would extend the window to day 1 If day 1 were also missing the window would be extended to day 2 for that firm Eventus keeps trying to extend the window until it has obt
97. d Mean Adjusted Returns Comparison period mean adjusted returns are computed by subtracting the arithmetic mean return of the common stock of the jt firm computed over the estimation period R from its return on day t Aj Rj Rj The definitions of the average abnormal return and cumulative average ab normal return follow those for market model abnormal returns above A 2 Event Study Test Statistics Standardized Abnormal Return Method Eventus uses the standardized abnormal return method only for market model abnormal returns unless you specify STDALL on the EVTSTUDY statement Event Studies Centered on a Single Date Eventus test statistics using standardized abnormal returns follow Patell 1976 Many published studies use the Patell test see for example Linn and McConnell 1983 Schipper and Smith 1986 and Haw Pastena and Lilien 1990 Under the null hypothesis each Aj has mean zero and variance Cre The maximum likelihood estimate of the variance is 1 Rint Rm Rig Dj LUk Tp Rmk Rm De 2 2 SAj SA 14 105 where Tp 2 D Aj 2 k Tp sa i De Rmt is the observed return on the market index on day t Rm is the mean market return over the estimation period and D is the number of non missing trading day returns in the D day interval Tp through Tp used to estimate the parameters for firm 7 Define the standardized abnormal return or standardized prediction error
98. d creates new reports by merging two or three OUTSAS data sets from previous runs into a single sample ALLDAYS If the number of days covered by the event study PRE POST 1 exceeds 98 not all the individual day portfolio returns are listed by default Day 0 is always listed together with a proportionate number of contiguous PRE and POST days All the windows you request are printed regardless of their date ranges To obtain a complete printout of all the individual days portfolio returns specify ALLDAYS This option is ignored for TWIN event studies 138 BOOT Requests bootstrap tests for window abnormal returns Bootstrap tests are in addition to the normally reported parametric and nonpara metric tests BOOT implies the BUYHOLD and STDCSECT options It is not possible in the current version of Eventus to bootstrap TWIN event stud ies The option is not available unless it was specified on the original EVTSTUDY statement BTAIL 1 2 Selects one or two tailed bootstrap tests The default is 2 This option has no effect unless the BOOT option also is specified BUYHOLD Specifies buy and hold compounded return computation for win dows instead of the default additive cumulation The option is not available unless it was specified on the original EVTSTUDY statement CDCSI Requests the Collins and Dent 1984 test assuming cross sectional independence instead of the default standardized test The option is not available unless it or th
99. d cumulative re turn windows a choice of raw comparison period mean adjusted market adjusted or market model abnormal returns Simple statements allow the re searcher to run a complete event study from reading the CRSP stock database to printing results with a program as short as four lines This manual is not intended to be a textbook on event study methods al though citations to relevant literature are provided whenever possible For an overview of event study methods please see Peterson 1989 Binder 1998 and Mackinlay 1997 For a rigorous justification of standard event study procedures see Prabhala 1997 Please visit our Web site www eventstudy com for additional usage ex amples frequently asked question lists FAQs upgrade announcements documentation and software sales and technical support Eventus is analytical software and does not include data While Eventus can deal with all the CRSP stock databases mentioned in this manual your institution or firm may not subscribe to all of them If not some features of Eventus may not be of use to you Users can supply stock and index returns from non CRSP sources as well The symbols that appear in the statement descriptions in this manual have the following meanings A vertical bar indicates that only one of the words it joins may be chosen Anything within square brackets is optional Boldface type indicates that you enter the word exactly as it appears in the s
100. d merging saved event st di S oho a coa a g o a a e a ee oe ee 40 Sample Eventus program using OLDSTUDY to merge two saved event Studies ocea oi a E a a e ee Ee a 42 Eventus statements for an event study centered around a single vii 4 2 4 3 4 4 4 5 4 6 4 7 4 8 5 1 5 2 5 3 6 1 6 2 6 3 6 4 6 5 6 6 7 1 T2 7 3 8 1 g1 Eventus program for an event study of the Too Big to Fail POLOJ o i doem s Ge soa a fe a e eae ee Ea Part of the request file for the Too Big to Fail study First 62 lines of the user supplied returns file for the Too Big to Fail Study e e QOS Se e a ew ER A ee ee The Eventus sample listing and input report The Eventus parameter estimate listing o oa oaa aa Event study results gt s 6 a ropise tikt tossa Eventus statements for event studies cumulating returns be tween paired event dates o oa o o a ee eee Eventus statements for extracting saved event study results Example using EXTRACT to organize firm by firm event study results for further analysis 2 22004 Contents of SAS data set abnormalreturns produced by Fig re 5 2 CONG ak E oe a a le a ee OE EE ESS Eventus statements for an event parameter approach event study Eventus program to assemble data for the event parameter eae a e a ee a en ee Request file for the event parameter approach demonstration Eventus program to complete the event parameter
101. data are arranged in columns with each row reporting the PERMNO identification variable if applicable CRSP trading day or month number and prices All the data for the first stock are listed with each trading day on its own row followed by all the data for the second stock and so on Specify the fileref sas file shortcut or a DDname on a mainframe system to which to write the prices with the EXTFILE option on the PRICES statement On mainframe systems that require files to be pre allocated the file pointed to by the fileref should have a logical record length of 80 or longer a fixed block format and a block size that conforms to your system s rules TRADES This option is ignored unless NMS is specified on the PRICES state ment Specifies that the number of trades on each date should be read from the Supplemental Nasdaq Data and included in the output file VSAS Creates a vertical format SAS data set A SAS data set cannot be read by non SAS software The following variables are included in the data set PERMNO CUSIP if the CUSIPERM option is in effect CRSPDAY1 DATE BIDLO and ASKHI if applicable PRICE SHARES when applicable TRADES when applicable and any identifying variable you list on the REQUEST statement CRSPDAY1 is the CRSP day or month etc number for the beginning date you request DATE is the actual date of each price recorded as a SAS date variable PRICE is the security price and SHARES is the number of shar
102. dex file header names shares nasdin dists prtnun if a size decile index is to be used and tsdata MONTHLY Indicates that Eventus is to read from the monthly CRSPAccess database or that the appropriate fileref or libref already points to a CRSP SFA database CRSP data stored in SAS data sets or USERSTOK file that contains monthly data When working with CRSP SFA old for mat data files a filename statement environment variable or control language must associate the fileref CRSPSTOK and CRSPST2 etc when applicable with the proper monthly file if you use this option NASDAQ The NASDAQ option tells Eventus that the fileref CRSPSTOK is as sociated with an old Nasdaq file instead of an old NYSE AMEX file A filename statement environment variable logical variable or con trol language must associate the fileref CRSPSTOK with the old CRSP Nasdaq database if you use this option for an event study or to pull returns prices or volume data The NASDAQ option is not supported for CRSPAccess format nor for CRSP data installed in SAS data sets NLIBNAME libref Gives the sas libref of the CRSP Supplemental Nasdaq data file This option is for systems that store the CRSP data in the form of SAS data sets and is needed only in runs where the NMS option appears on the PRICES statement In the NLIBNAME library Eventus expects to find SAS data sets named nmshdr and nmsdata NONCRSP Denotes that user extracted stock and index return data
103. distribution per share for which the stock went ex on that date Two or more cash distributions with the same date are added and the total reported If the SPLITADJ option also is specified the reported distribution total reflects the same split adjustment as the price The DISTRIB option is not available when the HSAS option is specified unless only one trading day or month per stock is being extracted The variables DISTCODE containing the four digit CRSP distribution code and DIVAMT contain ing the cash distribution per share are added to any output data set EXTFILE Gives the fileref of the external file in which Eventus is to store the prices This option is only valid with the default TEXT or optional BINARY file format options If EXTFILE is not specified the prices will be printed to the fileref USERDATA HSAS Creates a horizontal format SAS data set A SAS data set cannot be read by non SAS software The data set will contain as many of the fol lowing variables as applicable PRI1 through PRInnnn BIDL1 through BIDLnnnn ASKH1 through ASKHnnnn TRAD1 through TRADnnnn PERMNO 141 EVENTDA1 EVENTDA2 SHARES and any ID variable specified on the REQUEST statement You must specify the SAS data set name using the OUTSAS option EVENTDA1 and EVENTDA2 are SAS date variables representing the be ginning and ending trading days of the interval you request The one to three digit number nnnn is the maximum number of trading days
104. djusting for splits and other stock distributions By default Eventus does not adjust stock prices bid and ask quotations trading volume or shares outstanding for stock splits and stock dividends Returns on the CRSP files already are adjusted You can specify SPLITADJ on the PRICES or VOLUME statement to adjust for any stock splits or stock dividends that occur between the first and last dates of data For example suppose that a 2 for 1 stock split occurs on the 50th of 100 days of data Without the SPLITADJ option Eventus will report the actual data for each of the 100 days With the SPLITADJ option the trading volume and shares outstanding for the second 50 days will be divided by 2 and the prices and bid ask quotations for the second 50 days will be multiplied by 2 91 Selecting an output file location Text or ASCII binary output is directed by default to the fileref USERDATA If the fileref is undefined a file named userdata dat is usually created in the current working directory To select a specific disk file in which to store the data use a filename userdata statement before the EVENTUS statement to give the path and name of the file to be created To use a different fileref besides userdata specify the fileref with the EXTFILE option Vertical and horizontal SAS data sets require a libref to point to the aggregate storage location typically a folder or directory where the data set should go If you specify the VSAS or HSA
105. e as the main daily data and has the member name sizeindx 162 References Arshadi Nasser and Thomas H Eyssell Regulatory Deterrence and Reg istered Insider Trading The Case of Tender Offers Financial Manage ment 1991 20 2 30 39 Bhagat Sanjai M Wayne Marr and G Rodney Thompson The Rule 415 Experiment Equity Markets Journal of Finance 1985 40 5 1385 1401 Binder John J The Event Study Methodology Since 1969 Review of Quantitative Finance and Accounting 1998 11 2 111 137 Boehmer Ekkehart Jim Musumeci and Annette B Poulsen Event Study Methodology under Conditions of Event Induced Variance Journal of Financial Economics 1991 30 2 253 272 Bollerslev Tim Ray Y Chou and Kenneth F Kroner ARCH Modeling in Finance A Review of the Theory and Empirical Evidence Journal of Econometrics 1992 52 1 2 5 60 Brickley James A Frederick H Dark and Michael S Weisbach The Economic Effects of Franchise Termination Laws Journal of Law and Economics 1991 34 1 101 132 Brockett Patrick L Hwei Mei Chen and James R Garven A new stochastically flexible event methodology with application to Proposition 103 Insurance Mathematics and Economics 1999 25 2 197 217 Brown Stephen J and Jerold B Warner Measuring Security Price Per formance Journal of Financial Economics 1980 8 3 205 258 Brown Stephen J and Jerold B Warner
106. e do not specify the fileref request 101 102 Appendix A Technical Reference A 1 Event Study Prediction Errors Market Model Abnormal Returns Assume that security returns follow a single factor market model Rit Qj By Rmt Ejt where Rj is the rate of return of the common stock of the jt firm on day t Rm is the rate of return of a market index on day t jt is a random variable that by construction must have an expected value of zero and is assumed to be uncorrelated with Rmt uncorrelated with Ry kj not autocorrelated and homoscedastic 3 is a parameter that measures the sensitivity of Rj to the market index Define the abnormal return or prediction error for the common stock of the j firm on day t as Aje Rye j Bj Rm where the coefficients and B are ordinary least squares estimates of a and 3 The average abnormal return or average prediction error AAR is the sample mean N 2 An AAR ss N where t is defined in trading days relative to the event date e g t 60 means 60 trading days before the event 103 Over an interval of two or more trading days beginning with day T and ending with T gt the cumulative average abnormal return is 1 N R CAART 7 Fs XO Aj N z 1 t T Over an interval of two or more trading days beginning with day T and ending with 75 the average compounded abnormal return is ACART T 1 N T gt A T2 T 1 A ob wT Re
107. e EGLS option appeared on the original EVTSTUDY statement CSECTERR Substitutes a cross sectional standard deviation for the default time series standard deviation in non standardized t statistic compu tations DETAIL DETAIL FULL Sce EVTSTUDY above These need not have been specified on the original EVTSTUDY statement in order to work here ID variable Names the variable to be used as an observation event iden tifier The variable must exist on the INSAS data set INSAS 2 3 Jibref membername Selects the Eventus abnormal return data set to use INSAS2 and INSAS3 optionally select additional data sets to combine with the first data set to run a merged event study Each of these must first be created by the Eventus EVTSTUDY statement using OUTSAS and PACKAGE 1 All data sets to be combined must have had the same ESTLEN if any on the original REQUEST statement JACKNIFE Designates the jackknife test instead of the generalized sign test as the nonparametric test to accompany non standardized method parametric tests The option is valid only if it also appeared on the original EVTSTUDY statement 139 MEDIAN This option affects printed output but only when the PAGE WIDE option is specified on the EVENTUS statement The MEDIAN option causes the daily or monthly median abnormal return to be printed instead of the number positive and negative and nonparametric test The option affects only the daily or monthly output not the output
108. e ee Oe eo 4 Event Studies Using Non crsp Data 4 1 Event Studies Centered on a Single Event Date 4 2 An Event Study Example 04 4 3 Abnormal Returns between Paired Events The TWIN Option 5 Extracting Event Study Results for Further Analysis 5 The EVENTUS Statement lt s s sm saca oe we Raw RDS De The WINDOWS Statement lt 2 48 ca soeg aiea ew we Re RES 5 3 The EXTRACT Statement 44444 4454 edo eR a ee ROS OA Usage examipl lt n kee bb He cD SORE Y BEE So 6 Event Studies Using the Event Parameter Approach 6 1 Statements for the Event Parameter Approach 6 2 An Event Parameter Approach Example 7 Obtaining Returns Prices Volume Number of Trades and Shares Outstanding from the CRSP Database 7 1 The EVENTUS statement 2 464 446 eee Rw ew 13 13 31 37 40 43 43 54 59 63 63 64 64 67 83 T2 The REQUEST statement e sedane 406 o mari d Bae A ac a 84 7 3 The RETURNS PRICES and VOLUME statements 87 Converting Calendar Dates to CRSP Trading Day or Month Numbers Using DATECONV 93 8 1 The EVENTUS statement aoaaa a e a e 0000084 93 8 2 The DATECONV statement aooaa ea c a e aa 94 Converting CUSIP Identifiers Using CUSIPERM 99 9 1 The EVENTUS statement ooa a aa 100 9 2 The CUSIPERM statement 0004 100 Technical Reference 103 A 1 Event Study Prediction Errors 103 A 2 Event Study Test Statistics
109. ecifies a one day or one month window containing the second day or month after the event date When begin and end are equal no cumulation or compounding occurs the only purpose for such a specification would be to cause a single date to appear in the window output section for ease of reference If the WINDOWS statement is omitted Eventus reports three windows PRE 2 1 0 and 1 POST The results for the windows appear at the bottom of the same page on which the results for individual dates are printed or on a separate page if there is not enough space The EVTSTUDY statement The EVTSTUDY statement tells Eventus to run an event study The options permit you to change the default event period control the handling of ob servations with missing returns select market indices and methods for com puting abnormal returns and test statistics increase or decrease the amount of printed output and store results in a SAS data set 23 Selecting the amount of printed output EVTSTUDY normally produces a report of the results of searching the CRSP database for your sample The report lists the PERMNO identifying variable and event date from your request file together with the name of the firm from the CRSP database The report also tells for each firm how many returns Eventus found in the estimation period and in the event period If there was a problem for example if the date you asked for was outside the range of data for that f
110. ectional regression with statements like the following data regression_variables merge abnormalreturns explanatory by permno proc reg data regression_variables model wincar2 regressors weight weight2 where the word regressors is replaced by the name of one or more explana tory variables The weight statement can be omitted for ordinary least squares regression The procedure allows many additional options includ ing hypothesis tests using a heteroscedasticity consistent covariance matrix Please see sas documentation for further details 69 70 Chapter 6 Event Studies Using the Event Parameter Approach In the conventional approach the market model or other benchmark param eters are estimated over a period that excludes the event dates to be tested The abnormal returns on the event dates then are estimated in a second stage In the event parameter approach the market model is augmented by adding dummy variables to identify event periods allowing the joint es timation of the market model parameters and abnormal returns Karafiath 1988 provides a tutorial on the event parameter approach Malatesta 1986 describes the approach using joint generalized least squares estimation In principle there can be a unique dummy variable for each date in the event period a dummy variable for each of several event dates or a common dummy variable for a range of dates Eventus implements the event parameter approach by defining
111. ent or the operating system must associate the libref with the file or directory used as the library A related option PACKAGE determines the content of the file If you intend to use OLDSTUDY to reprint or merge event study results specify PACKAGE 1 If you plan to use EXTRACT to create dependent variables for cross sectional analysis of abnormal returns see Chapter 5 specify PACK AGE DG or just PACKAGE D if you do not need weights or standardized re turns Table B 1 on page 134 lists the complete set of PACKAGE specifications available 3 2 An Event Study Example O Hara and Shaw 1990 conduct an event study of the statement by the U S Comptroller of the Currency in 1984 that some banks are too big to be allowed to fail Figure 3 2 shows the Eventus program to replicate the study for the sample of 22 banks that were insolvent according to a measure proposed by Swary 1986 The EVENTUS statement needs no options for this study O Hara and Shaw state that their estimation period is days 55 through 6 relative to the event date of September 20 1984 To replicate their estimation period we specify EST 6 and ESTLEN 50 on the REQUEST statement We identify the individual bank stocks using the numbers from 31 Figure 3 2 Eventus program for an event study of the Too Big to Fail policy filename request C Research Banks permnos dat eventus title O Hara and Shaw JF Dec 1990 pp 1587 1600
112. entus reports a generalized sign test statistic of 0 04 for day 0 The statistic is based on the normal approximation to the binomial so it has an approximately standard normal distribution A table of the standard normal distribution shows that the probability of observing a value of z lt 0 04 is 0 5160 Thus O Hara and Shaw 1990 appear to use the generalized sign test The program does not include a WINDOWS statement so Eventus reports three windows that jointly cover the entire event period In this study the authors focus on day 0 because there is no reason for any other day to have a stock price reaction When you run Eventus the SAS log window or log file depending on the mode of SAS operation reports the completion of data steps and procedures that Eventus executes internally Most Eventus users will find that these reports have little meaning for them and they may ignore most of what is in the log However it is still a good idea to look briefly through the sas log for messages that begin with EVENTUS NOTE EVENTUS WARNING or EVENTUS ERROR If you contact us for technical support to help resolve a problem with an Eventus run please include the entire log window contents or log file resulting from a single run 3 3 Abnormal Returns between Paired Ev ents The TWIN Option The TWIN option provides a means of computing portfolio cumulative average abnormal returns over periods that vary in length from one firm to another
113. entus where to find saved event study SAS data sets The libref and membername parameters should match those on the OUTSAS options of the original EVTSTUDY statement s See page 31 The event studies to be merged all should have had the same PRE POST and abnormal return method options in the original programs If the original programs specified different variable names for ID or incompatible identifying vari able formats such as numeric and character omit the ID option from the OLDSTUDY statement Selecting benchmarks and standardized tests In general the OLDSTUDY statement detects the abnormal return methods used in the original event study and takes them as the defaults To omit mar ket adjusted returns specify NOMAR To omit market model returns specify NOMM To omit just the standardized abnormal return tests using the mar ket model specify NOSTD To omit everything except the market model with standardized abnormal return tests specify STDONLY To tell Eventus to report standardized abnormal return tests for non market model methods specify STDALL on the OLDSTUDY statement See Section 3 1 for a discussion of the DETAIL RANKTEST SERIAL and TAIL options Figure 3 10 displays a sample Eventus program to merge two event studies Notice that because Eventus does not need to read a request file or a CRSP database you do not use a REQUEST statement with OLDSTUDY 41 Figure 3 10 Sample Eventus program using OLDSTUDY
114. eries OUTSAS libref membername Specifies a SAS data set in which to save ab normal returns and other data for each event Eventus will create the SAS data set named if it does not already exist or replace it if it does The use of a different second level name for each data set permits a single SAS data library to contain multiple Eventus event study data sets See the PACKAGE option below for a detailed description of the contents of the event study data set OVERLAP Suppresses checking for overlapping estimation period and event period PACKAGE specifier specifier Determines the content of the sas data set created when OUTSAS is specified Table B 1 lists the available spec ifiers The OUTSAS file always contains at least the variables PERMNO NAME identifying variable if any CRSPDAY and EVENTDAT or CRSPDAY1 CRSPDAY2 EVENTDA1 and EVENTDA2 for TWIN event studies WEIGHT_ and RESTYPE When the CUSIPERM option appears on the REQUEST state ment the saved data set includes the CUSIP variable in addition to PERMNO When the SIZEINDX option appears on the REQUEST state ment the saved data set includes the variable __CAP__ the size based portfolio number CRSP assigns the number 1 to the smallest market capitalization portfolio and 10 to the largest Note that a stock s size portfolio membership changes from year to year and can differ between the now standard NYSE AMEX Nasdagq database and the formerly com mon exchange specific C
115. eriod select market indices and methods for computing abnormal returns and test statistics increase or de crease the amount of printed output and store results in a permanent SAS data set Selecting the amount of printed output EVTSTUDY produces a report of the results of reading your data The report lists the CUSIP identifying variable and event date from your request file and the name of the issuer if you specified the NAME option on the REQUEST statement The report also tells for each firm how many returns Eventus 48 found in the estimation period and in the event period If you prefer not to get this report specify the NONAMES option Unless you specify NOPLIST on the EVTSTUDY statement you also get a list of the estimation period return statistics for each observation and the mean and median for the sample The report includes market model alphas betas and residual standard deviations fraction of market model residuals positive and additional statistics if space permits The last set of output from EVTSTUDY contains the event time results By default one page is generated for each abnormal return method Each page lists the portfolio mean abnormal return test statistic number positive and negative and significance levels for the event date day 0 and as many as 34 surrounding days 94 if PAGE WIDE appears on the EVENTUS statement depending on PRE and POST At the bottom of the page for each window you req
116. ermine whether the CRSP database contains NYSE AMEX Nasdaq Nasdaq only or some other security universe the SP500 and COMPOSIT options are the same except for the way the output is labeled It is up to you to make sure the database you use is the one you intended SIZEINDX This option allows the use of size portfolio returns Eventus uses the size capitalization portfolio membership information on the CRSP stock file to determines which size portfolio return to use All tests described in this manual are conducted as they would normally be except that the index return is a size portfolio return rather than the return on a broader market index A SAS filename statement operating system command control lan guage statement or environment variable must associate the fileref sizeindx with the size portfolio index file The size portfolio index file must be in one of two specific formats either the CRSP SFA ASCII character Indices Decile File format or the Eventus size index file format The Eventus size index file format is binary or character depending on whether the current program is using binary SFA or CRSPAccess or character files If the sizeindx fileref points to a CRSP Indices Decile File specify SIZEINDX CRSP on the REQUEST statement When a CRSP Indices Decile File is used in conjunction with a CRSPAccess database the specific Indices Decile File to use depends on whether the CRSPAccess indices and portfolio assignments module was i
117. es specify NOCLOSE on the PRICES statement Thus the option combination BIDASK NOCLOSE reports only secondary prices from the main time series arrays An additional feature of the NOCLOSE option is that when combined with the NMS option it disables the mixing of Supplemental Nasdaq bid and ask quotations with secondary prices Instead Eventus reports both bid and ask prices from the Supplemental Nasdaq Data Arrays and secondary prices from the main time series structures as four separate variables when you specify BIDASK NMS NOCLOSE Reporting the number of shares outstanding with stock prices or volume The SHARES option is valid only on the PRICES and VOLUME statements The SHARES option tells Eventus to store the number of shares outstanding in thousands along with the share price or trading volume In a text format file the default output format the number of shares appears after the price or volume on the same line In a binary format output file the number of shares appears immediately following the price or volume in the final next to final if TRADES is specified four bytes or character positions of the output record in IB4 format like a FORTRAN INTEGER 4 variable In a VSAS or HSAS output file the number of shares is a variable named SHARES Deriving cash distributions The DISTRIB option on the PRICES statement tells Eventus to read dividends and other cash distributions from the CRSP distribution structure Daily or mo
118. es of the stock and index return file as sociated with the fileref USERSTOK The first line is a header line for the user s reference Eventus ignores this line because of the SKIP option on the EVTSTUDY statement The next 61 lines contain the daily data for the 50 day estimation period followed by the 11 day event period The file contains 1 61 x 22 1343 lines in all Eventus reads only the first four columns of data from this file We happen to have included the calendar date on each line for reference but Eventus does not use this information in any way Had we not specified the ID option on the REQUEST statement we would have needed only three columns in this file for the CUSIP stock return and index return The first page of Eventus output lists the estimation period describes the sample size explains the statistical significance symbols and may contain other information depending on the options selected For an example see Chapter 3 In figure 4 5 is a listing of the sample and a detailed report of the number of returns Eventus found for each observation Figure 4 6 presents the Eventus parameter estimate listing The user can suppress this listing by specifying NOPLIST on the EVENTUS statement Eventus prints more statistics including the first order autocorrelation of the market model residuals if the PAGE WIDE option appears on the EVENTUS statement The event study results appear in figure 4 7 The results are substantial
119. es outstanding in thousands There is one observation in the SAS data set for each trading day in the interval you request Use the VSAS option if you plan to process the data with a SAS program that uses BY variable techniques Specify the SAS data set name using the OUTSAS option 144 B 8 REQUEST Statement This statement is required after the EVENTUS statement and before the EVT STUDY RETURNS PRICES or VOLUME statement It is only used if the EVTSTUDY RETURNS PRICES or VOLUME statement is used It reads the user request file and the CRSP index file and merges the two The user may specify these options AUTODATE Specifies that a calendar date in the request file that is not a trading day should be converted to the following trading day Eventus converts a date in the request file that follows the last trading day listed in the crspindx file to a SAS missing value which may cause unexpected error messages AUTODATE BACK Specifies that a calendar date in the request file that is not a trading day is to be converted to the preceding trading day Eventus converts a date in the request file that precedes the first trading day listed in the crspindx file to a SAS missing value which may cause unexpected error messages COMPOSIT See SP500 CUSIP Specifies that the CRSP database being used is sorted either sorted or indexed if in SAS data set form by CUSIP and the request file contains CUSIPs instead of PERMNOs The default is that
120. esults as using a value weighted index O Hara and Shaw use a value weighted index for which we have collected data The POST 5 and PRE 5 specifications mean that the event period returns span days 5 through 5 Thus after the 50 days of estimation period data for each firm we must provide 11 days of event period data centered on September 20 1984 TAIL 2 means that significance level symbols will reflect two tailed tests Finally SKIP 1 instructs Eventus to ignore the first line of the USERSTOK returns file Figure 4 2 Eventus program for an event study of the Too Big to Fail policy filename request C Folder request_filename ext filename userstok C Folder userstok_filename ext eventus noncrsp title O Hara and Shaw JF Dec 1990 pp 1587 1600 title2 Replication of Table III C Insolvent Bank Sample request est 6 estlen 50 id Bankname idfmt 10 evtstudy nomar nostd buyhold value post 5 pre 5 tail 2 skip 1 59 Figure 4 3 Part of the request file for the Too Big to Fail study 56480910 840920 Manufactur 16161010 840920 ChaseManha 17303410 840920 Citicorp 06605010 840920 Bankameric Figure 4 3 displays the first few lines of the request file Note that Eventus automatically sorts the request file The CusIPs could be any eight character strings as long as they are consistent between the request file and the returns file Figure 4 4 shows the first 62 lin
121. event date event study The options are described below 43 Figure 4 1 Eventus statements for an event study centered around a single date filename request G Some Folder Filename extension filename userstok G Some Folder Filename extension EVENTUS NONCRSP MONTHLY WEEKLY QUARTERLY PAGE WIDE TITLE text TITLE2 text REQUEST ID variable IDFMT format DATEFMT MMDDYY YYMMDD DDMMYY DATE EST value value POOL ESTLEN n NAME WINDOWS begin end begin end EVTSTUDY NOPLIST DETAIL DETAIL FULL RAW CP NOMAR NOMM NOSTD STDONLY STDALL VALUE BOTH SW GARCH EGARCH PRE periods POST periods OVERLAP CSECTERR STDCSECT SERIAL RANKTEST MEDIAN BUYHOLD TAIL 1 2 OUTSAS libref membername PACKAGE specification 44 The EVENTUS statement The NONCRSP option on the EVENTUS statement indicates that you are supply ing the stock and index return data for the event study so that Eventus does not expect to search a CRSP database A SAS filename userstok statement or the operating system must associate the fileref USERSTOK with the file con taining the returns The content and format of this file is discussed under the REQUEST statement below MONTHLY WEEKLY QUARTERLY With non CRsP returns input the return in terval options provide only a description of the data for labeling purposes they have no effect on the analysis Eventus uses the descript
122. example First part of the results for the event parameter approach demonstratio os 60 684 6 RG RS ROL ERR RR OHS Remaining results for the event parameter approach demon SUTAHOM o 6 geek MAR eG ee e eS BG OS Eventus statements to read stock returns from a CRSP database Eventus statements to read stock prices from the CRSP database Eventus statements to read trading volume data from the CRSP GA eee eis ees oe A se Oe ee we we Eventus statements for converting calendar dates to CRSP trad Ing day OF month numbers eo e rosa kero ee ew RS Eventus statements to convert CUSIP identifiers to CRSP per manent identification numbers s sooo a a viii 59 56 57 58 59 60 61 64 68 69 72 77 78 79 80 81 84 85 86 94 Chapter 1 Introduction Eventus performs event studies using data read directly from CRSP stock databases or pre extracted from any source Eventus can also read raw re turns prices bid and ask quotations trading volume number of trades and shares outstanding from the CRSP database Eventus can print the raw data or store them in several disk file formats The Eventus system includes util ity programs to convert calendar dates to CRSP trading day numbers con vert CUSIP identifiers to CRSP permanent identification numbers and extract event study cumulative or compounded abnormal returns for cross sectional analysis Eventus provides user control over estimation periods an
123. ey The Specification and Power of the Sign Test in Event Study Hypothesis Tests Using Daily Stock Returns Journal of Financial and Quantitative Analysis 1992 27 3 465 478 Cowan Arnold R Nonparametric Event Study Tests Review of Quanti tative Finance and Accounting 1992 2 4 343 358 Cowan Arnold R Tests for Cumulative Abnormal Returns over Long Peri ods Simulation Evidence International Review of Financial Analysis 1993 2 1 51 68 Cowan Arnold R Nandkumar Nayar and Ajai K Singh Stock Returns Before and After Calls of Convertible Bonds Journal of Financial and Quantitative Analysis 1990 25 4 549 554 164 Cowan Arnold R and Anne M A Sergeant Trading Frequency and Event Study Test Specification Journal of Banking and Finance 1996 10 1731 1757 Dopuch Nicholas Robert W Holthausen and Richard W Leftwich Ab normal Stock Returns Associated with Media Disclosures of Subject to Qualified Audit Opinions Journal of Accounting and Economics 1986 8 2 93 117 Draper Norman R and Harry Smith Applied Regression Analysis Second Edition John Wiley amp Sons New York 1981 Giaccotto Carmelo and James M Sfiridis Hypothesis Testing in Event Studies The Case of Variance Changes Journal of Economics and Business 1996 48 4 349 370 Haw In Mu Victor S Pastena and Steven B Lilien Market Manifesta tion of Nonpublic Information
124. f missing days is q the market model abnormal return for the first post missing day t is 6a Ast Ry la Jg 1 a a bi 5 Rent h 0 while the maximum likelihood estimate of the variance of A is eae 1 b Feta q Ran T D DEA Rmk E Rm 118 Table A 3 Eventus special SAS missing values for missing returns from the CRSP database Reason CRSP missing Special SAS return code missing value No portfolio assignment 44 Xx for excess return file Missing delisting price 55 D gt 10 trading days between 66 G non missing prices No trading on Nasdaq 77 oT Date outside return 88 R date range No price available 99 B Event dropped NA M Date outside Eventus NA A search range Day in estimation period after NA N a missing return day 119 120 Appendix B Reference Guide to Eventus Statements B 1 CUSIPERM Statement This statement is used after the EVENTUS statement to convert CUSIP iden tifiers to PERMNOs The statement reads the user request file and performs the conversion The user may specify these options COLUMN n Use when the CUSIP is not the first item on each line of the request file Substitute the starting column for the CusIP for n This option is not needed if only blanks precede the CUSIP on each line EXTFILE Gives the fileref of the external file in which Eventus is to store the updated copy of the request file The default is request On mainf
125. for NONCRSP event studies the user supplied returns file The request file is an external file you use it to supply the PERMNO or CUSIP identifiers for your sample the event dates and optional other information Each line of the request file contains data for a single security base date combination The base date is used as day or month 0 in the event study The base date in the event parameter approach is not necessarily an event 72 date It is simply a reference point around which the event windows will be built At a minimum each line must contain a PERMNO or CUSIP and base date For example a line of the request file could look like this 10123 19900512 Spacing does not matter as long as the PERMNO or CUSIP comes first then the base date with at least one space between If the request file contains CUSIPs you need to specify either the CUSIP option or the CUSIPERM option Use the CUSIPERM option when directly accessing a CRSP database that is sorted by PERMNO to instruct Eventus to convert the CUSIPs in the request file to PERMNOs Use the CUSIP option when Eventus is not reading directly from a CRSP database or when the CRSP data are sorted by CUSIP The rest of this section describes REQUEST statement options for the pro cessing of dates and market and stock returns and for the construction of the estimation period A SAS filename statement or the operating system must associate the fileref REQUEST with the request file Inc
126. for each value of WEIGHT When CRSP excess returns based upon beta or standard deviation ranks are in use WEIGHT_ takes the value EXCES The RESTYPE variable takes a three character value of raw for unad justed raw returns CP for comparison period mean adjusted MAR for market adjusted MM for OLS GARCH or EGARCH market model adjusted SW for Scholes Williams market model adjusted returns or XS for CRSP Excess Return File excess returns A 5 Missing Returns CRSP codes any missing return on its files as an integer strictly less than 1 0 Eventus internally converts each of these CRSP missing return codes to the SAS special missing value Special missing values work the same way as the regular SAS missing value in all arithmetic operations When an estimation period contains a sequence of one or more missing values Eventus does not use the first succeeding non missing return The reason is that the first non missing return is a multi period return Permit ting multi period returns could have unexpected consequences for parameter estimates The first non missing return following a sequence of missing esti mation period returns is replaced by the special missing value N When a sequence of one or more returns is missing in the event period Eventus adjusts the abnormal return computation procedure to account for the multi day character of the first post missing return For example if the number o
127. for windows NOPLIST Suppresses the printing of the estimation period statistics that normally appear between the input report and the event study results NOSTD STDONLY NOSTD omits standardized abnormal returns from the output provided that STDONLY was not specified with the original event study STDONLY prints only market model standardized abnormal re turn tests provided market model abnormal returns were included in the original event study RANKTEST Designates the rank test instead of the generalized sign test as the nonparametric statistic to appear with the non standardized parametric tests The option is valid even if it was not specified on the original EVTSTUDY statement STDALL Specifies that the standardized test statistic be computed for the non market model benchmark s as well as for the market model The option is ignored in TWIN event studies All applicable standardized test options then affect all benchmarks in use For example the SERIAL STDCSECT and BOOT options affect standardized tests for market ad justed returns as well as market model returns when STDALL is in effect STDCSECT Specifies that the standardized cross sectional test Boehmer Musumeci and Poulsen 1991 be substituted for the Patell z test in the standardized method This option is valid only if it was specified on the original EVTSTUDY statement This option implies SERIAL option SW Produces Scholes Williams 1977 market model results in add
128. format stock files The filerefs of the files must be CRSPSTOK and CRSPST2 The user must supply two request files corresponding to the two stock files DUAL is the same as REQFILES 2 This option is obsolescent because of the ubiquity of merged NYSE AMEX Nasdagq data The DUAL option is not supported for CRSPAccess format nor for CRSP data installed in SAS data sets 126 ELIBNAME libref This option is only used in situations when both the ESTINTER and LIBNAME options also are specified Gives the SAS libref of the CRSP stock files to be searched for the estimation period return data This option is for systems that store the CRSP data in the form of SAS data sets In the ELIBNAME library Eventus expects to find SAS data sets named cal the CRSP stock index file for the estimation period header and tsdata ESTINTER interval The holding interval of the estimation period returns if different from the holding interval of event period returns Valid values of interval are YEAR QUARTER MONTH WEEK and DAY EXCESS When working with an SFA format CRSP database specifies that the CRSP excess returns file is associated with the fileref CRSPSTOK When working with a CRSPAccess format database specifies that Even tus should substitute a beta matched standard deviation matched if SPORT is also specified portfolio return for the market index return Requires an add on CRSP data subscription FBIN 0 2 When your site receives the Eventus i
129. g In Figure 2 4 the name of Integrated Brands is truncated to make room for share class letter A Full firm names can be printed by using landscape orientation described in the next chapter The third page presents firm by firm and sample mean and median statis tics of returns and estimated market model parameters By default the market model is estimated by ordinary least squares using all available data from a 255 trading day estimation period ending 46 trading days before the event date In Figure 2 5 the alpha beta and residual standard deviation are the estimated market model intercept slope and root mean squared error respectively The final three pages of output present the event study results The event period is defined by default as from 30 days before through 30 days after the event date and is broken into three windows for abnormal return cumu lation the pre event period days 30 through 2 days 1 and 0 which is the period most commonly examined for the immediate impact of the event and the post event period days 1 through 30 The researcher can se lect different event period and window definitions using options described in later chapters By default Eventus reports three alternative return bench mark and standard error combinations Figure 2 6 presents market adjusted returns which are simple differences between the stock return and the mar ket index return that do not use the market model Figure 2 7
130. he TEXT format This type of file like the TEXT file can be read by any major computer language including FORTRAN SAS and C EXTFILE Gives the name or fileref of the external file in which Eventus is to store the volume data If the argument of EXTFILE is a file name it may include the path but must not include any blank or period On most systems a dat extension will be added to the file name automatically On mainframe systems where applicable the file should have a disposition of NEW or OLD a fixed block format and a logical record length of 80 The EXTFILE option is not for use with the VSAS and HSAS options which select SAS data set rather than external file output HSAS Creates a horizontal format SAS data set A SAS data set cannot be read by non SAS software The data set will contain as many of the following variables as applicable VOL1 through VOLnnn PERMNO EVENTDA1 EVENTDA2 SHARES and any identifying variable you list on the REQUEST statement You must specify the SAS data set name using the OUTSAS option 155 EVENTDA1 and EVENTDA2 are SAS date variables representing the begin ning and ending trading days of the interval you request The one to three digit number nnn is the maximum number of trading days in any of the intervals you requested For example if the number of trading days you ask for ranges from 2 for some firms to 90 for others nnn is 90 For a firm in this example with only two trading days V
131. he results Eventus will create a new data set unless the one named already exists in the SAS data library in which case the existing one will be overwritten A SAS libname statement or the operating system must associate the libref with the aggregate storage location a folder or directory on most systems used as the library A related option PACKAGE determines the content of the file If you intend to use OLDSTUDY to reprint or merge event study results specify PACKAGE 1 If you plan to use EXTRACT to create dependent variables for cross sectional analysis of abnormal returns see Chapter 5 specify PACK AGE DG or just PACKAGE D if you do not need weights or standardized re turns Table B 1 on page 134 lists the complete set of PACKAGE specifications available 4 2 An Event Study Example In this section we replicate part of the O Hara and Shaw 1990 study as in Chapter 3 using non CRSP return data Figure 4 2 shows the Eventus program to run the replication The EVENTUS statement needs only the NONCRSP op tion O Hara and Shaw state that their estimation period is days 55 through 6 relative to the event date of September 20 1984 To replicate their esti mation period we specify EST 6 and ESTLEN 50 on the REQUEST statement This means that the first 50 days of returns data for each firm must cor respond to this estimation period We identify the individual bank stocks using an abbreviated firm name calling the identifying
132. ics Note that the serial dependence that the SERIAL option corrects is not due to any presumed dependence in the true market model error term but occurs because all of the abnormal return estimators being cumulated are functions of the same estimators of the market model parameters The derivation of the corrected standard error used by Mikkelson and Partch 1988 requires that the abnormal return be interpreted as a forecast error 52 Alternative nonparametric tests Normally Eventus prints the generalized sign test statistic see Cowan 1992 for each day in the event period and for each window with both standardized and non standardized tests If you specify RANKTEST on the EVTSTUDY state ment Eventus instead will report the nonparametric rank test introduced by Corrado 1989 with non standardized tests The JACKNIFE option produces the jackknife test developed by Giaccotto and Sfiridis 1996 The RANKTEST and JACKNIFE tests are mutually exclusive The generalized sign test still will be reported with standardized abnormal return method results Reporting median abnormal returns This option is needed only when the PAGE WIDE option appears on the EVENTUS statement With the default output format PAGE TALL Eventus always reports median and mean abnormal returns With the optional wide output format Eventus reports median abnormal returns in place of the num ber positive and negative if you specify MEDIAN on the EVTSTUDY statement
133. ies The Options Eventus lets you configure many details of your event study to suit you This chapter tells how 3 1 Event Studies Centered on a Single Event Date The typical event study analyzes returns around one type of event date at a time If you need to cumulate returns between two paired event dates when the number of periods between them varies from firm to firm see section 3 3 after reading this section Figure 3 1 displays the Eventus statements to run an event study with a single event date The options are described below The EVENTUS statement Use options on the EVENTUS statement when you need to provide Eventus with special information about the CRSP stock database to be used and to set the orientation of printed output By default Eventus assumes that you are using a CRSP daily database MONTHLY To use the monthly CRSP database instead of the default daily specify MONTHLY on the EVENTUS statement For example EVENTUS MONTHLY 13 Figure 3 1 Eventus statements for an event study centered around a single date filename request G Some Folder Filename extension EVENTUS MONTHLY EXCESS ESTINTER DAY MONTH PAGE WIDE TITLE text TITLE2 text 3 REQUEST CUSIP CUSIPERM ID variable IDFMT format GROUP variable GRWEIGHT DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP AUTODATE BACK NODIVIDX SP500 COMPOSIT SIZEINDX CRSP EST value value POOL ESTLEN n MINESTN n
134. ificant at 001 38 Figure 3 8 Eventus statements for event studies cumulating returns between two firm specific event dates filename request G Some Folder Filename extension EVENTUS TWIN LIBNAME CRSP MONTHLY EXCESS ESTINTER DAY MONTH PAGE WIDE TITLE text TITLE2 text 3 REQUEST CUSIP CUSIPERM ID variable IDFMT format GROUP variable GRWEIGHT DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP AUTODATE NODIVIDX SP500 COMPOSIT EST value value POOL ESTLEN n MINESTN 1 SHORT WINDOWS EVENT1 descriptor EVENT2 descriptor2 EVTSTUDY NONAMES NOPLIST SIC DETAIL NOSTD VALUE BOTH SPORT SW GARCH EGARCH CSECTERR STDCSECT EGLS CDCS j SERIAL MEDIAN BUYHOLD TAIL 1 2 OUTSAS libref membername PACKAGE spccifier on the EVENTUS statement and the specification of the WINDOWS statement The descriptors in the WINDOWS statement are 1 11 character names by which you want Eventus to designate the two event dates on the output Blank spaces are not allowed within either descriptor See Section 3 1 for descriptions of the remaining options Each line of the request file must contain a pair of event dates separated by one or more blank spaces in the location where the single event date goes in a normal event study request file 39 Figure 3 9 Eventus statements for reprinting and merging saved event studies EVENTUS TWIN MONTHLY EXCESS PAGE WIDE
135. ight brackets mean that the ratio is more positive than in the estimation period NOTE Useable returns means all nonmissing returns except the first day after a missing estimation period return statement The event study results appear in figure 3 7 The results are substantially the same as O Hara and Shaw 1990 report The authors report average and median abnormal returns on day 0 of 0 42 and 0 24 while Eventus re ports 0 43 and 0 24 If we use the OUTSAS option to save the event study then run a program with the EXTRACT statement discussed in Chap ter 5 to examine the results we find that 0 0042 is the truncated value of the mean 0 0042566 Eventus correctly rounds the result to 0 43 for reporting only not computation Thus truncation may explain the 0 01 discrep ancy in the reported average abnormal return O Hara and Shaw report a t statistic of 0 78 and Eventus reports 0 77 O Hara and Shaw compute the es timation period residual standard deviation by dividing the sum of squared differences by 49 or one less than the number of days in the estimation period Eventus divides by two less than the number of days because we es timate two market model parameters The difference in computation of the 34 Replication of Table III C Results of Daily Stock Returns Input Swarynum PERMNO 1 oono AN 10 12 13 14 15 16 17 18 19 20 21 22 48223 41718 47079 58827 47896 46842 5167
136. in any of the intervals you requested For example if the number of trading days you ask for ranges from 2 for some firms to 90 for others nnnn is 90 For a firm in this example with only two trading days PRI3 through PRI90 contain missing values Because the SAS system stores a large amount of information on each variable for its own use a horizontal format data set occupies considerably more disk storage space than a vertical format data set see the VSAS option NMS Instructs Eventus to obtain bid and ask quotes if you also specify BIDASK and the number of trades if you also specify TRADES from the Supplemental Nasdaq Data Arrays When available quotes from the Supplemental Nasdaq Data Arrays CRSP names Bid and Ask replace secondary prices CRSP names Bid or Low Price and Ask or High Price To prevent the replacement of secondary prices and in stead have Eventus report both secondary prices and Supplemental Nas daq quotes when the latter are available specify NMS NOCLOSE on the PRICES statement When both NMS and NOCLOSE appear on the PRICES statement Eventus produces an output file or data set that contains both the secondary prices and the Supplemental Nasdaq quotes The secondary prices are identified as BID OR INTRADAY LOW and ASK OR INTRADAY HIGH in an external file or by the variable names BIDLO and ASKHI in a vertical SAS data set or BIDL1 through BIDLnnnn and ASKH1 through ASKHnnnn in a horizontal SAS data set The Supple
137. ing variable so that each observation is weighted equally GRWEIGHT Valid only if the GROUP option is specified Denotes that the request file contains a group weight variable This variable expressed as a decimal specifies the weight to be given the individual stock within its group portfolio All the weights for a single group should sum to 1 If the request files is a SAS data set the variable name of the group weight must be grweight ID variable Names the variable to be used as an observation event iden tifier The identifying variable may be of any data type If you also specify INSAS the identifying variable must exist on the permanent SAS data set specified Specify IDFMT see below also IDFMT format Gives the format of the identifying variable in the external request file For example if your identifying variable is a one to four digit integer specify IDFMT 4 INSAS libref membername Used when the request file is a SAS data set The data set must contain these variables 147 PERMNO numeric variable containing the 5 digit CRSP permanent number unless the CUSIPERM option or CUSIP option is specified or the NONCRSP option is in effect CUSIP character variable of length eight containing the CUSIP identifier for CRSP input or containing any identifier for non CRSP input The CUSIP variable on the input data set is ignored unless the CUSIPERM option or CUSIP option is specified or the NONCRSP option is in effect
138. ion must be either a valid SAS date informat or the word CRSP The word CRSP tells Eventus to look for a 1 to 4 digit integer representing a CRSP trading day or month number 1 July 2 1962 on the daily calendar except 1 December 14 1972 for early versions 1985 and 1987 of the Nasdaq database Leading zeroes need not but may be included in the CRSP day number Any format other than CRSP must be a valid SAS date format although the period at the end is optional The default DATEFMT YYMMDD accommodates both two digit and four digit years automatically EST periods and POOL These options are provided for compatibility with the REQUEST statement they have no effect when specified on the DATECONV statement ESTLEN n This option is provided for compatibility with the REQUEST statement EXTFILE fileref file name Gives the fileref or name of the external file in which Eventus is to store the converted request file If the argument is a file name it may include the path but must not include any blank 122 or period On most systems a dat extension will be added to the file name automatically The default is USERDATA On mainframe systems where the file is pre allocated by control language or in the filename statement the file should be given a disposition of NEW or OLD a fixed block format and a logical record length of 80 GROUP variable Names a grouping variable to be used in an event study to combine multiple observations into
139. ion to label the results but otherwise the options have no effect You also can specify ANNUAL for annual data PAGE WIDE Normally Eventus prints portrait oriented or tall pages The default output is easy to view on a display screen before printing and is suit able for almost any printer The PAGE WIDE option on the EVENTUS statement allows Eventus to print landscape oriented or wide output With wide out put more information will fit on one page The disadvantage of wide output is that it produces a line width of 132 characters To print wide output you need a printer capable of landscape orientation or a small font The REQUEST statement The REQUEST statement instructs Eventus to read the request file and the user supplied returns file The request file is an external file you use it to supply the identifiers for your sample the event dates and optional other information Eventus does not compare the identifiers and event dates in the request file to any external database The identifier is referred to as CUSIPs in this chapter but it need not be a true CUSIP it can be any one to eight character alphanumeric value padded by blanks if it is shorter than eight characters The date must be a valid calendar date but is not subject to any other limitation Each line of the request file contains data for a single security event date combination At a minimum each line must contain a CUSIP and event date The ALLDAYS option
140. irm the report will state exactly what the problem was If you prefer not to have Eventus print this report specify NONAMES on the EVTSTUDY statement Unless you specify NOPLIST on the EVTSTUDY statement you also get a list of the estimation period return statistics for each observation and the mean and median for the sample The report includes market model alphas betas and residual standard deviations fraction of market model residuals positive and additional statistics if space permits The last set of output from EVTSTUDY contains the event time results The SIC option causes Eventus to report SIC industry classification codes from the CRSP database on the list of estimation period return statistics If an output SAS data set is being created see the OUTSAS option below the character variable SICCODE is included in the data set The sic code for the last reporting date on or before the event date is listed unless the earliest reporting date precedes the event date then the first SIC code is listed See the CRSP release notes for information about the accuracy of the SIC code data By default one page is generated for each abnormal return method in effect Each page lists the portfolio mean abnormal return test statistic number positive and negative and significance levels for the event date day 0 and as many as 34 surrounding days 94 if PAGE WIDE appears on the EVENTUS statement depending on PRE and POST At the botto
141. is 3 Eventus assumes that you are specifying the number of months for a monthly event study weeks for a weekly event study or days for a daily event study Tt also is possible to specify the estimation period by calendar date instead of relatively see EST SPECIFIC on page 146 in Appendix B 74 MINESTN Specify MINESTN n to remove an observation from the sample if the stock has fewer than n days or months of return data in the estimation period For example MINESTN 60 means that if the stock has fewer than 60 usable returns in the estimation period that observation will be dropped Eventus tells you when observations drop out of the sample MINESTN must be less than or equal to ESTLEN The WINDOWS statement The WINDOWS statement is required for the event parameter approach Use it to specify ranges of dates windows relative to the base date over which Eventus should compute abnormal return parameters and statistical significance tests For example when daily returns are used the window 2 2 defines a five trading day period from two days before through two days after day 0 Up to 99 windows can be specified Single date windows may be specified by repeating the relative date For example 3 3 establishes a window containing only day 3 For each window Eventus enters a dummy variable in the event parameter regression model The dummy variable is equal to one on each date within the window and equal to zero on
142. ith PAGE WIDE see the ALLDAYS option below page 51 49 time series standard errors and market model abnormal returns using stan dardized abnormal return tests To add comparison period mean adjusted returns specify CP on the EVTSTUDY statement To add total unadjusted or raw returns specify RAW To omit market adjusted returns specify NOMAR To omit market model returns altogether specify NOMM To omit just the standardized abnormal return tests using the market model specify NOSTD To omit everything except the market model with standardized abnormal return tests specify STDONLY These options can be combined For example CP NOMAR NOMM causes EVTSTUDY to analyze comparison period mean adjusted returns and omit all other types of abnormal returns To tell Eventus to report standardized abnormal return tests for non market model methods specify STDALL on the EVISTUDY statement Selecting the value weighted market index Normally Eventus labels the index return as equally weighted Specify VALUE to label the index as value weighted index or BOTH if you supplied two index return series and want results for each Scholes Williams and GARCH market model estimation Eventus reports market model results using both ordinary least squares and Scholes Williams 1977 beta estimation when you specify SW When the GARCH option is specified Eventus estimates the market model assuming a GARCH 1 1 error structure For exponential
143. ition to the OLS results but only if the option also appeared on the original EVTSTUDY statement TAIL 1 2 Specifies the significance levels of the reported test statistics are to be based on one or two tailed tests The default is TAIL 1 The TAIL option does not set the tails for bootstrapped tests please see the BTAIL option 140 B 7 PRICES Statement Reads prices from CRSP daily or monthly stock databases Eventus writes the prices and associate data to the SAS output window or stores them in a SAS data set or external file Choose from the following options BIDASK Specifies that the secondary price variables are to be read and included in the output Secondary prices are either bid and ask quotes or intraday intramonth high and low transaction prices see the CRSP Data Description Guide for details BINARY This is identical to the TEXT format below except that the prices themselves are written in real binary FLOAT4 RB4 or S370RB4 depending on the system format Real binary format uses about one third as much disk space for storing prices as the TEXT format A binary file like a TEXT file can be read by any computer language including FORTRAN C and SAS DISTRIB Requests that dividends and other cash distributions CRSP dis tribution code lt 4999 be extracted When the DISTRIB option is in effect the printed or SAS data set output includes one zero or nonzero value for each trading date containing the total cash
144. k or multimonth periods All results are reported in terms of multiday periods as well However Eventus interprets the EST ESTLEN MINESTN and other REQUEST statement options in terms of the original days weeks months or quarters VALUE BOTH By default Eventus uses only equally weighted market index returns in the market model and market adjusted returns event studies Specify VALUE to change to the value weighted index or the alternative index indicated by SP500 or COMPOSIT specified on the REQUEST state ment or BOTH to produce separate event studies using both indexes B 5 EXTRACT Statement Extracts window cumulative or compounded abnormal returns and optional weighted least squares regression weights from a SAS data set previously saved by the Eventus EVTSTUDY statement with the OUTSAS option The following parameter is mandatory INSAS libref membername Points to the SAS data set containing the saved event study results 136 The following options can be specified on the EXTRACT statement BUYHOLD Specifies buy and hold compounded return computation for win dows instead of the default additive cumulation CDCSI Causes Eventus to compute regression weights to reflect the weight ing implied by the Collins and Dent 1984 test assuming cross sectional independence This option affects only the weight so it is not applica ble unless the WPREFIX option also is specified EXTEND n If there are missing returns in a window fo
145. le the event dates and depending on the options you specify see below possibly other information Chapter 2 describes the basic structure of the request file For monthly event studies the dates in the request file can be any day of the month For weekly event studies use the last trading day in the week or specify any day up to the last trading day and use the AUTODATE option described below The rest of this section describes REQUEST statement options for the pro cessing of dates and market and stock returns and for the construction of the estimation period A SAS filename statement or the operating system must The ALLDAYS option see page 28 below is available with either wide or tall output to force Eventus to use additional pages to avoid the limitation on the number of days 16 associate the fileref REQUEST with the request file Option to convert CUSIPs to PERMNOs If you have CUSIPs in the request file you can specify the CUSIPERM option to have Eventus internally match them to the corresponding CRSP PERMNOs but not alter the request file To create a permanent request file with CUSIPs replaced by PERMNOs please see Chapter 9 The CUSIPERM option works properly only if the PermnoUp program de scribed in the Eventus installation instructions is run during Eventus instal lation and after each annual or quarterly update of the CRSP database Only those CUSIPs present in the CRSP database will be matched If you s
146. le and a 15 filename estindx statement should point to the calendar index file On a system that permanently stores the CRSP data in SAS data sets the option ELIBNAME 1libname should appear on the EVENTUS statement to point to the aggregate storage location typically a folder or directory containing the CRSP data with the frequency to be used for the estimation period When working with CRSPAccess format databases Eventus uses the CRSPAccess environment variables or logicals to locate the data PAGE WIDE Normally Eventus prints portrait oriented or tall pages The default output is easy to view on a display screen before printing and is suitable for almost any printer The PAGE WIDE option on the EVENTUS statement allows Eventus to print landscape oriented or wide output With wide output more information will fit on one page For example Eventus will print the complete firm names from the CRSP database for the stocks in the sample the day of the week of the event dates and up to 95 days of abnormal returns in the event period Wide output produces a line width of 132 characters It may be necessary to manually select landscape orientation when printing it The REQUEST statement The REQUEST statement instructs Eventus to read the request file and the calendar of the CRSP database then match the dates in the request file to CRSP dates The request file is an external file you use it to supply the PERMNO identifiers for your samp
147. lid only on the RETURNS statement Normally no market index returns are included The INDEX option specifies that you want returns on a market index for the same days as stock returns By default the index is the CRSP equally weighted index To get value weighted or both equally and value weighted index returns also specify VALUE or BOTH The Standard and Poor s index return or the Nasdaq Composite Index return replaces the value weighted index when you specify SP500 or COMPOSIT on the REQUEST statement Obtaining SIC codes The SIC option causes Eventus to report SIC industry classification codes from the CRSP database If an output SAS data set is being created see the OUTSAS option below the numeric variable SICCODE is included in the data set The sic code for the last reporting date on or before the starting date of data to retrieve is listed An exception occurs when the earliest reporting date follows the starting date of data to retrieve then the earliest SIC code is listed See the latest CRSP manual for information about the accuracy of the SIC code data 88 Obtaining share type codes To store the CRSP two digit share type code in an output SAS data set being created specify the SHRCODE option Normally the share code output will be the last share code reported on the CRSP database on or before the starting date of data to retrieve In rare cases the earliest share code reporting date for a stock may be later than the sta
148. lso includes a nonparametric generalized sign test which uses the fraction of positive ab normal returns in the event period as the normal instead of assuming 0 5 Appendix A presents further details of the methods PERMNO 10506 10914 36150 67652 72100 75111 75241 76263 76369 76754 77142 77170 77446 77833 79739 83447 91732 Sample Eventus output Second page Eventus R Software from Cowan Research L C Figure 2 4 Results of Daily Stock Returns Input Name on Event Date INTEGRATED BRAN A ALLWASTE INC CO CORE INDUSTRIES I AMERIWOOD INDUSTR IMO INDUSTRIES IN AMAX GOLD INC RP PIONEER NATURAL R ILLINOIS CENTRAL HEALTH MANAGEMENT BAY NETWORKS INC GREYHOUND LINES I CHRYSALIS INTERNA ESKIMO PIE CORP PEAK TECHNOLOGIES SEDA SPECIALTY PA EXCEL COMMUNICATI P M C INTERNATION Event Date 10 13 1997 03 06 1997 06 16 1997 03 30 1998 06 26 1997 02 09 1998 08 05 1998 02 10 1998 08 14 1997 06 15 1998 10 16 1998 11 18 1998 11 17 1998 03 31 1997 06 17 1997 06 15 1998 07 08 1998 Esti mation Period Returns lt 255 255 255 255 255 255 255 253 255 255 255 255 255 255 255 255 255 Event Period Returns lt 61 61 61 61 54 61 61 61 61 61 61 61 60 61 61 59 61 Reason if no useable returns No data on date No useable returns found Or beyond estimation period Figure 2 5 Sample Eventus output Third page Eventus R Software fro
149. luding an identification variable Each line of your request file may include an identification variable following the base date Eventus will read it and print it on subsequent output if you name it with ID ID ID works or you can choose another name Use IDFMT to tell Eventus what format to use for reading and printing the identifying variable IDFMT 4 means a 1 4 digit integer while IDFMT 4 means a four character string Other lengths and other SAS formats are permitted also With IDFMT 4 specified a line of your request file might look like this 16372210 19840920 AY4E Using date formats other than YYMMDD The REQUEST statement allows you to specify the format of dates in your request file You may list calendar dates in nearly any conventional format If you use either six or eight digit YYMMDD the default you need not specify DATEFMT Besides MMDDYY and DDMMYY you can use the SAS date format DATE An example of a date in DATE format is 19OCT97 lTf you are familiar with permanent SAS data sets you might want to read about the INSAS option on page 147 73 Options for constructing the estimation period The estimation period is simply a time period that contains no event dates Eventus will combine the estimation period with the event period to estimate the augmented market model The estimation and event periods may be defined so that they are discontinuous if desired EST and POOL By default Eventus determines
150. ly the same as O Hara and Shaw 1990 report Chapter 3 has a detailed discussion 56 Figure 4 4 First 62 lines of the user supplied returns file for the Too Big to Fail study CUSIP Bank name RETURN INDEX Date 56480910 Manufactur 0 010152 0 003478 840703 56480910 Manufactur 0 020513 0 005392 840705 56480910 Manufactur 0 041885 0 003201 840706 56480910 Manufactur 0 043716 0 006032 840709 56480910 Manufactur 0 005236 0 002348 840710 56480910 Manufactur 0 015625 0 013716 840711 56480910 Manufactur 0 020513 0 003855 840712 56480910 Manufactur 0 035176 0 005025 840713 56480910 Manufactur 0 004854 0 003065 840716 0 56480910 Manufactur 038647 0 004453 840717 56480910 Manufactur 0 009302 0 005560 840718 56480910 Manufactur 0 014085 0 006433 840719 56480910 Manufactur 0 038095 0 005262 840720 56480910 Manufactur 0 004950 0 005737 840723 56480910 Manufactur 0 039409 0 006702 840724 56480910 Manufactur 0 010256 0 005587 840725 56480910 Manufactur 0 015228 0 008092 840726 56480910 Manufactur 0 035000 0 007644 840727 56480910 Manufactur 0 009662 0 004767 840730 56480910 Manufactur 0 000000 0 002720 840731 56480910 Manufactur 0 004785 0 021275 840801 56480910 Manufactur 0 009524 0 024754 840802 56480910 Manufactur 0 018868 0 028282 840803 56480910 Manufactur 0 050926 0 002992 840806 56480910 Manufactur 0 017621 0 000459 840807 56480910 Manufactur 0 008969 0 004584 840808 56480910 Manufactur 0 027149 0 021686
151. m of the page for each window you request are the portfolio cumulative or compounded abnormal return the test statistics the number positive and negative and significance symbols You may want to see window or daily abnormal returns for individual firms EVTSTUDY will print individual firm abnormal returns only for the standardized abnormal return method The standardized abnormal return To print all the dates in the range defined by PRE and POST when the number exceeds 35 or 95 with PAGE WIDE see the ALLDAYS option page 28 24 method can be extended to non market model benchmarks by specify the STDALL option page 25 To obtain window cumulative abnormal returns for every firm specify DETAIL on the EVTSTUDY statement As many firms as can fit will appear on a single page To obtain complete daily abnormal returns for each firm specify DETAIL FULL Full detail takes at least one page for each firm so be prepared for lots of output If you need to look at only a few individual days use the following trick Suppose you want to see day 1 for every firm On the WINDOWS statement include 1 1 Then specify DETAIL not DETAIL FULL on the EVTSTUDY statement This way you get the day you need without all the extra output DETAIL specifications are ignored if you specify NOMM or NOSTD unless you also specify STDALL To save or export individual firm raw or abnormal returns for further analysis use the OUTSAS option page 31 to
152. m Cowan Research L C 3 Parameter Estimates and Estimation Period Statistics Market Residual Mean Model Res Standard PERMNO Alpha Beta Return iduals gt 0 Deviation 10506 0 00033 1 13 0 00192 42 74 0 07700 10914 0 00056 0 65 0 00021 41 56 0 02483 36150 0 00019 0 75 0 00043 42 74 0 01934 67652 0 00358 1 44 0 00206 50 19 0 02716 72100 0 00265 1 12 0 00192 51 76 0 03949 75111 0 00469 0 76 0 00367 56 86 0 03000 75241 0 00310 1 27 0 00137 45 45 0 01967 76263 0 00093 1 14 0 00056 45 49 0 01278 76369 0 00627 1 68 0 00510 51 76 0 10974 76754 0 00287 2 87 0 00203 46 27 0 03338 77142 0 00064 1 04 0 00109 43 13 0 02837 77170 0 00141 1 04 0 00168 48 62 0 07034 77446 0 00041 0 62 0 00058 45 49 0 04138 77833 0 00467 2 05 0 00186 53 33 0 05053 79739 0 00093 1 20 0 00003 47 45 0 02780 83447 0 00140 2 25 0 00244 42 35 0 03508 MEAN 0 00204 1 31 0 00060 47 20 0 04043 MEDIAN 0 00141 1 13 0 00030 45 88 0 03169 Figure 2 6 Sample Eventus output Fourth page Eventus R Software from Cowan Research L C Market Adjusted Returns EW Index N Positive Negative Generalized Sign 11 Days 30 2 1 0 1 30 Average Median Abnormal Abnormal Return Return 0 69 0 17 1 20 1 79 0 48 0 70 0 29 0 36 0 39 0 31 1 74 0 13 6 47 0 34 0 61 0 07 0 88 0 18 3 07 0 85 0 99 0 12 2 47 0 36 4 22 0 76 4 74 3 96 18 25 13 34 0 07 0 20 2 10
153. mation period follows the event date and begins that number of trading days months after day month zero For example EST 61 means that the estimation period will begin 61 trading days after the event date while EST 80 tells Eventus to end the estima tion period on event time date 80 For a TWIN event study Eventus considers negative values relative to the first event date positive values relative to the second e If you specify POOL Eventus constructs the estimation period us ing equal numbers of returns from two periods one ending 61 days before the first event date and one beginning 61 days after the second event date You can control the length of the com bined period with the ESTLEN parameter described below For example ESTLEN 200 with POOL gives you an estimation period consisting of days 160 through 61 and 61 through 160 Do not specify POOL with EST SPECIFIC e If you specify EST SPECIFIC Eventus looks for a second date in the request file or a third date in a TWIN event study and ends the estimation period on the date specified for each firm This allows you to choose estimation periods that fall a varying number of days from the event date according to the circumstances of each case When using EST SPECIFIC with external request files make sure that the estimation date is the next item after the event date on each line If the request file is a SAS data set specified by INSAS the specific estim
154. me CUSIP is required the variable itself can contain any character string up to eight characters in length for example a Datastream mnemonic code For each observation in 131 the request file there must be ESTLEN PRE POST 1 observations in the INSAS file The default if the INSAS option is not specified is to read the user supplied data from an external file associated with the fileref USERSTOK instead of a SAS data set When the INSAS option is in effect there is not need to define the fileref USERSTOK In this case the input SAS data set is called the USERSTOK file even though it is not a plain text file ITSUR Causes Eventus to use the iterated seemingly unrelated regressions approach where the market model augmented by a dummy variable for each window specified on the WINDOWS statement is estimated by iterated joint generalized least squares over the combined estimation period and event period return series JACKNIFE Designates the jackknife test instead of the generalized sign test as the nonparametric test to accompany non standardized method parametric tests MAXMISS n Specifies to remove from the sample any firm event with more than n days or months of missing returns in the event period MEDIAN Selects printing of the median abnormal return in place of the number positive and negative and nonparametric significance level ex cept in the standardized method The option is ignored when PAGE TALL is in effect because b
155. mental Nasdaq quotes are iden tified as BID FROM NMS FILE and ASK FROM NMS FILE in an external file or by variable names BIDNM and ASKNM or BIDN1 through BIDNnnnn and ASKN1 through ASKNnnnn in a vertical or horizontal SAS data set respectively When the option combination BIDASK NMS is in effect without NOCLOSE Eventus reports only one pair of merged secondary price variables con taining Supplemental Nasdaq bid and ask quotes for those stocks and 142 date ranges for which the supplemental quotes are available and the Bid or Low Price and Ask or High Price data items from the main CRSP secondary prices arrays in all other cases The merged sec ondary price variables are identified as BID OR INTRADAY LOW and ASK OR INTRADAY HIGH or by variable names BIDLO and ASKHI or BIDL1 through BIDLnnnn and ASKH1 through ASKHnnnn The CRSP Data De scription Guide for the stock database provides a full explanation of the underlying stock data NOCLOSE Specifies that primary closing prices are to be omitted used to gether with the BIDASK option when only secondary prices are desired Note the special interaction explained above of this option and the NMS option OUTSAS Specifies a two level SAS name libref membername under which to create the SAS data set containing the prices This option is valid only with the HSAS or VSAS file format options SHARES Specifies that the number of shares outstanding on the event date in thousands should be
156. ments The file need not be sorted by PERMNO If the CUSIPERM option or the CUSIP option is specified the first variable on the line should be CUSIP instead of PERMNO REQFILE2 fileref Specifies a second request file to be used in the same job The format required for the second request file is the same as for the first SHIFT1 n SHIFT2 n The first date in the request file is shifted by n periods and the second date is shifted by n periods For the monthly file the periods are months For the daily file the periods are trading days if DATEFMT CRSP and calendar days otherwise Both n and n may be specified as any integer value For example SHIFT1 1 shifts June 1 2000 back to May 31 2000 SHIFT1 and SHIFT2 may be specified singly as well as together Using these options with calendar dates may result in invalid date messages unless AUTODATE is also specified SHORT This option is provided for compatibility with the REQUEST state ment The request file must contain an S or L code which is copied to the converted request file at the end of each line SORTBYID Specifies that the converted request file is to be sorted by the variable listed in the ID option The default is to sort by PERMNO unless the CUSIP option is specified in which case the default is to sort by CUSIP 125 UPCUSIP This option is valid only when CUSIP also is in effect Specifies to try to update CUSIP identifiers in the request file to match the latest
157. n The Agency Costs of Free Cash Flow Acquisition Activity and Equity Issues Journal of Business 1991 64 2 213 227 Mikkelson Wayne H and M Megan Partch Valuation Effects of Security Offerings and the Issuance Process Journal of Financial Economics 1986 15 1 2 31 60 Mikkelson Wayne H and M Megan Partch Withdrawn Security Of ferings Journal of Financial and Quantitative Analysis 1988 23 2 119 134 Errata 1988 23 4 487 Neter John William Wasserman and Michael H Kutner Applied Linear Regression Models Richard D Irwin Inc Homewood Illinois 1983 O Hara Maureen and Wayne Shaw Deposit Insurance and Wealth Ef fects The Value of Being Too Big to Fail Journal of Finance 1990 45 5 1587 1600 Patell James M Corporate Forecasts of Earnings Per Share and Stock Price Behavior Empirical Tests Journal of Accounting Research 1976 14 2 246 274 Peterson Pamela P Event Studies A Review of Issues and Methodology Quarterly Journal of Business and Economics 1989 28 3 36 66 Pilotte Eugene Growth Opportunities and the Stock Price Response to New Financing Journal of Business 1992 65 3 371 394 Prabhala N R Conditional Methods in Event Studies and an Equilibrium Justification for Standard Event Study Procedures Review of Finan cial Studies 1997 10 1 1 38 Rozeff Michael S and Mir A Zaman Market Efficiency
158. n stalled to integrate it with the main stock data With the module installed use dsic dat or msic dat for daily or monthly data respectively and spec ify PORTYPE 2 on the REQUEST statement Without the module installed use dsix dat or msix dat for daily or monthly data respectively and do not use the PORTYPE option 4Tn the Eventus binary format the date expressed as a YYMMDD or CCYYMMDD integer occupies the first four bytes of each record The ten size portfolio returns for that date follow in the next 40 bytes each represented as a four byte real value For an Eventus character format file each line begins with a single blank space followed by the date as a six digit YYMMDD or eight digit CCYYMMDD integer followed by a single blank space followed by the first size portfolio return in scientific notation with six decimal places E13 6 format the remaining size portfolio returns follow each separated by a single blank space Note that portfolio 1 contains the stocks with the smallest capitalizations while portfolio 10 contains the stocks with the largest capitalizations A program to generate an index file to these specifications is included with Eventus 21 Computing continuously compounded returns If you want to analyze returns in continuously compounded form specify LOG on the REQUEST statement Eventus then transforms each firm and index return rj to log 1 rjz as it is read from the CRSP database By default Eventus
159. n 130 VAXRB4 SAS informat 130 VOLUME statement 155 default output 157 options BINARY 87 155 EXTFILE 155 HSAS 88 NASDINFO 91 156 OUTSAS 92 156 SHARES 90 92 156 SHRCODE 89 156 SIC 156 SPLITADJ 91 156 TRADES 91 157 VSAS 88 157 REQUEST required before 145 Warner Jerold B 111 weight firms within subgroups 147 negative 19 weighted least squares 65 example 69 with EVTSTUDY 25 27 49 51 TWIN 39 61 158 with EXTRACT 64 158 with OLDSTUDY 40 Zaman Mir A 19 Zivney Terry L 112 with serial correlation adjustment 67 Weisbach Michael S 111 Williams Joseph T 26 50 104 Willinger G Lee 32 55 window variable length 37 WINDOWS statement 23 39 40 48 61 158 event parameter approach 75 versus TIMEUNIT 27 177
160. nd CRSPDAY2 If 148 EST SPECIFIC is also in effect the ending day or month number of the estimation period must be included as variable CRSPEST INSAS2 libref membername Applicable only when you specify DUAL on the EVENTUS statement used in place of REQFILE2 when the request files are SAS data sets The format of the second request file must be exactly the same as the format of the first request file see the INSAS option above IX2Y Specifies that the SFA or Eventus format size index file being read in the current run uses two digit years The default is to expect four digit years LOG Causes daily or monthly returns to be transformed to continuously compounded returns by taking In R for stock returns and In Rypz for market index returns MINESTN n Optionally specifies the minimum number of usable trading days in the estimation period Eventus considers an estimation period return usable if it is non missing except for the first return following a sequence of one or more missing returns NAME Indicates that a firm name appears as the last item on each line of the request file The option is ignored everywhere but non CRSP event studies NDAYS n An option on REQUEST statements in conjunction with RETURNS PRICES and VOLUME statements and TWIN event studies Specifies that you want to retrieve n consecutive trading days of data or months in cluding days or months with missing data values for each observation If you spe
161. ne run 129 SAS data sets 161 SFA 160 Supplemental Nasdaq 89 91 142 157 CRSP share type code 89 CRSPDAY 148 CRSPDAY1 148 CRSPDAY2 148 cumulative returns 30 53 CUSIP option 145 CUSIPERM option 17 145 CUSIPERM statement 121 options COLUMN 101 121 EXTFILE 101 121 REQFILE 121 Dark Frederick H 111 DATECONV statement options AUTODATE 96 122 CUSIP 122 CUSIPERM 95 122 DATEFMT 96 122 EXTFILE 98 122 GROUP 95 123 GRWEIGHT 95 123 ID 123 IDFMT 123 INSAS2 124 NDAYS 97 OUTDTFMT 97 124 OUTSAS 98 124 REQFILE 125 REQUEST2 125 SHIFTn 97 125 SHORT 96 SORTBYID 97 UPCUSIP 126 dates converting to CRSP calendar 93 CRSP day numbers 3 20 96 format of 20 46 73 85 96 97 122 145 monthly data 93 multiperiod 26 non trading 20 86 96 122 145 old Nasdaq files 145 paired 37 59 shifting 150 weekly data 93 Dent Warren T 29 52 DISTRIB option 90 dividends 90 141 Dopuch Nicholas 111 dummy variables 71 EGARCH errors 26 estimation 103 estimation period 22 46 74 146 different return interval from event period 15 115 event parameter approach 74 firm specific 146 holding fixed over multiple event studies 146 length 22 47 74 147 missing returns in 23 75 118 149 split pre and post event 22 74 146 170 ESTLEN option 22 47 74 event dates SIZEIND 129 converting to CRSP calendar 93 TWIN 37 40 59 97 124 129 CRSP day numbers 3 20 96 158 non trading 20 86 96 122 145 TWIN with
162. no WINDOWS statement happens to appear before the EVTSTUDY statement so Eventus generates default windows as explained in Chapter 3 The EVTSTUDY statement includes two option specifications to save needed data in a SAS data set The option OUTSAS WORK INTERMEDIATE names the SAS data set to be created Using WORK as the first part of a two part data set name specifies a temporary data set A temporary data set ceases to exist after the user closes SAS in the case of interactive use or after execution com pletes in the case of a batch run Eventus currently requires the two part name but WORK INTERMEDIATE and the one part name INTERMEDIATE are completely interchangeable in ordinary SAS language and procedures A per manent data set could have been specified by replacing WORK with a libref previously defined in a libname statement or the Add New Library dialog The option PACKAGE DG specifies that the contents of the data set will include the necessary data to build firm by firm cumulative abnormal re turns D and the corresponding weighted least squares WLS weights G A complete list of PACKAGE option values appears in Table B 1 in Appendix B If Eventus were running in SAS version 6 12 or earlier the user would have to choose a shorter data set name INTERMEDIATE would trigger an error because it is longer than eight characters 67 Figure 5 2 Example using EXTRACT to organize firm by firm event study results for further analysi
163. nstallation files the de fault for this option is set according to the type of CRSP files you have On some systems a fixed length binary format file formerly had to be used instead of the variable length format generated from a character format file by CRSP subroutines FBIN specifies a fixed length file in native binary format FBIN 2 specifies a fixed length file in IBM binary format Specify this option in an Eventus program only when the type of file you want to use is not the default for your site This option is now obsolete GETDATA Specify this option only in a program that uses the RETURNS or PRICES statement to read and store CRSP stock data HOSTBIN 0 Indicates that the input CRSP database is SFA binary format with either two digit or four digit years Specifying 0 turns the option off Normally set at Eventus installation but may be specified at run time IDXLEAD n Indicates the number of bytes to skip at the beginning of each record of an input SFA binary CRSP calendar index file to compensate 127 for record length blocks Defaults to zero and ignored with CRSPAc cess format and non CRSP data Normally set at Eventus installation but may be specified at run time for experimentation with alternative values LIBNAME libref Gives the SAS libref of the CRSP stock data sets This option is for systems that store the CRSP data in the form of sas data sets Eventus expects to find SAS data sets named cal the CRSP stock in
164. nthly per share cash distributions are reported with zero reported on any 90 non ex dividend date If the SPLITADJ option is in effect distributions are split adjusted in the same manner as the price In a VSAS or HSAS output file the cash distribution amount is a variable named DIVAMT and the CRSP distribution code is a variable named DISTCODE Reporting the number of trades for Nasdaq stocks The TRADES option is valid on the PRICES and VOLUME statements Eventus reads the number of trades from the Supplemental Nasdaq File which must be associated with the fileref CRSPNMS When you use the PRICES statement you must also specify the NMS option for TRADES to work In a VSAS output file the number of trades is a variable named TRADES in an HSAS file the variable names are TRAD1 TRADnnnn Reporting Nasdaq specific data items The NASDINFO option is valid only on the VOLUME statement and only when the SHARES option is valid NASDINFO automatically specifies SHARES also The option reads the trading status trait code the National Market System indicator and the number of market makers from the Nasdaq information structure array of the CRSP Nasdaq stock file Eventus stores the data if the HSAS or VSAS option appears on the VOLUME statement The variable names in the SAS data set are TRAIT NMS and MAKERS Eventus does not place the NASDINFO data in an external file if the default external text file output or BINARY output is selected A
165. nths long when using weekly or monthly data You can change the estimation period length with ESTLEN The largest It is also possible to specify the estimation period by calendar date or trading day or month number see EST SPECIFIC on page 146 in Appendix B 22 number you can use is 999 and the smallest is 3 Eventus assumes that you are specifying the number of months for a monthly event study weeks for a weekly event study or days for a daily event study MINESTN Specify MINESTN n to remove an observation from the sample if the stock has fewer than n days or months of return data in the estimation period For example MINESTN 60 means that if the stock has fewer than 60 usable returns in the estimation period that observation will be dropped Eventus tells you when observations drop out of the sample By default the only constraint on estimation period data is the requirement of three returns for running a market model regression Don t specify a MINESTN greater than ESTLEN or you won t have any events left at all The WINDOWS statement The WINDOWS statement is optional Use it to specify ranges of dates windows in event time over which Eventus should report cumulative or compounded abnormal returns and associated test statistics Each window specification requires a begin and end which can be the same if desired For example 3 0 specifies a four day or four month window ending with the event date 2 2 sp
166. ntire 111 days 78 Figure 6 4 Eventus program to complete the event parameter example libname project c Research Demo Project varies by system eventus noncrsp title Event Parameter Approach Demonstration request insas project request est 105 estlen 100 windows 5 2 1 0 1 5 evtstudy insas project stockret pre 5 post 5 olsparam factors 2 The event study results appear in figures 6 5 and 6 6 Eventus prints the results on three pages which are condensed into two here For each window Eventus reports an F test for each of three null hypotheses the conventional event study hypothesis that the cumulative average abnormal return is equal to zero the hypothesis that all the firm firms in the sample jointly have cumulative abnormal returns equal to zero and the hypothesis that all the firm firms in the sample have equal cumulative abnormal returns but not necessarily equal to zero When you run Eventus the SAS log window or log file depending on the mode of SAS operation reports the completion of data steps and procedures that Eventus executes internally Most Eventus users will find that these reports have little meaning for them and they may ignore most of what is in the log However it is still a good idea to look briefly through the sas log for messages that begin with EVENTUS NOTE EVENTUS WARNING or EVENTUS ERROR If you contact us for technical support to help resolve a problem with an Even
167. ock prices from the CRSP database filename request G Some Folder Filename extension EVENTUS GETDATA PAGE WIDE MONTHLY REQUEST ID variable IDFMT format DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP AUTODATE NDAYS 1 PRICES BINARY VSAS HSAS SIC SHRCODE SHARES DISTRIB BIDASK NOCLOSE NMS TRADES SPLITADJ EXTFILE fileref OUTSAS libref SAS_dataset Including an identification variable Each line of your request file may include an identification variable following the event date Eventus will read it and store it with the returns prices or volume if you name it with ID ID ID works or you can choose another name Use IDFMT to tell Eventus the format to use for reading and printing the identifying variable IDFMT 4 means a 1 4 digit integer while IDFMT 4 means a four letter word Other lengths and other sas formats also are valid Options for processing dates The REQUEST statement allows you to specify how Eventus should handle the dates in your request file DATEFMT Calendar dates You can list calendar dates in nearly any conventional format If you use either six or eight digit YYMMDD the default you do not 85 Figure 7 3 Eventus statements to read trading volume data from the CRSP database filename request G Some Folder Filename extension EVENTUS GETDATA PAGE WIDE MONTHLY REQUEST ID variable IDFMT format DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP AUTODATE ND
168. on some lines and no blank on others Any other information you want may appear on the same line CUSIPERM duplicates your original file exactly except that it replaces any CUSIP with the corresponding five digit PERMNO followed by three blanks and sorts the lines in PERMNO order However any information beyond column 80 will not be copied Make sure that no leading zeros are omitted from the CUSIPs CUSIPs can contain letters as well as digits so CUSIPERM treats the CUSIP as an eight character string not a numeric variable The following two options may be specified on the CUSIPERM statement 100 In which column does the CUSIP identifier begin The COLUMN option The COLUMN option is needed only when non blank characters precede the CUSIP identifier in the request file Specify the column number of the first character of the CUSIP Selecting an output file location The original request file remains unchanged The converted request file con taining PERMNOs is directed by default to the fileref USERDATA If the fileref is undefined a file named userdata dat is usually created in the current working directory To select a specific disk file in which to store the con verted request file use a filename userdata statement before the EVENTUS statement to give the path and name of the file to be created To use a different fileref besides userdata specify the fileref with the EXTFILE op tion To avoid overwriting the input request fil
169. onthly event study weeks for a weekly event study or days for a daily event study Reading firm names from the request file To have Eventus print security issuer names as it does when using CRSP data base data include the name as the last item on each line of the request file and specify the NAME option on the REQUEST statement Up to 33 characters will be used How to construct the USERSTOK return data file To manually build a USERSTOK file place security and index return data in an external file using one line for each trading day or month or other period for each security Each line in the request file must have a corresponding set of lines in the USERSTOK external file The number of lines for each security must be the same For example with the default 255 day estimation period and 61 day event period each line in the request file must correspond to 316 consecutive lines in the USERSTOK file On each line include the following information CUSIP identifying variable if you specified ID rate of return for the security rate of return for the market index If you specified BOTH on the EVTSTUDY statement see below list the equally weighted index return first then the value weighted index return When a security or index return is missing use the SAS missing value code a decimal point Be sure to leave at least one blank space on each side of the missing value code 47 In many situations an easier approach is to use SA
170. onthly usage the user needs to specify the EVENTUS statement option MONTHLY and also change the data files to which the above filerefs point To use a different data frequency in the estimation period of an event study from that used in the event period in conjunction with the EVENTUS statement option ESTINTER the filerefs crspes and estindx must point to the CRSP SFA standard stock data file and stock associated calendar indices file from which Eventus is to read the estimation period returns C 3 Size index files used with both CRSPAc cess and SFA format databases To use the CRSP Indices Decile product in conjunction with the REQUEST statement option SIZEINDX CRSP the fileref sizeindx should point to the CRSP Stock File Decile Indices file To use a size index file built by a program included for the purpose in the Eventus installation kit in conjunction with the REQUEST statement option SIZEINDX the fileref sizeindx must point to it 160 C 4 CRSP data permanently stored in SAS data sets Eventus can use CRSP data stored in permanent SAS data sets provided that a given SAS data library usually a single folder or directory contains only daily data or only monthly data not a mixture of the two When the EVENTUS statement option SASCRSP is set by default or specified at run time Eventus expects the SAS data library to be used for the current run to be associated with the libref CRSP For example libname crsp H SASC
171. or months in the event period EVTSTUDY computes and reports abnormal returns for the event period which is defined by default as days or weeks 30 through 30 or months 12 through 12 To change from these defaults specify the number of days before and after on the EVTSTUDY statement using PRE and POST These options may be used singly or in combination For example PRE 60 means that the abnormal returns are to start with day 60 unless POST is also specified the latter retains its default value If you specify PRE or POST you may need to change the estimation period from the default To change the estimation period specify EST on the REQUEST statement Eventus will stop and print an error message if the event period defined by PRE and POST overlaps the default or specified estimation period Selecting an estimation method To invoke the event parameter method specify one of the EVTSTUDY state ment options OLSPARAM SUR or ITSUR The options indicate estimation using ordinary least squares seemingly unrelated regressions also called joint gen eralized least squares or iterated seemingly unrelated regressions methods respectively 76 Figure 6 2 Eventus program to assemble data for the event parameter example filename request c myproject request dat filename trates c myproject rates dat libname project c myproject sasfiles data project request infile request input cusip 8 crspday
172. or the second window 2 2 will be a variable named DEP2 and so on The default is VPREFIX WINAR The WPREFIX option specifies that you want the output to include a weight variable and gives the prefix for the weight variable name The weight vari able is the reciprocal of the variance of the market model cumulative abnor mal return Running a squares regression with the weight variable specified in the WEIGHT statement of PROC REG is equivalent to estimating an ordinary least squares regression with all the variables including the vector of ones used for the intercept multiplied by the square root of the weight variable You can specify WPREFIX if you use TYPE MM see below and the original EVTSTUDY statement did not specify NOMM or NOSTD or with another TYPE value if the original EVTSTUDY statement specified STDALL When saving cumulative abnormal returns in an external file instead of a SAS data set omit VPREFIX However it is still necessary to specify WPREFIX W the argument is arbitrary to generate weights even though the weight prefix will not appear in the external file In the external file the cumulative abnormal returns for the first firm appear on one line continuing See Draper and Smith 1981 Section 2 11 or Neter Wasserman and Kutner 1983 pp 171 172 65 onto additional lines as needed Then the weights for the first firm appear on a new line or more than one line if necessary The returns for the se
173. oth the median and the number positive and neg ative are always printed in this situation Affects printed output only and affects only the daily or monthly output not the output for win dows NONAMES Suppresses printing of the list of PERMNOs or CUSIPs firm names and number of estimation and event period returns found The default is to print the list NOPLIST Prevents the printing of the estimation period statistics that nor mally appear between the input report and the event study results NOPRINT implies NOPLIST NOPRINT A report of the CRSP database input operation is produced but all other event study printed output is suppressed An output data set still is created if the OUTSAS option is specified This option was 132 designed primarily for a batch environment and can be unstable in an interactive SAS session The NONAMES and NOPLIST options can be combined to reduce output in an interactive setting NUMFM SAS_numeric_informat Specifies the numeric format of stock and market index returns in an input USERSTOK file for a non CRSP event study Must be a valid SAS numeric informat Defaults to 11 6 11 character positions wide including sign decimal point and 6 digits to the right of the decimal point OLSPARAM Selects the event parameter approach where the market model augmented by a dummy variable for each window specified on the WINDOWS statement is estimated on the combined estimation period and event period return s
174. presents one set of market model results The t statistics in Figures 2 6 and 2 7 use standard errors estimated from the time series of sample mean portfolio ab normal returns this approach avoids biases due to cross correlated abnormal returns at the expense of somewhat reduced power Figure 2 8 displays the Figure 2 3 Sample Eventus output First page Eventus R Software from Cowan Research L C 1 Eventus R software is produced by Cowan Research L C http www eventstudy com ESTIMATION PERIOD Ends 46 days before the event date 255 days in length TOTAL NUMBER OF EVENTS 17 EVENTS WITH USEABLE RETURNS 16 EVENTS DROPPED 1 STATISTICAL SIGNIFICANCE LEVELS 1 tailed NONPARAMETRIC TEST Generalized sign test in all event studies For nonparametric tests significance levels of 10 05 01 and 001 are denoted by lt lt lt lt lt lt or gt gt gt gt gt gt respectively Left brackets lt appear when the ratio of positive to negative is less than in the parameter estimation period Right brackets mean that the ratio is more positive than in the estimation period NOTE Useable returns means all nonmissing returns except the first day after a missing estimation period return market model results with a Z test using the standardized abnormal return approach which estimates a separate standard error for each firm event and assumes cross sectional independence The default output a
175. produce weights that provide regression weighting consistent with 137 the portfolio weighting implicit in the non dependence corrected mul tiperiod Patell 1976 test No weights are actually created unless the WPREFIX option also is specified TYPE CP MAR MM RAW SW Selects comparison period mean adjusted market adjusted market model adjusted raw or Scholes Williams market model adjusted returns for output The selected type must exist on the input SAS data set The default is TYPE MM VALUE Selects value weighted index or S amp P 500 or Nasdaq Composite Index based abnormal returns when the EVTSTUDY statement that cre ated the input SAS specified the BOTH option The default is to use equally weighted index based abnormal returns in this situation VPREFIX prefix Sets the variable name prefix to use for the window ab normal returns Acceptable values of prefix are valid SAS names up to six characters long Longer values will be truncated to six characters The default prefix is WINAR WPREFIX prefix Causes Eventus to include WLS regression weights in the output and sets the variable name prefix to use for the weights Ac ceptable values of prefix are valid SAS names up to six characters long or up to seven characters if fewer than ten windows are to be output The default is not to output weights there is no default prefix B 6 OLDSTUDY Statement The OLDSTUDY statement reprints the reports from previous runs of Eventus an
176. r a security Eventus attempts to fill out the window by searching n trading days or weeks months quarters or years following each window for nonmissing re turns The default is EXTEND 0 indicating no filling out EXTFILE fileref file name Gives the fileref or name of the external file in which Eventus is to store the converted request file If the argument is a file name it may include the path but must not include any blank or period On most systems a dat extension will be added to the file name automatically The default is USERDATA On mainframe systems where the file is pre allocated by control language or in the filename statement the file should be given a disposition of NEW or OLD a fixed block format and a logical record length of 80 ID variable Names the variable to be used as an observation event iden tifier The variable must exist on the INSAS data set Specify IDFMT see below also IDFMT format Gives the output format of the identifying variable named in theID option Any SAS format is acceptable OUTSAS libref membername Indicates that the output should be in the form of a SAS data set and names the data set to be created within the aggregate storage location associated with libref The default is to create an external file ASCII or EBCDIC text or binary file SERIAL Selects WLS regression weights based on the serial dependence corrected standard error in equation A 1 on page 108 The default is to
177. rame systems where applicable the file should have a disposition of NEW or OLD and a logical record length of 80 REQFILE fileref The request file for CUSIPERM must be an external non SAS formatted file such as a card image format file Replace fileref with the fileref associated with your file The REQFILE specification may be omitted if the fileref of the request file is REQUEST The file need not be sorted by CUSIP 121 B 2 DATECONV Statement The DATECONV statement is used after the EVENTUS statement to convert dates between calendar and CRSP formats It reads the user request file and performs the date conversion operation The user may specify these options AUTODATE Specifies that any calendar date in the request file that is not a trading day should be converted to the following trading day CUSIP Specifies that the CRSP database or USERSTOK file to be read is sorted by CUSIP and the request file contains CUSIPs instead of PERMNOs This option is provided for compatibility with the REQUEST statement CUSIPERM Specifies that the request file contains CUSIPs but a CRSP data base sorted by PERMNO is to be read Eventus will attempt to convert CUSIPs to CRSP PERMNOs The option requires a conversion table a component of Eventus created by the PermnoUp program at Eventus installation and after each update of the CRSP database DATEFMT format Specifies the input format of the dates being read from the request file The specificat
178. read the SIC code from the CRSP database and to add the variable SICCODE to any output SAS data set TEXT This is the default Eventus writes the prices on a file that people can read The data are arranged in columns with each row reporting the PERMNO identification variable if applicable crsp date and return All the data for the first stock are listed with each trading day or month on its own row followed by all the data for the second stock and so on Specify the file name or fileref a SAS file shortcut or on mainframe systems a DDname to which to write the returns with the EXTFILE option on the RETURNS statement If the argument of EXTFILE is a file name it may include the path but must not include any blank or period On most systems a dat extension will be added to the file name automatically On mainframe systems where files must be pre allocated the file should have a logical record length of 80 or longer a fixed block format and a block size that conforms to your system s rules VALUE BOTH In the absence of one of these options Eventus includes only equally weighted market index returns in the output when the INDEX option is specified Specify VALUE to change to the value weighted index or the alternative index indicated by SP500 or COMPOSIT specified on the REQUEST statement or BOTH to get both 153 VSAS Creates a vertical format SAS data set A SAS data set cannot be read by non SAS software The following
179. returns for multiperiod windows spec ify BUYHOLD on the EVTSTUDY statement The option applies only to non standardized abnormal return methods By default Eventus reports non standardized results for each abnormal return method used in an event study and standardized results in addition for the market model abnormal returns Eventus computes buy and hold abnormal returns by compounding successive daily or other period raw returns and market index returns then adjusting the raw returns according to the abnormal return method used Comparison period mean returns and market model alphas are adjusted for the window length The Eventus output labels the mean buy and hold window returns 30 Average Compounded Abnormal Return while the default additive win dow abnormal returns appear as Cumulative Average Abnormal Return Reporting one instead of two tailed tests Eventus reports all significance levels for one tailed tests by default TAIL 2 changes to two tailed tests Saving results to a file By default none of the event study results is saved in a disk file To store firm by firm results specify the SAS file to store them in using the OUTSAS option The libref parameter specifies the libref of the sas data library in which to store the results Eventus will create a new data set unless the one named already exists in the SAS data library in which case the existing data set will be overwritten A SAS libname statem
180. returns when you specify STDCSECT on the EVTSTUDY statement Boehmer Musumeci and Poulsen 1991 introduce the method and document its empirical properties The method is the same as the Patell standardized method described above except that there is a final empirical cross sectional variance adjustment in place of the analytical variance of the total stan dardized prediction error For additional discussion of event date variance increases and related tests see Sanders and Robins 1991 For day t in the event period the test statistic is TSAR ars e N SgAR where 2 1 y iA j 5SSAR ZTN 2 SARu N SAR Eventus extends the cross sectional standardized method to multiperiod windows using the correction for serial correlation described above Thus the STDCSECT option implies the SERIAL option Define the standardized cumulative abnormal return for stock j as SCART 4 To CART T ScARn ma 109 where SCART T2 is as defined in equation A 1 Then the standardized cross sectional test statistic for the null hypothesis that CAAR 0 is N gt CARN Ta 2t I N28 5CAR where 1 N 1 2 5 SCARet N 1 2 SCART Toi m N 2 SCART To j Time Series Standard Deviation Method By default the time series standard deviation method is applied to all bench mark methods raw returns comparison period mean adjusted market ad justed and market model being estimated to produce the results not labeled
181. rting date of data to retrieve In this case Eventus uses the earliest share code Please see the latest CRSP Data Description Guide for further details on the share type code Selecting bid and ask or intraday high and low prices The BIDASK option valid only on the PRICES statement reads the secondary price variables Bid or Low Price and Ask or High Price from the CRSP stock database CRSP documentation indicates that these prices may be either closing bid and ask prices or intraday high and low transaction prices Eventus does not distinguish between the two types of data when executing a PRICES BIDASK statement However the nature of the secondary prices can be inferred from the sign of the primary price variable when using daily data CRSP uses a negative price to indicate a bid ask average in this case the secondary price variables are closing bid and ask price quotes When the primary price is positive it is a closing trade price and the secondary price variables are the lowest and highest trade prices during the day The secondary price data are stored in variables named BIDLO and ASKHI if you specify VSAS or BIDL1 BIDLnnnn and ASKH1 ASKHnnnn if you specify HSAS If you specify BINARY or use the default text output format each firm s secondary prices follow its primary prices on separate lines identified by the words BIDLO and ASKHI The NMS option on the PRICES statement causes Eventus to read the Sup plemental Nasdaq
182. s filename request F Any Folder Filename extension eventus title US Targets of Canadian Acquirers 1997 1998 request evtstudy outsas work intermediate package DG windows 30 2 1 0 extract type MM vprefix wincar wprefix weight insas work intermediate outsas work abnormalreturns A new WINDOWS statement must come between the EVTSTUDY and EXTRACT statements whether or not there is a WINDOWS statement before EVTSTUDY The windows listed on the new WINDOWS statement can be different from those on any preceding WINDOWS statement The EXTRACT statement includes options to specify the type of abnormal return benchmark to use market model the prefix wincar to use in build ing variable names for the cumulative abnormal returns CARs the prefix for variable names for the WLS weights weight the name of a SAS data set previously built by an EVTSTUDY statement work intermediate and the name of the output SAS data set to create work abnormalreturns Figure 5 3 displays the contents of work abnormalreturns produced by the SAS statements proc print data abnormalreturns id permno The _weight_ variable has the value Equal for all observations in the example because only the event study use the default equal weighted market index Had the BOTH option appeared on the EVTSTUDY statement there would have been two observations in work abnormalreturns for each PERMNO one with _weight_ of Equal and one with Value The C
183. s and index returns for that event are multiplied by 1 before any analysis SIZEINDX CRSP Causes Eventus to use a size index file for the market index return in event studies The size decile is matched to the size portfolio number in the annual data structure of the CRSP stock file If SIZEINDX CRSP is specified Eventus expects the fileref sizeindx to be associated with the character not binary CRSP Indices Decile file If only SIZEINDX is specified Eventus expects the fileref sizeindx to be associated with a binary or character file containing the date followed by the returns on size decile portfolios 1 10 Such a file is generated by a program included with the Eventus installation kit If the main CRSP database and size index file are in the form of SAS data sets please see the EVENTUS statement option SIZEIND SP500 COMPOSIT SP500 specifies to read the return on the Standard and Poor s 500 Composite Index instead of the value weighted index of all stocks regardless of whether NODIVIDX is specified COMPOSIT tells Eventus to read the return on the Nasdaq Composite Index instead of the value weighted index of all stocks regardless of whether NODIVIDX is specified With SFA CRSP files if the index file associated with the fileref crspindx is a NYSE AMEX or NYSE AMEX Nasdaq index file the Standard and Poor s Composite Index is read instead but Eventus output will incorrectly identify it as the Nasdaq Index UPCUSIP This option v
184. second date from your request file OUTDTFMT format Gives the date format for Eventus to use in creating the converted request file Valid formats are all SAS date formats and CRSP The default is YYMMDD8 OUTSAS Gives the libref and member name of the SAS data set to create for the converted request file This option should be used only when the converted request file is desired in a SAS data set format as opposed to an ASCII or EBCDIC text file The latter is created by the EXTFILE option or default Do not specify both EXTFILE and OUTSAS 124 REQFILE fileref The most often used method of storing the request infor mation PERMNO identifiers dates etc is to use an external non SAS formatted file typically an ASCII text file Replace fileref with the fileref associated with your file The REQFILE specification may be omitted if the fileref of the request file is REQUEST Each line of the request file should contain the following variables in order PERMNO event event specific ID number grouping group S or date date 2 estimation date or string variable weight L Each value must be separated by at least one blank but the exact position of the values is unimportant as long as they appear in the order shown The square brackets simply indicate items that need not always appear do not include them in the file Whether an optional item should appear is determined by the options specified on the EVENTUS and DATECONV state
185. statement allows the use of different return intervals for parameter estimation and event testing For example it is possible to conduct a daily return event study but estimate the market model slope intercept and standard error from monthly returns Eventus adjusts the estimated intercept and standard error for the difference in the return interval The market model intercept mean estimation period market index return and comparison period mean stock return are divided by the adjustment factor shown in table A 1 the standard error is divided by the square root of the adjustment factor The number of returns in the esti mation period D is multiplied by the adjustment factor Other reported statistics such as the estimation period standard deviation of raw returns reflect appropriate adjustments using the same factor 115 Table A 1 Adjustments factors for different estimation and event return intervals Event Period Estimation Period Interval Day Week Month Quarter Year Day 1 00 4 88 21 08 63 25 253 00 Week 0 21 1 00 4 32 12 96 51 84 Month 0 05 0 23 1 00 3 00 12 00 Quarter 0 02 0 08 0 33 1 00 4 00 Year 0 01 0 02 0 08 0 25 1 00 Note Values are rounded to two decimal places for display Internally values for event period intervals longer than a day are calculated from the values for daily event periods and are not rounded A 4 Variable Names In Eventus Output SAS Data Sets
186. store intermediate results and also see the EXTRACT statement in Chapter 5 Specifying benchmarks and standardized or non standardized tests The default is to compute three sets of abnormal returns and test statis tics market adjusted returns market model abnormal returns using portfolio time series standard errors and market model abnormal returns using stan dardized abnormal return tests To add comparison period mean adjusted returns specify CP on the EVTSTUDY statement To add total unadjusted or raw returns specify RAW To omit market adjusted returns specify NOMAR To omit market model returns altogether specify NOMM To omit just the standardized abnormal return tests using the market model specify NOSTD To omit everything except the market model with standardized abnormal return tests specify STDONLY These options can be combined For example CP NOMAR NOMM causes EVTSTUDY to analyze comparison period mean adjusted returns and omit all other types of abnormal returns To tell Eventus to report standardized abnormal return tests for non market model methods specify STDALL on the EVTSTUDY statement Specification of a market index Normally Eventus uses the CRSP equally weighted market index Specify VALUE to change to the value weighted index or BOTH to produce a set of results using each 25 Using standard deviation portfolio excess returns The SPORT option applies only when EXCESS is specified on the EVEN
187. tatement description Replace a word or symbol set in slanted Roman type with a word or numeral of your choice An ellipse indicates that you can continue with additional specifications Within the text Eventus statement names and variable names are set in typewriter style EVENTUS refers to the statement of that name while Eventus refers to the software Chapter 2 Event Studies The Essentials Much has been learned from the body of research based on event study methodology As one moves forward it is expected that event studies will continue to be a valuable and widely used tool in economics and finance MacKinlay 1997 p 38 Eventus offers many options that let you tailor event study output to your needs To start with though you might run an Eventus program in the simplest possible form with almost no options This may help you better understand the basic structure of the program Figure 2 1 shows the minimum set of statements needed to run an event study The user can type the statements into the SAS Editor window for interactive submission all at once or one line at a time or create and save a separate program file in plain EBCDIC text format on IBM mainframes and plain ASCII text on all other systems for batch submission Except for items inside quotation marks the program statements are not case sensitive The example assumes the use of a CRSP daily stock database Later chapters describe the usage of monthly
188. tatement is not present Eventus automatically supplies three windows including 1 0 the part of the event period preceding date 1 and the part following date 0 A WINDOWS statement with no windows listed suppresses the window output With a two date event study TWIN option on the EVENTUS statement the usual windows are not valid Instead specify descriptors for the two event dates on the WINDOWS statement as shown above These descriptors may be up to 14 characters long they need not be valid SAS names but may not contain blank spaces When used with the EXTRACT statement WINDOWS specifies the event win dows to cumulate and store for further analysis In this situation 1 0 is not automatically included The WINDOWS statement syntax in this usage is the same as for the type of event study single date or TWIN EXTRACT is processing 158 Appendix C How Eventus Finds the CRSP Stock Database C 1 CRSPAccess Format Eventus looks for a Windows environment variable Unix environment vari able or OpenvMS logical variable named CRSP_DSTK to automatically identify the location of the daily CRSPAccess database and the environment variable or logical CRSP_MSTK to identify the monthly CRSPAccess database The CRSP document CRSP Access Release Notes describes how to set these environment variables or logicals on supported systems However if for some reason the appropriate environment variable or logical is not set it is possi
189. tatements that the user enters in the SAS editor window The statements could also be saved in a plain text program file for batch execution The filename request statement points to the request file the quoted string shown as an example is in Windows syntax but all other aspects of the statement are identical across operating environments Figure 3 3 displays the request file Note that it is sorted by Swarynum here but in general it need not be sorted at all Figure 3 4 shows the first page of Eventus output The output will be in the SAS output window or procedure output file in the case of a batch run 32 Figure 3 3 Request file for the Too Big to Fail study 48223 840920 01 41718 840920 02 47079 840920 04 58827 840920 05 47896 840920 06 46842 840920 07 51676 840920 08 48071 840920 09 48354 840920 10 50024 840920 11 52863 840920 12 19043 840920 13 53858 840920 14 53903 840920 15 60839 840920 16 51772 840920 17 59379 840920 18 53209 840920 19 59109 840920 20 53938 840920 21 56805 840920 22 26550 840920 69 The first page lists the estimation period summarizes the results of searching the CRSP file for the sample explains the statistical significance symbols and may contain other information depending on the options selected In figure 3 5 is a listing of the sample and a detailed report of the number of returns Eventus found for each observation The firm names listed are cur rent as of the event date
190. te For example Figure 2 2 displays a request file that is perfectly acceptable even though it includes irregular spacing and a mixture of two digit and four digit years The request file need not be sorted in any particular order The EVENTUS statement gets the package started Here the statement consists simply of the word EVENTUS and a semicolon Later chapters describe options that can be used on the statement The REQUEST statement in the Eventus interprets two digit years from 00 24 as 2000 2024 and two digit years from 25 99 as 1925 1999 next line tells Eventus to read the request file In the last line EVTSTUDY produces the actual event study After the user submits all the statements the results appear in the SAS Output window or the procedure output file in the case of a batch run Figures 2 3 and following present the results for the above request file and Eventus statements The first page of output summarizes the sample and some default settings The second page presents a firm by firm listing of the sample indicating how many daily returns Eventus found for each event The firm name comes from the name history array of the CRSP database and is the name in effect on the event date By default Eventus produces portrait oriented output and abbreviates the name for page formatting reasons If a share class letter exists in the database it is reported as the last character of the abbreviated name column in the sample listin
191. ted for serial dependence Mikkelson and Partch 1988 and others perform such a correction on cumulative returns The SERIAL option causes Eventus to apply the correction to standardized abnormal returns only Karafiath and Spencer 1991 and Cowan 1993 provide simulation evidence of the properties of the corrected and uncorrected test statistics Note that the serial dependence that the SERIAL option corrects is not due to any presumed dependence in the true market model error term but occurs because all of the abnormal return estimators being cumulated are functions of the same estimators of the market model parameters The derivation of the corrected standard error used by Mikkelson and Partch 1988 requires that the abnormal return be interpreted as a forecast error Alternative nonparametric tests Normally Eventus prints the generalized sign test statistic see Cowan 1992 for each day in the event period and for each window with both standardized and non standardized tests If you specify RANKTEST on the EVTSTUDY state ment Eventus instead will report the nonparametric rank test introduced by Corrado 1989 with non standardized tests The JACKNIFE option produces the jackknife test developed by Giaccotto and Sfiridis 1996 The RANKTEST and JACKNIFE tests are mutually exclusive The generalized sign test still will be reported with standardized abnormal return method results 29 Bootstrapped versions of parametric tests The
192. till treat the sample as an equally weighted portfolio precision weighted portfolio in the case of standardized abnormal return tests of stocks held long That is after performing the sign reversal Eventus makes no further distinction between S and L stocks This allows you to create an arbitrage portfolio by specifying a short position for half the observations and a long position for the other half The portfolio weights sum to zero in an arbitrage portfolio Studies of insider trading often use combinations of short and long posi tions in abnormal return tests For examples see Rozeff and Zaman 1988 and Arshadi and Eyssell 1991 Options for processing dates The REQUEST statement allows you to specify how Eventus should handle the dates in your request file 19 DATEFMT Calendar dates You may list calendar dates in nearly any conventional format The default is YYMMDD which automatically accommodates both eight digit four digit year and six digit two digit year dates Besides MMDDYY and DDMMYY you can use the SAs date format DATE CRSP Trading Day Numbers Eventus never requires you to man ually convert calendar dates to CRSP day numbers If you have CRSP day or week or month numbers you can use them in your request file Specify DATEFMT CRSP on the REQUEST statement AUTODATE If some of the calendar dates in the request file are potential non trading days you have two choices Your first choice is to run your
193. timates and Estimation Period Statistics Market Residual Mean Model Res Standard BANKNAME CUSIP Alpha Beta Return iduals gt O Deviation Bankameric 06605010 0 00042 1 394 0 00185 52 00 0 01748 BankersTru 06636510 0 00273 1 089 0 00451 50 00 0 01107 BankofBost 06071610 0 00134 1 007 0 00299 48 00 0 01261 lines omitted ChaseManha 16161010 0 00081 1 095 0 00260 48 00 0 01216 WellsFargo 94974010 0 00395 0 952 0 00550 52 00 0 01283 MEAN 0 00176 0 939 0 00330 46 00 0 01788 MEDIAN 0 00153 1 025 0 00298 46 00 0 01681 When you run Eventus the SAS log window or log file depending on the mode of SAS operation reports the completion of data steps and procedures that Eventus executes internally Most Eventus users will find that these reports have little meaning for them and they may ignore most of what is in the log However it is still a good idea to look briefly through the sas log for messages that begin with EVENTUS NOTE EVENTUS WARNING or EVENTUS ERROR If you contact us for technical support to help resolve a problem with an Eventus run please include the entire log window contents or log file resulting from a single run 4 3 Abnormal Returns between Paired Ev ents The TWIN Option The TWIN option provides a means of computing portfolio cumulative average abnormal returns over periods that vary in length from one firm to another The Eventus statements in figure 4 8 run a TWIN event study For the most part the options
194. tion 28 McConnell John J 105 mean adjusted returns 25 50 66 105 MEDIAN option 30 53 Mikkelson Wayne H 29 52 MINESTN option 23 75 missing returns code 118 152 estimation period 23 75 118 147 149 event period 28 66 118 Mitchell Mark 108 Moore William T 108 Moriarity Shane 32 55 multiperiod returns 26 Musumeci Jim 28 52 109 111 135 140 National Market System indicator 91 156 Nayar Nandkumar 108 112 NDAYS option 86 97 NODIVIDX option 20 with RETURNS 87 NOMAR option 25 41 50 NOMM option 25 41 50 non trading dates 20 96 122 145 NONCRSP option 45 nonparametric tests 29 53 112 NOPLIST option 24 49 NOSTD option 25 41 50 number of trades 91 O Hara Maureen 31 32 34 54 55 56 OLDSTUDY statement 138 options ALLDAYS 138 BOOT 139 BUYHOLD 139 CDCSI 139 CSECTERR 139 ID 139 JACKNIFE 139 MEDIAN 140 NOMAR 41 NOMM 41 NOPLIST 140 NOSTD 41 RANKTEST 140 STDALL 41 140 STDCSECT 140 STDONLY 41 SW 140 TAIL 140 output data file 92 97 98 101 CAARS 63 CUSIPERM 101 for cross sectional analysis 63 individual days 63 PRICES 144 173 RETURNS 153 VOLUME 157 OUTSAS file contents controlling 133 event studies variable names 116 interpreting 116 OVERLAP option 28 51 PACKAGE specifications 31 54 133 Partch M Megan 29 52 Pastena Victor S 105 Patell James 28 52 105 Peterson James D 112 Peterson Pamela P 1 Pilotte Eugene 28 51 POOL option 22 74 POST option 27 51
195. to merge two saved event studies If you create and save data sets with the following programs EVENTUS EVENTUS REQUEST REQUEST EVTSTUDY OUTSAS MYLIB RES1 EVISTUDY OUTSAS YOURSAS RES2 you can merge the two portfolios and print the combined event study with a third EVENTUS TITLE text TITLE2 text WINDOWS 45 2 1 5 OLDSTUDY INSAS MYLIB RES1 INSAS2 YOURSAS RES2 TAIL 2 42 Chapter 4 Event Studies Using Non CRSP Data While the CRSP database is the only data source from which Eventus auto matically retrieves and assembles data you can run an Eventus event study with other data too This chapter describes how to conduct an event study where Eventus reads stock return data that the researcher has already re trieved and organized If your data source provides only price and index level data you will need to calculate returns before running Eventus Datastream users can automate most of the process using the EventStream package from Cowan Research L C EventStream accepts Eventus like op tions and request files generates ready to run DSWindows macros for use with the Datastream terminal application pre processes downloaded data builds USERSTOK files and generates Eventus statements Please see our www eventstudy com www eventstudy com web site for details 4 1 Event Studies Centered on a Single Event Date Figure 4 1 displays the Eventus statements to run a single
196. to run an event study ALLDAYS If the number of days covered by the event study PRE POST 1 exceeds 98 not all the individual day portfolio returns are listed by default Day 0 is always listed together with a proportionate number of contiguous PRE and POST days All the windows you request are printed regardless of their date ranges To obtain a complete printout of all the individual days portfolio returns specify ALLDAYS This option is ignored for TWIN event studies BOOT Requests bootstrap tests for window abnormal returns Bootstrap tests are in addition to the normally reported parametric and nonpara metric tests BOOT implies the BUYHOLD and STDCSECT options It is not possible in the current version of Eventus to bootstrap TWIN event studies BOTH Please see VALUE BTAIL 1 2 Selects one or two tailed bootstrap tests The default is 2 This option is ignored unless the BOOT option also is specified BUYHOLD Specifies buy and hold compounded return computation for win dows instead of the default additive cumulation CDCSI Requests the Collins and Dent 1984 test assuming cross sectional independence Sanders and Robins 1991 instead of the default stan dardized test 130 CP Requests comparison period mean adjusted returns Does not affect market adjusted and market model returns CSECTERR Substitutes a cross sectional standard deviation for the default time series standard deviation in non standardized t statisti
197. tores them in a disk file It is intended mainly for those researchers who want to analyze returns entirely with their own programs EXTRACT is useful for those who want to take advantage of the event study features of Eventus then perform their own cross sectional analyses Figure 5 1 displays the Eventus statements needed to extract and cu mulate saved event study results The statements and their options are explained below 5 1 The EVENTUS Statement The EVENTUS statement is required It sets up internal parameters for your Eventus run No options are needed on the EVENTUS statement 63 Figure 5 1 Eventus statements for extracting saved event study results EVENTUS WINDOWS begin end begin end EXTRACT INSAS libref membername ID variable IDFMT format VWPREFIX prefix WPREFIX prefix TYPE CP RAW MAR SW VALUE EXTEND n SERIAL BUYHOLD TEXT BINARY OUTSAS libref membername EXTFILE fileref 2 ee 5 2 The WINDOWS Statement Use the WINDOWS statement to specify one or more intervals or windows of days weeks or months the data frequency that the event study used The cumulative or compounded abnormal return for each window is cal culated and stored for each firm You can list up to 99 windows on one WINDOWS statement The windows must be within the PRE and POST limits of the original event study Within the limits the windows need not be the same as those listed in the WINDOWS st
198. trading day or month numbers filename request G Some Folder Filename extension filename userdata G Some Folder Filename extension EVENTUS TWIN MONTHLY DATECONV CUSIPERM ID variable IDFMT format INSAS libref membername INSAS2 libref membername DATEFMT MMDDYY YYMMDD DDMMYY DATE CRSP AUTODATE NDAYS n SHIFT1 n SHIFT2 n OUTDTFMT MMDDYY YYMMDD DDMMYY DATE OUTSAS libref membername EXTFILE fileref 8 2 The DATECONV statement The DATECONV statement instructs Eventus to read your request file which is an external file that contains PERMNO identifiers dates and sometimes other information A SAS filename statement or the operating system should associate the fileref REQUEST with your request file Each line of the request file for a DATECONV program should contain the following items PERMNO first date second date if the option TWIN appears on the EVENTUS statement and the identifying variable value if any Omit the second date if you specify NDAYS explained below on the DATECONV statement When converting calendar dates to month numbers the dates in the request file may be any day of the month When converting to week numbers each calendar date can be the last trading day in the week or with the AUTODATE option see below any prior date in the week Tf you are familiar with permanent SAs data sets you may want to read about the INSAS option on page 147 94 Searching by CU
199. tter retains its default value If the event period length PRE POST 1 exceeds 35 days weeks or months only 35 normally are printed 95 if the PAGE WIDE option is in ef fect Any windows specified on a WINDOWS statement are printed however 27 regardless of the range of dates covered If you want to print results for all the dates individually specify ALLDAYS on the EVISTUDY statement For further discussion see page 130 in Appendix B If you specify PRE or POST you may need to change the estimation period from the default To change the estimation period specify EST on the REQUEST statement Eventus will stop and print an error message if the event period defined by PRE and POST overlaps the default or specified estimation period If for some reason you want to override overlap checking and allow the estimation period and the event period to have dates in common specify OVERLAP on the EVTSTUDY statement Dropping observations with missing event period returns Specifying MAXMISS n on the EVTSTUDY statement causes Eventus to exclude from the analysis any firm that has more than n missing returns in the event period PRE POST To keep only those observations with no missing event period returns specify MAXMISS 0 Computing cross sectional standard errors instead of time series standard errors Normally Eventus computes portfolio standard errors from the time series of estimation period portfolio abnormal returns The CSECT
200. tus normally computes for the standardized abnor mal return method is the widely used statistic described by Patell 1976 The STDCSECT option substitutes the standardized cross sectional test This is an extension of the Patell test introduced by Boehmer Musumeci and Poulsen 1991 The standardized cross sectional test compensates for a pos sible variance increase on an event date by performing a simple cross sectional variance adjustment The SERIAL option explained below is automatically invoked by the STDCSECT option EGLS and Collins Dent tests The EGLS and CDCSI options replace the default standardized test with the estimated generalized least squares test and Collins Dent test assuming cross sectional independence respectively These tests are discussed in detail by Sanders and Robins 1991 The EGLS and CDCSI options automatically invoke the SERIAL option Adjusting window z tests for serial dependence The SERIAL option applies to standardized abnormal returns only Normally the test statistics for abnormal returns cumulated over intervals you define in the WINDOWS statement are not adjusted for serial dependence Mikkelson and Partch 1988 and others perform such a correction on cumulative returns The SERIAL option causes Eventus to apply the correction to standardized abnormal returns only Karafiath and Spencer 1991 and Cowan 1993 provide simulation evidence of the properties of the corrected and uncorrected test statist
201. tus run please include the entire log window contents or log file resulting from a single run 79 Figure 6 5 First part of the results for the event parameter approach demonstration Event Parameter Approach Demonstration 2 Cumulative Median Average Cumulative Abnormal Abnormal Window Return Return 5 2 1 07 1 68 1 0 4 30 3 28 1 5 3 05 4 21 Test 5 2 CAAR 0 Numerator 0 366193 DF 1 F Value 0 3662 Denominator 1 DF 388 Prob gt F 0 5454 Test 5 2 CAR 0 for all firm events Numerator 0 460447 DF 4 F Value 0 4604 Denominator 1 DF 388 Prob gt F 0 7648 Test 5 2 CAR is equal across firm events Numerator 0 608711 DF 3 F Value 0 6087 Denominator 1 DF 388 Prob gt F 0 6097 Test 1 0 CAAR 0 Numerator 12 08822 DF 1 F Value 12 0882 Denominator 1 DF 388 Prob gt F 0 0006 Test 1 0 CAR 0 for all firm events Numerator 4 846007 DF 4 F Value 4 8460 Denominator 1 DF 388 Prob gt F 0 0008 Test 1 0 CAR is equal across firm events Numerator 3 614898 DF 3 F Value 3 6149 Denominator 1 DF 388 Prob gt F 0 0134 Test 1 5 CAAR 0 Numerator 2 326272 DF 1 F Value 2 3263 Denominator 1 DF 388 Prob gt F 0 1280 Test 1 5 CAR 0 for all firm events Numerator 3 19089 DF 4 F Value 3 1909 Denominator 1 DF 388 Prob gt F 0 0135 Test 1 5 CAR is equal across firm events Numerator 3 065976 DF 3 F Value 3 0660 Denominator 1 DF 388 Prob gt F 0 0280
202. uest are the portfolio cumulative abnormal return the test statistics the number positive and negative and significance symbols You may want to see window or daily abnormal returns for individual firms EVTSTUDY will print individual firm abnormal returns only for the standardized abnormal return method To obtain window cumulative ab normal returns for every firm specify DETAIL on the EVTSTUDY statement As many firms will appear on a single page as there is room for To obtain complete daily abnormal returns for each firm specify DETAITL FULL Full detail takes at least one page for each firm so be prepared for lots of output If you need to look at only a few individual days use the following trick Suppose you want to see day 1 for every firm On the WINDOWS statement include 1 1 Then specify DETAIL not DETAIL FULL on the EVTSTUDY statement This way you get the day you need without all the extra output DETAIL specifications are ignored if you specify NOMM or NOSTD unless you also specify STDALL For individual raw or abnormal returns from non standardized method event studies see the EXTRACT statement in Chapter 5 Specifying benchmarks and standardized or non standardized tests The default is to compute three sets of abnormal returns and test statis tics market adjusted returns market model abnormal returns using portfolio To print all the dates in the range defined by PRE and POST when the number exceeds 35 or 95 w
203. ult value If the event period length PRE POST 1 exceeds 35 days weeks or months only 35 normally are printed 95 if the PAGE WIDE option is in ef fect Any windows specified on a WINDOWS statement are printed however regardless of the range of dates covered If you want to print results for all the dates individually specify ALLDAYS on the EVITSTUDY statement For further discussion see page 130 in Appendix B If you specify PRE or POST you may need to change the estimation period from the default To change the estimation period specify EST on the REQUEST statement Eventus will stop and print an error message if the event period defined by PRE and POST overlaps the default or specified estimation period If for some reason you want to override overlap checking and allow the estimation period and the event period to have dates in common specify OVERLAP on the EVTSTUDY statement Computing cross sectional standard errors instead of time series standard errors Normally Eventus computes portfolio standard errors from the time series of estimation period portfolio abnormal returns The CSECTERR option specifies that the standard error for each date in event time should be computed across securities instead For an example of the application of the cross sectional method see Pilotte 1992 This option has no effect on the standardized abnormal return method 51 The standardized cross sectional test The z statistic that Even
204. variables are included in the data set PERMNO CUSIP if the CUSIPERM option is in effect CRSPDAY1 DATE RETURN and any identifying variable you list on the REQUEST statement If you specify INDEX on the RETURNS statement MARKET and _WEIGHT_ are also included CRSPDAY1 is the CRSP trading day number for the beginning date you requested DATE is the actual date of the return recorded as a SAS date variable RETURN is the security return and MARKET is index return including dividends The variable WEIGHT_ is a character variable of length 5 equal to Equal Value NASDQ or SP500 There is one observation in the data set for each trading day in the interval you request Use the VSAS option if you plan to process the data with a SAS program that uses By variable techniques Specify the SAS data set name using the OUTSAS option B 10 TITLE and TITLE2 Statements TITLE statements are optional they specify page headings mainly for event study output Enter title statements after the EVENTUS statement TITLE and TITLE2 are regular SAS statements not part of Eventus as such The default title is EVENTUS Eventus does not use additional title statements TITLES etc if you specify them they are ignored B 11 UPCUSIP Statement This statement is used after the EVENTUS statement to update CUSIP identi fiers to match the latest version of the CRSP stock file installed at your site Now that CRSP stock files are sorted
205. ve output formats are available text horizontal text binary vertical SAS data set and horizontal SAS data set The default text format can be printed viewed on a display screen and processed by SAS and other programs The binary format specified by the word BINARY on the RETURNS PRICES or VOLUME statement can be processed by programs The vertical and horizontal 87 SAS data sets can be read only by SAS procedures and data steps In a vertical data set data are stored under a single variable name RETURN PRICE or VOLUM with one SAS observation per trading day or month per firm In a horizontal format data set there is a separate variable for each trading day or month so that there is one SAS observation per firm See Appendix B pages 141 and following for more detailed descriptions of the formats Obtaining standard deviation portfolio excess returns The SPORT option applies only to the RETURNS statement not the PRICES and VOLUME statements and only when the EXCESS option appears on the EVENTUS statement and the appropriate CRSP add on module is installed There are two kinds of returns The first which Eventus selects by default is based on portfolios of stocks ranked by beta The other is based on a ranking by standard deviation of return To use standard deviation excess returns instead of beta excess returns specify SPORT on the RETURNS statement Reporting market index returns with stock returns These options are va
206. vent dates The number of trading days between the two event dates varies from firm to firm Let the cumulative abnormal return for firm j be T2 CART To XS Aje t T where Tij 75 are the two event dates specific to firm j Let L be the length of the event period in trading days L Toj Tiz 1 The z statistic for testing the significance of CAR r 7 is Toj BAR t T 2er D JD Assuming cross sectional and time series independence the test statistic for N CAAR X CART To j 1 107 is i N zaa N3 z j 1 Correction for Correlation of Abnormal Returns When the SERIAL option appears on the EVTSTUDY statement Eventus uses a corrected version of the Patell test The correction affects only windows not single period test statistics Following Mikkelson and Partch 1988 the corrected test statistic for the null hypothesis that CAAR 0 is N Jaana Ne A CAAR j l SCART 7 where 2 2 j SCART Taj SA Lj 1 T D A 1 For an event study centered on a single event date Tij T2 and L are equal across firms and the subscript 7 can be dropped from them The corrected test accounts for the fact that within the window the ab normal returns for each stock are serially correlated The serial correlation occurs because all the abnormal returns are functions of the same market model intercept and slope estimators Applications of the corrected test in addition to Mikkelson
207. ventus run Use this version to create a copy of your request file that contains PERMNOs There also is a CUSIPERM option on the REQUEST statement that converts CUSIPs on the fly during an event study or data extraction job However the REQUEST statement option version does not produce a converted version of the request file for use in future Eventus jobs Figure 9 1 lists the Eventus statements required to convert CUSIP identi fiers The statements are discussed below 99 Figure 9 1 Eventus statements to convert CUSIP identifiers to CRSP permanent identification numbers EVENTUS CUSIPERM COLUMN 1 EXTFILE fileref 9 1 The EVENTUS statement The EVENTUS statement is mandatory No options are needed on the EVENTUS statement for a CUSIPERM run 9 2 The CUSIPERM statement The CUSIPERM statement instructs Eventus to read your request file which contains CUSIP identifiers A SAS filename statement or an operating sys tem control language statement must associate the fileref REQUEST with the request file In a CUSIPERM program the request file need not conform to the request file format described elsewhere in this guide An eight character CUSIP identifier must appear within the first 80 columns of each line The CUSIP may begin in any column position on the line If the CUSIP is the first non blank item on each line it need not even start in the same column on each line For example it could be preceded by two blanks
208. with a common grouping variable value receive equal weights in the group portfolio To change this specify the GRWEIGHT option Each line of the request file then must contain a within group weight on each line after the grouping variable value The weights must sum to 1 For example suppose a researcher specifies this REQUEST statement request id Event group GROUP grweight The following request file provides the required information 56899 19850131 00 84 08333 56899 19860131 01 85 16666 56899 19870211 02 86 20 56899 19880205 03 87 20 60038 19850227 04 84 08333 60038 19860226 05 85 16666 60038 19870302 06 86 20 60038 19880223 07 87 20 28804 19850130 08 84 08333 28804 19860314 09 85 16666 28804 19870206 10 86 20 28804 19880209 11 87 20 42083 19841018 12 84 25 42083 19851016 13 85 16666 42083 19861016 14 86 20 42083 19871022 15 87 20 18817 19850213 16 84 25 18817 19860307 17 85 16666 18817 19870310 18 86 20 18817 19871221 19 87 20 44740 19850222 20 84 25 44740 19860225 21 85 1667 Following the PERMNO and event date each line lists the identifying number Event the group number GROUP and the weight for the observation within its group The example happens to have several repeated PERMNOs but this has no particular relation to the grouping feature there could just as well have been 21 unique PERMNOs Event 00 is part of GROUP 84 and is to receive weight 08333 while Events 04 08 12 16 and 20 in the

Eventus User's Guide

Contents

Download Pdf Manuals

Related Search

Related Contents