Capital Stock User Guide
Contents
1. Imputation 2 also interpolates but provides momentum extrapolation at the edges Imputation 3 creates an average capex figure over the real values and then uses this number to fill in the missing cells At this stage the three capex series already paint different pictures which will impact on the estimated capital stock series produced by the PIM It is critical to stress the importance of the net position i e negative values are likely The graph below shows what happens to the imputed series when net capex becomes negative in some years Ga imputed 1 imputed 2 imputed 3 The capex values given above are typical of the values we see in the microdata and reflect the volatile nature of firm investment patterns The firm shown above is a net investor 10k over the four years for which we have data it is not clear which gt This technique is used in Martin 2003 4 The use of net investment partially counters criticism of constructing capital stock estimates at the Reporting Unit level i e firms which close local units should report a negative investment However as illustrated by Harris 2005 firms will not be allocated additional capital when they acquire new local units This is another reason why the capital stock series may underestimate the true figures imputation technique if any could be deemed to be best given the volatile nature of firm investment Methods 1 and 2 give strong investment2. 2003 each asset should be weighted according to its asset s price Bibliography Attanasio O P L Pacelli I Reduto dos Reis 2003 Investment Patterns in UK manufacturing Establishments Doms M and T Dunne 1998 Capital Adjustment Patterns in Manufacturing Plants Review of Economic Dynamics n 1 409 429 Oulton N and S Srinivasan 2003 Capital stocks capital services and depreciation an integrated framework Bank of England Working Paper n 192 Tevlin S and K Whelan 1999 Explaining the Equipment Investment Boom of the 1990s Division of Research and Statistics Federal Reserve Board December 29 mimeo Vaze P 2003 Estimates of the volume of capital services Economic Trends n 600 Construction of Capital stock Description of programme The construction of the capital stock involves the use of the following do files 1 prepare_deflators do 2 make_row do 3 pim do 1 prepare_deflators do Input deflators from VICS xls these are implied deflators from historic investment series calculated by the ONS for the construction of 2003 Volume Index of Capital Services VICS The Volume Index of Capital Service measure the flow of capital services derived from all the capital assets that exist in a sector see Vaze 2003 Deflators are split by asset industrial sector and year Assets b Buildings poth Plant other plant and machinery including purchased software a
3. 5 1986 1987 1988 1989 1 1991 1992 While it is possible to impute for any number of missing values the quality of the imputation will suffer when we have fewer real values to base our calculations on i e we are likely to impute disposals of capital for all years for some firms In practice it is prudent to set a tolerance level Currently the maximum ratio of imputed to real values is set at a maximum of 1 1 Lowering our intolerance to imputed values will create estimates of firm capital stock for many more firms and is likely to pick up more SMEs due to the stratified sampling framework within the ABI The table below outlines the number of firms for which we calculate capital stock given different tolerance levels Users can easily alter the calc_ratio in the syntax file and re run to suit their needs Tolerance level max number of Number of firms with Calc_ratio imputed to real values capital stock series 0 33 1 3 13131 0 66 2 3 21809 1 1 1 41782 1 5 3 2 50622 2 2 1 74356 3 Allocating capital to firms The first stage in the PIM is to calculate the share of total industry capital which should be allocated to the firms in your sample We do this by assuming that the ratio of firms investment i e sum of individual firms investment to industry investment corresponds to their share of industry capital The calculation of the initial capital allocation is not precise and allocating firms too little capita
4. IC Total code letter of Method letter code purchases Employment TP Employment capital m1 1 a al 20 50 0 50 0 10 0 05 m1 2 a al 20 25 0 50 0 10 0 05 m1 3 a a2 50 200 0 33 0 90 0 30 m1 4 a a2 100 500 0 67 0 90 0 60 SIC SIC code by Share Firm SIC SIC Total code letter Total of Method letter code purchases Employment Employ purchases capital m2 1 a al 20 50 0 67 0 21 0 14 m2 2 a al 20 25 0 33 0 21 0 07 m2 3 a a2 50 200 0 29 0 79 0 23 m2 4 a a2 100 500 0 71 0 79 0 56 The flexible set up of the capital stock syntax allows a user to specify which method m0 ml or m2 they would like to use in their capital stock Investigation by the MAUS team indicates that Method 1 produces the best results See table below Counts of the number of negative capital stock series created under different methods method All assets Plant amp Machiner Vehicles Buildings m0_totpurch 229 173 1719 207 m0_matfuel 622 444 4888 943 m1_totpurch 216 164 1652 189 m1_matfuel 613 446 4719 853 m2_totpurch 293 220 2596 333 m2_matfuel 393 282 4525 689 This table illustrates one of the major problems in constructing the capital stock data set The volatility in the net investment figures results in over 10 of the capital stock series for vehicles becoming negative Obviously it is impossible for a firm to hold a negative stock of vehicles and this points to a combination of the main problems not allocating enough initial capital an inability to split capital at a de
5. UK Data Archive Study Number 6698 Capital Stock Dataset 1979 2005 Secure Data Service Access Quick Guide for Users Dataset Capital Stock Dates available 1980 2001 Source Businesses Coverage All sectors Collected by ONS Link fields dlink_ref2 DBR Legal restrictions STA data Quick summary The ARD has no capital stock information The suite of programs above use industry investment data and the ARD s own capital expenditure figures to create an estimate of capital stock for each reporting unit in each year 1980 2001 The estimates are calculated using the Permanent Inventory Method PIM Separate depreciation rates are used for plant and machinery vehicles and buildings but an aggregate capital stock is reported The document building the capital stock_oct2002_02 pdf contains a description of the method particularly in respect of the treatment of missing values initial codes and firm births It also shows the impact of some alternative assumptions The variables in the files are year year of observation siclett 2 letter sic92 all_def95 average investment deflator average over p amp m buildings and vehicles sel_idx sel_idx 1 if capital stock could be computed for reporting unit rcapstk95 estimate of real capital stock with 1995 base year dlink_ref2 reporting unit identifier Sampling frame The unit is the reporting unit from the ARD In its current form the data is available fr
6. aggregated into one asset type called pm e drop observation lt 1979 e transform capital stock s unit from millions to thousands pound to make it consistent with ARD investment data which are in thousands e assign 0 if capital stock is negative this is to avoid negative shares e saved as prabstock02 dta Siclett Year ind_capstk95b ind_capstk95pm ind_capstk95v e Input sectoral investment in constant price from VICS xls This is the Gross Fixed Capital Formation GFCF in 1995 prices used by the ONS for the construction of 2003 VICS The GFCF consists of resident producers acquisitions less disposals on fixed capital assets as above split by asset type industrial sector and year e poth pcom and int are aggregated into one asset type called pm e drop values lt 1979 e If invest is 0 it is replaced with a missing value zero values would not allow us to estimate the sectoral share of capital stock to be assigned to the sample of selected reporting units later on in the programme e Missing values are linearly interpolated e Take absolute values of investment this is to avoid negative shares e Transform investment s unit from millions to thousands pound e Save as prabinvest dta Siclett Year ind_ invest95b ind_invest 95pm Ind_ invest 95v e Input reporting unit level information from the ARD Standard Variables dlink ref2 reporting unit refer
7. are sse rncapex Firm Capital Stock year Firm Capital Stock 1 asset depreciation rncapex asset year Firm Total Capital Stock asset year gt Firm capital years year asset We choose to use geometric depreciation because the asset price depreciation and physical deterioration age efficiency profile of the assets coincide gt See Fraumeni 1997 See ONS Productivity Handbook 2007 5 Dealing with the remaining negative capital stock series Given that a negative capital stock series is impossible in reality we can address this problem by injecting additional capital at a firm level This can be done by stipulating that a firm s capital stock must not go below zero increasing the investment figures in the proceeding years and then re running the PIM It is advisable to inject the capital close to the year in which the capital stock turns negative consider the huge capital injection which would be required if a 20 year vehicles series turned negative in the last year i e a 1000 figure would need 87 000 1000 1 0 8 20 injection in the first year to ensure that an entirely positive series is created For this reason we localize the injection to at most three years previously Asset Investment Injection negative capital srock Hf aeppecus son The table below illustrates how a firm s vehicle stock may turn negative assuming 20 depreciation and how a localized injection
8. are is estimated assuming the following hypothesis Reporting unit s initial capital share Reporting unit s average material share e Calculate material share P reporting unit purchase of goods and services totpurch e By year and sector creates the sum of total purchases across all units sum_totpurchjj gt Pi if sel_ idx ie jt e Reporting unit material share P jti ie jt e Calculate reporting unit average material share average across life span of the unit o Dm M te T e Calculate reporting unit average material share relative to year totals 1 mM TM tei Yit A 1 2M Y ym icjt ie jt tei save as cappanel_raw 3 pim do e Set local depreciation rates Local pm 0 06 e Set depreciation rates by asset d_all pm b v 3 d pm 0 06 d_b 0 02 d v 0 2 e Generate empty capital stock at reporting unit level gt rcapstk_aa 95 gt Kit e Estimate total capital stock to be assigned to the selected sample in each year sector This is estimated multiplying VICS capital stock by the investment share of selected unit in corresponding year sector rem_ aa ind _capstck95 aa invest_inflator aa rem_ aa K i Pa make it equal to 0 if negative Estimate initial capital stock at reporting unit level This is obtained multiplying the capital stock of all selected units in sector j and year t age 1 year in which the unit first appears in the sample by the average materia
9. correctly depreciated and deflated Worth investigating with the ABI team the possibility of constructing separate investment data for computer Depreciation rates should be updated using the information available in VICS xls An average depreciation rate for poth pcom and int should be calculated taking into account the relative importance of each asset Furthermore depreciation rates should be kept with a split by asset and sector current version is split by assets only 5 When constructing the initial capital stock at enterprise level the sectoral capital stock is split among all reporting unit in sector j and year 1 Then the investment is added to the initial capital stock This causes a double counting problem because the investment in the initial year is already accounted for in the sectoral capital stock One way to solve this problem would be to lag the sectoral investment and relative shares by one period K i Kja Paja Yje T I iaji t age 1 g share of sectoral capital stock to be assigned to selected sample y share of selected sample capital stock to be assigned to the individual reporting unit 6 At the end of the programme the capital stock by asset is aggregated to obtain an overall measure of capital stock rcapstk95 When aggregating different assets we should use appropriate weights currently we just sum up the 3 measures of capital stock by asset According to Oulton and Srinivasan
10. d those which are selected only in some years hence drop units which are never selected In years in which the small enterprises aren t selected we will interpolate the missing investment values Merge Register Panel variables with Standard Variables merge rpanel_cap dta using dlink_ref2 Keep following variables dlink ref2 year ncapex renamed ncapex_all ncapex_pm ncapex_b ncapex_v sel_id marker which indicates if reporting unit was selected in a specific year sel emp go totpurch sic92 region Correct wrong asset types in 1994 amp 1998 by assigning missing values Create age variable _n this variable will be used to identify when the reporting unit enters the dataset for the first time This information is required when estimating the initial capital stock at reporting unit level Create SIC92 with 2 digit from SIC92 4 digit SIC 92 2 digit are converted into SUT classifications using numlettlook_7 See VICS xls for info This conversion is necessary to make ARD industrial classification consistent with VICS industrial classification Input deflators e Create an average deflator across all assets all_def95 e Use deflators to create real investment at reporting unit level rncapex_all95 ncapex_all all_def95 rncapex_b95 ncapex_b b_def95 rmcapex_pm95 ncapex_pm pm_def95 rmcapex_v95 ncapex_v v_def95 Interpolate investment missing data when missing at the edge take previous or first non missing and
11. dology implies that firms follow a smooth capital adjustment pattern over time Despite this assumption is in line with neoclassical investment models there is evidence that firms adjust their capital stocks in a lumpy manner Investment seems to occur in large and concentrated episodes Doms and Dunne 1998 and Attanasio et Al 2003 Worth investigating investment patterns to see if assumption behind interpolations are correct Worth considering alternative methods ie regressing missing investment The sectoral capital stock produced for the construction of 2003 Volume Index of Capital Services Vaze 2003 separates computer from plant and machinery and apply the appropriate deflators and depreciation rates Keeping computers separate is quite important because i ICT assets have a shorter life span than other type of assets ii they are characterised by rapid fall in relative price due to quality improvements see Oulton and Srinivasan 2003 and Tevlin and Whelan 2000 When constructing the capital stock at reporting unit level we loose this level of detail In fact since in the reporting unit investment in plant amp machinery includes computer software and hardware when entering VICS data we aggregate the sectoral information for plant and machinery poth with hardware pcom and software int in order to make it consistent with the ARD information As a consequence when running the PIM the capital stock for plant and machinery is in
12. ence number key consistent over time based g on post 94 idbr_ref totpurch total purchases of goods and services ncapex total net capital expenditure ncapex_pm net capital expenditure for plant and machinery ncapex_b net capital expenditure for land amp buildings ncapex_v net capital expenditure for vehicles go Gross output used for checks only employment figures held to sampled enterprises used for checks only In this context net capital expenditure means acquisitions less disposal proceeds of fixed assets It excludes any allowances for depreciation See code amp derivation file for more details on capex variables sel_emp e ncapex pm is not reliable thus it is replaced with a residual figure calculated from total net capital expenditures as follows ncapex_pm ncapex ncapex_b ncapex_v ncapex derived from total acquisition q600 ABI less disposal includes land buildings plant machinery hence computer hardware purchased computer software and computer software developed by own staff see ABI survey form for more detail Consequently ncapex_pm represents an estimate for investment in plant and machinery including hardware and software e Save as rpanel_cap dta Load the Register Panel dataset reg_pan dta The Register Panel which contains selected and non selected reporting unit all in one file allows us to identify and keep reporting units which are always selected an
13. eries which are created and are negative The capital stock data set created by the MAUS team indicates that the best results for all years are obtained by specifying the key variable to be total purchases A basic allocation of firm level capital can be created using E key var mO_Share py key VAL siclett J We shall refer to the basic formulae above as Method O m0 in the syntax file Method 0 is only able to split capital at the level of SIC letters which is another reason why capital stock series may become negative Consider firms in the same SIC letter but different SIC codes e g al Civil engineering a2 plumbling SIC SIC total Share of Method Firm Year letter code purchases capital mo 1 1998 a al 20 0 11 m0 2 1998 a al 20 0 11 m0 3 1998 a a2 50 0 26 mo 4 1998 a a2 100 0 53 We can make an attempt to partially remedy this situation by using a more advanced method of allocating capital to firms We can weight the allocation of capital by both Total Purchases and Employment or an alternative variable The formulae below outline two alternative methods for creating a share measure ml_Share totpurch P D employment ieza year gt totpurch 34 year x employment eter year mI Sharez employment rn D totpurch s4 year SY employment eza year YS totpurch sen year The tables below illustrate how the 2 different methods create different shares compared to the original methodology SIC SIC code by Share Firm SIC S
14. figures but the true values could equally be negative It should be noted that almost all imputation techniques will smooth the estimated investment patterns below the volatile true series which we would see if we had the real investment figures for the missing years The MAUS team decided to use the employment series to help with the imputation of capex It is hoped that this gives the estimated capex figures more real information compared to just using one of the techniques outlined above The latest capital stock data set thus creates a capex series which inputes using an average of net capital expenditure per employee for the real data and then applies this to the missing values This is similar to imputation method 3 above and while it produces low investment figures for missing years it does ensure some consistency in the net position i e net positive investors continue to invest over the years This technique is increasingly valuable in highly volatile series where the other methods often impute large negative figures The table and graph below illustrate the imputation for a firm which is a net investor for the four years which we have real data year capex employment Imputed 1985 100 1282 1986 145 1859 1987 20000 200 20000 1988 13000 190 13000 1989 250 3205 1990 17000 200 17000 1991 6000 190 6000 1992 230 2949 Averages 2500 195 n a Average capex per employee 13 m Imputed e employment 198
15. g entries for firms when they do not appear in the ARD but are still trading e g if a firm is in the ARD in 1997 and 2000 the ARDPD ensures it is present for 1998 and 1999 also This process creates a lot of missing values which must be imputed in the previous example we would not have any investment figures for 1998 and 1999 to feed into the PIM and section 2 outlines the approach we have taken to address this issue To create a PIM it is necessary to allocate shares of the aggregate capital stock series to firms for the first year that they appear in the survey basically the PIM depreciates this value adds the net investment the following year and repeats these steps for all years The aggregate capital stock series used is the Volume in Capital Services VICS Capital services are a flow measure which reflects the input of capital into production thus are more suitable for analysing productivity than the National Accounts wealth estimates of capital stock The main difference between the National Accounts measure of capital stock and the capital services estimate is how the net stock estimates are weighted Capital stocks are weighted using asset purchase prices while capital services are weighted using asset rental prices The rental price is sometimes referred to as the user cost of capital and it captures three components depreciation of the asset nominal rate of return and the change in the purchase price of the asset If a f
16. generate rncapex_aa95 old to distinguish original data from the interpolated one rmcapex_aa95 aa asset In order to identify reliable interpolation and original values generate sel_idx 1 if the maximum number of interpolations per reporting unit are less than 6 and investment values ncapex_all95 are non missing This marker will be used at the end of the programme to restrict the calculation of capital stocks to reliable investment figures only Input investment and capital stock data at sector level prabstock02 dta Calculate the initial share of sectoral capital stock to be assigned to the sample of selected units from now on we will refer selected as selected by the survey and interpolated In fact since our sample of units represent only a certain percentage of enterprises in the sector we need to calculate how much of the sectoral capital stock we need to allocate to our sample of units in each sector and year We calculate this share using the following hypothesis Capital share of selected sample Investment share of selected sample Q Calculate the investment share of selected sample in every sector year as a percentage of sectoral investment in corresponding sector year i reporting unit aa asset j industrial sector I investment in real terms t year I 2 icajt I iajt Pajt invest_inflator aa j ajt Calculate reporting unit s share of initial capital stock This sh
17. irm is to rent an asset the lease price must cover the loss in asset value due to depreciation the more an asset depreciates the greater its rental price Asset prices also affect the rental price as such movements reflect a capital gain or loss for the owner Similar to previous a declining asset price must be covered by a higher rental price to compensate the owner Both of these key components can be illustrated by ICT assets Prices of ICT goods continue to fall sharply and a capital stock measure would result in such good being given less weight however this Detailed information about the data sets held in the VML can be found at http www ons gov uk about who we are our services vml about the vml datasets available dataset downloads index html More information about VICS can be found at http www statistics gov uk StatBase Product asp vInk 14205 amp Pos amp ColRank 2 amp Rank 1000 implies a rising rental price and a capital services measure would give such assets a higher weight As noted by Wallis amp Dey Chowdhury 2007 there is a divergence between the volume of capital service and volume of capital stock series after 1980 which becomes increasingly stark after 1990 This divergence is caused by the move towards short lived and more productive assets such as computers for which the flow of capital services is high The standard capital stock measure does not capture this shift and so understates growth in the pr
18. l is one reason for the implausible generation of negative capital stock series see section 5 It seems plausible that the stratified sampling framework of the ABI ARD results in too low an allocation of industry capital The largest firms presumably with most capital represent a disproportionate number of the firms in the microdata The formulae for the first stage of capital stock allocation is given by Industry Capital Allocation firm Capital Firm Investment Share asser asset asset Firm Investment Share asset gt rncapex See Industry Investment EE aed Now that we have decided how much capital should be allocated to the firms we are interested in the second step is to devise a method for splitting this capital amongst the firms At this junction we want to choose a variable which is plausibly related to capital stock and is available for all the years The main options are total purchases and materials amp fuels both could be seen as proxies for the capital stock and firms are asked these questions in all years Alternatives which are only available from 1997 onwards include number of local units Spending on Insurance and Spending on Road Transport As noted above the new flexible do file allows users to specify which variable they would like to use and within this do file there is another do file which compares the alternative options against the number of capital stock s
19. l share of the individual reporting unit then adding the reporting unit s investment in same year if t age 1 amp sel_idx 1 rceapstk_aa 95 rem aa mat_sh_av_shtrncapex_aa 95 _ Kan Kaj Pan Vin tA iajl Apply Perpetual Inventory Method for all years capital stock in previous year less capital consumption depreciation plus investment in current year If sel_idx 1 Kia Kia l d 1 iajt iajt rcapstck_pm95 net capital stock for plant and machinery including comp hardware amp software rcapstk_b95 net capital stock for land and buildings rcapstk_v95 net capital stock for vehicles rcapstk95 rcapstck_pm95 rcapstk_b95 rcapstk_v95 save capstock pm dep
20. lysis amp User Support MAUS team has recently made a new version of the Capital Stock data set available within the VML It spans the years 1979 2005 and creates a firm level estimate of capital stock for approximately 42 000 enterprises This data set contains methodological improvements compared to previous versions and specifically addresses the negative capital stock series which occurred in previous constructions of the data set The major development in the latest version of the capital stock data set is the ability for users to tailor the data set for their own purposes The syntax file available within the VML has been designed to be very flexible users can now easily create their own capital stock estimates by selecting from a range of key variables amp methods and re running the file in their own work area For example by selecting a higher tolerance ratio of imputed to real investment figures a user can significantly expand the number of enterprises for which a capital stock series are calculated This Technical Guide is intended to serve as a guide to the construction of the capital stock data set It covers e The data sets used Dealing with missing values and imputation techniques Allocation of the aggregate capital stock series to firms first appearance The Perpetual Inventory Model PIM Addressing negative capital stock series and injecting capital The capital stock data set is unlike the majority of the other data
21. nd excluding hardware pcom Computer hardware v Vehicles int Intangibles dominated by own account software excluding purchased software Industrial sectors The industrial desaggregation is identical to the Supply Use Table SUT s classification for Gross Fixed Capital Formation with a breakdown of 36 industries see information in VICS xls for details poth pcom and int are averaged to create one asset type called pm representing plant and machinery assets In fact we need to aggregate the 3 assets in order to make VICS information consistent with the enterprise level information from the ARD If deflator is missing it is replaced with the average deflators across sectors by year Drop observations lt 1979 Finally create one file per asset type each spit by sector and year b_def95 dta deflator for buildings pm_def95 dta deflator for plant and machinery v_def95 dta deflator for vehicles 2 make_row do Input sectoral net capital stock from VICS xls This is the capital stock net of geometric depreciation estimated by the ONS for the construction of 2003 VICS Net Capital Stock of an asset is a wealth measure calculated summing up the past history of gross investment in that asset in constant prices and subtracting estimated depreciation and retirements Depreciation is what must be spent to maintain the value o capital stock at the existing level e poth pcom and int are
22. oductive input of capital in the UK economy A real firm level capital stock series is constructed by MAUS using deflators to adjust for price changes in asset values over time Deflators are constructed by SIC letter and asset type for each year using 1995 as the base The deflators are then applied to the investment figures net capital expenditure for each asset before they are used in the PIM Real input and output series are required to remove price affects and reveal changes to firms production efficiency 2 Missing values and imputation techniques Some of the most important decisions to make in the construction of the capital stock data set are with regards to the treatment of missing cell values and the cleaning of the data set before the PIM is run There are many different approaches that we could take to this problem and this section outlines how MAUS have chosen to impute cell values Employment is used as the spine for many of the imputations plus the allocation of the aggregate capital share thus the first stage is to ensure that we have a complete employment series Generally we can be confident using employment as this information is available from the Inter Departmental Business Register IDBR hence even if a firm does not respond to the questions in the ABI survey we are likely to have an employment figure When a cell value is missing between years we simply interpolate the value Missing usually coded 0 cell value
23. of capital can ensure that stocks do not fall below zero year 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 net capex_vehicles 10 000 12 000 5 000 2 000 20 000 40 000 20 000 50 000 20 000 5 000 5 000 Capital allocation 50 000 O OO0O0OOOOOGOODO oO capital stock 60 000 60 000 43 000 32 400 45 920 76 736 41 389 16 889 6 489 10 191 13 153 adjusted capex 10 000 12 000 5 000 2 000 52 986 40 000 20 000 50 000 20 000 5 000 5 000 adjusted capital stock 60 000 60 000 43 000 32 400 78 906 103 125 62 500 0 20 000 21 000 21 800 In practice the calculated capital stock series often becomes negative for many observations within a series hence it is often necessary to run the above injection several times The current syntax file calls in the localised_boost do file several times in a row to deal with this problem 1 2 3 4 Capital stock Methodological and data Issues Deflators for poth pcom and int are aggregated with a simple mean A weighted mean should be more appropriate We could use VICS investment as weights Small reporting units are sampled on a random basis hence the ARD displays missing values in those years in which small firms are not selected In order to avoid underestimating the accumulation of capital by small firms before applying the PIM missing investment values are linearly interpolated This metho
24. om 1980 2001 Organisation of files Ga Files are created as capstock_xx dta in the cap_stock folder of the data drive The xx relates to the depreciation level on plant and machinery As there is some dispute about the appropriate rate to use various datasets have been created All years and sectors are aggregated into one file Known data issues A change in the statistical treatment of leasing investment around 1988 occurred but this seems not to have had any impact on the estimates Disinvestment is taken as evidence of prior investment when allocating initial shares This may lead to a disproportionately high share of initial capital being allocated to declining or restructuring businesses Other issues Investigation of 1970 s characteristics needs to be carried out Val Office for Nee National Statistics Virtual Micro Data Laboratory Technical Guide Estimating capital stock Robert Gilhooly The Office for National Statistics ONS does not ask firms what their capital stocks are hence it is necessary to construct capital stock series and then link these to the other confidential data sets held within the Virtual Microdata Laboratory VML This technical guide is designed to be used in conjunction with the syntax file do which created the capital stock data set It provides a step by step guide to the key stages as set out in the syntax file and follows the same ordering to aid understanding The Microdata Ana
25. s are more common in the first years of a company s existence and these are filled in rolling back a three year average of the employment figures available This is illustrated in the table below Original cells Final cells Year Employment Adjusted Employment 1980 0 108 1981 0 111 1982 0 113 1983 100 100 1984 120 120 1985 120 120 1986 114 114 1987 132 1988 150 150 1989 170 170 1990 165 165 We require a complete series of asset investment figures to compute the PIM and must also impute these missing values The investment series in the ARD are net capital expenditure capex by asset It is worth repeating that these figures are net i e acquisitions minus disposals While we do ultimately require the net investment figure for PIM the volatile nature of firm investment and missing values provide a challenge for imputation we are put further into the dark as acquisition and disposals series which might provide more information on underlying investment patterns are not available The following graphs illustrate the problems we face Ga imputed 1 imputed 2 imputed 3 The graph above shows 3 methods of imputing missing values labeled while the figures on the x axis represent real values e g 15 000 for 1988 for a firm which has four positive net capex cells from 1985 to 1992 Imputation 1 interpolates the 1989 value and provides a constant extrapolation at the edges 1985 1986 amp 1992
26. sets held in the VML because it is a constructed data set Therefore there are many decisions needed to overcome the issues outlined above and there is not a definitive approach to take The user of this data set is advised to read this document and then decide whether to use the file as it stands or adjust it using the options outlined 1 Data sets The capital stock data set is constructed from the Annual Respondents Database Panel Data ARDPD set which is itself constructed from the Annual Respondents Database ARD The ARD holds firms responses to the Annual Business Inquiry ABI while the MAUS team does some work on the raw ABI files to ensure that there is longitudinal consistency in variables and firm reference numbers The ABI is the most comprehensive business survey conducted by ONS and covers over 100 key economic variables such as turnover production and operation costs employment industry classification and investment The ABI has surveyed all sectors since 1997 while firms in the manufacturing and construction sectors returned questionnaires in from 1973 and 1994 respectively It is a census of large businesses and a stratified sample of small and medium sized enterprises The stratified sampling framework means that smaller firms move in and out of the survey and to create capital stock estimates it is necessary to have a panel which accounts for such missing years The ARDPD solves this problem by creatin
27. tailed enough level imputation techniques failing to create a large enough investment to assets e g vehicles when there are missing cell values and possibly using too high a depreciation rate In the current methodology all firms are allocated a share of the aggregate capital stock when they first enter the data set The PIM which we specify also adds on the firm s investment which it undertakes in its first year It could be argued that if an investment figure is available we should just use that as the firm s first year capital stock however it seems highly probable that the firm was in existence before it is picked up by the ABI survey thus allocating a share of the aggregate capital stock figure in its first year provides a proxy for a firm s pre ABI capital stock 4 Perpetual Inventory Model After selecting your key variable total purchases methodology capital share allocations and dealing with missing values running the PIM is relatively uncontentious All that is left for the user to do is to select the depreciation rates they would like to use The MAUS team follow the depreciation rates set out by ONS Plant amp Machinery 6 Buildings 2 Vehicles 20 However alternatives such as those based on BEA estimates include Plant amp Machinery 13 Buildings 2 5 Vehicles 25 The PIM uses geometric depreciation as follows Firm Capital Stock g co1 yea Allocation firm Capital sse Firm Investment Sh
Download Pdf Manuals
Related Search