2011 Skills for life survey: small area estimation user guide
Contents
1. Department for Business Innovation amp Skills 2011 Skills for Life Survey Small Area Estimation User Guide The views expressed in this report are the authors and do not necessarily reflect those of the Department for Business Innovation and Skills Department for Business Innovation and Skills 1 Victoria Street London SW1H OET www bis gov uk Research paper number 81D November 2012 2 2011 Skills for Life Survey Small Area Estimation User Guide Contents Lll om iceetay 2 Report Authors and Acknowledgements eeeeeeeeeeeeeeeeeeeee esee enne enhn nnn nnn nnn nnn nennen nn 4 Brel m 5 ludi m 8 SU INrOdUCtO Masset theese eA ech ee cee T T T EITHER OE 8 S Ney Methodology sasirnane p iiia te A 2det bere E E OT o T 8 BE mee Bins eat ees den ges ote ese ea sete Dees apace eee E A euaaeees 8 POW ON NER E RN T EDT mn 9 Data weighting and imputellof uicit tores nues rcu che RE RR KE Rb RR AME RSENEE RH MN EAD BRE MM EE 9 Modelling occisi simia dai ci odo dccesackadonel Kaos ce GG UO a OU wa Ond ROI OR Ga 10 Top line geographical area estimates ceeeseeeeeieeeeeeee nennen nmenme nennen tnis tna ned anna o teneas 10 Local Area SUS se iE epe FX In eR MU Uie M UM MR M M UE M M IM RED M MIS 10 Modelling Local Area ESUGBs iocos2. social and motivational factors related to skills levels in today s population of 16 65 year olds The SfL201 1 headline findings paper is available at http www bis gov uk assets biscore further education skills docs 0 9 11 1367 201 1 skills for life survey findings pdf accessed 27 7 2012 The full SfL2011 survey report will be published in Summer 2012 Survey methodology The interview The interview comprised a background questionnaire followed by a pre assigned random combination of two of the three skills assessments literacy numeracy and ICT The background questionnaire was designed to collect a broad set of relevant demographic and behavioural data A refined and updated version of the 2003 background questionnaire was used redundant items were removed and some new questions were added The literacy and numeracy assessment tools were the same as those used in the 2003 Skills for Life survey SfL2003 to ensure absolute comparability with the 2003 survey Respondents who completed the questions allocated to the literacy and numeracy assessments were assigned to one of the five lowest levels of the National Qualifications Framework Entry Level 1 or below Entry Level 2 Entry Level 3 Level 1 or Level 2 or above A new ICT assessment tool was however included Major advances in software had made possible the development of a more valid and reliable assessment tool for ICT using real office desktop applications The I
3. RAE Consulting and Innovation and Research Fellow at the University of Plymouth and Paul Hewson Associate Professor in Statistics at the University of Plymouth 2011 Skills for Life Survey Small Area Estimation User Guide Data Sets Two sets of small area estimates are provided for seven geographies Middle Layer Super Output Areas MSOAs Standard Table Wards 2005 Statistical Wards 2011 Council Wards 2011 Parliamentary Constituencies Local Authorities and Local Enterprise Partnership areas The first set of estimates concern the number and proportion of adults falling into each of the reported adult literacy numeracy and ICT skill levels as well as of the number and proportion of adults for whom English is or is not a first language see Table 1 below The second set conflates the adult skill estimates into just two categories namely Entry Level and below and Level 1 and above All estimates are accompanied by a measure of uncertainty These 95 Credible Intervals Cls literally describing the range within which we are 95 certain the true number or rate will lie have been drawn directly from each estimate s posterior distribution see Section 0 below If estimates are required for other combinations of skill levels it will be necessary to algebraically approximate the Cls as described in Section 6 of the Technical Report although the estimates themselves can be simply summed Table1 Estimates provided in Exc
4. Skills for Life Survey Small Area Estimation User Guide than 1 000 residents or 400 households were merged into neighbouring wards ST wards were used to report the small area estimates derived from the SfL2003 2005 Statistical Wards n 7 972 In 2003 a policy was introduced across National Statistics to minimise the statistical impact of frequent electoral ward boundary changes particularly in England The wards resulting from this policy which were fixed annually were known as statistical wards A change of policy meant that the last set of statistical wards were for 2005 Statistical wards are now rarely used although postcodes are still assigned to 2005 statistical wards in the National Statistics Postcode Directory 2011 Council Wards n 7 618 Also known as Electoral Wards Divisions these are sub divisions of Local Authorities and are the key building block of UK administrative geography being the spatial units used to elect local government councillors in metropolitan and non metropolitan districts unitary authorities and the London boroughs in England The ward geography used in this report is that current as of the 1 January 2011 and is detailed in the February 2011 ONSPD Council wards have a GSS code starting E05 2011 Parliamentary Constituencies n 533 The English Parliamentary Constituencies used in this report relate to those defined by the Parliamentary Constituencies England Order 2007 and the Parliamenta
5. assume that local populations have the same socio economic composition as they had in 2001 Demographic age sex changes since the 2001 Census have been accounted for by adjusting the estimates to align with the ONS s experimental quinary population estimates for mid 2009 This is likely to be a generally reasonable assumption but as described in Section 4 the Technical Report in some areas this assumption will not hold It should also be recognised that these estimates only apply to people living in households communal establishments were excluded from the SfL2011 sample and the model parameters derived from our analysis of the SfL201 1 have been applied to microsimulated populations describing the socio demographic composition of local household populations e tis important to view the estimates in the light of local knowledge partly because of the data limitation noted above but also because small area estimation cannot capture locally anomalous or unusual patterns in the distribution of adult basic skills Thus if a particular group of people nationally is found to have particularly high skill levels then it is assumed that this will apply to all local areas The multilevel nature of the model will reveal whether this relationship is mediated by any MSOA level effects but the same principle applies these MSOA level effects are also based on modelled relationships derived from an analysis of the dataset as a whole Local knowledge
6. base weight or starting position for an iterative sequence of rim weighting to population totals In keeping with the 2003 survey these population totals were a the cross classification of sex and age and b region and were drawn from the most recent Office for National Statistics ONS mid year population estimates 2009 In line with SfL2003 the assessment scores for some respondents were imputed where they did not complete the assessment For the literacy and numeracy assessments the approach taken was identical to that in 2003 to maintain comparability respondents who either a said they could not read English or b assessed their reading level as poor and chose not to attempt the assessments are given Entry Level 1 or below values for all assessments In addition respondents who had never used a computer are given the lowest value for all the ICT assessments below Entry Level 1 for multiple choice and word processing below Entry Level 2 for email below Entry Level 3 for spreadsheets Any other cases missing a skill Level are excluded from analysis that covers that skill Level Modelling Top line geographical area estimates 2011 Skills for Life Survey Small Area Estimation User Guide The weighted SfL201 1 survey provides a sound basis upon which to estimate levels of literacy numeracy and ICT skills in England as a whole and for individual regions Some of the key findings for literacy and numeracy ar
7. inevitable that in some areas there will have been significant changes to the socio economic composition of populations which cannot be captured until the results of the 2011 Census are published 14 2011 Skills for Life Survey Small Area Estimation User Guide Table5 2001 Census tables used to microsimulate MSOA populations Census Area Statistics CAS Tables CASO001 Age by sex and whether living in a household or communal establishment CAS016 Sex and age by general health and limiting long term illness CASO17 Tenure and age by general health and limiting long term illness CAS021 Economic activity by sex and limiting long term illness CAS026 Sex and economic activity by general health and provision of unpaid care CAS032 Sex and age and level of qualifications by economic activity CASO33 Sex and occupation by age CAS034 Former occupation by age CAS061 Tenure and car or van availability by economic activity CAS105 Age by highest level of qualification CAS113 Occupation by highest level of qualification CAS Theme Tables CASTO3 Theme table ethnic group cross tabulated by a sex b ageband c birthplace d economic activity e limiting long term illness and f resident type Key Statistics Tables KS05 Country of birth UV03 Sex UV04 Age UV09 Ethnic group England and Wales UV22 Limiting long term illness UV24 Qualifications England and Wales UV28 Economic activity UV30 Occupation UV43 Tenure p
8. literacy numeracy and ICT skill levels for every MSOA in England Aggregation of these estimates to regional or national level will produce alternatives to the direct estimates used in this report For the most part the differences between the two are slight but there are several instances in which one of the estimates lies outside of the credible range of the other This prompts the question which is the more accurate It is important to stress that the two methods treat the survey data in a different way With direct estimation it is being treated as a sample drawn from the population With modelled estimation it is being used as a source for identifying and quantifying relationships between demographic characteristics and literacy numeracy or ICT skill levels It is also important to recognise that both sets of estimates hold only in a conditional way For the direct estimates used in the S L201 1 report the conditions include a that the weights employed effectively counter any systematic non coverage or non response bias for all population sub groups and b that the sampling distribution has a normal shape with a mean equal to the true population value For the modelled estimates the conditions include a that the small area population data largely drawn from the 2001 Census are accurate b that the survey data allow unbiased estimates of model parameters to be extracted c that the population and survey da
9. 011 2011 Skills for Life Survey Heaaline findings Department for Business Innovation and Skills Research Paper Number 57 available online at http www bis gov uk assets biscore further education skills docs 0 9 1 1 1367 201 1 skills for life survey findings pdf accessed on 18 01 12 Harding C et al forthcoming 2011 Skills for Life Survey a survey of Literacy Numeracy and ICT Levels in England Department for Business Innovation and Skills Research Paper Martyn Plummer 2011 rjags Bayesian graphical models using MCMC R package version 3 3 Available at http CRAN R project org package rjags accessed on 12 2 2012 McLennan D et al 2011 The English Indices of Deprivation 2010 Department for Communities and Local Government Available online at http www communities gov uk publications corporate statistics indices2010 accessed 12 1 2012 Williams J S Clemens S Oleinikova and K Tarvin 2003 The Skills for Life Survey a National Needs and Impact Survey of Literacy Numeracy and ICT skills Department for Education and Skills Research Report 490 available online at https Awww education gov uk publications eOrderingDownload RR490 pdf accessed on 27 10 11 23 2011 Skills for Life Survey Small Area Estimation User Guide Annex Definition of Skills Levels Entry Level 1 is the national school curriculum equivalent for attainment at age 5 7 Adults below Entry Level 1 may not be able to write s
10. 15 5000 If you require this publication in an alternative format email enquiries bis gsi gov uk or call 020 7215 5000 URN 12 1316
11. CT assessment consisted of four components word processing emailing spreadsheets and a set of multiple choice questions assessing ICT skills such as internet use Respondents allocated to the ICT assessment were assigned to one of five skills levels mentioned above for each of the four components a total aggregated ICT level is not calculated 2011 Skills for Life Survey Small Area Estimation User Guide Level 1 is broadly comparable with a low level GCSE D G while Level 2 is comparable with a high level GCSE A C Entry Level is therefore below a low level GCSE The Annex to this report offers a brief definition of these levels Fieldwork The survey population was all adults aged 16 65 normally resident in England Residents of institutions were excluded for practical reasons The interview was carried out via Computer Assisted Personal Interviewing CAPI and on average lasted 70 minutes Fieldwork was carried out between May 2010 and February 2011 In total 7 230 interviews were conducted with 57 per cent of assumed eligible households taking part in the survey Of the 7 230 completed interviews 6 049 respondents were assigned to the literacy assessment 6 053 were assigned to the numeracy assessment and 2 358 to the ICT Data weighting and imputation The weighting strategy used was matched to that used for the SfL2003 survey A sampling weight was calculated for each respondent and this sampling weight was then used as the
12. D This provides a count of residential addresses in each postcode and lists within which higher geographies MSOAs Local Authorities Parliamentary Constituencies etc these postcodes lie On this basis it is possible to derive address weighted lookup tables relating MSOAs to each of the other geographies of interest These lookup tables were used to allocate the 2009 MSOA level posterior estimates as appropriate and mean estimates and 95 Cls for the new geographies were derived directly from the re distributed and re aggregated sets of posterior estimates In this way full information is retained concerning model uncertainty Whilst a perfectly adequate approach with respect to Local Authorities Parliamentary Constituencies and Local Enterprise Partnership areas all of which are essentially aggregations of MSOAs the degree of spatial overlap with the various ward geographies raises concerns Geographical variations in household size and composition means this use of addresses will provide only an imperfect guide to the distribution of adult residential populations between different spatial units The various ward estimates must therefore be considered secondary to those produced for MSOAs Summary of results As illustrated by Figure 1 below MSOA level results exhibit impressive granularity and serve to highlight and quantify the extent to which adult skills vary at the very local level For instance whilst Entry Level and below lit
13. ch has thus been adopted 10 2011 Skills for Life Survey Small Area Estimation User Guide The idea here is that survey data which includes information on the socio demographic characteristics of individuals are interrogated in order to derive models which best describe how a dependent variable for instance numeracy skills responds to individual and area level predictor variables e g age sex educational qualifications and at an area level the proportion of people with low incomes in the MSOA in which the survey respondent lives Local area estimates for the dependent variable are then calculated by applying the model s parameter estimates to the corresponding covariate values for the local areas In effect the goal is to pool evidence from across the wider sample in order to generate local estimates Modelling Local Area Estimates Multilevel models have been specified in order to capture the hierarchical structure within which processes are presumed to operate whereby for instance it is assumed that individuals literacy skills can best be predicted with reference to both their individual social demographic characteristics and aspects of the group of which they are a part The dependent variables modelled were a whether or not individuals speak English as a first language for which a multilevel binary logistic regression model was specified and b a series of skills levels for literacy numeracy and the four ICT skill
14. cteristics and many of the socio demographic questions asked in the SfL2011 mirror precisely those asked in the 2001 Census Questions regarding benefit status meanwhile can be related to Lower Layer Super Output Area LSOA level data on the number of people claiming benefits in the 4 Quarter of 2010 as recorded in the DWP s Work and Pensions Longitudinal Study WPLS dataset available via NOMIS http www nomisweb co uk accessed 27 7 12 Parameter selection including a systematic search for possible interaction effects was undertaken using standard automatic and manual selection procedures yielding a parsimonious This approach is described in more detail in the Technical Report which also includes a glossary defining terms such as posterior distribution 11 2011 Skills for Life Survey Small Area Estimation User Guide set of candidate predictor variables These were fitted using approximate Bayesian methods to ensure that each model was suitable and simultaneously to consider a variety of potential Middle Layer Super Output Area MSOA level predictor variables On this basis low income i e the proportion individuals in each MSOA deemed to be income deprived according to the authors of the 2010 Index of Multiple Deprivation was identified as being a sufficient proxy for underlying differences between MSOA populations Having obtained a parsimonious set of candidate variables for each model as listed in Table 4 b
15. domains for which multilevel ordinal logistic regression models were specified The latter were constructed as cumulative link models relating an assumed continuous latent variable to specific skills levels using modelled cut points between skill Levels As detailed in the Technical Report a Bayesian approach has been adopted Arguably more straightforward and flexible than classical or frequentist approaches this provides comprehensive information about the estimates generated In other words by modelling the full posterior predictive distribution of estimates in effect a large number of independent estimates of for instance the number of adults with Level 1 literacy skills it is possible to derive empirically both a point estimate of the number of adults with Level 1 literacy skills the mean of the posterior distribution and a 95 per cent credible interval around that point estimate the range within which 95 per cent of those posterior estimates lie This defines the range within which we are 95 per cent certain the true value lies Variable Selection and Model Fitting A key constraint in small area estimation is that individual level variables must be identically defined in both the survey from which the model was derived and the census or other source from which area level covariate data are drawn Fortunately it is now standard practice for most surveys and censuses to elicit a well defined set of socio demographic chara
16. e e 29 0 million 16 65 year olds 85 per cent were at Level 1 or above in literacy with 5 1 million 15 per cent at Entry Level 3 or below e 26 0 million 16 65 year olds 76 per cent were at Entry Level 3 or above in numeracy with 8 1 million 24 per cent at Entry Level 2 or below e ICT overall distributions are shown in Table 3 Table3 ICT Levels in England Word Processing Email Spreadsheet Multiple Choice n 2 253 n 2 247 n22 228 n 2 274 Adult Adult Adult Adult Population Population Population Population Estimate Estimate Estimate Estimate millions millions millions millions Below Entry Level 1 15 5 0 8 2 6 30 10 4 Entry Level 1 12 4 0 39 13 2 0 2 Entry Level 2 17 5 8 1 0 3 1 0 4 Entry Level 3 16 5 6 9 2 9 27 9 3 12 4 2 Level 1 15 5 2 8 2 6 17 5 8 26 8 8 Level 2 or above 25 8 6 52 17 9 17 5 8 53 17 9 Unweighted Total 34 1 34 1 34 1 34 1 Base SfL2011 All aged 16 65 with word processing scores emailing scores spreadsheet scores multiple choice scores Estimates presented at national and regional level are direct survey estimates Local Area Estimates The SfL201 1 is however clearly unable to support direct estimation at the sub regional level It is not just that the sample size is too small but that many of the wards and MSOAs for which estimates are required are not even represented in the national survey A model based approa
17. el files reporting all skill level estimates EL1 amp L2 amp Literacy a counts amp b proportion below El EL3 L1 Sone 1 EL1 amp L2 amp Numeracy a counts amp b proportion below EL2 EL3 L1 above EL1 amp L2 amp Email a counts amp b proportion below EL2 EL3 L1 dibus Word Processing a counts amp l Below EL1 EL2 EL3 L1 L2 amp b proportion EL above EL2 amp L2 amp Spreadsheets a counts amp b proportion below EL3 L1 Shows ICT Multiple Choice a counts amp Below EL1 EL2 EL3 L1 L2 amp b proportion EL above English not a first language a counts amp Not ESOL ESOL English Spoken as an b proportion ESOL other language Each counts table also gives the reference population for each local area which is the number of people aged 16 64 living in households In addition to the above seven geographies regional estimates have also been provided however unlike the other geographies these estimates are based on direct survey estimates and not modelled estimates The estimates are made available through a total of 17 Excel files The file naming convention describes a the geography b the population base and c whether the file refers to all reported skill levels suffix _A or just Entry Level and below and Level 1 and above suffix Gibson A and Hewson P forthcoming 201 1 Skills for Life Sur
18. elow the public domain and widely used rjags software a program for Bayesian analysis of complex statistical models using Markov chain Monte Carlo McMC techniques was used to obtain the posterior distributions of all parameters McLennan D et al 2011 The English Indices of Deprivation 2010 Department for Communities and Local Government Available at http www communities gov uk publications corporate statistics indices201 0 Accessed 12 1 2012 Martyn Plummer 2011 rags Bayesian graphical models using MCMC R package version 3 3 Available at http CRAN R project org package rjags Accessed on 12 2 2012 12 2011 Skills for Life Survey Small Area Estimation User Guide Table4 Individual level variables used in each model ICT English not Word Spread Multiple as first Literacy Numeracy Email Processing sheets Choice LEE Tee VET 3 Variable Factors Model Model Model Model Model Model Model Sex 2 factors Vv Vv Vv Vv Vv V Ageband 4 V V V Vv Vv Ethnicity 6 Vv Vv Vv Vv Vv Vv Birth Place 2 Vv V Highest Qualification 6 Vv Vv Vv Vv V Limiting Long term Illness 2 Vv Vv Vv Vv Tenure 3 Vv Vv Vv Vv Vv Vv V Occupation SOC2000 10 V V V V Vv Vv Vv Economic Activity Status 4 Vv Vv Vv Vv Vv V Benefit Status IB or SDA 2 Vv Vv Vv Vv Vv V Vv Note These models are predictive rather than interpretative Thus whilst highest qual
19. eople England Wales and Northern Ireland UV71 Communal establishment residents Generating local area predictions MSOAs are the de facto standard geography for which statistical data are published Up to date covariate data thus tend to be available for MSOAs and as a result this provided the obvious upper level for the multilevel models MSOA populations were thus microsimulated on the basis of 2001 Census and DWP WPLS data and the posterior model parameter estimates were applied to derive 1 000 independent estimates of the likelihood that each microsimulated individual would have the various literacy numeracy and ICT skill levels Aggregating these likelihood estimates to MSOA level provides a direct estimate of the number and proportion of adults in each of the skill levels or who do do not speak English as a first language These estimates were then weighted to align with the ONS s 2009 age sex population estimates This accounts for known demographic changes although the estimates still assume socio economic characteristics of MSOAs are as reported in the 2001 Census Estimates were also required for a number of other geographies for which little or no relevant covariate data are available For these geographies it has been necessary to attribute MSOA 15 2011 Skills for Life Survey Small Area Estimation User Guide level estimates on the basis of the February 2011 Open National Statistics Postcode Directory ONSP
20. eracy rates vary from 11 3 Wokingham to 21 6 Knowsley at local authority level at MSOA level the proportion ranges from 8 8 Basingstoke amp Deane 021 to 33 6 Liverpool 022 As described in the Technical Report this pattern is repeated across the numeracy and ICT skills domains and it is clear that areas of adult skill deficits can be very localised It is also apparent that there are marked variations in skill levels within individual local authority areas This is pronounced for many large urban authorities for instance Liverpool with 20 8 of adults with Entry Level or below literacy skills includes individual MSOAs with rates varying from 11 6 to 33 6 but rural areas are not exempt Thus North East Lincolnshire with an overall rate of 17 3 contains individual MSOAs with rates varying from 13 7 to 28 3 This suggests that any policy response to skills deficits will also need to be spatially fine grained Figure 1 also draws attention to the fact that adult skills deficits are not a purely urban phenomenon although the densest concentrations of poor literacy numeracy and ICT skills do appear to affect England s principal conurbations As detailed in the Technical Report this rural dimension appears particularly marked with respect to ICT skills with relatively high rates of Entry Level 1 and below skill levels emerging in many south western and northern counties as well as around the Wash Turning to the distr
21. es for the best buy or work out a household budget Adults with ICT Level 2 skills are able to use a variety of appropriate sources of information and evaluate its fitness for purpose as well as evaluate and use different methods of organising and presenting information taking into account fitness for purpose and audience Level definitions adapted from National Audit Office 2008 Skills for Life progress in improving adult literacy and numeracy available online at http www nao org uk publications 0708 skills for life progress in i aspx accessed on 17 8 11 p 10 ICT skills levels adapted from the Qualifications and Curriculum Authority January 2007 CT Skills for Life curriculum Department for Education and Skills available online at http archive niace org uk Research ICT ICT Skill for Life curriculum Jan07 pdf accessed on 17 8 11 p 19 24 Crown copyright 2012 You may re use this information not including logos free of charge in any format or medium under the terms of the Open Government Licence Visit www nationalarchives gov uk doc open government licence write to the Information Policy Team The National Archives Kew London TW9 4DU or email si nationalarchives gsi gov uk This publication is also available on our website at www bis gov uk Any enquiries regarding this publication should be sent to Department for Business Innovation and Skills 1 Victoria Street London SW1H OET Tel 020 72
22. he de facto standard geography for which most ONS and other administrative and statistical data are published MSOAs have a Government Statistical Service GSS code starting E02 2001 population base Estimates have been produced by applying the model to population covariates derived from the 2001 Census and 4 Quarter 2010 DWP benefit data weighted to match the 2001 Census population These figures have not been adjusted to account for any population change since 2001 2009 population base Estimates have been produced by applying the model to population covariates derived from the 2001 Census and 4 Quarter 2010 DWP benefit data weighted to fit ONS mid year age sex population estimates for 2009 These figures have thus been adjusted to account for known demographic population changes since 2001 although it is necessary to assume that MSOA populations have remained static in terms of their socio economic composition Estimates for all other geographies assume a 2009 population base and are derived from the 2009 MSOA estimates being proportionally attributed on the basis of addresses in the February 2011 Open National Statistics Postcode Directory ONSPD Standard Table ST Wards n 7 932 ST wards were used to release 2001 Census outputs with a higher disclosure risk than was acceptable using Census Area Statistics CAS wards ST wards were based on the 2003 statistical ward set though a total of 113 wards those with fewer 6 2011
23. hort messages to family or select floor numbers in lifts Adults with ICT Entry Level 1 skills are able to get information from an ICT based source and follow recommended safe practices Entry Level 2 is the national school curriculum equivalent for attainment at age 7 9 Adults with below Entry Level 2 may not be able to describe a child s symptoms to a doctor or use a cash point to withdraw cash Adults with ICT Entry Level 2 skills are able to use ICT to communicate as well as enter and edit small amounts of information in ways that are fit for purpose and audience Entry Level 3 is the national school curriculum equivalent for attainment at age 9 11 Adults with skills below Entry Level 3 may not be able to understand price labels on pre packaged food or pay household bills Adults with ICT Entry Level 3 skills are able to interact with and use an ICT system to meet needs as well as present information in ways that are fit for purpose and audience Level 1 is equivalent to GCSE grades D G Adults with skills below Level 1 may not be able to read bus or train timetables or check the pay and deductions on a wage slip Adults with ICT Level 1 skills are able to select and use a variety of appropriate sources of information as well as enter organise develop format and bring together information to suit content and purpose Level 2 is equivalent to GCSE grades A C Adults with skills below Level 2 may not be able to compare products and servic
24. ibution of adults who do not speak English as a first language Figure 2 the pattern is much as might be expected with high rates across most of London as well as in the centres of many of England s other major towns and cities notably the old cotton towns of Lancashire The glaring anomaly on the map concerns an improbably high proportion of adults not speaking English as a first language in a small part of rural East Anglia Choropleth maps illustrating the proportion of adults with low literacy numeracy and ICT skills at both MSOA and local authority level are included in the Technical Report A local authority level map of the proportion of adults not speaking English as a first language is also included in the Technical Report 16 2011 Skills for Life Survey Small Area Estimation User Guide Thus the model predicts that over 3096 of adults in the Forest Heath 002 MSOA E02006239 do not speak English as a first language and that the rate exceeds 1096 in another four Forest Heath MSOAs In fact Forest Heath is distinctive because of the very high proportion of non UK born household residents recorded in the 2001 Census 50 3 Elsewhere in the country such high percentages are invariably associated with immigrant populations but here it is due to the presence of two large American airbases at Lakenheath and Mildenhall In the SfL201 1 as in the country as a whole being non UK born is strongly associated with not speaking Engli
25. ifferently however and the upper and lower 95 Cls cannot be summed in the same way If 9596Cls are required for combined skill levels and or aggregated geographic units then they will have to be approximated as described in Section 6 of the Technical Report As noted in Section 0 aggregation of MSOA level literacy numeracy and ICT estimates to the regional or national level will produce alternatives to the direct estimates from the survey itself The Department has chosen to use direct national and regional estimates and modelled estimates for all sub regional geographies Finally the estimates are necessarily constrained by the data collected as part of the SfL201 1 survey and the extent to which they could be reconciled with local covariate data Thus a broader range of appropriate predictor variables would at least in theory have enabled us to develop better models and thus produce estimates that were more sensitive to the specific characteristics of some local populations 22 2011 Skills for Life Survey Small Area Estimation User Guide References Gibson A and Hewson P forthcoming 2011 Skills for Life Survey Small Area Estimation Technical Heport Department for Business Innovation and Skills Research Paper Gibson A Bailey T and Fraser D 2004 Demographic mapping of the 2003 Skills for Life Survey to local areas Technical Report for the Department for Education and Skills December 2004 Harding C et al 2
26. ification is highly correlated with literacy it does not form part of the final model because an alternative set of predictor variables offer a better fit with the data 13 2011 Skills for Life Survey Small Area Estimation User Guide Having derived the posterior distributions comprising 1 000 simulated values for each model parameter the next stage is to apply them to individuals and to then aggregate and summarize the resulting responses at MSOA level This is more challenging then the approach used to generate local area estimates from the SfL2003 where model parameter posterior distributions were applied to aggregated area level data because it requires information on the composition of local populations The idea though is that this new approach should significantly improve estimate accuracy as valuable information about local areas will not have been lost through aggregation Determining the composition of local populations Unfortunately direct information concerning the detailed composition of local populations does not exist It is however possible to use what is known about the aggregate characteristics of any given population i e how many males and females how many people with or without a limiting long term illness how many people in each age band etc in order to microsimulate the likely number of people with each unique combination of characteristics for instance the number of 16 34 year old males with a
27. itional LEPs will be formed over the coming years and basic skills estimates for these as well as for the Northamptonshire and Buckinghamshire Thames valley LEPs can be approximated using the method described in Section 6 of the Technical Report Gibson A Bailey T and Fraser D 2004 Demographic mapping of the 2003 Skills for Life Survey to local areas Technical Report for the Department for Education and Skills December 2004 7 2011 Skills for Life Survey Small Area Estimation User Guide Survey Background Survey introduction The first Skills for Life Survey was commissioned by the Department for Education and Skills DfES in 2001 The survey aimed to produce a national profile of adult literacy numeracy and Information and Communication Technology ICT skills and to assess the impact different skills had on people s lives In 2010 the Department for Business Innovation and Skills BIS commissioned a follow up to the initial survey with the key purpose of updating the baseline information collected about adult literacy and numeracy as well as setting a new functional baseline for ICT skills A key consideration in designing the new 2011 survey was maintaining comparability with the baseline 2003 survey to enable measurement and analysis of changes in skills levels amongst the population between the respective periods A further objective of the 2011 Skills for Life Survey SfL201 1 was to understand the demographic
28. limiting long term illnesses the number of 16 34 year old females with a limiting long term illnesses and so on The defining characteristic of a successfully microsimulated population is that when aggregated it will match in all respects what is known about the overall characteristics of that population The data used to generate microsimulated populations for all 6 781 MSOAs in England have been drawn from a series of univariate and multivariate 2001 Census tables listed in Table 5 below as well as from the WPLS dataset regarding 4th Quarter 2010 claimants of Incapacity Benefit and Severe Disablement Allowance As detailed in the Technical Report the goal is to iteratively assign people to cells in the simulated populations until the resulting marginal totals match the known marginal totals derived from the 2001 Census and DWP WPLS dataset Disclosure control measures used to ensure confidentiality of both census and DWP data mean that microsimulation cannot precisely reconstitute the socio demographic composition of individual MSOA populations but there is no question that it provides a close approximation More problematic is the fact that the 2001 Census was undertaken the best part of a decade before the SfL2011 Demographic changes since the 2001 Census have been accounted for by adjusting the estimates so that they align with the most recent 2009 ONS mid year age sex population estimates available at the time of analysis but it is
29. nitas tuer aan ta datd bandas ortu trn M ruit n diui 2 daca b E RERM P xata LERRA E RARE 11 Variable Selection and Model Fitting sesssisces ierit rr ln ta vas eli gas eter o Rede acusa re eei ir Prin aS Eire eR reae 11 Determining the composition of local populations sssssssem Ree 14 Generating local area predictiOriS o ie pont Roster Mte n ERIS Mtas RARR UN Lx NESSUN MM EI EEEE ERU RM E MA RU AEU MEE 15 Su umMary Of TeESUltS RO S STO 16 Relationship between direct survey estimates and modelled small area estimates 20 CI o 21 hiurlpljcr e MM 23 Annex Definition of Skills Levels nent riti ranae ee sanus nenas una nana EE tux naa REIR R Ri RR ERAS 24 2011 Skills for Life Survey Small Area Estimation User Guide Report Authors and Acknowledgements The 2011 Skills for Life Survey was conducted by a partnership of two complementary agencies TNS BMRB a research agency and AlphaPlus Consultancy Ltd TNS BMRB was responsible for all data collection and primary data processing whilst AlphaPlus provided advice on Skills for Life policy and related issues throughout the survey Both agencies were responsible for the main survey analysis and reporting The Small Area Estimation project was undertaken by Alex Gibson Principal Researcher and Director of
30. ry Constituencies England Amendment Order 2008 They came into effect at the May 2010 General Election Their composition is detailed in the February 2011 ONSPD and no further changes are envisaged until 2014 2015 Parliamentary constituencies have a GSS code starting E14 Local Authorities n 326 The local authorities to which this study refers are those district level or lower tier authorities current as of the 1 January 2011 thereby including the structural changes effected on 1 April 2009 There are a total of 326 local authorities comprising 36 metropolitan districts with a GSS code starting E08 201 non metropolitan districts E07 56 unitary authorities E06 32 London boroughs and the Corporation of the City of London E09 Their composition is detailed in the February 2011 ONSPD Local Enterprise Partnership LEP areas n 37 Each LEP comprises a number of local authorities and some Local Authorities are in more than one LEP The composition of LEPs in terms of their constituent Local Authorities is detailed in www bis gov uk policies economic development leps statistics accessed 22 04 2012 When the estimates in this report were produced there were 37 LEPs Shortly thereafter in late September 2011 a 38 LEP Northamptonshire was announced A further LEP has been announced more recently Buckinghamshire Thames Valley These two LEPs have not been included in the analysis It is likely that add
31. sh as a first language In Forest Heath this relationship will almost certainly not apply but without sampling a number of individuals from that particular MSOA it is impossible for the model to recognise and reflect Forest Health s uniqueness In this respect the methodological strength of small area estimation the fact that it pools the evidence of modelled relationships derived from an analysis of the dataset as a whole is also its principal weakness It cannot isolate and reflect genuine local exceptions such as Forest Heath This is inevitable when a survey of 7 230 individuals is being used to derive estimates across a total of 6 781 MSOAs 17 2011 Skills for Life Survey Small Area Estimation User Guide Figure 1 Entry Level and Below Literacy Skills English MSOAs nz6 781 Percent of 16 64 yr olds at Entry Level or below 2096 and above 1896 2096 1696 1896 1496 1696 Below1 496 18 2011 Skills for Life Survey Small Area Estimation User Guide Figure 2 Proportion of adults not speaking English as a first language English MSOAs nz6 781 Percent of 16 64 yr olds not speaking English as a first language 996 and above HE ENME s a 3 5 2 3 Below 2 19 2011 Skills for Life Survey Small Area Estimation User Guide Relationship between direct survey estimates and modelled small area estimates The SfL201 1 dataset has been used to generate estimates of
32. should be used to decide whether or not an area is likely to be atypical e Given the above it is important to recognise that small area estimation is effective at identifying and highlighting those areas where adult literacy numeracy and ICT skill levels are likely to be relatively poor and where therefore it is ikely that policy initiatives would be best directed The results cannot and should not be used as the basis for some sort of performance league table ranking individual MSOAs LAs or other geographic or policy units on the basis of say predicted adult literacy rates e As noted in Section 0 small area estimation was undertaken with respect to MSOA populations These estimates were then attributed to other geographies on the basis of addresses listed in the February 2011 Open National Statistics Postcode Directory For this reason as well as the fact that MSOAs are today the de facto standard geography for which very local spatial administrative and statistical data are published the various ward level estimates must be considered secondary to those produced for MSOAs e As noted in Section 0 estimates can be summed across skill categories and or across a number of individual geographic units e g to combine a series of LA level estimates to produce a new estimate for a composite Local Enterprise partnership area Credible 21 2011 Skills for Life Survey Small Area Estimation User Guide Intervals must be treated d
33. ta are comparable and d that the processes used to microsimulate the composition of local populations and to simulate parameter posterior distributions are appropriate to the task and correctly implemented Given the variety of conditions placed on both sets of estimates some conflicts are to be expected even if as here the majority of estimates are close matches for one another Ultimately it is impossible to answer the question which is the more accurate without full and current population data The choice between estimation methods depends on the degree of belief about the assumptions that underpin each method The Department has chosen to use the modelled estimates only for sub regional geographies for which direct estimates are unreliable due to sample size limitations however direct survey results are used for national and regional level estimates There are two reasons for this First the direct estimates are simpler to present because they are based on traditional survey methods Second because the modelled estimates have been computed at MSOA level and only then aggregated to regions Any small systematic biases at MSOA level may be magnified at higher order geographies 20 2011 Skills for Life Survey Small Area Estimation User Guide Caveats When reviewing and interpreting small area estimation data it is essential that the following caveats are borne in mind e As noted in Section 0 all estimates necessarily
34. vey Small Area Estimation Technical Report Department for Business Innovation and Skills Research Paper 5 2011 Skills for Life Survey Small Area Estimation User Guide B All estimates assume the socio economic characteristics of local populations to be as detailed in the 2001 Census although all except a single set of MSOA level estimates have been updated to reflect the ONS s mid year 2009 age sex population estimates Table2 Small Area Estimation Data Excel Files and their contents Filenames all have xIsx suffixes Population All Skill Level EL amp below and L1 amp Geography Base Estimates above categories Middle Layer Super Output Areas 2001 MSOA 2001 A Middle Layer Super Output Areas 2009 MSOA 2009 A MSOA 2009 B Standard Table wards 2009 STWards 2009 A STWards 2009 B 2005 Statistical wards 2009 StatWards 2009 A StatWards 2009 B 2011 Council wards 2009 CouncilWards 2009 A CouncilWards 2009 B 2011 Parliamentary Constituencies 2009 ParlCons 2009 A ParlCons 2009 B Local Authorities 2009 LAs 2009 A LAs 2009 B Local Enterprise Partnership areas 2009 LEPs 2009 A LEPs 2009 B Middle Layer Super Output Areas n 6 781 MSOAs were devised by the ONS as part of a hierarchy of units specifically designed for the collection and publication of small area statistics They are of broadly consistent size of about 7 200 people and are not subject to boundary changes They are t
Download Pdf Manuals
Related Search