An Introduction to the Analysis of WERS 2004

1
An Introduction to the Analysis of WERS 2004

• John Forth Lucy Stokes
• WERS 2004 Information Advice Service

2
Aims

• Introduce the publicly-available data files
• Content
• Availability and access procedures
• Key features
• Analytical issues
• First steps
• Weighting and statistical inference
• Linking data files
• Where to get help and advice

3
Assumptions

• Know something about nature and content of WERS
  2004
• Yet to obtain or use the data, or
• At the very early stages in your analysis

4
Existing research using WERS

• Primary analysis
• 40-page booklet of First Findings (July 05)
  http//www.dti.gov.uk/employment/research- evalua
  tion/wers-2004/index.html
• 400-page sourcebook (July 06) www.routledge.com/
  textbooks/0415378133
• 120-page report on SMEs (July 06)
  http//www.dti.gov.uk/employment/research- evalua
  tion/wers-2004/index.html
• Compendium of regional tabulations (Oct 06)

5
Existing research using WERS

• Secondary analysis
• Bibliography of research using WERS 1980-1998
• Searchable on-line database of research using
  WERS 2004
• Both available at
• http//www.wers2004.info/research/search.php

6
Overview of WERS 2004 data
7
Cross-Section Managers

• Workforce composition
• Management of personnel and employment relations
• Recruitment and training
• Workplace flexibility and the organisation of
  work
• Consultation and information
• Employee representation
• Payment systems and pay determination
• Grievance, disciplinary and dispute procedures
• Equal opportunities, work-life balance
• Workplace performance

8
Cross-Section Employee Reps

• Structure of representation at the workplace
• Time spent on representative duties
• Means of communication with employees
• Incidence of negotiation and consultation over
  pay and other matters
• Involvement in redundancies, discipline and
  grievance matters
• Incidence of collective disputes and industrial
  action
• Relations with managers
• Union recruitment

9
Cross-Section Employees

• Working hours
• Job influence
• Job satisfaction
• Working arrangements
• Training and skills
• Information and consultation
• Employee representation
• Pay

10
Cross-Section FPQ / ABI

• Turnover
• Employment costs
• Purchases
• Capital stocks
• Capital expenditure (acquisitions and disposals)
• RD activity

11
Longitudinal analysis

• Each XS independent samples no overlap between
  surveys
• Time-series with 5 data points
• Changes to questionnaires over time
• 1998 a major break point
• Expansion of population in 1998 and 2004 to
  include smaller workplaces

12
Longitudinal analysis

• Two-wave Panel Surveys provide longitudinal data
  on individual workplaces
• 1984-1990 (trading sector only)
• 1990-1998 (all workplaces with 25)
• 1998-2004 (all workplaces with 10)
• Survival status of original x-section
• Changes in practice in continuing workplaces
  (headline practices only)

13
Overview of WERS 2004 data
938 (76)
166 matched for 1998 and 2004 (18)
14
Data availability

• General release data
• Restricted until April 2007
• Region identifiers
• Detailed industry (below SIC(2003) Section level)
• Financial Performance Questionnaire
• Permanently restricted
• Annual Business Inquiry
• Not available
• Names addresses of respondents / workplaces and
  the wider organisation

15
Obtaining general release data

• Where
• UK Data Archive (http//www.data-archive.ac.uk)
• Study Number 5295
• What
• Data and core documentation (questionnaires,
  codebooks, technical report, introductory note)
• How
• Athens ID required
• Download or CD
• Local use

16
Obtaining financial data

• Where
• ONS Virtual Micro-data Lab
• (http//www.statistics.gov.uk/about/bdl/)
• London, Newport, Titchfield, Southport
• What
• General release data FPQ ABI
• Core documentation (limited for ABI)
• How
• Application to ONS Micro-data Release Panel
• Site access only
• Withdrawal limited to non-disclosive results

17
Timed release of restricted data

• In April 2007
• FPQ to the UK Data Archive
• Region codes and detailed industry codes to UKDA
  and ONS

18
Analysis first steps

• Read core documentation
• Survey questionnaires
• Technical report
• Introductory note
• Check latest WIAS guidance
• http//www.wers2004.info
• Variable notes
• Derived variables
• Errata in primary analysis
• FAQs

19
Key features of the data files

• Layout
• Variable naming convention
• SqnameN (e.g. ASTATUS1)
• S Section letter
• Qname descriptive name
• N numbered response

20
Key features of the data files

• Multiple-response sets (e.g. CFACTORS)
• CFACTORS
• Which of the following factors are important when
  recruiting new employees?
• PROBE Which others? UNTIL 'None'
• 1) References,
• 2) Availability,
• 3) Recommended by another employee,
• 4) Skills
• 5) Age
• 6) Qualifications,
• 7) Experience,
• 8) Motivation,
• 9) Other (please specify CFACTOTH)

21
Key features of the data files

• Multiple-response sets
• CFACTOR1 1st response given
• CFACTOR2 2nd response given, etc
• XCFACT1-3 codes for verbatim responses using
  other, please specify code
• Convert to dummies using ANY command (SPSS) or
  EGEN command with EQANY option (Stata)

22
Key features of the data files

• T variables (e.g. FMEASPR, FMEASPRT)
• FMEASPR
• What proportion of non-managerial employees at
  this workplace have their performance formally
  appraised?
• INTERVIEWER If respondent gives answer as an
  exact number you can code 97 here and record
  the number of the next question
• 1) All (100),
• 2) Almost all (80-99),
• 3) Most (60-79),
• 4) Around half (40-59),
• 5) Some (20-39),
• 6) Just a few (1-19),
• 97) Number
• If giving exact number
• FMEASPRT
• How many non-managerial employees here have their
  performance formally appraised?
• ENTER NUMBER

23
Key features of the data files

• In this case, code FMEASPRT into FMEASPR using
  total non-managerial employees (ZALLEMPS
  ZMNG_TOT)
• Syntax available for all T variables at
• http//www.wers2004.info/FAQ.phpsyntax

24
Key features of the data files

• Missing values
• -9 Not answered / refused
• -8 Dont know
• -1 Not applicable
• Treatment in data files
• SPSS Assigned as user-missing values
• Stata Not assigned as missing values (valid
  values)

25
Producing reliable estimates

• Sample bias ? Weights
• Less precision than SRS ? Survey-adjusted
  variance estimation

26
Importance of weighting

• Sample of workplaces not SRS
• Unequal probabilities of selection by workplace
  size and industry (IDBR)
• Large workplaces and small industries
  over-represented vs population
• Also variations in response rates by size and
  industry (at least)
• Weight 1 / p(selection and response)
• Weighted estimates free of known biases (i.e.
  representative of wider population)

27
Correctly estimating variances

• Textbook formulae assume SRSWR
• WERS not sampled according to SRSWR
• Unequal p(selection) clustering of employee
  sample ? larger standard errors than SRSWR
  (50-60 larger, on average)
• Textbook formulae ? Type I or II errors
• Linearization or replication methods ? SEs that
  account for the survey design

28
Software options

• Stata version 5 onwards
• svy suite of commands (included)
• svyset informs Stata about the sample design
• svy prefix can be used with wide range of
  statistical procedures
• iweights will remove bias but incorrectly
  estimate variances (SEs)
• Syntax examples at http//www.wers2004.info/FAQ.p
  hpstata

29
Software options

• SPSS version 12 onwards
• Complex Samples module (add-on)
• CSPLAN ANALYSIS informs SPSS about the sample
  design
• Limited range of CS procedures then available
  (descriptives, x-tabs, logit, ordinal, GLM)
• WEIGHT BY will remove bias but incorrectly
  estimate variances (SEs)
• Syntax examples at http//www.wers2004.info/FAQ.p
  hpspss

30
Linking data files

• Combining data from different questionnaires for
  linked analysis
• Examples
• Using data on payment practices from MQ in
  analysis of employees wages
• Comparing managers and employee representatives
  ratings of climate
• Linking 1998 and 2004 observations in Panel
• Link via unique workplace identifier (SERNO)

31
Linking data files (cross-section)

• One-to-one match FPQ ? MQ

32
Linking data files (cross-section)

• One-to-many match SEQ, ERQ ? MQ

33
Software options

• SPSS
• MATCH FILES FILEmaster file
• /TABLEsecondary file
• /BY serno
• Stata
• get filemaster file
• merge serno using secondary file
• drop _merge2

34
Linking data files (cross-section)

• Many-to-one match MQ? SEQ, ERQ

35
Software options

• SPSS
• AGGREGATE then MATCH FILES
• Stata
• collapse then merge
• Issue
• summary data item from SEQ may be measured with
  error (sampling error)
• errors in variables regression?

36
Linking data files (panel)

• One-to-one match 1998 ? 2004
• Wide form one record per workplace

37
Linking data files (panel)

• Long form one record per workplace per year
• Syntax for wide and long forms available at
  http//www.wers2004.info/FAQ.phpconstruct

38
Aims

• Introduce the publicly-available data files
• Content
• Availability and access procedures
• Key features
• Analytical issues
• First steps
• Weighting and statistical inference
• Linking data files
• Where to get help and advice

39
Further info and advice

• WERS 2004 Information and Advice Service
• Website http//www.wers2004.info
• Email wers2004_at_niesr.ac.uk
• Telephone [phone]