Methodology
This page summarises the design and administration of the 2011 Workplace Employment Relations Survey (WERS). It provides basic information on the sample design, response rates, sample sizes and weighting methodologies. It also provides information about the statistical precision of the survey results.
The material presented here is an abridged version of 'The Design and Administration of the 2011 Workplace Employment Relations Survey' (deposited with the UK Data Service along with the survey data).
The Design and Administration of the 2011 WERS
(PDF file, via UKDS website)
Overview
WERS 2011 is the sixth survey in a series of workplace surveys. The survey population for the 2011 survey is all workplaces in Britain that have 5 or more employees and are operating in Sections C-S of the Standard Industrial Classification (2007). This population accounts for 35% of all workplaces and 90% of all employees in Britain. A workplace is defined as comprising ‘the activities of a single employer at a single set of premises’. A branch of a high street bank, a factory and the head office of a local council are all considered to be workplaces in their own right.
The 2011 WERS has the following components:
• Survey of Managers - comprising the Employee Profile Questionnaire (EPQ) and the Management Questionnaire (MQ)
• Survey of Worker Representatives (WRQ)
• Survey of Employees (SEQ)
• Financial Performance Questionnaire (for workplaces in the trading sectors only) (FPQ)
At each participating workplace, the study commences with an interview with the most senior manager with responsibility for employment relations, human resources or personnel. The manager is asked to provide a demographic profile of the workforce prior to the interview and, if the workplace is in the private sector or a trading public sector corporation, they are also asked after the interview to provide financial performance information about the workplace. Permission is also sought from the manager to distribute a self-completion questionnaire to a maximum of 25 employees at the workplace. If a union or non-union employee representative is present, an interview with them is sought. If a workplace had both union and non-union representatives, an interview with each is sought. The union interview is conducted with the most senior lay representative of the largest recognised union at the workplace, or the largest non-recognised union if none are recognised. The non-union representative interview is conducted with the most senior non-union employee sitting on a joint workplace consultative committee. Where there is no such committee, a “stand-alone” non-union employee representative is interviewed.
Workplace managers and employee representatives are asked to act primarily as informants about their workplace, and so the vast majority of the data collected in those interviews relates to the features of the sampled workplace rather than to the particular characteristics of the individual respondent.
A major feature of the design of previous surveys in the WERS series (including the 2004 survey) was the use of two entirely separate samples. The first was a fresh cross-section sample of workplaces that were surveyed to provide representative results for the population of workplaces in existence at the time of the survey. The second was a panel sample which consisted of all participating workplaces from the previous cross-section (for example, the 1998 cross-section sample in the case of the 2004 study) who remained in existence at the time of the new WERS. That panel sample was used primarily to assess the extent to which individual workplaces changed their behaviour over time. The full set of questionnaires was administered in the fresh sample, whilst panel workplaces only received an abridged version of the management interview. The samples were always analysed separately.
A key design innovation of the 2011 WERS was to integrate the fresh cross-section and the panel samples. In the field, workplaces in the panel sample were eligible for all four components of the 2011 WERS. Users of the survey data therefore have access to management, employee and worker representative data for both waves of the 2004-2011 panel.* Moreover, weights have been devised to enable the panel sample to be combined with the fresh sample to form a combined sample that is cross-sectionally representative of all workplaces in 2011. This more than compensates for the smaller size of the fresh sample in 2011, creating a cross-sectionally representative combined sample that is larger than the cross-section sample that was available in 2004. The remainder of this discussion will frequently refer to the ‘panel sample’, the ‘refreshment sample’ and the ‘combined sample’, these being the core elements of the design of the 2011 survey.
* Users should note, however, that there is no tracking of individual respondents within these workplaces across the two waves.
Sampling
There were two overarching aims for the sample design in 2011:
1. to obtain interviews at 900 of those 2,295 workplaces which participated in the WERS 2004 Cross-Section; and
2. to obtain interviews at a further 1,800 workplaces from a new (and independent) refreshment sample, which could then be combined with the 900 panel workplaces to form a cross-sectionally representative combined sample of 2,700 workplaces.
The issued panel sample was the full set of 2,295 achieved cases from the 2004 WERS Cross-Section, minus nine workplaces. A total of 2,286 addresses were therefore issued to the field. Each workplace was traced in order to establish whether it continued to be in existence in 2011 and, if so, to establish how many employees it had in 2011. Workplaces that continued in existence through to 2011 and which employed five or more employees in 2011 were pursued for interview.
The refreshment sample in 2011 was selected as a stratified random sample from the Inter Departmental Business Register (the IDBR) which is maintained by the Office for National Statistics. A sample of 4,848 local units was selected from the IDBR in July 2010. The sample was drawn from the population of local units with five or more employees, operating in Sections C-S of the Standard Industrial Classification (2007) and located within Great Britain. Local units on the IDBR were first categorised into 153 sampling strata, formed from the cross-tabulation of nine categories of workplace size (5-9 employees; 10-24; 25-49; 50-99; 100-199; 200-499; 500-999; 1000-1999; and 2000+) and 17 industry sectors (Sections C-S of the Standard Industrial Classification (2007)). Sampling fractions were then set in each stratum so that the panel and refreshment samples, when combined at the end of fieldwork, would yield an achieved combined sample that offered at least 250 workplaces in each size band up to and including 200-499 employees; at least 150 workplaces in the size bands 500-999 and above; and a minimum of 85 cases in each industry sector. The sampling fractions were larger among local units with greater numbers of employees and higher than average within small industry sectors, most notably SIC (2007) Sections D and E.
Survey instruments
Copies of the survey questionnaires can be accessed via the Survey Documentation section of this website.
The Employee Profile Questionnaire (EPQ)
The EPQ contains a total of 16 questions which collect key information about the size and structure of the workforce at the workplace. These questions ask about the numbers of staff in specific categories, including occupational groupings, part-time versus full-time work, age groups and pay bands. Some of this information is used to route the management respondent through the MQ (e.g. routing them around questions asked about union membership if they have reported on the EPQ that the workplace has no union members). The data from the EPQ was collected by the interviewer at the beginning of the Management Questionnaire.
The Management Questionnaire (MQ)
The Management Questionnaire is the hub of WERS and – along with Question 1 on the Employee Profile Questionnaire – is the one element that is required in order for a workplace to count as having participated in the survey. The management interview was conducted on-site by a trained interviewer, using Computer Assisted Personal Interview (CAPI) technology, and lasted 90 minutes on average. The questionnaire covered a broad range of topics, as summarized below:
• 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 & experience of the recession
The Worker Representative Questionnaire (WRQ)
Up to two worker representatives were eligible for the Worker Representative Questionnaire (WRQ) at each workplace:
• A representative of a trade union. The representative selected would be the most senior representative of the largest recognised union (in terms of members at the workplace), or where there was no recognised union, the most senior representative of the largest non-recognised union.
• A non-union representative. The representative selected would be the most senior non-union representative sitting on the Joint Consultative Committee that covered the widest range of issues or, if there was no JCC, a stand-alone non-union representative.
The number of worker representative interviews that were sought could therefore be two, one or none, depending upon which structures for employee representation were found to be present at the workplace during Sections D and E of the MQ (specifically at questions: ESTEWARD, EOTHUREP, DISSUES, EUJCC and EOTHREPS). The WRQ covered the following topics:
• The structure of representation at the workplace
• The operation of joint consultative committees
• Negotiation, consultation and information provision
• The role of the employee representative
• Collective disputes and procedures
• Grievance and disciplinary procedures
• The involvement of employee representatives in workplace change
The Financial Performance Questionnaire (FPQ)
The short Financial Performance Questionnaire (FPQ) contained 10 questions which gathered basic financial details about the workplace (or about the performance of the whole organisation where workplace-level data was not available). The FPQ was offered as a paper questionnaire or online and was given to the management respondent at the end of the MQ interview, either for them to complete or for them to pass on to someone else who would be able to provide the necessary detail. All workplaces had been eligible for the FPQ in 2004 but, as the data collected from public sector workplaces was rarely used, the 2011 questionnaire was issued only to workplaces in the trading sector (ASTATUS1<=8).
The Survey of Employees Questionnaire (SEQ)
The Survey of Employees Questionnaire (SEQ) was administered to up to 25 staff at each workplace. The manager was asked at the end of the management interview whether it was permissible to distribute SEQs among employees at the workplace. If they agreed, 25 employees were selected at random by the interviewer from a list of all employees at the workplace, provided by the management respondent. If the workplace had 25 or fewer employees, all were selected to participate. Paper questionnaires were left for distribution to each selected employee. Login details were provided on a label on the front of the personally-addressed paper questionnaire, so that employees could chose to complete the questionnaire online if they wished. Freepost envelopes were provided so that those wishing to complete the paper questionnaire could return it directly to NatCen.
Response rates
Response rates and numbers of productive interviews for each instrument are outlined below. The later section on Weighting includes a discussion of how observable non-response biases were addressed during the development of the survey weights.
The EPQ/MQ:
The response rate for the EPQ/MQ was 46% in the combined sample, yielding a total of 2,680 productive interviews.
Productive interviews were obtained from 989 workplaces in the panel sample (response rate: 52%) and from 1,691 workplaces in the refreshment sample (response rate: 43%). Workplaces in the panel sample were expected to be more predisposed towards responding by virtue of having already participated in WERS 2004.
The WRQ:
Among the 1,153 workplaces that had an eligible union representative, 797 workplaces generated a productive interview, giving a response rate of 69% among union representatives. Among the 415 workplaces with an eligible non-union representative, 205 generated a productive interview, giving a response rate of 49% among non-union representatives. Among the 229 workplaces with eligible union and non-union representatives, interviews were obtained with both types of representatives in 82 workplaces (36%).
The FPQ:
A total of 1,941 workplaces were eligible for the Financial Performance Questionnaire by virtue of the management respondent having categorised the workplace as belonging to the trading sector (ASTATUS1<=8) in the MQ. The management respondent allowed the interviewer to leave an FPQ in 1,713 (88%) of these workplaces. Useable data was provided for 545 of these workplaces, representing an overall response rate of 28%.
The SEQ:
Managers gave permission for interviewers to select a sample for the SEQ in 2,170 workplaces (81%). Interviewers then placed a total of 44,371 SEQs in these workplaces and a total of 21,981 questionnaires were returned, giving a response rate of 50% among all sampled employees. However 247 workplaces returned no questionnaires at all, it is possible that the 3,858 questionnaires which interviewers left for distribution at these workplaces did not ever reach the sampled employees. If one assumes that these questionnaires were not distributed, then the response rate among employees who are assumed to have received a questionnaire is 54%.
Weighting
When drawing a simple random sample, each member of the population has the same probability of selection and so, in the absence of non-response biases, the achieved sample will inevitably resemble the population from which it was drawn. However, as noted above under 'Sampling', the WERS sample design purposefully gives an above-average probability of selection to larger workplaces and those from less populated industries. As a consequence, the profile of the issued sample of workplaces is out of kilter with the population at large, since large workplaces and those from small industries are proportionately over-represented.
Similarly in the employee survey, once an employee’s workplace has been selected to participate in WERS, a member of staff in a small workplace has a higher probability of receiving an SEQ than an employee in a large workplace (since questionnaires were distributed to all employees in workplaces with 5-25 employees and to only 25 employees in larger workplaces). So employees from small workplaces are over-represented in the employee sample when compared with the population for the employee survey (i.e. all employees in workplaces participating in WERS).
On top of biases introduced purposefully as part of sampling, variable rates of non-response can also cause the achieved sample to depart from the population it is intended to represent.
Weights broadly equal to 1/(probability of selection & response) have therefore been devised in order to bring the profiles of the achieved samples of workplaces and employees into line with the profiles of the respective populations, thereby removing known biases introduced by the sample selection and response process. Failure to use the weights – either in univariate or multivariate analyses – will typically lead to biased estimates.
In WERS 2011, weights were devised for each of the instruments (the EPQ/MQ, the WRQ, the FPQ and the SEQ). Separate sets of weights were devised for the panel sample and the combined sample. A third set of weights was also devised for the refreshment sample, but these were primarily derived for the purpose of evaluating the combined sample weights and will not typically be required for analysis.
The names of the weighting variables that have been included on the 2011 WERS dataset are listed in the 'Introductory Note to Accompany the Deposited Data' (deposited with the UK Data Service along with the survey data).
Introductory Note to Accompany the Deposited Data (PDF file; UK Data Service website)
Precision
The standard errors of survey estimates are affected by the sample design, particularly by the effect of sampling weights, clustering, and stratification. As described above under 'Sampling', the WERS workplace samples are based on a stratified design with unequal selection probabilities per establishment (which are dealt with by applying sampling weights). The employee sample is clustered within participating workplaces. These factors (sampling weights, stratification and clustering) have to be taken into account when calculating standard errors (and, as a consequence, p-values and confidence intervals). Ignoring the sample design (i.e. assuming simple random sampling and taking into account only sample size and population variance) gives estimated standard errors that will be almost always too small.
Analysis suggests that ignoring the sample design will typically generate standard errors that are around two-thirds of their true size in the case of the EPQ/MQ data from the combined sample and around half of their true size in the case of the SEQ data from the combined sample. However, this ‘design factor’ will vary from estimate to estimate. Most statistical packages are now able to accommodate these issues and it is recommended that analysts carry out all their analyses using complex-sample methods to provide unbiased estimates with design-adjusted standard errors.
