WERS98 Data Dissemination Service
TECHNICAL PAPERS
In addition to the Guide to the
Analysis of the Workplace Employee Relations Survey 1998, the WERS98 Data
Dissemination Service also publishes Technical Papers relating to the analysis of the
survey data. At present two Technical Papers are available. Both may be viewed or
downloaded by following the links at the bottom of this page.
Abstract:
"The use of sampling weights in survey analysis has been the subject of a number of academic papers in recent years, but the implications for data users are not widely known. The tendency has been for analysts to ignore sampling weights for regression analyses, the justification being that, although point estimates, such as means and percentages, may be biased if weights are not used, associations between variables will be approximately unbiased, and the interpretation of models, whether or not weighted, will be broadly the same. For surveys with weights that are largely unrelated to the variables of interest this standpoint may be legitimate. WERS is, however, different. Not only are the weights highly correlated with most of the key survey variables, but also the range of the weights is such that using the weights can significantly affect how models are interpreted. This paper describes why and when weights are needed for WERS analyses. The intention is to demonstrate that the arguments for using weights for point estimates extend through to the more analytic techniques, such as regression analysis. Only under very special circumstances can the weights be ignored. The paper concludes with a short review of the available software and an annotated bibliography of books and papers relating to the analysis of complex survey data."Commissioned by the WERS98 Data Dissemination Service and presented at the Second Meeting of the WERS98 User Group, NIESR, March 2001.
Users who are interested in the impact of the complex sample design should also consult our recalculated design factors (DEFTs).
Abstract:
"The availability of longitudinal survey data from the Workplace Industrial Relations Survey (WIRS) series provides many opportunities for the analysis of change in British employment relations. In particular, the data available from the 1998 Workplace Employee Relations Survey offers unprecedented opportunities for researchers to discover more about the forces that are driving change in aggregate behaviour. The paper focuses on the particular question of how change in the aggregate behaviour of workplaces may come about, aiming to show how the technique of shift-share analysis may be used to separately identify the impact of compositional and behavioural change.The paper first illustrates the method using time-series data on the decline in the incidence of trade union recognition at workplace level between 1990 and 1998. The paper then extends the technique to data from the WERS98 Panel Survey and data on closures and new workplaces, in order to find out more about the decline in union recognition. It finds that the primary reason for the fall in union recognition between 1990 and 1998 was the lower rate of recognition among workplaces joining each of the three broad sectors of the economy when compared with workplaces leaving each sector. Other factors of minor importance included changes in composition and changes in within-group behaviour among continuing workplaces"
Note: The calculations used in preparing the paper are also available within a Microsoft Excel spreadsheet. Please note that the spreadsheet can only be viewed or downloaded with Microsoft Internet Explorer. Users without access to Microsoft Internet Explorer should e-mail the WERS98 Data Dissemination Service at the following address (wers98@niesr.ac.uk) to request that a copy of the spreadsheet be sent to you.
| Abstract: Addresses three common technical questions that have arisen among users following recent discussions about the use of sampling weights. These questions concern: the origin of the sampling weights and their meaning; the analysis of sub-samples; and the implications for variance estimation. |
Instructions for Viewing and Downloading the Technical Papers:
The Technical Papers listed above are made available in Adobe Acrobat® (pdf) format. Users without this software may download it free of charge from the Adobe web-site by following the pictorial link below. To view a document below, follow its hyperlink. To download the document, right-click on the hyperlink and this will give you the Save As option.
To view a Technical Paper, follow its hyperlink. To download a Paper, right click on the hyperlink and this will give you the Save As option.
WERS98 Data Dissemination
Service
National Institute of Economic and Social
Research, 2 Dean Trench St, London SW1P 3HE
URL: http://www.niesr.ac.uk/niesr/wers98/
E-mail: wers98@niesr.ac.uk
This site is maintained by
Simon Kirby
Page last updated: 12 December 2001