This chapter discusses the potentials and constraints of using a volunteer Web survey as a worldwide data collection tool for wages. It provides a detailed description of the bias related to individual-level wages and core socio-demographic and employment-related variables across selected developed and developing countries and evaluates the efficiency of post-stratification weights in adjusting these biases. The results confirm that Web samples are particularly attractive to younger persons, part-timers, and persons working in non-manual occupations. This can be observed across countries, although the strength of the bias differs between them. With respect to the efficiency of post-stratification weights, the results are inconclusive. Whereas it is advisable to implement weights for descriptive purposes of socio-demographic variables, the contrary holds in case of wages. Additionally, weights can have the opposite effect by (moderately) increasing the difference in the estimated parameters between the reference and the Web sample.
Top1. Introduction
Wages are central to the world of work. Living standards and the livelihood of wage earners and their families depend on the level of wages, and on when and how they are adjusted and paid. Wages affect job and life satisfaction, working conditions as well as migration within and across countries (ILO, 1982). Moreover, they are a major component of overall consumption and a central factor in the economic performance of countries. Accordingly, it is no surprise that wages are one of the key variables in socio-economic research (Freeman, 1995, 1996; ILO, 2010).
However, collecting information on wages is no easy undertaking. Besides the high rates of people who do not answer wage‐related survey questions (Plasman et al., 2002), measurement issues are also relevant especially when wages are compared across countries (Rodgers et al., 1993). In Europe, for instance, wage information is primarily based on administrative data gathered through companies’ personnel records or social security records. While those data sets offer wage information, they offer only limited information on working conditions, socio-demographic characteristics, or attitudes. Where wage information is collected in surveys, it is often measured in an aggregated and standardized way to allow cross-national comparisons. These difficulties make the worldwide collection of valid wage data a huge challenge. And due to economic, social and political globalization, gathering such global data is a momentous task. In this context, Web surveys seem to offer a lot of advantages such as worldwide coverage, low cost, and a fast data collection process. They also allow multi-country and multilingual homogenized surveys that are crucial in times of globalization. Moreover, for sensitive questions such as income, they can provide more reliable results as the often observed social desirability effects can be reduced or even eliminated (Joinson et al., 2008). The drawback of Web surveys, however, is that they comprise many methodological challenges. A core problem is related to the representativeness of the data as the sub-population with Internet access might be quite specific. Particularly non-probability based Web surveys are problematic because respondents are not selected at random, and the target population forms a convenience rather than a probability sample.
Against this backdrop, the chapter aims to provide an assessment of the potentials and the constraints of using a volunteer Web survey—the WageIndicator—as a worldwide data collection tool for wages. It provides not only a detailed discussion of the bias related to wages and core socio-demographic and employment-related variables across selected developed and developing countries, but also evaluates the efficiency of one adjustment technique (post-stratification weighting) to correct the biases arising from non-randomized sample selection. The outline of the chapter is as follows. Section 2 provides an overview of the problems associated with measuring and collecting wages across countries and with using non-probability Web surveys for data collection. It also evaluates possible solutions, like weighting, to correct the observed bias between the Web sample and the population of interest. Section 3 introduces the data sets used for the analysis, describes the observed bias with respect to wages and the core socio-demographic and employment-related variables, explains the applied weighting technique, and presents the results. Section 4 concludes and discusses the findings. Particular attention is devoted to the advantages and disadvantages of collecting wages on a worldwide scale via a volunteer Web survey and of post-adjustment measures to deal with the methodological challenges. In addition, future research directions are addressed.