Tuesday 16th July
Wednesday 17th July
Thursday 18th July
Friday 19th July
Download the conference book
Explanatory and independent variables in social surveys across countries 1 |
|
| Convenor | Dr Uwe Warner (CEPS/INSTEAD) |
| Coordinator 1 | Professor Juergen H.p. Hoffmeyer-zlotnik (University Giessen) |
This session deals with the national standardization and the international harmonization of explanatory and independent variables in social surveys.
National standards for socio-demographic variables allow to compare surveys with other national survey data from official statistics, market and academic research. This is of importance because data from official statistics are often used to describe the quality of social survey outcomes.
International harmonization of background variables allows the comparison of survey data across countries.
For this session we welcome contributions on
- theories about the comparative approach in survey methodology in particular during the data collection,
- the strength and weakness of standardization on national and the harmonization on international level,
- the "best practice" to make social and demographic variables in register data comparable across countries,
- new and innovative measurements of explanatory and independent variables in social surveys,
- ongoing research discussing the explanatory power of social-demographic background variables in comparative surveys,
- studies exploring the total survey error and the measurement quality of independent variables in international surveys,
- and other topics on national standards and international harmonization of survey measurements.
Age, Gender and Education (AGE) are among the most used explanatory items in modern survey research. Others are Household Composition, Income or ISCO Code. One can find the respecitive variables in in many multi-variate analyses as quasi routine checks. However, one does hardly find theoretical reasoning why they were taken and, even more, which concepts and constructs they stand for in a given analysis.
For instance "sex", that somewhat surprising dichotomy "female-male", often ascribed by interviewers, does it measure "gender" (which is not a dichotomous concept nowadays), or does it measure a general division of labour between human beeings along the line of X or Y chromosones, or what else, maybe discrimination of people according to a societal accord dichotomised as women and men?
Of, for instance, "age in years", is it bound to the life-cycle concept or the cohort concept? If cohorts, why are cohorts categorised via 10-year steps and not historical societal circumstances? Are these comparable across time and countries at all?
Finally, are AGE and the others really "independent" on a conceptual level? Take for example the "astonishing" low educational level of elderly "women". Not astonishing, if one looks back in time and the educational opportunities for men, women, poor or rich.
This presentation will emphasise the necessity that only a proper conceptual definition will lead to constructs, indicators and items for measuring really explanatory measures.
In the years 2005 until 2009, the ISSP background variables (BV) underwent a fundamental documentation, review and renovation process. In the ISSP 2010 module the ISSP members were obliged to provide the renovated background variables for the first time. The integration of the ISSP 2010 data offers us now the chance to do the last step in the renovation process: to assess the quality of the new ISSP background variables. For this presentation we will evaluate these modified background variables focusing on variables related to respondent's work situation. We will present the results of a quality assessment in technical terms investigating coverage and non-response of ISSP BV, both with respect to variables in general and with respect to single categories. Additionally, we will check whether the problems found in the former ISSP background variables were solved by the revision.
European Union Statistics on Income and Living Conditions (EU-SILC) is obviously very challenging source of information for dealing with explanatory and independent variables in various socio-economic analysis. It should be stressed that in EU-SILC in some countries information is obtained from registers ('registers countries') and in others (like in Central and Eastern European countries) the information is obtained from surveys (such countries are called 'survey countries'). These various methods of getting "the same" information need to be considered and understood by users in their analysis of secondary data, especially when they are dealing with harmonization and comparison of results between countries. It is challenging as the EU-SILC surveys carried out by national statistical offices are difficult and complex in nature and even the basic demographic information could be asked in different way (e.g. 'age at the end of the income reference period' and 'year of birth'). However, some more detailed examples could be indicated. For example, some statements which are used in constructing the concept of material deprivation based on data from EU-SILC are not always relevant and important in all countries, like heating in Greece, Italy or Bulgaria comparing to heating during winter time in much colder Nordic countries. More examples will be given and discussed showing that topic taken in this session needs more attention of users with more in-depth analysis which allow to compare the results of the same issues between similar surveys in various countries.
The usual approach to survey data is that all data is collected during a specific time point, but the reality is far from it, with fieldworks lasting up to half a year or even more. During that time events may occur that influence responses. This has been recognized by the European Social Survey(ESS) with the collection of contextual data parallel to the fieldwork process. As of ESS Round 6, there are clear guidelines provided on how to recognize possible influential events and how to organize this information. But how can one measure if the collected contextual data has had influence on responses?
Previous researches on media and event effects have mostly used simple statistical methods like mean comparison and modelling. Current paper aims to develop a new and a more revealing method of looking at data using cumulative frequencies of scale points.
Usually we handle categorical study variables as single vectors with values from a discrete interval and then look at means or frequencies. Here we take variables as matrices, with scale points of the variable in columns. To incorporate time we find frequencies of the scale points by days and then cumulative frequencies. We assume that if a variable is independence of time and event effects, then all cumulative frequencies of its scale points should rise steadily. If this is not the case, then we have reason to believe that an event triggered a strong reaction in respondents.