Variables

A big problem in social science research is outlining the manifest variables so that they, in an adequate way, measure the latent ones, that is, the theoretical constructs that our study uses. How do we, for example, turn antisocial potential, self-control, or informal social control into something we can actually measure? After all, there exist no such universal variables, nor any universal agreement on how such a variable should be constructed and/or what it should include. So, how do we do it?

We usually call this the problem of operationalization, and while we almost always encounter it, why we do so differs from study to study. If we conduct a self-report study, one reason why this problem occurs could simply be ethical: we cannot always ask what we want. However, much more common is the fact that we have to use data where the existing variables do not fully mirror the theoretical dimensions we want to include. One reason for this, of course, is that we often have to rely on data which is quite old, or in the case of register data, has been collected for a purpose other than the researcher’s. At any rate, such issues lead to the inevitable fact of compro- mise when it comes to validity – for example, ‘Do we really measure self-control? Maybe we have to settle for studying one dimension, or simply one or a few indica- tors of the phenomenon we are after? ’

The problem of operationalization is likely to increase if the study has a longitu- dinal character: the same phenomenon (such as social adjustment, for example) must be operationalized differently at different ages. In the case of social adjustment, it is highly logical and relevant to include questions such as ‘do you work and, if so, how many hours a week? ’ To ask the same question to a 13-year-old (or a 5-year-old! ) would be somewhat strange. However, we still want to know how socially adjusted our sample was at the age of 13, so the problem does not go away. Thus, as we go through our study, we have to be sensitive to the social process of ageing and the flow of the life course, and re-operationalize many of our theoretical constructs to capture what is actually going on in our participants’ lives. Another problem concerns those contextual, social changes that we also have to consider: the times are changing. So, within our measure of social adjustment, we may want to include something that captures economic status. One way to capture this in 1985 or 1990 would be to ask whether or not the respondent has a mobile phone. Today, that question is essentially meaningless for any measure of status.

Additionally, the way we choose to formulate our questions and our interests in certain theoretical constructs and not others, may give rise to issues for long-term

studies: what the researchers may have wanted to know when they initiated their research project, when the respondents were children or teenagers; when they are adults and time has passed, the questions we can now answer may not be considered relevant or interesting (although such words are highly loaded with value judgements) by the science community or policy makers. The reverse is also true, of course, that new research questions emerge, due to a new social and political climate, that nobody considered relevant when we initiated our study. In our own study, violence is a strik- ing example of this. In Sweden, at the beginning of the 1960s, violence was considered a ‘natural way’ for young boys and men to solve conflicts. Theft, however, was a very serious crime and an indicator of a highly problematic background and high-risk life circumstances. Thus, among the older samples in the Stockholm Life Course Project, we have practically no variables that capture the use of violence in youth, but a large number of variables that attempt to capture theft. Since the 1960s, of course, there has been a change in how society in general, and the criminal justice system in par- ticular, perceives teenage violence. Today, it is a serious social problem and, at the individual level, an indicator of future problem behaviours.

There is no simple solution to this problem of operationalization (nor indeed to

many other, related problems). However, a broad and encompassing data collection in the initial stage of the research project, where one also collects data that does not feel immediately relevant, could make future analyses possible. At the same time, that kind of data collection may not be recommended for other reasons: first of all, data, in many ways, costs money and, second, the need for data always has to be weighed against individuals’ personal integrity.