The macro-variable approach

The macro variable approach takes quite the opposite methodological stance to the previous strategy. This approach to comparative research is based on the study of a large number of countries, and on quantitative examination of the (direct) effect of specific institutional characteristics on social and economic behaviour and outcomes. According to this approach, country names (or other social units of interest) are replaced by scores on particular macro-level variables. This approach thus offers a means of evaluating the direct effects of various institutional arrangements, such as specific policies or country characteristics, on country variations in outcomes. An example of studies that used macro-micro linkages in their analysis is the work of Gornick et al. [Gor98], who used a variety of indicators to construct an index of family policies, which was then used to explain country variations in the effect of young children’s presence on their mothers’ employment. In another study using more sophisticated techniques involving multi-level statistical models, Mandel and Semyonov [Man05] tested the consequences of family-supportive policies on women’s labour force participation, their concentration in female-type occupations and gender inequality in pay. Similarly, Uunk et al. [Uun05] tested for the net effect of childcare arrangements on women’s labour supply, controlling for alternative explanations (i.e. economic affluence and egalitarian gender ideology). Several studies examined the effect of policies, country characteristics, and other macro-level variables on the household’s division of labour [Fuw04] [Hoo06] [Sti07].

This approach allows researchers to explain country variations in particular behaviour (e.g. women’s labour force participation or women’s economic standing following a divorce) as a product of country-specific institutional arrangements (e.g. policies, family structure, labour market characteristics and more). Moreover, using data on countries as well as individuals within countries makes it possible to test for macro-micro interactions. This approach thus allows one to examine how institutional contexts mediate the relationship between two individual-level characteristics, for example, the effect of gender on earnings [Man05].

The advantage of this method is clear, since it enables the influence of individual-level and country-level characteristics on behaviour and outcomes to be disentangled [Van02], and provides quantitative estimates of the extent to which institutional arrangements (at the country level) account for differences in the outcome variable after taking into account the effect of individual-level characteristics.

The macro-variable approach, however, also has its limitations. They result from two sources: first, the measurement of institutional arrangements is at times difficult to operationalise precisely and it is even more difficult to find comparable indicators for a large number of countries. This limits, often arbitrarily, the number and the specific list of countries studied. It is often difficult to obtain comparable individual-level data for a large number of settings (e.g. to measure labour force involvement, familial arrangements or gender ideology in a comparable fashion). Consequently, some information, which could otherwise be obtained from in-depth individual-level studies of a small number of countries, is typically unavailable when dealing with many countries.

The macro-variable strategy is able to accommodate a large number of countries by representing them as a profile of scores for any number of variables that are theoretically relevant to the issue under investigation. In fact, for some analysis techniques a large number of cases is actually necessary. This is the source of the second limitation. In a multi-level analysis, which utilises fixed and random effects, a small number of countries may result in unstable estimations, and only a limited number of country indicators could be included in the empirical model. Increasing the number of countries is not always a viable solution, since it is not easy to combine macro and micro-level data for a large enough number of countries. The use of indices, such as the one proposed by Gornick et al. [Gor97], Gornick and Meyers [Gor03], and Mandel and Semyonov [Man05], can provide a partial solution to the problem. However, as argued below, indices may also create new problems, especially when different components of an index affect the outcome variable differently.

Go to next page >>