The Happiness Workout

Abstract

This study tests for causality from exercise and physical activity to life satisfaction (LS) by applying an instrumental variable approach with the respondents’ perceived benefits of exercise participation as instruments. Using data across 25 countries from the Eurobarometer survey, our results confirm the positive association between exercise and LS. In terms of causality, the results indicate that being active increases LS for both gender, though more for men than women. One main reason for this relationship is that exercise is perceived as being pleasurable, something that policy-makers should keep in mind when designing programmes to get us off the sofa.

Appendix

This section presents the results of the factor analysis carried out for the perceived benefit indicators included in the survey. The methods carried out here are well documented in textbooks and handbooks—see for example Manly (2005) and Nardo et al. (2008). In brief, we first perform a factor analysis where the suggested number of factors is identified based on the Kaiser criterion, which drops factors with eigenvalues below unity. Second, the Kaiser–Meyer–Olkin (KMO) measure for sampling adequacy is calculated; both for individual variables and for the overall analysis. KMO values vary between zero and one, where values equal to 0.6 or above are desirable, since large values imply that a certain variable has enough information in common to produce a reliable factor analysis. Third, in order to obtain an interpretable pattern of factor loadings a varimax rotation is usually performed. Finally, factor scores are estimated.

The factor analysis here is based on tetrachoric correlations of the underlying variables (i.e. perceived benefits). An initial factor analysis on all 14 variables suggests that three factors are enough to summarize these benefit measures. These are the factors with a loading greater or equal to 1 (Kaiser criterion). However, subsequent estimation of the KMO measure of sampling adequacy suggests that the last variable, “other” (KMO _Other  = 0.492) should be discarded from the analysis since any value below 0.5 is considered to be unacceptable. This also makes intuitive sense given the unspecific nature of this variable. Repeating the factor analysis on the remaining 13 variables indicates that the benefit indicators can be summarized by two factors, as reported in Table 5. Furthermore, we now obtain KMOs well above 0.75 and an overall KMO of 0.929—see Table 6.

Table 5 Factor analysis and eigenvalues

Table 6 KMO measures

Going back to Table 5, we observe that the third factor has an eigenvalue of 0.956 and does not fulfil the Kaiser criterion at the margin. Deriving the rotated factor loadings under the strictly suggested two factor model however does not yield a clear distinction of benefit factors, since under that specification the health benefit would be included together with the ‘purposeful’ benefits (e.g. ‘develop new skills’) leading to a mixed interpretation of the factors. Thus, we will allow for three factors to be extracted from the model. The varimax rotated factor loadings are reported in Table 7. Bold figures indicate the benefits with the highest loading across the three factors.

Table 7 Rotated factor loadings

Hence, benefit 1 can be summarised in Factor 1, benefits 2–5 in Factor 2 and benefits 6–13 in Factor 3. We label these factors ‘health’, ‘pleasure’ and ‘purpose’, respectively. The last column labelled ‘uniqueness’ indicates the proportion of the common variance of each benefit indicator not well-represented by the three factors, where lower values are desirable.