How Satisfied are the Self-Employed? A Life Domain View

We (and much of the literature) use the terms self-employment and entrepreneurship interchangeably for ease of reference (on this practice see also Carter 2011). Some might consider there to be differences between entrepreneurship and self-employment, e.g. in terms of innovation, growth ambition, etc. Bear in mind that when talking about entrepreneurship in this paper, we refer to self-employment.

See, however, Hanglberger and Merz (2011) for some evidence that this positive effect is driven in parts by anticipation and adaptation effects and largely disappears when controlling for those.

Typical returns to self-employment are lower than compared to earnings for being employed, but there are issues of underreporting of returns of the self-employed for tax reasons, as well as other considerations that warrant further research on this front (Carter 2011).

Interestingly, the latter has been identified to decrease job satisfaction in a recent study of European self-employment (Millán et al. 2013).

Also, the positive effect of being self-employed on job satisfaction diminishes markedly when taking into account the heterogeneity of the control group of the employed in terms of the size of the firm they are working in Benz and Frey (2008a, p. 374), as well as when accounting for the afore-mentioned anticipation and adaptation effects (Hanglberger and Merz 2011).

Cooper and Artz (1995) found that entrepreneurs with initially high expectations for their business venture performance turned out to be more satisfied than other entrepreneurs, suggesting that these more satisfied individuals have some more optimistic personality traits that influence their subsequent job satisfaction. Wether there are certain personality traits that facilitate self-employment and the related satisfaction derived from it is still a matter of active research (e.g. Caliendo and Kritikos 2012).

Block and Koellinger (2009) find a similar difference in terms of satisfaction with the startup, i.e. necessity entrepreneurs do not receive satisfaction from what they are forced to be doing.

We treat this measure as cardinal at various instances throughout our analysis since it was shown in the literature that this does not alter findings substantially and it allows us to use a fixed-effects regression framework Ferrer-i-Carbonell and Frijters (2004).

Of these, a minority reports work domain satisfaction judgements, which are summarized in Table 1. We interpret these to refer to individuals’ dissatisfaction with not having a job.

Excluding “non-working” individuals and focussing only on individuals whose labor-force status is “unemployed” doubles the coefficient size to \(-0.60\)***.

We focus on three lags, mostly because taking longer lags would give us a lower number of observations (i.e. data limitations) and for allowing easy comparability with previous investigations (e.g. Binder and Coad 2013), and because we suspect that 3 years is long enough to pick up the most interesting effects immediately associated with a transition into self-employment. In a five-lag model not shown here, the fourth and fifth lag show extremely decreased numbers of people (only 42 unemployment to self-employment cases in \(t+4\) and 25 in \(t+5\)), which decreases the reliability of statistical inference. In the longer lags, no statistically significant effects are found beyond the fourth lag for the opportunity case (PSM coefficient: \(0.29^{*}\)). While hedonic adaptation is still ill understood, assessing longer lags would be desirable, given better data, and from Clark et al. (2008), we know that some life events have a more lasting influence over time. Whether self-employment belongs into that category seems questionable but should be further analyzed in future research.

One exception to this constitutes the use of caliper matching, which increases the variance of the estimates (Caliendo and Kopeinig 2008, pp. 42–45) and leads to a loss of significance in our model for life satisfaction. Coefficients are positive in lags \(t+2\) and \(t+3\) in this case, however, as well.

Note that matching estimates refer to total effects on subjective well-being while regression coefficients are ceteris paribus effect sizes, holding all other variables of interest constant (Oakes and Kaufman 2006, p. 382). They should not be directly compared with each other, thus, and comparison of our results with other studies using multivariate regression is not straightforward.

The opposite pattern was found for unemployment, where the unemployed exhibited higher satisfaction with their amount of leisure time in a British sample (Powdthavee 2012).

Our results cannot confirm the finding that self-employment creates work–home–conflict, e.g. by decreasing family satisfaction (this was found by Parasuraman and Simmers 2001): while the coefficient for family satisfaction is negative in our sample as well, in most cases it is not statistically significant.

We are happy to provide these more detailed diagnostic analyses on request.