19
to use their time. Hence, one would expect that the interaction of the time-composition and
the saddening effect causes a negative difference in experienced utility between the employed
and the unemployed on weekdays, but a positive difference on weekends.10 Indeed, Figure 1
shows that the employed report a duration-weighted episode satisfaction of 7.15 on weekdays,
while the unemployed report 6.99. On weekends, both groups report higher episode
satisfactions, but the employed experience a much larger increase in their well-being than the
unemployed. The employed report an episode satisfaction score of 7.69, while the
unemployed’s episode satisfaction rises to only 7.27. When we look at net affect, the
subjective well-being of the employed is lower on weekdays than that of the unemployed
(4.00 vs 4.14). On the weekend, however, the ranking is turned around. The employed reach
an average net affect of 5.58, whereas the unemployed, although enjoying the weekend
somewhat better than weekdays too, report only an average net affect of 4.68. In the case of
the U-index, the employed report a slightly lower percentage of time spent on unpleasant
activities than the unemployed (16.2% vs 17.3%). On the weekends, however, this gap widens
tremendously. The employed spend only 8.5 percent of their time in unpleasant states, while
the unemployed still report this share to be 15.7 percent. This shows that the time-
composition effect plays a crucial role in explaining the vanishing difference in experienced
utility between the employed and the unemployed.
3.4 Regression analysis
The differences in experienced utility (or the absence thereof) between the employed and the
unemployed could have various causes. Besides a genuine relationship between employment
status and experienced utility, it could be that other factors that are correlated both with
experienced utility and employment status are the true causes of any correlation between the
two variables. To control for such factors, we conduct a regression analysis to estimate the
impact of employment on experienced utility and to compare it to its impact on life
satisfaction.
Table 5 contains the results of regressing both life satisfaction and the three measures of
experienced utility on a set of socio-economic characteristics, including the respondent’s own
employment status, income, age, family status, number of children, and measures of how the
10 While this procedure is suggestive of a decomposition in a saddening and a time-composition effect, it cannot
provide a full decomposition. While the well-being difference on weekends can be attributed to the saddening
effect alone, differences on weekdays still consist of saddening and time-composition effects.