positive effect. However, the studies they include appear to use mainly linear specifications (years
of schooling), whereas Hartog and Oosterbeek (1998) stress the importance of allowing for a nonlinear
relationship. Also Witter et al.’s meta-analysis is based on pre-1984 studies, and Veenhoven
(1996) has suggested that the effect of education has changed to become negative over time. Even
though Witter et al. (1984) identify a generally positive effect, Michalos (2007, p.13) notes that
educational attainment accounts for only one to three per cent of the variation in adult wellbeing in
the 90 studies they reviewed. Analysing US data from 1972 to 2006, Stevenson and Wolfers (2008)
do enter education as a series of dummy categories (college graduate, some college, high school
graduate, did not graduate from high school) and find a steady increase in happiness with education
from one category to the next. Moreover, the premium associated with higher education seems to
have increased over time. Ross and Van Willigen (1997) investigate the relationship between
education and a range of measures of subjective quality of life. They find education guards against
all tested measures of distress, such as anxiety and depression, but no significant total effect of
education on job dissatisfaction. After controlling for job characteristics, the more-educated are
found to be less satisfied with their work, leading the authors to conclude that the more-educated
experience less distress, primarily because education provides access to non-alienating work, but the
higher expectations of the more-educated result in no increase in overall job satisfaction.
The more recent studies that indicate a negative relationship between education and wellbeing
include a number of Australian studies, of which several are based on the Household, Income and
Labour Dynamics in Australian Survey (HILDA) (Dockery 2003; Headey & Wooden 2004;
Hickson & Dockery 2008). Using data from the 1995 Year 9 cohort of LSAY, Dockery (2005) also
finds that school leavers who had higher school achievement scores in Year 9 were less happy. This
is consistent with earlier findings by Marks and Fleming (1999) for the school leavers in the late
1970s and early 1980s, where the longitudinal Youth in Transition data were used.
Since education has a positive effect on other life domains, such as health and income, the inclusion
of these variables along with education among the independent variables means that the coefficient
on the education variable must be interpreted as some sort of ‘residual’ effect of education, after
allowing for its impact on other life domains. If these intermediary variables were excluded, then
surely the ‘full’ effect of education would be to increase happiness. However, using data from the
2002 wave of HILDA, Hickson and Dockery (2008) show that, even in a reduced model including
only age, gender, marital and disability status, education has a negative and highly significant effect
on life satisfaction. Their findings were largely unchanged whether a linear ‘years of schooling’
variable or a more flexible specification with a series of dummy variables representing the different
education levels was used. The simple mean ratings of life satisfaction (on a scale from 0–10) for
Australians who had only completed Year 11 or below was 8.03, compared with 7.79 for those with
a university degree.
Hickson and Dockery’s (2008) attempts to account for the negative association between education
and life satisfaction by modelling the formation of higher expectations, based primarily on parental
education, were largely unsuccessful. Clark and Oswald (1994) found that, among unemployed
persons, those with higher levels of education were far less happy than those with lower levels of
education. This may be interpreted as their state of unemployment incurring higher opportunity
costs in terms of foregone earnings, but is also consistent with higher expectations among the moreeducated
playing a role in determining happiness. Estimates on the effects of ‘over-education’ and
‘under-education’ that have recently been applied in wage equations may be an interesting avenue of
investigation. For Australia, Mavromaras, McGuinness and Fok (2009) and Voon and Miller (2005)
have found that years of ‘over-education’ relative to the typical educational requirements of a
worker’s job have a lower impact on wages. This may similarly impact upon happiness.
From the theoretical considerations discussed above it can be seen that the explanation for the
negative relationship between educational attainment and happiness may lie in the differential effect
of education upon satisfaction within individual life domains, and in turn the relationship between
satisfaction within these domains and overall happiness. Some evidence on this can be gleaned
from existing empirical studies. Headey and Wooden (2004) report multivariate models for overall