For an attempt to marshal all the possible factors see Bartolini,
Bilancini, and Sarracino (2016)— also discussed in Chapter 8. Another
factor limiting the benefits of national income growth is boom and
bust. As De Neve, Ward, De Keulenaer, et al. (forthcoming) show, the
losses in happiness when income falls exceed the gain in happiness
from an equal rise in income.
The key table on social comparisons is the pooled cross-
sectional table shown in online Full Table 2.3. This table includes all
comparisons simultaneously.
Card et al. (2012).
One issue is the relevant reference group. Two common choices
are neighbors (broadly defined, as here) or coworkers (or those similar
to the individual on the labor market). See Clark and Senik (2010), and
Layard, Mayraz, and Nickell (2010). As regards neighbors, well- being
has been found to fall with average incomes in the local area (Ferrer- i-
Carbonell [2005]; Luttmer [2005]; Kingdon and Knight [2007]). As
regards coworkers, a number of papers have shown that well- being
is negatively correlated with others’ earnings (G. Brown et al. [2008];
Cappelli and Sherer [1988]; Card et al. [2012]; and Godechot and
Senik [2015]). See also evidence from neuroeconomics (Fließbach et
al. [2007]) and hypothetical preference questions (Solnick and Hemen-
way [2005]).
However in some studies well- being has been found to rise with
others’ income. This could be for reasons related to local public goods,
to the tunnel effect (whereby others’ good fortune informs you about
your own future prospects), or altruism: see, for example, Clark, Kris-
tensen, and Westergård- Nielsen (2009), Dunn, Aknin, and Norton (2008),
Senik (2004), and see also Clark and D’Ambrosio (2015).
For comparator effects we focus on the results using pooled
cross- sections. The reason is that in a fixed- effects analysis, the effect
of comparator incomes (as measured) depends heavily in our sample
on information for people who move between regions— a quite small
number of people.
This assumes that average comparator income is measured by
absolute mean income rather than the mean of log income. To check
on this we estimated an equation that included simultaneously log Y ̄
and log Y. The effects of log Y ̄ far exceeded that of log Y, which was in
three out of four countries insignificant.
See online Table A2.2, second column for each country.
