Sammendrag
In my thesis I discuss and investigate whether there is an effect of
politicians' gender on policy outcomes. The main part consists of an
empirical investigation of this question, using a rich set of panel
data on Norwegian municipalities.
In my empirical analysis I employ data on municipal budget shares
for eight different service sectors and the share of women in the
local council as well as various control variables. The data is from
the period 1972-1999. I argue that it is particularly suitable for
the subject of study because female representation in politics
increased rapidly in Norway during this period, partly due to
women's activism within the party system. This implies that the
changes in female representation it the local councils can be at
least partly considered as driven by forces outside of each
municipality.
Women in politics has been a popular subject of study in political
economics during recent years, and investigating gender differences
in politics is of high interest in itself, both from a political and
an academic point of view. However, my study is also related to a
central question in the field of political economics, namely whether
(or to what extent) politicians are able to pursue their own
interests when deciding on policies.
Theoretical models differ in their predictions of the effects of
party representation and politicians' personal ideology on policy
outcomes, and the evidence from empirical investigations is also
mixed. The traditional median voter theorem states that who
eventually gets elected does not matter at all for policies. Several
empirical studies lead to rejection of this 'full convergence'
hypothesis. I offer a brief introduction to the theoretical
discussion of this subject and a somewhat more detailed review of
the empirical literature, some of which focuses explicitly on female
politicians. I also review a few other studies which do not concern
female politicians but gender differences in politics more
generally.
A popular model showing that personal ideology could matter for
politics is the citizen-candidate model, developed contemporaneously
by Osborne and Slivinski and Besley and Case. Using a slightly
altered version of this model, I show that if the barriers facing
female candidates in the political system are lowered, this could
make it more likely that a women is elected. This could again lead
to different policy outcomes than when gender barriers are high,
even though voter preferences have not changed. At the same time,
the model shows that if voter preferences change towards a more
'female' stand on policies, this could yield higher support for
female candidates and ultimately affect the policy outcome. The
model hence shows the importance of separating between these two
effects.
My way of doing this in the panel data regression is to add a rich
set of electoral, demographical and socio-economic variables which
should pick up effects related to voter preferences. In addition, I
add municipality- and time- fixed effects (FE) to account for
unobservable characteristics of each municipality and factors which
change over time affecting the spending decisions in all
municipalities.
The regressions are carried out using Stata 9. My results show a
positive estimated effect of female representation in the local
council on the share of the budget devoted to childcare with a high
degree of statistical significance. This estimated effect does not
depend much on the choice of control variables, nor on the time
period of study. I also find a positive effect on the share spent on
culture and a negative effect on administrative spending, though
especially the latter is less robust. The estimated effects on other
purposes like for instance education and elderly care are either
statistically insignificant or highly unstable across
specifications.
The estimated effects are very small in magnitude. More precisely,
the results imply that a ten percentage point increase in the number
of women in the local council only leads to an increase in the
budget share spent on childcare of 0.08 percentage points. The
effect on cultural spending is of similar magnitude.
I argue that the robustness of the results to the inclusion of more
control variables could imply that there is not much of a problem
with omitted variables related to voter preferences. As argued
above, changes in female council representation are likely to be
driven by external factors and could hence be considered exogenous
in this setting. In an attempt to investigate this further, I try
estimating the model by two-stage least squares (2SLS) using the
introduction of gender quotation rules in the Norwegian Labour Party
as an instrument for the share of women in the municipality council.
Instrumental variables using gender quotation rules in other
political parties seem to be irrelevant.
Unfortunately, my instrument is to weak to give precise estimates of
the effects on local public spending, and the 2SLS results therefore
tell us nothing about the validity of the results from the
regression without instrumental variables.
The results from the ordinary fixed-effects regression hence stand
uncontested: Increased female representation in Norwegian local
politics seems to have had a positive impact on the composition of
local public spending. The most robust effect is found for spending
on childcare, which is a sector where we might expect there to be an
conflict of interest between women and men. This and the other
effects found are however very modest, possibly reflecting a large
degree of consensus rule in the Norwegian local democracy or heavy
regulation of local public spending by the central government.