15(27)
treatment group are, on average, older than the single parents in the
control group and the average age of their youngest child is, for this
reason, likely also higher. Despite this, and despite their living space being
larger (households in the treatment group, on average, lost approximately
SEK 3200 per year in housing allowance due to the size limit), the
treatment and control groups on average seem to be very similar—which is
what we need in order to credibly rely on the DD estimator.
3.3
Empirical model
The basic DD framework for this study can be described as follows:
Let
)(
t
i
be the risk hazard of recipient household i moving into
overcrowded conditions at time t. The households are observed in a pre-
treatment period, t = 0, and in a post-treatment period, t = 1. Between
these two periods, a fraction of the households, i.e., the treatment group is
exposed to the imposed limit on floor space. Looking at Fig. 3, which shows
the distribution of the calculated decrease in housing allowance for
households in the treatment group, it is obvious that this can be a fairly
heterogeneous group. The treatment variable,
T
i
(usually a dummy
variable), in our case will therefore be a continuous variable that assumes
the value of 0 for households in the control group. For households in the
treatment group, this variable represents the potential decrease in housing
allowance [tSEK/year] attributable to the size limit. Furthermore, the
reform variable, i.e., the interaction term
T d
97
, will then also be a
continuous variable that is zero for all households before 1997 and then 0
for the control group, but a continuous variable representing the decrease
in housing allowance for those households affected by the size limit.
10
Figure 3. Decrease in housing allowance [tSEK/year] attributable to the
size limit.
10
We also performed the analysis using the usual dummy variable approach, i.e., without
considering that households in the treatment group were affected by the size limit to varying
degrees. The results of this analysis suggest a reform effect of approximately 60%, i.e., the risk
of moving into overcrowded conditions is 60% higher after 1997 for the treatment group, overall,
than for the control group.
Decrease in housing allowance [tSEK/year] due to size limit
0.00
5.00
10.00
15.00
20.00
25.00
0
500
1000
1500
2000
2500
Frequency
Mean = 3.1765
Std. dev. = 2.92218
n
= 22 773
