12(27)
exposed to a treatment, often via a regime shift, an untreated control
group can be used to identify temporal variations in the outcome variable
not due to the treatment exposure. Any comparison of pre- and post-
treatment outcomes for the treatment group would likely be contaminated
by, for example, temporal trends in the outcome variable or events other
than the treatment that occurred between the pre- and post-treatment
periods. Furthermore, the DD estimator is intended to measure how
changes in the outcome variable over time differ between the treatment
and control groups. If this difference is significant, we can conclude that the
reform did influence the housing consumption of the affected households
and that the housing allowance may have some real effects on household
housing consumption.
The DD estimator assumes that all trend effects are common to both the
treatment and control groups (Abadie, 2005, p.1–2). Thereby, we want to
eliminate the possibility of trend effects differing between the treatment
and control groups, which could bias the DD estimate, i.e., all relevant
differences in a natural experiment between the treatment and control
groups should be controlled for to credibly invoke the independence
assumption between groups. To reduce the risk of this assumption not
being fulfilled, we have limited this study to single-parent housing
allowance recipients living in rental apartments. Couples with children also
experienced other changes in the 1997 reform, the impacts of which may
be difficult to separate from each other. Furthermore, households in rental
apartments are chosen since transactions costs—i.e., search costs and
actual moving costs, which create sluggishness in the housing adjustment—
have been found to be greater for homeowners than renters. Moreover,
eligible housing expenses are calculated differently for different tenures,
and these calculation rules also changed for homeowners in the 1997
reform. These heterogeneities contributed to the decision to focus on single
parents living in rental apartments, which is the group of households
overrepresented in the housing allowance system today. However, beyond
the data limitations, limiting oneself to studying housing allowance
recipients implies, after all, a certain homogeneity in the surveyed
population, at least in terms of housing conditions and income.
The sample used here consists of single-parent households (with children
not older than 18 living full time with their parents) who lived in rental
apartments and received a housing allowance in May 1996 and May 1997
and did not live in overcrowded conditions in May 1994 or, if they did not
become recipients until 1995 or 1996, at the time they entered the housing
allowance system. We follow these households until they fulfill the
requirements for overcrowding according to norm 3, or at most, until May
1999.
The treatment group consists of those households that had useful floor
space exceeding the size limit in May 1996, just before the size limit was
imposed, while the control group consists of those households that had
useful floor space below this limit in May 1996. Figure 2 shows the “flow”
into overcrowding for the treatment and control groups between 1994 and
1999.
9
9
Households covered by the guarantee level introduced in November 1997 (see Table 2) have
been classified as belonging to the control group. We also conducted the analysis including these
households in the treatment group; this had no effect on our results.
