11(27)
3
Data and empirical strategy
3.1
Data
The dataset used here was extracted from the Swedish National Insurance
Agency’s register on housing allowance recipients. This database contains
information on all Swedish recipients between 1994 and 2002, i.e., almost
700,000 households. When analyzing the impacts of the housing allowance,
such register data are, according to Shroder (2002), preferred to survey
data, which have been demonstrated to often include biases from reporting
errors regarding assistance status. Besides the fact that this database has
previously been largely unavailable for research, it also provides important
information on the housing conditions of low-income households in Sweden.
Since the
census
ceased in 1990, few
Swedish datasets contain detailed
information on households’ housing conditions.
We have information on the recipients’ actual housing conditions in May
each year, i.e., at most we have one observation per recipient household
each year (however, for each year, we also have information on whether
recipient households received a housing allowance in January–April, or any
of these months). The database contains information constituting the basis
for determining how much housing allowance a specific household is
entitled to: household composition (i.e., coupled, single, number of
children, and age), dwelling characteristics (i.e., living space, number of
rooms, tenure, and housing expenses), income data, address, and amount
of granted housing allowance.
To control for differences between housing and labor markets, local values
of Tobin’s
q
have been added to the data together with municipal
unemployment rates. Tobin’s
q
(Tobin, 1969) is based on the relationship
between the marginal value of new investment and the cost of
reinvestment. It has been estimated by Tommy Berger (Institute for
Housing and Urban Research, Uppsala University) at the municipal level
and is defined as the ratio between the transaction price and the cost of
reinvestment (Berger, 2000). A ratio above 1 indicates that it is profitable
to make a housing investment in the municipality, while a value of less
than 1 indicates the opposite. Furthermore, municipalities have been
classified into nine categories according to the Swedish Association of Local
Authorities,
8
based on structural parameters such as population,
commuting pattern, and economic structure.
3.2
Empirical strategy and design
In this study, a quasi-experimental dimension of the 1997 housing
allowance reform, i.e., the imposed limit on floor space, is exploited by
applying the DD estimator (see, e.g., Abadie, 2005; Meyer, 1995). The DD
estimator is based on the idea that when only a fraction of a population is
accessed October 29, 2009.
