17(27)
4
Results
Table 3 reports the results of the complementary log-log model without
frailty. The table presents the exponentiated coefficients rather than the
coefficients themselves and states the risk relative to the base category.
11
An exponentiated coefficient of 1 indicates no difference in risk between the
two categories, while an exponentiated coefficient >1 (<1) indicates an
increased (decreased) risk. Standard errors are similarly transformed. The
test of significance is whether the risk is 1. The variable of main interest
here is the reform variable
T*[d97]
at the top of Table 4. The simplest
model in Table 4, Model 1, includes the reform variable, the year dummy
for 1997–1999, the continuous variable controlling for differences between
the control and treatment groups (i.e., the differences between households
depending on the extent to which they were affected by the imposed size
limit), and controls for quadratic duration dependence. Model 2 adds
individual and household characteristics. In Models 3–5, controls for local
housing and labor market conditions are added. Models 1–5 have also been
estimated including controls for continuous, instead of quadratic, duration
dependence; this did not alter the results.
11
For a continuous variable, an exponentiated coefficient gives the percentage change in the
hazard rate, given a unit change in the explanatory variable.
