287 -
measures this. The variable HAHE is the respondent's
average wage rate in 1981. It is here used to measure the
respondent's resources in the manner suggested by the
conception of the search process as a job shift process.
This wage rate will be partly endogenous to the unemployment
process, and a better measure would be the respondent's
predicted wage rate (Atkinson, Gomulka, and Micklewright,
1984). An attempt was made to obtain predicted wage rates
using variables (education and experience) available in the
present analysis to estimate a wage equation for the whole
sample of male heads. The resulting predicted wage rates
for the unemployed did not have a significant effect on the
reemployment rates and the apparent unreliability bias other
coefficients. Hence, despite the possible endogeniety, the
actual wage rates were used.
In some spells, 32 to be exact, the unemployed lost
unemployment compensation during the spell.
This was taken
into account in computing the hourly unemployment compensa¬
tion. In addition, the unemployment compensation variable
is treated as a time varying co-variate in the analysis.
Finally, the time dependent co-variate NBF was constructed
to indicate that unemployment compensation was lost.
The measure of the duration of the unemployment spell was
obtained using information about the duration in weeks of
the most recent spell in 1981 or the number of weeks of
unemployment completed for those unemployed at the time of
interview. As shown in Table 1, there is a considerable
amount of censoring present. Given the censoring, the most
informative description of these data is provided by the
Kaplan-Meier estimate of the survivor function providing an
estimate in the presence of censoring of F(t), that is the
proportion remaining unemployed by time t. Selected values
of the Survivor function are presented in Table 2.