23.07.2013 Views

samlet årgang - Økonomisk Institut - Københavns Universitet

samlet årgang - Økonomisk Institut - Københavns Universitet

samlet årgang - Økonomisk Institut - Københavns Universitet

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

422<br />

NATIONALØKONOMISK TIDSSKRIFT 2005. NR. 3<br />

age, between 3 and 6 years of age, between 7 and 9 years of age, between 10 and 17<br />

years of age, an indicator variable for the presence of stepchildren, and finally an indicator<br />

variable that takes the value 1 if the couple had a child together before marriage.<br />

I also include variables measuring the age of the couple and the age difference. The variable,<br />

sickness, is an indicator variable taking the value 1 if the individual receives<br />

sickness benefits during the year. As a general rule sickness benefits are received if a<br />

person has a spell of illness for more than 13 weeks. I distinguish between individuals<br />

living in the Copenhagen metropolitan area and individulas living in the provinces by<br />

the indicator variable province. Each individual's degree of unemployment during the<br />

year is defined as the number of hours of unemployment divided by the number of potential<br />

supplied working hours.<br />

4. Econometric specification<br />

Our empirical model closely follows Lillard et al. (1995). That is, I specify a model<br />

of the risk of divorce which explicitly incorporates both the direct effect of premarital<br />

cohabitation and a potential self-selection of more divorce-prone individuals into premarital<br />

cohabitation. Our model differs slightly from Lillard et al.’s as our data are<br />

observed on a yearly basis. Hence, I model the transition out of marriage as a grouped<br />

duration model. In addition, I impose a different distributional assumption on the<br />

unobserved heterogeneity components. The correlation structure between the two processes;<br />

the transition out of marriage and the decision to cohabit before marriage, determines<br />

the endogeneity of cohabitation, i.e. the self-selection effect. Therefore,<br />

if unobserved factors both influence the individual risk of divorce and the decision to<br />

cohabit, then they are correlated and cohabitation is endogenous. Below I briefly sketch<br />

the empirical model employed in this study.<br />

4.1. Empirical model<br />

The hazard function is specified as a mixed proportional hazard. That is, it is a product<br />

of a function of time spent being married (the baseline hazard), a function of observed<br />

time-varying characteristics, x d , and a function of unobserved characteristics,<br />

v d ;<br />

h(t|x d t ,vd )= (t) · (x d t ,vd ) , (4.1)<br />

where (t) is the baseline hazard and (x d t ,vd ) is the scaling function specified as<br />

exp(x d t d + v d ). I follow Lillard et al. (1995) and assume that the unobserved component,<br />

v d , is individual-specific to the unit of observation. The unit of observation is the<br />

person in our sample that is followed throughout the period from 1980 to 1995.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!