Estimation of Length of Job Search by Survival Analysis

摘要: This paper is a report of an empirical test the job search model, which holds that probability exiting unemployment related both to duration and costs remaining unemployed. Cox survival regression analysis used estimate conditional (either finding employment or leaving labor force). The predicted for all unemployed heads households in Panel Study Income Dynamics (Institute Social Research) 1986 data set. measure length spell, defined as number weeks continuous unemployment. Two measures opportunity cost are used: first reservation wage; second ratio wage final at worker's last job. In addition cost, other explanatory variables are: gender, race, age, years schooling, county rate, receipt compensation any time (ascertained 1987). study investigates effects gender race on through separate estimates models each four gender-race groups. investigation effect duration, this uses sample persons only part population received compensation. LITERATURE REVIEW model was introduced by Stigler 1962. Since then, research has been reported by: Akerloff Main; Carlson Horrigan; Flinn; Frank Freeman; Kiefer; McDonald Butler; Perry; Salant; Topel. Salant (1977), Main (1980, 1981a), found completed spells less than in-progress spells. They attributed with longer being over-represented Horrigan (1983) proportion those increased words, person unemployed, likely it s/he will become employed. Using National Longitudinal Survey sets, Freeman (1978) frequency were relatively uncorrelated. also age men women. agreement this, Perry (1972) youths have shorter adults. Flinn (1986) discussed estimation distribution when following characteristics: 1) Observations right-censored, since currently individuals know how long they cannot tell their spell end. 2) length-biased, "longer more be interrupted point