College dropout and graduation with endogenous waiting time to enrollment

Using data from the 1979 and 1997 National Longitudinal Surveys of Youth (NLSY79, NLSY97), we investigate the college attendance, dropout, and graduation behavior of U.S. high school graduates. To estimate the determinants of the duration of college attendance, while accounting for the effect of (possibly) endogenous waiting time until enrollment, two bivariate discrete-time hazard models with competing risks are estimated. First, we jointly model the duration until college enrollment and the duration of college attendance until exit. For the duration until college enrollment, no distinction is made between those who initially enroll in a two-year versus a four-year institution. College exit can occur because of dropout or graduation and is modelled using a discrete-time competing risks approach. To allow for endogeneity of wait time, the unobservable determinants of wait time are allowed to be correlated with the unobservable determinants of dropout and graduation risks. In the second model we explicitly distinguish between those who first enroll in a two-year institution and those who first enroll in a four-year institution using a competing risks approach. Thus we model both durations as competing risks. Moreover, when analyzing the enrollment duration we estimate separate competing risks models for those who enter two-year institutions and those who enter four-year institutions. Again, we account for possible endogenous waiting time until enrollment by allowing the unobservable determinants to be correlated both within and across wait time and college attendance durations.

Delaying college entry significantly increases the chances of college dropout and reduces the probability of attaining a four year degree. For those who first enroll in four-year institutions, delaying entry by one year after high school graduation reduces the probability of graduating with a four-year degree by about 32 percent in NLSY79 and 12 percent in NLSY97 in models that account for the endogeneity of delaying enrollment. We suggest that the decreased impact of delay is due to the increase in resources dedicated to student support since the 1980s. There is also empirical evidence that the negative impact of delayed enrollment on college graduation probabilities varies by ability measured by the Armed Forces Qualifying Test (AFQT) score with the largest estimated impact of delayed enrollment occurring for those with low AFQT scores. We also show that delay is associated with lower starting wages.

In ongoing work we endogenize the choice of a more selective institution and investigate whether the impact of delayed entry varies by selectivity.