with below average credentials, but charges low tuition, attrites a
In addition, law schools would likely have to attrite students who
In the BGLW test, the value of y at the initial wave of the survey, which we denote by [y.sub.0], is regressed on x and on future A (namely, whether the individual later attrites).
where the outcome variable, [A.sub.it], equals 1 if the individual attrites at time t, conditional on still being a respondent at t - 1.
It relates closely to the issue of whether transitory events affect later attrition, although we cannot be sure of that interpretation because we cannot, by definition, determine whether recent events will persist in the future or not if the individual attrites (and hence whether the events will, in retrospect, be seen to be permanent or transitory shocks).
For our models we pool all observations on individuals 25-64 in original 1968 sample families for all years 1970-89 for which they are observed.(51) We estimate logits for whether the individual attrites in the next period as a function of the four summary measures discussed above defined as of the current period.
where [Y.sub.ct] is an outcome variable of interest for child c at time t; [X.sub.ct] is a vector of the child's observed characteristics at time t (when the child is an adult), [X.sub.p[Tau]] is a vector of parental characteristics at some prior time [Tau] [less than] t, and [[Epsilon].sub.t] is a vector of unobservables.(4) [A.sub.t] is an indicator variable equal to 1 if the child attrites by time t and zero if not, and [Mathematical Expression Omitted] is its latent index.(5) [Z.sub.t] is a vector of observable characteristics (including [X.sub.ct] and [X.sub.p[Tau]]) that are not necessarily independent of [[Epsilon].sub.t].(6)
Second, selection on a right-hand-side variable in (1) does not lead to attrition bias.(18) For example, if [X.sub.p1] is in [Z.sub.t] (namely, a parental characteristic affects both the child's outcome and the likelihood that the child attrites), then this selection by itself causes no bias; bias only occurs if [Y.sub.ct] differs for attritors and non-attritors, holding [X.sub.p1] fixed.
When we say the "child" has attrited, we include the case where the entire parental family attrites before the child has left the household.
A worker is assigned an outcome value of one if he remains in the PSID in a given year and assigned a zero in the year he attrites. A worker could then contribute at most 22 years of data and had to contribute at least one year of data.
They emphasized that ignoring the relation between the decision to attrite and latent heterogeneity could lead to inconsistent estimates of the earnings equation in general and of the treatment effect in particular.
If the decision to attrite comes from unobserved preferences to work (earn income) then labor supply parameters and subsequent deadweight loss calculations are inconsistently estimated if attrition is not included in the structure of the econometric model.