We propose a new test for the presence of job-market signaling in the sense of Spence (1973), based on an equation in which log-wages are explained by two endogenous variables: the student’s degree and the student’s time to degree, not simply by years of education. Log-wages are regressed on a measure of education, which is a position on a scale of certificates and degrees, and a measure of the student delay, defined as the difference between the individual’s school-leaving age and the average school-leaving age of students holding the same certificate or degree. We use past school-opening instruments, and distance-to-the-nearest-college, also measured in the past, when students were entering grade 6, to identify the parameters. We find a robust, significant and negative impact of the delay variable on wages, averaged over the first five years of career. A year of delay causes a 9% decrease of the student’s wage. The only reasonable explanation for this effect is the fact that longer delays signal unobserved characteristics with a negative productivity value. We finally estimate a nonlinear model of education choices and cannot reject the assumption that the data is generated by a job-market signaling equilibrium.