Abstract
Causes and consequences of teacher moonlighting are investigated, using a nationwide survey of US teachers conducted in the mid-1980s. A model of teacher time allocation is estimated by maximum likelihood methods. Moonlighting is shown to be highly insensitive to levels of teacher pay, even when controlling for variations in costs of living and local labor market conditions. However, teachers who moonlight do not appear to shortchange their students when it comes to preparing lessons, grading papers or assigning homework.