Studies of injury morbidity often rely on self-reported survey data. In designing these surveys, researchers must chose between a shorter recall period to minimize recall bias and a longer period to maximize the precision of rate estimates. Using data from the Health and Retirement Study, which employed a recall period of 1 year, we examined the effect of the recall period on rates of occupational injuries among older workers as well as upon rate ratios of these injuries for nine risk factors. We fit a stochastic model to the occupational injury rates as a function of time before the interview and used this model to estimate what the injury rates would have been had we used a 4-week recall period. The adjusted occupational injury rate of 5.9 injuries per 100 workers per year was 36% higher than the rate based on a 1-year recall period. Adjustment for recall period had much less effect on rate ratios, which typically varied by < 10%. Our work suggests that self-reported surveys with longer recall periods may be used to estimate occupational injury rates and also may be useful in studying the associations between occupational injuries and a variety of risk factors.