To meet the demand for remanufactured products, accurate forecasting of used product returns is needed. The quantity and timing of the returns of used products for remanufacturing depend on the quantity of new products sold in previous periods. Conventional time series forecasting techniques are not able to capture the relationship between past sales and future returns and hence cannot be used to predict used product returns for remanufacturing. Distributed lag models (DLMs) which can model the dependence of future returns on past sales has been proposed in previous studies for forecasting used product returns. However, the choice of an appropriate lag function for the DLM and estimation of the parameters of the lag function were the main challenges in previous studies. In this research, a DLM with a negative binomial lag function is proposed for forecasting used product returns, and Bayesian Markov Chain Monte Carlo simulation is used to estimate of the parameters of the lag function. To validate the forecasting model, the mean absolute error (MAE) and the mean absolute percent error (MAPE) are computed. Numerical experiments were conducted to illustrate the proposed forecasting model and the parameter estimation approach. The results showed the proposed forecasting model predicts used product returns with good accuracy.