Roohollah Zare - Department of Economics, Beyza branch, Islamic Azad University, Beyza, Iran

The present paper examines the role of stock market returns in forecasting real output growth for 1, 2, 4 and 8 quarter ahead forecast horizons at Iran. The forecasting experiment in the present paper is based on autoregressive distributed lags models. Recursively estimating the models, we generate h-step-ahead forecasts for the periods of 1, 2, 4, and 8 quarters and calculate the Mean Squared Forecast Errors (MSFE) to assess the forecasting performance of the various models. To test the statistical differences in forecasting accuracy we use the out-of-sample-F (OOS-F) statistics recently proposed by McCracken (2007). By employing a quarterly dataset spanning from 1992Q1 to 2007Q4 and recursively estimating the models from 2008Q1 onwards as the out-of-sample period, the results show that stock returns do contain useful information in predicting future output growth with a limited predictive ability only in the short-run, especially at less than 1 quarter horizon.

forecast evaluation, stock returns, real output growth, Iran