The purpose of this paper is to examine the size and nature of the errors in GDP estimates in the G-7 countnes and Spain from 1968 to 1995. These GDP estimates are published by the Organization for Economic Cooperation and Development in the OECD Economic Outlook. Two altemative hypotheses about why errors arise are distinguished and tested. The first is that the preliminary announcement is simply the true variable measured with some degree of error. In this case, the preliminary announcement is an unconditionally unbiased but irrational forecast of the true value. That is, the error is correlated with known variables; in particular, it is correlated with the preliminary announcement itself. The second is that errors arise because the eariy GDP estimates are a rational forecast of the true variable. In this case, the error is a rational forecast error and is therefore uncorrelated with known information.

Keywords: forecast, economic growth, prediction interval, forecast error, OECD.