Many earnings forecasts studies are conducted for the Anglo-Saxon countries, but to date very few have focused on continental Europe. This paperexamines whether there is a relationship between discretionary accounting and the accuracy of long-run forecasts of annual earnings voluntarilydisclosed by managers in the directors' report, which is a mandatory chapter of the annual report in the Netherlands. Long-run forecasts mean forecastsmade at least seven months before year-end. First, we discuss the rationale for management's incentive to improve the accuracy of earnings forecastsvoluntarily disclosed (in the directors' report). We then provide some information with respect to theinstitutional background in the Netherlands. Afterwards we conduct empirical tests to determine whether qualitative earnings forecasts provideadditional information on future performance over and above the information from auto-regressive models i.e. reported EPS on t-1 and EPS on t-1 afterthe correction for discretionary accounting and also whether there is support for the hypothesized relationship between discretionary accounting andqualitative management forecast accuracy. Our empirical results indicate that qualitative earnings forecasts provide a statistically significant betterprediction of the actual earnings than reported EPS on t-l and EPS on t-l after the correction for discretionary accounting and that there is arelationship between high pre-discretionary forecast errors and the adoption of discretionary accounting. After adopting discretionary accounting, theforecasterrors are reduced. Overall, the study shows that management has an incentive to engage in discretionary accounting to improve the accuracy of thedisclosed qualitative earnings forecasts in their directors' reports. However, one should bear in mind that `either discretionary accounting is used toimprove the accuracy of earnings forecasts disclosed in the directors reports or managers' earnings forecasts are sometimes disclosed in anticipation of