Insight into the investment behaviour of firms is central in understanding economic dynamics. A critical question, however, is whether firms provide sufficiently reliable data to enable them to make plausible forecasts at the meso (regional or sectoral) level. This paper analyses Dutch investment forecasts at different levels of aggregation. The central research question is whether entrepreneurs, individually or as a group, make systematic errors in their investment forecasts. A statistical test reveals that investment forecasts are not biased at the aggregated (regional and sectoral) level. At the micro level, however, there is a significant bias. Hence, using aggregated (regional and sectoral) data to test the lack of bias (unbiasedness) of forecasts may lead to the wrong conclusions. Moreover, aggregated investment forecasts may then be an inappropriate source for policy recommendations, despite their seemingly high reliability. This finding may in principle be valid for many European countries, since data collection on investment is organized in similar ways throughout Europe.