Cognitive Biases in Macroeconomic Forecasts: – A Pooled Post-Mortem Study of OECD Forecasts

Abstract: This thesis has two purposes: 1) To evaluate the forecast quality of the OECD and 2) to analyze whether the OECD forecasts are afflicted by cognitive bias. Cognitive bias refers to a wide range of basic judgmental errors that are common to humans. Specifically we examine if the OECD forecasters 1) are too optimistic or pessimistic, 2) become more pessimistic the closer they are to a target year, 3) anchor too much on certain reference points, or that they 4) make forecasts that are too positive or negative after consistent change. We choose to look at the forecasters at the OECD during the period 1985 - 2004. Forecasts of five different horizons for GDP and its final domestic demand components represent the database for 23 developed countries. We find support for too optimistic projections of government consumption, while long-term forecasts are more optimistic than short-term forecasts for GDP, government consumption and fixed investment. Furthermore, forecasters make too negative forecasts after positive trends and vice versa. Regarding forecast quality, we find that OECD is accurate, which supports past findings. However, in contrast to past studies, we observe more irrationality in the forecasters.