The current state of affairs concerning experts adjusting model-based forecasts is summarized in the previous two chapters. Chapter 3 showed that experts adjust model-based forecasts very frequently, whether the data concern business or economic variables, and perhaps more often than is really necessary, at least from an econometric perspective outlined in Chapter 2. Such adjustments by experts can be large in size, and this might perhaps be associated with biased model-based forecasts. Indeed, even the elaborate CPB macroeconomic model seems to deliver biased forecasts for various important macroeconomic variables. It was also found that expert adjustment is more often upwards than downwards, which may be a sign of (over-)optimism, or perhaps of alternative loss functions, as will be discussed in the present chapter. In individual cases it can be that more upwards than downwards adjustment may be a useful strategy, but on average, across many cases, there is no need (at least econometrically) to have more upward than downward adjustments.

The empirical results summarized in Chapter 4 showed that the apparent downside of the tendency to frequently modify model forecasts seems to be that the scrutinized expert-adjusted forecasts currently available are not very much better than the model forecasts, at least on average. Again, in individual cases it can occur that expert adjustment leads to substantial forecast improvement, but on average current insights suggest that expert adjustment as it is carried out now is not very useful. It seems in fact that when expert-adjusted forecasts are better they are only a little better, and when they are worse, they are much worse. For the CPB forecasts we learn that when experts really do have expertise concerning a certain variable, their added value can lead to substantial improvements. On the other hand, for the pharmaceutical sales forecasts we learn that most experts seem to be inclined to deviate too much from the model forecasts, which on average leads to no improvement.