Twice a year, the National Bank of Belgium (NBB) conducts large-scale macroeconomic projection exercises in close collaboration with the European Central Bank (ECB) and the other national central banks in the Eurosystem. The projections, which are compiled independently by the central banks, and the associated economic analysis are produced as a source of information for the euro area’s monetary policy. So, the Governing Council usually takes note of them in its decision-making. The NBB makes its macroeconomic projections on the basis of analytical econometric models and expert assessments. The projections are not just a quantitative exercise as they are backed up by a consistent economic scenario targeting a wide range of macroeconomic variables.

This article analyses the track record and accuracy of the macroeconomic projections for the Belgian economy since 2001 and focuses mainly on forecasts for GDP and the components of aggregate demand. The results for inflation and for employment growth are also covered briefly. As we have a reduced sample only running from 2001 to 2017, the average error is still strongly influenced by certain outliers, and in this case, by observations from the great recession.

The accuracy of the projections needs to be analysed in relative terms because it is hard to establish any absolute scale of magnitude. In other words, it is worth comparing the NBB’s projections with those from other institutions, comparing forecast horizons and comparing change over time. In comparison with other institutions, when considering projections made for more or less the same period of time, the NBB’s tend to be at least as accurate. At national level, the accuracy of the NBB’s projections seems to be in line with that of the Federal Planning Bureau. Compared with the projections from the European Commission, the OECD and the IMF, the NBB’s are slightly better for forecasts with a shorter horizon and comparable if a relatively long horizon is considered. As regards the forecast horizon, it is notable that the forecast errors generally tend to creep up as the horizon gets longer and, for longer horizons in particular, it is still difficult to detect the turning points in the business cycle. As regards change over time, the period after the great recession appears to have more accurate projections from the NBB than the pre-crisis period. Moreover, since the introduction of specific nowcasting models, the NBB’s very short-term forecasts seem to have improved. That said, as the more recent period has been less volatile, it is still hard to draw any definitive conclusions.

The NBB’s forecasts for Belgian GDP naturally follow a consistent economic scenario both with regard to the various macroeconomic aggregates and the different projections coming from other Eurosystem countries. As a result, the NBB’s projections are conditional on some assumptions that are set jointly with the other Eurosystem central banks. It turns out that the forecast errors over the period under consideration seem to be mainly attributed to errors connected with these common assumptions and in particular to those on Belgium’s export markets. Another factor worth noting is that the GDP forecast errors tend to be smaller than those for its demand components, implying that the errors made in forecasting certain demand components tend to be offset by errors of an opposite sign made in forecasting other components. For example, forecasts for private consumption are characterised by a relatively high average error and a tendency to over-estimate unlike business-investment-related forecasts.

The NBB’s projections for inflation seem to fare better than those from international institutions and especially when shorter forecasting horizons are analysed. As far as labour market projections are concerned, they are generally no less accurate than those from international organisations. For the more recent period, employment growth tends to be underestimated, potentially due to the impact of structural reforms and labour cost moderation policies.