Using probit models of downturn risk to calibrate GDP fan charts for New Zealand

Macroeconomic forecasters typically forecast fewer recessions than the number experienced, which means economic growth tends to be over-predicted on average. Consequently, forecast errors are not normally distributed, making it difficult to convey the uncertainty and risks based on the historical forecast track record. To characterise this risk, recent OECD work constructed fan charts parameterised on historical forecast errors and the probability of a future downturn estimated from a probit model comprising a range of potential macroeconomic and financial early warning indicators. As the probability of a downturn increases the associated fan chart is wider, reflecting increased uncertainty, and more skewed to the downside, reflecting greater downside risks. This paper applies this methodology to New Zealand; although one important difference compared to other OECD economies is that the time span of macroeconomic data without major structural change is significantly shorter. Forecast errors for GDP by the OECD, Reserve Bank of New Zealand and New Zealand Treasury all appear to be non-normally distributed. Fan charts for GDP forecasts from the mid-year 2018 OECD Economic Outlook are symmetric due to the low probability of a downturn. Fan charts estimated for the period preceding the global financial crisis using currently-available data have a downwards skew. However, those estimated using data only available in the lead up to the crisis have many insignificant coefficients, likely due to the structural changes that have occurred in the New Zealand economy since the 1980s.