The perils of approximating fixed-horizon inflation forecasts with fixed-event forecasts

BIS Working Papers  |  No 700  | 
07 February 2018
PDF full text
 |  24 pages



Inflation forecasts are either "fixed event" (eg forecasts made each month for inflation in 2017) or "fixed horizon" (eg forecasts made each month for inflation over the coming 12 months). Fixed-event forecasts are preferable for some purposes, and fixed-horizon forecasts for others. However, fixed-event forecasts are generally more readily available than fixed-horizon ones. A common practice is therefore to convert fixed-event forecasts to fixed-horizon ones. We examine the effect of this conversion on the properties of the forecasts.


We focus on US forecast data that we can use to produce both actual and approximate fixed-horizon inflation forecasts with a 12-month horizon. We then compare the properties of the two. We find that there are quantitatively important differences between them. We also propose a partial fix for the conversion rule to improve the performance of the approximation.


We show that the approximation results in significant errors compared with the actual forecasts, equal to about 10% of the inflation rate. This error is more severe during recessions. In addition, measures of how much forecasters tend to disagree with each other, and how persistent forecasts are, may be misleading when we look at approximate forecasts instead of actual ones. We propose an improvement to the approximation, based on the idea that longer-horizon forecasts are more heavily "anchored" to a fixed number while shorter-horizon forecasts more strongly reflect recent inflation. In practical terms, the improvement means that we put a higher weight on longer-horizon forecasts, and a smaller weight on shorter-horizon forecasts, than the standard conversion rule suggests.



A common practice in studies using inflation forecasts is to approximate fixed-horizon forecasts with fixed-event ones. Here we show that this may be problematic. In a panel of US inflation forecast data that allows us to compare the two, the approximation results in a mean absolute approximation error of around 0.2-0.3 percentage points (around 10% of the level of inflation), and statistically significant differences in both the variances and persistence of the approximate inflation forecasts relative to the actual forecasts. To reduce these problems, we propose an adjustment to the approximation, consistent with a model where longer-horizon forecasts are more heavily "anchored", while shorter-horizon forecasts more closely reflect current inflation levels.

JEL classification: C43, E31

Keywords: fixed-event forecasts, fixed-horizon forecasts, inflation expectations