Working Paper: CEPR ID: DP10815
Authors: Xiaoming Cai; Wouter Den Haan; Jonathan Pinder
Abstract: Should an unexpected change in real GNP of x% lead to an x% change in the forecasts of future GNP? The answer could be no even if GNP is a random walk. We show that US economic downturns often go together with predictable short-term recoveries and with changes in long-term GNP forecasts that are substantially smaller than the initial drop. But not always! Essential for our results is that GNP forecasts are not based on a univariate time series model, which is not uncommon. Our alternative forecasts are based on a simple multivariate representation of GNP?s expenditure components.
Keywords: Business cycles; Forecasting; Unit root
JEL Codes: C53; E32; E37
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| unexpected changes in real GNP (F69) | future forecasts (F17) |
| initial shocks (E32) | long-term GNP forecasts (H68) |
| economic downturns (F44) | predictable short-term recoveries (E32) |
| univariate model (C29) | permanent effects of unexpected changes in GNP (F69) |
| multivariate model (C39) | captures faster-than-normal growth following downturns (E32) |
| type of shock (D80) | recovery patterns (C22) |