Working Paper: CEPR ID: DP2424
Authors: Francis X. Diebold; Lutz Kilian
Abstract: We propose a measure of predictability based on the ratio of the expected loss of a short-run forecast to the expected loss of a long-run forecast. This predictability measure can be tailored to the forecast horizons of interest, and it allows for general loss functions, univariate or multivariate information sets, and covariance stationary or difference stationary processes. We propose a simple estimator, and we suggest resampling methods for inference. We then provide several macroeconomic applications. First, we illustrate the implementation of predictability measures based on fitted parametric models for several US macroeconomic time series. Second, we analyse the internal propagation mechanism of a standard dynamic macroeconomic model by comparing the predictability of model inputs and model outputs. Third, we use predictability as a metric for assessing the similarity of data simulated from the model and actual data. Finally, we outline several nonparametric extensions of our approach.
Keywords: model evaluation; propagation mechanism; forecasting
JEL Codes: C22; C52; E32
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Predictability can be effectively measured through the proposed ratio of expected losses of forecasts (C53) | Economic series are inherently more predictable than others (E32) |
| Predictability measures can differentiate between the internal propagation mechanisms of economic models and their outputs (C53) | Predictability of model outputs varies significantly from that of inputs (C53) |
| Predictability of model outputs varies significantly from that of inputs (C53) | Strong internal propagation mechanism in the models analyzed (C69) |