Abstract and Keywords
This article, which reviews the recent literature on forecasting when there are unanticipated structural breaks, or parameter nonconstancies, and the forecasting model is misspecified in unknown ways, is organized as follows. Section 2 briefly reviews approaches to assessing forecast accuracy. Section 3 describes a general framework for analyzing various aspects of the forecasting problem which may contribute to forecast error: at the most basic level, these are that the model may not capture important features of the data generating process (DGP); the DGP may itself be nonconstant; and the form and unknown coefficients of a forecasting model will need to be specified, selected, and estimated from the available data. Section 4 focuses on the forecasting models that have been popular in time-series econometric modeling in the last quarter of a century, and fleshes out the possible sources of forecast error when there are various forms of structural break. Section 5 and 6 allow that macrovariables are usually aggregates, but may be available at a higher frequency, so it is considered whether the effects of breaks on forecasts can be alleviated by forecasting the disaggregated components or by time disaggregation. Section 7 reviews recent empirical research on forecasting that averages over many different forecasts to try and obtain more accurate forecasts in the presence of general forms of instability and model misspecification. Section 8 provides an analysis based on location shifts coming from data revisions. Finally, Section 9 offers some concluding remarks.
Access to the complete content on Oxford Handbooks Online requires a subscription or purchase. Public users are able to search the site and view the abstracts and keywords for each book and chapter without a subscription.
If you have purchased a print title that contains an access token, please see the token for information about how to register your code.