Reducing uncertainty in short‐term projections: Linkage of monthly and quarterly models

Abstract

This paper shows how monthly data and forecasts can be used in a systematic way to improve the predictive accuracy of a quarterly macroeconometric model. The problem is formulated as a model pooling procedure (equivalent to non-recursive Kalman filtering) where a baseline quarterly model forecast is modified through 'add-factors' or 'constant adjustments'. The procedure 'automatically' constructs these adjustments in a covariance-minimizing fashion to reflect the revised expectation of the quarterly model's forecast errors, conditional on the monthly information set. Results obtained using Federal Reserve Board models indicate the potential for significant reduction in forecast error variance through application of these procedures. K E Y \\ORUSForecasting Forecast pooling Macro-econometric models Kalman filteringThe typical process of generating a model-based forecast of economic activity involves more than mechanical simulation of a macro-econometric model forward in time. A model-based forecast is conditioned on values supplied for exogenous variables and on assumptions concerning the time paths of policy instruments. In addition, forecasters typically use information outside the model's data base-often from myriad and diverse sources-in formulating their final projection. This information may consist of timely high frequency measurements or data revisions, anticipated events not captured by the model structure, and e\en the predictions of other forecasters. The use of such information to alter a model's 5olution is often termed 'judgemental modification'.In surveying published descriptions of how private forecasters use their models, McNees reported that they attribute anywhere from 20 to 50 percent of the final forecast to the process of judgemental adjustments (1981, pp. 6-7). At the Federal Reserve Board (FRB), the staff outlook is anchored about a consensus projection of the most probable outcome associated with given settings of the policy instruments. As described in Kalchbrenner and Tinsley (1977, pp. 40-44), the consensus projection involves a considerable judgemental component.'' The opinions express'ed are rhose of the authors and do nor necessai-ily rellecr the view, of the Board of Governois of the Federal Reserve System or other mernherc of iis $!aft'.

Rational Partisan Theory, Uncertainty, and Spatial Voting: Evidence for the Bank of England's MPC

“…4 It is straightforward to cast this problem of optimally combining data of different frequencies as a filtering problem. See, for example, Corrado and Green (1988). 5 It is being assumed here that the policy horizon corresponds to two years (or so) into the future (see King, 2002).…”

Prediction of the Netherlands' money stock

“…The additional information, which makes the use of combined forecasts so fruitful, may relate either to different forecasting procedures (or models) using the same data or to the use of different sources of information, such as the opinions of experts (see Figlewski andUrich, 1983, andAshton andAshton, 1985; see also Winkler, 1984). A mixture of these two aspects in the combination of forecasts can be found in the forecasting procedure of the Federal Reserve Board (FRB) where a combination is made both of forecasts of different models and of forecasts at different aggregation levels (monthly and quarterly data, aggregates and components) (see Fuhrer andHaltmaier, 1988, Corrado andGreene, 1988). It must be noted that the forecasting procedure of the FRB specifically aims at utilizing scattered new information on economic developments in the best possible manner.…”

Creating high‐frequency national accounts with state‐space modelling: a Monte Carlo experiment

“…The idea of using the state-space framework to construct an optimal estimate of economic indicators has been adopted by Watson (1989, 1991) and Garratt and Hall (1996). Corrado and Greene (1988) demonstrated the applicability of Kalman filtering to improve the predicative accuracy of a quarterly model. Corrado and Haltmaier (1988) described an overview of the underlying structure of the model linkage project at the Federal Reserve Board.…”