Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This paper studies models for forecasting residential investment. It includes standard univariate and multivariate models, and proposes an error correction model (ECM) based on the stock-flow relationship of housing starts, completions and units under construction. All models are estimated on real-time data, and the root mean squared prediction errors (RMSPEs) of the models are compared, along with the RMSPEs of the Survey of Professional Forecasters (SPF) and the Federal Reserve’s Greenbook. For the 1981:Q3 to 2013:Q2 sample, the ECM improves upon the competing models, with the largest improvements on the univariate models coming from the current quarter forecasts and those on the multivariate models coming from the multi-step forecasts. Further, the ECM makes modest improvements to the SPF, and performs comparably to the Greenbook from 1981:Q3 to 2007:Q4. Relative to the current state of professional forecasting, the ECM performs best at multi-step forecast horizons and in volatile economic periods.