Working Paper: CEPR ID: DP7426
Authors: Kevin Lee; Nils Olekalns; Kalvinder K. Shields
Abstract: A canonical model is described which reflects the real-time informational context of decision-making. Comparisons are drawn with ?conventional? models that incorrectly omit market-informed insights on future macroeconomic conditions and inappropriately incorporate information that was not available at the time. It is argued that conventional models are misspecified and misinterpret news but that these deficiencies will not be exposed either by diagnostic tests applied to the conventional models or by typical impulse response analyses. This is demonstrated through an analysis of quarterly US data 1968q4-2008q4. However, estimated real-time models considerably improve out-ofsample forecasting performance, provide more accurate ?nowcasts? of the current state of the macroeconomy and provide more timely indicators of the business cycle. The point is illustrated through an analysis of the US recessions of 1990q3-1991q2 and 2001q1-2001q4 and the most recent experiences of 2008.
Keywords: business cycles; nowcasting; real-time data; structural modelling
JEL Codes: E52; E58
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Mis-specification (C50) | Not exposed through standard diagnostics (Y50) |
| Conventional models (C59) | Misinterpretation of macroeconomic conditions (E66) |
| Conventional models (C59) | Mis-specification (C50) |
| Real-time models (C32) | Enhance out-of-sample forecasting accuracy (C53) |
| Real-time models (C32) | Timely indicators of macroeconomic state (E32) |
| Real-time models (C32) | Better nowcasts and forecasts (C53) |
| Direct measures of expectations (D84) | Improvements in forecasting performance (C53) |
| Real-time data (Y10) | Understanding business cycle dynamics (E32) |
| Real-time data (Y10) | Accurately predicting business cycle dynamics (E37) |