Confidence Risk and Asset Prices

Abstract

Asset price movements in many cases seem de-linked from aggregate economic fundamentals. Forexample, RaviBansal andIvanShaliastovich (2008a) show that frequent large moves in asset prices, i.e. jumps, on average are not correlated with movements in macro-variables (see Table 1 below). Motivated by this, we present a general equilibrium model in which variation in investor confidence about expected growth determines risk premia and hence asset prices. This confidence risk channel can account for (i) the lack of connection between large asset-price moves and macro-variables such as consumption, (ii)large declinesinassetprices, thatis, the left tail of the return distribution, and (iii) observed predictability of equity returns and consumption growth by the price to dividend ratio. In essence, we present a model in which behaviorally motivated shifts in expectations play an important role for the asset prices. Our economy set-up follows a standard longrun risks specification of Ravi Bansal and Amir Yaron (2004), and features Gaussian consumption growth process with time-varying expected growth and volatility; there are no large moves orjumpsintheunderlyingconsumptionanddividenddynamics. Expectedgrowth isnotdirectly observable, and investors learn about it using the cross-section of signals. The time-varying cross-sectional varianceof thesignals determines the quality of the information, and therefore the confidence that investors place in their growth forecast. In the long-run risks framework, the fluctuations in confidence risk determines risk premia and asset prices. We model investors as being recency-biased in their expectation formation, that is, they overweigh recent observations as in Werner De Bondt and Richard Thaler (1990). This is important, as in the standard Kalman-Filter based expectation formation, periods of low information quality get down-weighted, which diminishes the role of the confidence risk channel.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 89 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.