We characterize an information cliff in the stock market: the supply of information on aggregate cash flows drops precipitously beyond a one-year horizon, and so does analyst forecast accuracy. We use a generalized state-space model to explore the implications for expected cash-flow growth and expected returns. Identifying the state-space dimensionality is the only necessary step for sharpening the model structure. Once done, the information cliff has a direct mathematical representation: the expected cash-flow component of the state space must be non-persistent. Furthermore, the expected market returns only depend on the valuation wedge between the total market and one-year dividend strip.

We study when rapid asset price appreciation becomes unsustainable. Analyzing U.S. stock market data from 1926 to 2022, we document that rapid price increases in individual stocks systematically predict subsequent crashes and negative returns. When returns exceed 200% over a three-month period, stocks experience an average decline of 29% over two years, with a 55% probability of crashing by more than 50%. By systematically varying both return thresholds and formation periods across over one thousand parameter combinations, we demonstrate that the likelihood and magnitude of subsequent crashes are primarily determined by the speed of the price increase, measured as average daily cumulative excess return. This finding documents a robust relationship consistent with a “speed limit” in asset prices—if returns exceed the speed limit, subsequent two-year crash probability rises sharply. Notably, this pattern exhibits a striking asymmetry: rapid price declines show no systematic tendency to reverse over the subsequent two-year horizon, suggesting potentially different mechanisms drive price increases versus decreases. Our findings raise questions about weak-form efficiency at the individual stock level and introduce “speed limits” as a simple yet powerful predictor of unsustainable valuations and potential bubble-like episodes.

We identify a common misconception that expected future changes in short-term interest rates predict corresponding future changes in long-term interest rates. People forecast similar shapes for the paths of short and long rates over the next four quarters. This is a mistake because long rates should already incorporate public information about future short rates and do not positively comove with expected changes in short rates. We hypothesize that people group short- and long-term interest rates into the coarse category of “interest rates,” leading to overestimation of their comovement. We show that this categorical thinking persists even among professional forecasters and distorts the real behavior of borrowers and investors. Expectations of rising short rates drive households and firms to rush to lock in long-term debt before further increases in long rates, reducing the effectiveness of forward guidance in monetary policy. Investors sell long-term bonds because they anticipate future increases in long rates. The result- ing increase in supply and decrease in demand for long-term debt cause long rates to overreact to expected changes in short rates, and can help explain the excess volatility puzzle in long rates.

I document a robust pattern in how Treasury market participants’ yield curve expectations respond to new information: forecasts for short-term rates underreact to news while forecasts for long-term rates overreact. I propose a new explanation of this based on ``autocorrelation averaging,’’ whereby, due to limited processing capacity, forecasters’ estimate of the autocorrelation of a given process is biased toward the average autocorrelation of all related processes. Consistent with this view, forecasters overestimate the autocorrelation of the less persistent term-premium component of interest rates and underestimate the autocorrelation of the more persistent short-rate component; a calibrated model quantitatively matches the documented pattern of misreaction. Moreover, banks’ allocations to Treasuries vary positively with their expectations of bond returns and misreaction proxies can strongly predict future short- and long-term bond returns, respectively.

The ratio of long- to short-term dividend prices, “price ratio” ($p{r}_t$), predicts annual market return with an out-of-sample $R^2$ of 19%, subsuming the predictive power of price-dividend ratio ($p{d}_t$). After controlling for $p{r}_t$, $p{d}_t$ predicts dividend growth with an out-of-sample $R^2$ of 30%. Our results hold outside the U.S. An exponential-affine model shows that the key to our findings is the (lack of) persistence of expected dividend growth. We find the expected return is countercyclical and responds strongly to monetary policy shocks. As implied by ICAPM, shocks to $p{r}_t$, the expected-return proxy, are priced in the cross-section.

Delegation bears an intrinsic form of uncertainty. Investors hire managers for their superior models of asset markets, but delegation outcome is uncertain precisely because managers’ model is unknown to investors. We model investors’ delegation decision as a trade-off between asset return uncertainty and delegation uncertainty. Our theory explains several puzzles on fund performances. It also delivers asset pricing implications supported by our empirical analysis: (1) because investors partially delegate and hedge against delegation uncertainty, CAPM alpha arises; (2) the cross-section dispersion of alpha increases in uncertainty; (3) managers bet on alpha, engaging in factor timing, but factors’ alpha is immune to the rise of their arbitrage capital - when investors delegate more, delegation hedging becomes stronger. Finally, we offer a novel approach to extract model uncertainty from asset returns, delegation, and survey expectations.