We put forward a model in which analysts are uncertain about a firm's earnings process. Faced with the possibility of using a misspecified model, analysts issue forecasts that are robust to model misspecification. We estimate that this mechanism explains approximately 60% of the autocorrelation...

Information and learning environments shape the dynamics of our beliefs that determine asset prices. When an agent jointly learns about consumption and dividend, her beliefs on them inter-temporally co-vary with each other, decoupled from their true underlying relationship. Such...

Agents are generally uncertain about multiple, and possibly time-varying, structural parameters that drive consumption and financial payoffs but learn through noisy correlated signals, such as aggregate or macroeconomic news. We find that dynamic learning of multivariate time-varying parameters...

For optimal asset allocation, mean-variance investors must learn about the joint dynamics of new and existing asset classes, not only their profitability. Bitcoin's digital gold narrative provides a unique laboratory to test this hypothesis. We find that a decrease in investors' estimate on...

When a new asset keeps changing its narrative, investors find difficulty in classifying and understanding the new asset. Rational investors therefore face unprecedented uncertainty and learn about the joint dynamics to optimize their portfolio accordingly. Bitcoin's "digital gold" narrative, as...

Quantile and least-absolute deviations (LAD) methods are popular robust statistical methods but have not generally been applied to state filtering and sequential parameter learning. This paper introduces robust state space models whose error structure coincides with quantile estimation...

Out-of-sample R2-hacking problems can arise even without multiple testing if a researcher constructs a prediction model using the intuition derived from empirical properties that appear only in the test sample. We provide a machine-learning solution for this problem in the context of robust...

Out-of-sample tests are subject to look-ahead bias when a forecaster constructs a model using an intuition derived from empirical patterns in the test sample. Even if model parameters are estimated without the test sample, information from it affects a forecaster's model choice. Since such...

The out-of-sample R2 is designed to measure forecasting performance without look-ahead bias. However, researchers can hack this performance metric even without multiple tests by constructing a prediction model using the intuition derived from empirical properties that appear only in the test...