Researchers from the University of Chicago propose a novel reinforcement learning approach to extract high-frequency aggregate growth expectations from asset prices. While much expectations-based research in macroeconomics and finance relies on low-frequency surveys, the multitude of events that pass between survey dates renders identification of causal effects on expectations difficult.
Their method allows the construction of a daily time-series of the cross-sectional mean of a panel of GDP growth forecasts. The high-frequency nature of the series enables clean identification in event studies. In particular, they use their estimated daily growth expectations series to test the “Fed information effect” and find little evidence to support its existence.
Also, the method can be used to construct firm-specific cash flow expectations at a daily frequency. Instead of using value-weighted equity and bond returns, the firm’s equity returns and corporate bond returns can be used. They can also use quarterly expectations from analyst forecasts to conduct the same exercise and obtain daily expectations.
Testing theories of expectations formation is an application of the framework. For example, testing the hypothesis that agents update their expectations about growth after observing commodity prices. One way to empirically test this hypothesis is to include commodity returns in the state vector and fit the optimal policy for updating growth expectations. The coefficient on commodity returns in the optimal policy reflects if an agent would find it optimal to use commodity returns in updating his expectation of growth.