We investigate the strategic role of a recommender who cares about accuracy and whose recommendations influence product quality. In the presence of such feedback effects, recommendations have a self-fulling property: the recommendation agent can select any firm that will end up being the firm with the best quality. This produces important inefficiencies that include (a) a lack of incentive to acquire valuable information, (b) a status quo bias, and (c) the avoidance of risky innovations. Monetary payments from firms may work in mitigating these inefficiencies, while competition between recommenders and monetary payments from consumers are ineffective.

Schumpeter (1939) claims that recessions are periods of “creative destruction,” concentrating innovation that is useful for the long-term growth of the economy. However previous research finds that standard measures of firms’ innovation, such as R&D expenditures or raw patent counts, concentrate in booms. We argue that these measures do not capture shifts in firms’ innovative search strategies. We contemplate firms’ choice between exploration versus exploitation over the business cycle and find evidence with more nuanced measures of patent characteristics that firms shift toward exploration during contractions and exploitation during expansions, with a stronger effect for firms in more cyclical industries.

We construct a neural network algorithm that generates price predictions for art at auction, relying on both visual and nonvisual object characteristics. We find that higher automated valuations relative to auction house presale estimates are associated with substantially higher price‐to‐estimate ratios and lower buy‐in rates, pointing to estimates' informational inefficiency. The relative contribution of machine learning is higher for artists with less dispersed and lower average prices. Furthermore, we show that auctioneers' prediction errors are persistent both at the artist and at the auction house level, and hence directly predictable themselves using information on past errors.