UC Berkeley

What explains the cross-sectional variation of expected returns? This dissertation contains two essays that study this question both theoretically and empirically for two popular puzzles in the asset pricing literature.

In the first chapter of the dissertation, I study the asset pricing implications of learning-by-doing in a partial equilibrium model where firms optimally choose to adopt the newest technology. Firms are heterogeneous in the sense that they have different learning curves. Adopting the newest technology is costly, however, because the firms forgo the experience they accumulated in the past. The model implies that firms with (1) obsolete technology, (2) little accumulated experience, (3) low forgetting rate, or (4) high learning rate are more likely to adopt earlier, which I label as 'early innovators'. I show that these firms load more on growth options as opposed to assets-in-place, are more exposed to technological shocks, and earn a lower risk premium. The model can match the magnitude of the value premium as well as the size premium.

In the second chapter of the dissertation, I document that idiosyncratic volatility and future returns are not simply negatively related. Past performance of the market predicts whether high or low idiosyncratic volatility stocks generate positive returns. A signed idiosyncratic volatility (SIV) factor, which is long high idiosyncratic volatility and short low idiosyncratic volatility following bull markets and vice versa following bear markets, produces significant positive risk-adjusted returns. A model with extrapolative agents and market segmentation can capture these facts.