When forecasters are optimistic about the future value of an asset, they react to negative news by engaging in optimism shifting, i.e., they are not only likely to remain optimistic but do so by shifting their optimism to a longer forecast horizon. We document this novel fact about expectation formation and belief updating in a large sample of predictions from CAPS, a social-finance platform. We show that optimism shifting can lead to large underperformance. Our cross-sectional and cross-forecaster analysis indicates that optimism shifting can be due to forecasters’ motivation to retain optimistic beliefs about (i) their skill (a confidence channel); (ii) the value of their financial assets (a physical-stakes channel); (iii) the value of their accrued knowledge about an asset (an intangible-stakes channel)

Using natural language processing techniques and millions of anonymous employee reviews, I propose a novel measure of workplace harassment in U.S. firms and investigate its causal effects on innovation and other firm-level outcomes. I exploit the quasi-experimental reduction in workplace harassment caused by changes in non-disclosure agreement (NDA) laws across U.S. states and document a negative impact of workplace harassment on innovation. Firms with previously higher levels of workplace harassment experience a significant increase in their innovation output following these regulations. The documented effect is significantly more pronounced for firms with heterogeneous inventor teams along the gender, race, and ethnicity dimensions. Underlying these effects are improvements in team capital and collaborative dynamics, as the positive changes in the workplace climate lead to significant increases in inventor productivity, as well as retention and entry rates of skilled workers.

This paper shows that early-stage VC-backed startups help develop the next generation of entrepreneurs, a benefit that extends beyond their own growth and successful exits. We follow 1.3 million employees without prior entrepreneurial experience across 22,200 U.S. VC-backed startups from 2000 to 2024. Employees who join within a startup’s first three years are approximately 50% more likely to subsequently found a VC-backed startup themselves than employees who join the same firm later. This outcome is stronger among employees who stay at least a year and those with STEM training, consistent with early-stage experience helping scientifically skilled workers acquire complementary entrepreneurial skills. The relationship is also stronger among employees who experience recessions or organizational changes following the startup’s first VC round, suggesting that navigating uncertainty contributes to entrepreneurial learning. Startups founded by early employees are more likely to receive funding from venture capital investors who backed their former startup employer. These findings suggest that early-stage startups serve as schools for entrepreneurship, developing skills and investor relationships that employees carry into new firms. Focusing only on startups that mature and exit successfully masks this broader benefit to the entrepreneurial ecosystem.

I study how local health shocks affect mortgage default and credit supply. Using variation from the 2010 reformulation of OxyContin, I show that increases in opioid abuse raise mortgage delinquency rates. These effects arise without corresponding declines in local income or employment and are concentrated in high loan-to-value areas, consistent with a collateral channel. I further show that lenders tighten credit supply in more exposed areas, highlighting how localized shocks can amplify financial distress through housing markets.