HOMBERT Johan

The thesis contains three essays. In the first essay, I investigate whether financial experts benefit from more machine-readable data in producing information in asset management. Exploiting an exogenous regulatory shock that makes firms' filings more machine-readable, I find that institutions with more financial experts have greater performance improvement than institutions with fewer financial experts, suggesting that financial experts benefit from more machine-readable data. This result allows for an assessment of the likelihood that algorithms will replace highly trained financial practitioners. In the second essay, I study the rationale and implications of the recent MiFID II regulation in Europe, which has made delegated asset managers' spending on sell-side analyst research more transparent to their clients. We show that transparency decreases the use of sell-side research but stimulates more buy-side research activity, which is consistent with the empirical results. Our model has additional predictions on managerial performance, liquidity, and social welfare. In the third essay, I study brokers in the private placement markets, who intermediate about 20% of the capital raised by non-financial firms in this market. I find that projects intermediated by brokers with better reputations are more likely to be fully funded. Contrary to existing theories about underwriters, projects sold through brokers are on average less likely to be fully funded and most issuers prefer direct sales. A model with both search frictions and asymmetric information suggests that these non-regularities may be due to the fact that the certification role of brokers is limited by competition between intermediated and direct sales. The model also explains some of the non-intuitive patterns of commission fees in the data. These results contribute to a better understanding of private equity markets and intermediaries in other financial markets.