Victor Lyonnet

Traditional and shadow banks interacted in similar ways in the 2007 and COVID-19 crises, when both assets and liabilities flew out of shadow banks and into traditional banks. We explain these facts in a model of the coexistence of traditional and shadow banks in which liabilities and assets flow from the former to the latter in good times, avoiding regulation, and move the other way in a crisis, alleviating fire sales. The model sheds light on how regulations for traditional banks have (unintended) consequences on the shadow banking sector.

Firms enter new sectors by either building on their resources or buying existing companies. Using French administrative data, we propose a measure of human capital distance between a firm and a sector of entry. Using a shift-share instrument, we show that firms build in close sectors and buy in distant sectors in terms of human capital distance. Firms build by hiring new workers, which becomes increasingly costly in distant sectors as it requires not only hiring more workers but also having more organizational capital to integrate these workers. Hence, firms buy in distant sectors to acquire already operational human capital.

What comes to mind when thinking about a successful entrepreneur? Belief formation models suggest that what comes to mind is an oversimplified picture of the characteristics of successful entrepreneurs—that is, stereotypes about successful entrepreneurs. Using French administrative data on 48,767 new firms, we show that some characteristics are stereotypical of success and have distributions that can generate miscalibrated beliefs. To illustrate how stereotypical thinking can lead to biased assessments, we report the discrepancies between the implied fraction of successful entrepreneurs under Bayesian versus stereotypical thinking for several stereotypes. We discuss the consequences of stereotyping for venture capital allocation.

We study how retail savings products can share market risk across investor cohorts, thereby completing financial markets. Financial intermediaries smooth returns by varying reserves, which are passed on between successive investor cohorts, thereby redistributing wealth across cohorts. Using data on euro contracts sold by life insurers in France, we estimate this redistribution to be large: 0.8% of GDP. We develop and provide evidence for a model in which low investor sophistication, while leading to individually suboptimal decisions, improves risk sharing by allowing intercohort risk sharing.

We examine the link between exporters' currency choices and their use of financial hedging instruments. Large firms are more likely to use hedging instruments, especially those pricing in a foreign currency. We provide suggestive evidence that access to hedging instruments increases the probability of pricing in a foreign currency. A model of invoicing currency choice augmented with hedging can rationalize these facts. In the model, large firms that would have chosen to price in their own currency in the absence of hedging instruments can decide to set prices in a foreign currency if they have access to such instruments.

We study early-stage venture capitalists' (VCs) decisions through the lens of a predictive model of venture success. Using French administrative data on VC-backed and non-VC-backed companies, we find that VCs invest in some companies that perform predictably poorly and pass on others that perform predictably well. VCs tend to select entrepreneurs whose features are representative of success—such as being male, graduates of elite schools, and based in Paris. Although entrepreneurs with these characteristics exhibit higher success rates, VCs exaggerate the importance of these features relative to their impact on performance, contributing to the narrowness of the VC industry.

This paper documents novel stylized facts linking the decline in new business formation to the rise of superstar firms using comprehensive French administrative data. Industries with larger increases in superstar firms' market share experience more pronounced decreases in new business creation. Rising concentration discourages low-ability, but not high-ability, entrepreneurs from starting businesses. This results in higher average firm quality as measured by a higher fraction of entrepreneurs who are highly educated, former executives, or serial entrepreneurs. Our findings help reconcile seemingly contradictory evidence in the literature and align with theories emphasizing technological changes that benefit the most productive firms while raising entry barriers.

We analyze 5,230 expert network calls using large language models (LLMs) to study how early-stage investors conduct due diligence. Applying a novel LLM-based topic modeling approach combined with SHAP analysis, we find that firms receiving expert calls have 44% higher odds of securing investment in the next quarter. The predictive content of these calls varies systematically with both discussion topics and firm characteristics: Positive discussions about technology integration and customer acquisition further increase deal odds by 31% and 15%, respectively. These effects are concentrated among younger firms, suggesting that expert validation can at least partially substitute for traditional financial metrics. However, while positive signals predict investment decisions, negative assessments on risk management are associated with 0.2 standard deviation lower long-term firm performance. This divergence between what predicts deals versus ultimate success is consistent with investors optimizing for power-law returns rather than success rates.