Looking at venture capitalists in the rearview mirror

The digital age has created massive amount of data that continue to grow exponentially. The International Data Corporation estimates that the world generates more data every two days than the entire humanity did between the dawn of time to the year 2003. This rapid evolution in data availability has been accompanied by advances in statistical techniques such as machine learning and Artificial Intelligence (AI), i.e. technology designed to identify statistical patterns in large datasets, outperform humans in many forecasting (or repetitive) tasks. This big data revolution has been reshaping the financial industry: AI techniques are being increasingly deployed in finance, in areas such as asset management, algorithmic trading, corporate banking, credit underwriting or blockchain-based finance. The deployment of AI technologies in finance is expected to drive competitive advantages for financial firms – for example by reducing their costs and increasing their productivity. In turn, these competitive advantages can benefit financial consumers by providing increased quality and personalized products.

In a recent article Ilustrate the latter. I examine the early adoption of new technologies by banks lending to small businesses, using the staggered deployment of high-speed internet in France from 1999 to 2007 as an exogenous shock to technology diffusion. I conclude that faster and easier data exchange between entrepreneurs seeking credit and small banks (lower search costs and improved customer–lender interaction) reduced the cost of borrowing faced by these entrepreneurs. In addition, the results show that banks are responding by changing their business practices, lending to small businesses located farther away, outside their local market, but without deteriorating the quality of their loan portfolio. This echoes recent work on fintech lending by Berg, Fuster and Puri (2022). Fintech lending is defined as the use of technology to provide lending products, with two main flavors: First, technology can be used to improve the customer–lender interaction (for example, with a fully online application process), giving rise to a better user experience, faster processing times, and lower operational costs; and second, it can be used in borrower screening or monitoring, for example, by using alternative data sources or machine learning methods. The authors find that the first channel seems to matter the most: the increase in convenience and speed have been more central to fintech lending's growth than improved screening or monitoring.

However, the adoption of AI technologies by financial intermediaries has also raised concerns regarding their effects on investment decisions and, more broadly, on the allocation of capital. A specific class of financial intermediaries has recently received a lot of attention with respect to their use of AI technologies: Venture Capitalists (VCs). Venture Capitalists are private equity investors that provide capital for startups with high growth potential and play a crucial role in the financing of innovation: among public firms founded within the last fifty years, VC-funded companies account for more than 92% of R&D spending and patent value! In a recent article, Bonelli (2022) shows that dozens of VCs have adopted AI technologies for screening startups over the last decade. These VCs employ AI algorithms to detect patterns in historical data from previous startups and extrapolate them to predict a new startup's outcome. Despite using cutting-edge advances in machine learning, these forecasting algorithms are in essence backward-looking-because they are trained using past data. As such, they may not be successful at screening startups that differ radically from past companies. The paper confirms this intuition: VCs adopting AI become better at identifying good quality startups, defined as startups that survive and receive follow-on funding, but only within the pool of startups whose business is similar to that developed by past companies. At the same time, VCs that adopt AI become less likely to invest in startups that achieve major success, like an Initial Public Offering (IPO), or an innovative breakthrough, for example a highly cited patent. This finding is associated with an increase in the share of their investments being oriented towards startups developing businesses closer to those already tried-and-tested. Overall, this study highlights a potentially adverse aspect of the adoption of Artificial Intelligence on firm financing, with less funds directed to very innovative - disruptive - companies. This raises two important questions of whether this might induce entrepreneurs to produce more backward-similar ventures at the expense of breakthrough innovations, and what could be the implications for long-term economic growth.