Working paper

We examine the productivity effects of a generative artificial intelligence technology—the assistive chatbot ChatGPT—in the context of mid-level professional writing tasks. In a preregistered online experiment, we assign occupation-specific, incentivized writing tasks to 444 college-educated professionals, and randomly expose half of them to ChatGPT. Our results show that ChatGPT substantially raises average productivity: time taken decreases by 0.8 SDs and output quality rises by 0.4 SDs.

In an RCT with US small businesses, we document that a large share of firms are not well-informed about bankruptcy. Many assume that bankruptcy necessarily entails the death of a business and do not know about Chapter 11 bankruptcy, where debts are renegotiated so that the business can continue operating. Small businesses are also unaware of a recent major reform that lowered the costs of bankruptcy procedures to enhance their protection.

Measuring the returns of advertising opportunities continues to be a challenge for many businesses. We design and run a field experiment in collaboration with Yelp across 18,294 firms in the restaurant industry to understand which types of businesses gain more from digital advertising. We randomly assign 7,209 restaurants to freely receive Yelp’s standard ads package for three months. The scale of the experiment gives us a unique opportunity to assess the heterogeneity in advertising effectiveness across a variety of business attributes.

The limited market size of many small emerging economies is a key constraint to the growth of innovative small and medium enterprises. Exporting offers a potential solution, but firms may struggle to locate and appeal to foreign buyers. A six-country randomized experiment was conducted with 225 firms in the Western Balkans to test the effectiveness of 30 hours of live group-based training and 5 hours of one-on-one remote consulting in overcoming these constraints.

We present novel evidence from the results of a randomized controlled trial on the role that information plays in the perceptions of the benefits and costs of exporting. We first present results from a baseline survey of approximately 1,000 UK manufacturing firms to show that non-exporters hold substantially more negative beliefs about the costs and benefits of exporting relative to exporters.

We study the allocation and productivity consequences of training production line supervisors in soft skills via a randomized controlled trial. Consistent with standard practice for training investments within firms, we asked middle managers – who sit above supervisors in the hierarchy – to nominate members of their supervisory team for training. Program access was randomized within these recommendation rankings. Highly recommended supervisors experienced no productivity gains; in contrast, less recommended supervisors’ productivity increased 12% relative to controls.

This paper tests whether providing more information on business practices can lead firms to seek out advice and improve their practices. The authors collaborated with a business advice provider in Brazil to implement a randomized experiment with 866 small firms. The treatment groups received different versions of an information sheet that benchmarked business practices to other firms and listed five practices to improve.

We analyze, benchmark, and run randomized controlled trials on a panel of 7,463 U.S. entrepreneurs making incentivized sales forecasts. We assess accuracy using a novel administrative dataset obtained in collaboration with a leading US payment processing firm. At baseline, only 13% of entrepreneurs can forecast their firm’s sales in the next three months within 10% of the realized value, with 7.3% of the mean squared error attributable to bias and the remaining 92.7% attributable to noise.

By enabling smaller loans, broader geographic reach, and less human bias in decision-making, process automation may reduce racial disparities in access to financial services. We find evidence for all three channels in a setting where private lenders faced no credit risk but decided who to serve: the Paycheck Protection Program (PPP), which provided loans to small businesses during COVID-19. Black-owned firms disproportionately obtained their PPP loans from fintech lenders, especially in areas with high racial animus.

Women are underrepresented in patenting and the gap is not closing quickly. One major roadblock to progress is a dearth of causal evidence on the potential effectiveness of policies to reduce the gender gap in patenting. Analyzing a randomized control trial at the United States Patent and Trademark Office that was designed to provide additional help to applicants who do not have legal representation, we find heterogeneous causal impacts across gender and technologies on the probability of obtaining patent rights.