I'm an Assistant Professor of Management at the Wharton School at the University of Pennsylvania. My research focuses on entrepreneurship and human capital, with a particular focus on emerging economies. My work draws on a variety of methodologies, with a focus on new computational methods and text as data. I received my Ph.D. at Columbia Business School.
Small unregistered firms contribute to a substantial proportion of global economic activity, particularly in developing regions. In explaining variation in productivity in these types of informal firms, research has focused primarily on the adoption of effective business practices and access to capital, with little focus on fundamental positioning. This article explores the nature of differentiation in microenterprises, introducing a text‐based measure of differentiation using state‐of‐the‐art sentence embeddings. Using a combined sample of nearly 10,000 microenterprises across eight developing countries, I examine whether (and which) microenterprises differentiate, whether differentiation is related to performance (and for whom), and whether any existing policy interventions affect differentiation.
We demonstrate how a novel synthesis of three methods—(a) unsupervised topic modeling of text data to generate new measures of textual variance, (b) sentiment analysis of text data, and (c) supervised ML coding of facial images with a cutting-edge convolutional neural network algorithm—can shed light on questions related to CEO oral communication. With videos and corresponding transcripts of interviews with emerging market CEOs, we use this synthesis of methods to discover five distinct communication styles that incorporate both verbal and nonverbal aspects of communication. Our data comprises interviews that represent unedited expressions and content, making them especially suitable as data sources for the measurement of an individual's communication style. We then perform a proof-of-concept analysis, correlating CEO communication styles to M&A outcomes, highlighting the value of combining text and videographic data to define styles. We also discuss the benefits of using our methods versus current research methods.
The emergence of large language models (LLMs) such as ChatGPT has opened new avenues for integrating artificial intelligence into the research process. One of the most promising early applications identified by social scientists is the use of LLMs for data annotation and text classification, a task traditionally performed either manually, with keyword methods, or through the development of customized machine learning tools. However, the benefits and opportunities of this particular use case of the technology remain poorly understood, and critically, the risks and potential downsides of the use of such technology in the research process have been largely unexplored. We examine the performance of ChatGPT in a research task for which we had previously developed a machine learning model at great effort and expense: identifying claims about sustainability in crowdfunding projects, based on the project text. We find that, with some prompt refinement, ChatGPT can easily match the performance of prior methods in annotating texts, at a great reduction in cost and time. However, our study also unearths that minor, seemingly inconsequential prompt variations can result in significantly different labels which, in turn, have implications for the robustness of downstream analyses and result interpretations. Through a sensitivity analysis of hypothetical downstream regressions, we demonstrate how prompt engineering can markedly alter result interpretations, and even enable the potential of unethical fishing for desired results. To guard against these risks, we develop a method we call Prompt Variance Estimation (PVE) to provide robustness to analyses that use LLM-generated labels, and provide instructions and code for its use.
In this study we examine how knowledge intermediaries — agents of knowledge institutions who communicate and encourage adoption of standardized best practices — affect practice adoption and productivity under conditions of resource constraint. We argue that while knowledge institutions are effective in identifying best practices and codifying recommendations, highly constrained firms such as microenterprises are often not able to adopt all recommended practices. We suggest that intermediaries, therefore, may serve a crucial role for highly constrained firms in helping them prioritize which codified best practices to adopt. We propose that knowledge intermediaries benefit firms by tailoring their recommendations to the firm’s level of constraint, prioritizing practices with a higher expected-benefit to expected-cost ratio. In examining these questions, we draw on a proprietary dataset of 1480 smallholder coconut farmers in the Philippines. We employ the geographic distance from agricultural extension offices as an instrument for receiving information from extension agents, who serve as the intermediaries for the agricultural authority and its eight “Good Agricultural Practices” (GAPs) for coconut production. We show that information from extension agents is associated with higher productivity (coconut yields) and higher awareness of specific GAP recommendations, but not higher blanket adoption of practices. Instead, farms receiving advice from extension agents adopt more “basic” and “intermediate” practices — those with higher average expected benefit and lower average expected cost — and fewer “advanced” practices, relative to farms who don’t receive agent advice. These effects are stronger for older farmers, who are more constrained than younger farmers in their ability to execute physically demanding farming practices. We suggest that knowledge intermediaries serve a crucial role in helping highly constrained firms optimize practice adoption given their limited time and resources.
This paper examines how skilled immigrants at headquarters (HQ) influence firms’ global product strategies by serving as informal coordination mechanisms. Using remarkably detailed data on product launches in the consumer packaged goods sector and confidential H-1B visa records, we employ an instrumental variable approach to isolate the causal impact of hiring immigrants. We find that hiring workers from a specific country leads firms to launch more new products in that country. This effect is significant only in countries where the firm has a foreign subsidiary, highlighting the complementary roles of informal (immigrant workers) and formal (subsidiaries) coordination mechanisms. Immigrant employees at HQ also tilt firms’ product strategy in their home countries toward standardization when a subsidiary is present, resulting in more products resembling those previously launched in the HQ market. This study makes a theoretical contribution by exploring how human capital at HQ functions as an informal coordination mechanism that complements formal structures and influences the balance between standardization and adaptation. It also makes an empirical contribution by offering causal evidence linking immigrant talent to the international product strategies of firms.