Carol Hengheng Lu

Hi! I’m a Senior Economist at Amazon, where I develop innovative solutions in advertiser incentive policies, ad auction design, and campaign recommendations. Prior to Amazon, I was a research intern at the Chief Economist's Office at Microsoft in 2019 and 2020. I earned my Ph.D. in Economics from Stanford University in 2022. My research specializes in empirical industrial organization and quantitative marketing, with a particular focus on online platforms and digital goods.

You can reach out to me at LinkedIn.

Research Papers

Revise and Resubmit, Journal of Political Economy

Abstract: We analyze 75,000 alternating-offer, business-to-business negotiations in the wholesale used-car market, each mediated (over the phone) by a third-party. We find that it matters who intermediates: high-performing mediators are 18.07 percentage points more likely to close a deal than low performers, and these differences are not due to statistical noise. Better mediators achieve trades quickly, and mediator skill improves with experience. We estimate buyer and seller value distributions using a structural model of two-sided incomplete-information in which mediators of different skill levels correspond to different equilibria. High-skilled mediators improve efficiency and overcome some of the inefficiency inherent in incomplete-information settings.

Revise and Resubmit, Quantitative Marketing and Economics

Abstract: Offering content via bundle-based subscription services is a prevalent business strategy for media platforms. Despite its popularity, there is limited empirical evidence on whether the subscription model can generate better outcomes for customers and content providers than the traditional sales model. We answer this question using a novel proprietary dataset from the Xbox video game platform and leveraging their introduction of the subscription service Game Pass. We develop and estimate a model of demand and supply for individual games and the subscription service. We find consumer surplus increases by 16% when Game Pass is introduced. A decomposition shows that the bundling and renting features of the subscription service contribute almost equally to the surplus change. We further simulate counterfactuals where Xbox only offers one “Grand subscription bundle” that includes all titles like most streaming video platforms and find consumer surplus is 38% lower than that under the traditional sales model. This outcome could be improved if Xbox offers multiple tiers of subscriptions, depending on the distribution of consumers’ title valuation and satiation rate. On the supply side, content providers are better off with the subscription-only model when the platform retains 30% or less of the subscription revenue.

Variety Seeking in High-Frequency Consumption: New Implications for Target Marketing, with Fan Zhang and Zhijie Lin (2023)

Abstract: We study consumers’ variety-seeking preference and explore its implications on target marketing using data from a food delivery platform. We estimate a demand model for restaurants accounting for customers’ heterogeneous state-dependent preferences and find that a substantial fraction of consumers are variety-seeking: they are willing to pay 19.9% more on average for switching to a different seller at next purchase. In the counterfactual analysis, we find that personalizing restaurant ranks based on consumers’ order history and their variety-seeking preference increases platform revenue by 7%.

Abstract: On top of the well-documented gender pay gap among wage earners, female-owned micro-enterprises are found to be less profitable. In this paper, we explore the gender gap in business performances among freelancers selling home-cooked meals, an area that is traditionally perceived as female-typed and involves complicated business strategy-making. Using high-quality proprietary data from a Chinese online food ordering platform, we document a 12% gender gap in sellers’ daily revenues. Factors highlighted in the existing literature such as working hours and experience narrow the gap by 63%. The remaining gap can be explained by male and female freelancers adopting different business strategies in pricing, positioning, and market expansion.

Journal of Mathematical Economics, 2018, 75: 67-70

Abstract: Arnold and Riley (2009) find that in the credit rationing model of Stiglitz and Weiss (1981), the expected revenue of a lender as a function of the loan rate cannot be globally hump-shaped, and thus credit rationing is hard to explain using the Stiglitz–Weiss model. However, Arnold and Riley base their analysis on the assumption that there is an upper bound of the returns of borrowers’ projects. We find that if unbounded returns of borrowers’ projects are allowed, then a lender’s expected revenue in the Stiglitz– Weiss model can in fact be globally hump-shaped. This also implies that credit rationing (with one equilibrium loan rate) can only arise in markets where the returns from investment are highly volatile.