Working Paper

Tacit Collusion by Pricing Algorithm with Rule-Based Rivals (Job Market Paper)

Abstract: Pricing algorithms, especially reinforcement learning algorithms, have been widely used in the past decades for firms in competitive markets, which are very helpful for firms to capture more information of the market and their rivals. Existing literature has shown that, however, reinforcement learning algorithms could lead to supracompetitive prices without any communication between firms. In a framework of price competition between two firms both starting with a rule-based algorithm, we show evidence that (1) if one firm is competing using a reinforcement learning algorithm with a rule-based rival, the price is always going to weakly increase, and (2) if both firms are adopting, the one that adopts later would benefit more, and supracompetitive prices are obtained but the price change is ambiguous compared to the initial case. Our findings contribute to the literature by highlighting the importance of the order of algorithm adoption and the transition from rule-based algorithm to reinforcement learning algorithm, which makes it possible to compare the result with previous steady state in a more general case.

Detecting Algorithmic Collusion by Algorithms

with Yonghong An and Yu Zhu

Abstract: Algorithms can facilitate collusive behaviors among competing firms. It is challenging for antitrust authority to monitor and detect algorithmic collusion due to complicated price patterns and frequent price changes. In this paper, we study two important issues assuming that antitrust authority employs algorithms: how firms respond to an algorithmic antitrust authority and what price patterns the algorithmic antitrust authority would detect. In a framework of quantity competition of two firms, we demonstrate numerically that antitrust authority’s algorithms can effectively boost firms’ quantities and reduce the possibility of algorithmic collusion. We also identify several important factors that affect the authority’s effectiveness in auditing collusion: cost uncertainty of firms, penalty on firms due to detected collusion, and the authority’s incentive to audit.

The Impact of Return Policies and Perception Heterogeneity on Sales Platforms for Experience Goods: An Experimental Study

with Jinliang Liu, Revise and Resubmit to Journal of Economic Behavior and Organization, SSRN.

Abstract: On sales platforms for experience goods, consumers only know the distribution of a good’s value before purchase, while its actual value is revealed only after the purchase. However, consumers’ perceptions of the distribution’s range may vary, and the presence of a return policy can also influence the behavior of both sellers and buyers. Utilizing a post-price framework and controlled laboratory experiments, this paper provides the first investigation of experience goods markets that examines (1) the presence of a return policy and (2) the level of consumers’ perception heterogeneity about the range of the value’s corresponding distribution. We find that return policies lead to lower uniform prices in markets for experience goods. In addition, a larger perception heterogeneity reduces prices and increases return rates. Our results highlight potential downsides of return policies in online sales platforms.

Let Them Eat Pie: Addressing Sample Selection in Multiparty Elections

with Ali Kagalwala, Thiago M. Q. Moreira and Guy D. Whitten

Abstract: Elections are central to the study of politics. When studying parties’ vote shares across districts, scholars are encouraged to use compositional-outcome models in order to test their theories about what factors shape the dynamics of electoral support. One problem with existing compositional modeling approaches is that they incorrectly deal with scenarios in which not all parties compete in every electoral district. Because unobserved factors that affect a party’s decisions to contest a district are likely correlated with its performance in districts where the party fielded candidates, failing to account for partial contestation is likely to result in sample selection bias. Addressing sample selection in a compositional setting is challenging because the outcomes are in log-ratio form, and thus the errors are not normally distributed. To deal with these issues, we introduce a novel maximum likelihood approach which accounts for this type of sample selection and demonstrate through simulations that our method outperforms commonly used solutions, including the conventional Heckman correction. We illustrate the utility of our approach by analyzing the 2017 and 2019 UK parliamentary elections in English constituencies.

Work In Progress

Historical Simulations of Unified China and Divided Europe: Insights from Reinforcement Learning

with Weiwen Yin and Li Zheng

Long Divided, Must Unite? Historical Simulations of Japan’s Political Evolution Using Reinforcement Learning