Research (selected)

Modern business practices frequently employ a blend of pricing strategies to segment markets effectively. As a result, consumers may encounter pricing schedules that are non-linear and multidimensional. This paper presents a structural approach for estimating multidimensional non-linear pricing models involving multiple decision variables in an energy market. Using a unique, rich panel dataset of Chinese household electricity consumption, we structurally estimate consumer preferences under the influence of an Increasing Block Price (IBP) and a Time-of-Use (ToU) system. Our structural approach allows us to distinguish and evaluate household-level price elasticities of demand, presenting a novel explanation for consumers’ feedback on marginal price changes. Through model-based simulations, we demonstrate that a 1% increase in price corresponds to a 0.7% reduction in total electricity demand. However, our analysis indicates that practical opportunities for optimization within multi-dimensional pricing systems are limited. Our findings offer distinct insights into the complex interplay between intricate pricing structures and energy consumption behavior, thereby providing valuable guidance for policymakers and regulators.

In this study, we explore the effect of peer recognition on content creation within a prominent Chinese Question-and-Answer (Q&A) platform, specifically focusing on whether votes from peers encourage influencers to engage in providing more answers. Using panel regression models with instrumental variables, our analysis reveals that peer votes have a substantial positive effect on content production. Additionally, we investigate the consequences of two distinct badge policies, the “self-authentication” and the “best-answerer” badge, on content production. Our results demonstrate that while badges aid users in recognizing the quality of an influencer, badges with strong connotations may constrain content creation due to concerns about reputation management and privacy. As such, strategies that enhance platform traffic by promoting voting could be counterproductive if they exacerbate privacy and reputation worries. Our findings provide valuable insights into the role of peer recognition and badge policies in shaping content contribution, bearing crucial policy implications for the design of Q&A platforms.

This paper investigates the complexities of digital product membership renewals. Using an extensive dataset from a Chinese content platform, we analyze the interplay of price changes and peer influences within referral networks and their collective impact on users’ decision to renew. Our study reveals several key insights: First, through regression modeling, we quantify the price elasticity and uncover a positive correlation between a user’s likelihood of renewal and both the renewal decisions of their referrers and the quantity of their referees. Second, we further examine the cascading effects of price changes across the network. One significant advantage of our structural model is its capacity to allow the referral network to be endogenously determined by users’ consideration and prediction of the renewal decisions of both upstream referrer and downstream referees. The findings reveal that targeted discounts for referees are more effective than either uniform discounts or targeted discounts aimed at referrers. Such marketing strategies would efficiently improve renewal rates while avoiding potential revenue loss. Last, further evidence sheds light on how different network structures affect overall renewal rates, suggesting that networks characterized by high level of connectivity yet low level of centrality are more conducive to sustaining customer loyalty.

This paper empirically examines the impact of the launch of a Government-initiated E-commerce Platform (GEP) on the sales of local specialties, with a particular focus on China’s Pu’er tea market. Employing two-way fixed effects (TWFE) regressions on a panel dataset of 983 local farmers over five years, which covers over 90% of local tea productions, we find that the introduction of the GEP leads to an average decline of 11.22% in offline household sales, while online sales experience an increase of 16.88%. This channel shift was observed at all levels of production and quality, yet the total volume of tea sales remained constant, suggesting that the GEP facilitated a more efficient allocation of sales between online and offline channels and increased profits. Further evidence shows that the platform was particularly beneficial to farmers who faced significant barriers to entry on traditional platforms, providing them with a more cost-effective alternative. Mechanism analysis suggests that the increase in online sales volumes is primarily due to an increase in product variety rather than an increase in the number of online sellers. Our study underscores the transformative potential of such E-commerce initiatives in specific markets and provides actionable insights for policymakers and practitioners.

We collect the data from Hong Kong’s major news media Facebook pages from 2019 to 2020 to examine the online ideological clashes between pro-democracy and pro-Beijing parties. We find that specific writing habits signify users’ backgrounds and elicit readers’ group segregation cues. Compared to Traditional Chinese comments, the increase of pro-police comments in Simplified Chinese induced a more polarized reaction from the opposite side with more anti-police comments and demand for regional independence. On the other hand, contents generated by the suspected bots alleviate the clashes. The results demonstrate the need for debiasing intervention and regulation in social media platforms.

This paper studies how network structure can affect the speed of adoption. In particular, we model the decisions to adopt Python 3 by software packages. Python 3 provides advanced features but is not backward compatible with Python 2, which implies adoption costs. Moreover, packages form input-output networks through dependency relationships with other packages, and they face additional adoption costs if the dependency packages lack Python 3 support. We build a dynamic model of technology adoption that incorporates such a network and estimate it using a complete dataset of Python packages. Estimation results show that the average cost of one incompatible dependency is roughly three times the cost to update one’s code. We conduct counterfactual policies of community-level targeted cost subsidy and show network structure is crucial to determine an optimal policy of cost subsidy.