Selected working papers

Social Media and Collective Action in China Abstract This paper studies how social media affects the dynamics of protests and strikes in China during 2009-2017. Based on 13.2 billion microblog posts, we use tweets and retweets to measure social media communication across cities and exploit its rapid expansion for identification. We find that, despite strict government censorship, Chinese social media has a sizeable effect on the geographical spread of protests and strikes. Furthermore, social media communication considerably expands the scope of protests by spreading events across different causes (e.g., from anti-corruption protests to environmental protests) and dramatically increases the probability of far-reaching protest waves with simultaneous events occurring in many cities. These effects arise even though Chinese social media barely circulates content that explicitly helps organize protests.
(with Bei Qin and David Strӧmberg)
2nd Round R&R at Econometrica

Social Media and Government Responsiveness: Evidence from Vaccine Procurement in China Abstract This paper studies how public opinion on social media affects local governments' procurement olvaccines in China during 2014-2019. To establish causality, we exploit city-level variation in theeruption of online opinion on vaccine safety, instrumented by quasi-random early penetration otsocial media. We find that governments in cities exposed to stronger social media shocksincreased the share of more-transparent procurement and shifted procurement from small localsuppliers to reputable nonlocal suppliers. The effect is driven by posts expressing anti-governmentsentiment instead of posts containing investigative information and is larger in cities where localofficials face higher top-down political pressure.
(with Yixin Mei)
Reject and resubmit at the American Economic Review

Neyman-Pearson and Equal Opportunity: When Efficiency Meets Fairness in Classification Abstract Organizations often rely on statistical algorithms to make socially and economically impactful decisions. We must address the fairness issues in these important automated decisions. On the otherhand, economic efficiency remains instrumental in organizations’ survival and success. Thereforea proper dual focus on fairness and efficiency is essential in promoting fairness in real-world datascience solutions. Among the first efforts towards this dual focus, we incorporate the equal op.portunity (EO) constraint into the Neyman-Pearson (NP) classification paradigm. Under this newNP-EO framework, we derive the oracle classifier, propose finite-sample based classifiers thatsatisfy population-level fairness and efficiency constraints with high probability, and demonstratestatistical and social effectiveness of our algorithms on simulated and real datasets.
(with Jianqing Fan, Xin Tong, and Shunan Yao)
Under review

Whither Journalism? The Impact of Social Media on News Production in China
(with Ruoyu Qian and David Strӧmberg)

Journal Publication

Street-Level Responsiveness of City Governments in China, Germany, and the United States Abstract This paper presents evidence from parallel field experiments in China, Germany, and the United States. We contacted the mayor’s office in over 6,000 cities asking for information about starting a new business. Chinese and German cities responded to 36-37 percent of requests while American cities responded only to 23 percent of requests. American and German cities were more responsive to requests from citizens than foreigners; Chinese cities were more responsive to requests from men than women. Chinese cities were more responsive to requests about starting a construction than a green business, and when the mayor was up for promotion. The evidence suggests that street-level bureaucrats exercise substantial discretion in all three countries, but use that discretion in different ways.
(with Ekkehard Köhler and John G. Matsusaka)
Journal of Comparative Economics, 51(2): 640-652, June, (2023)

Competition, Contracts, and Creativity: Evidence from Novel Writing in a Platform Market Abstract A growing number of people today are participating in the gig economy, working as independent contractors on short-term projects. We study the effects of competition on gig workers’ effort and creativity on a Chinese novel-writing platform. Authors pro-duce and sell their works chapter by chapter under a revenue-sharing or pay-by-the-word contract with the platform. Exploiting a regulation that induced a massive entry of novels in the romance genre but not other genres, we find that, on average, intensified competition led authors to produce content more quickly, whereas its effect on book novelty was weak. However, revenue-sharing books responded to competition substantially more than pay-by-the-word books, particularly regarding novelty. Moreover, the effect of competition on novelty is considerably stronger for books at earlier stages of the product life cycle. Finally, the platform increased the promotion of contracted books, which disproportionately favored pay-by-the-word books. We discuss the implications of these results for creative workers, platform firms, and policy makers in the gig economy.
(with Feng Zhu)
Management Science, 68(12):8613-8634, (2022)

Intentional Control of Type I Error over Unconscious Data Distortion: A Neyman-Pearson Approach to Text Classification Abstract This article addresses the challenges in classifying textual data obtained from open online platforms, which are vulnerable to distortion. Most existing classification methods minimize the overall classification error and may yield an undesirably large Type I error (relevant textual messages are classified as irrelevant), particularly when available data exhibit an asymmetry between relevant and irrelevant information. Data distortion exacerbates this situation and often leads to fallacious prediction. To deal with inestimable data distortion, we propose the use of the Neyman–Pearson (NP) classification paradigm, which minimizes Type II error under a user-specified Type I error constraint. Theoretically, we show that the NP oracle is unaffected by data distortion when the class conditional distributions remain the same. Empirically, we study a case of classifying posts about worker strikes obtained from a leading Chinese microblogging platform, which are frequently prone to extensive, unpredictable and inestimable censorship. We demonstrate that, even though the training and test data are susceptible to different distortion and therefore potentially follow different distributions, our proposed NP methods control the Type I error on test data at the targeted level. The methods and implementation pipeline proposed in our case study are applicable to many other problems involving data distortion.
(with Xin Tong, Lucy Xia, and Richard Zhao)
Journal of American Statistical Association, 116(533): 68-81, March, (2021).

Media Bias in China Abstract This paper examines whether and how market competition affected the political bias of government-owned newspapers in China from 1981 to 2011. We measure media bias based on newspaper coverage of government mouthpiece content (propaganda) relative to commercial content. We find three main results. A reform that forced newspaper exits (reduced competition) affected media bias - increasing product specialization with some papers focusing on propaganda and others on commercial content. Lower-level governments produce less-biased content and launch commercial newspapers earlier, eroding higher-level governments' political goals. Bottom-up competition intensifies the politico-economic tradeoff, leading to product proliferation and less audience exposure to propaganda.
(with Bei Qin and David Strӧmberg)
American Economic Review, 108(9): 2442-76, September, (2018).
AEA Research Highlight   |   VoxDev
Appendix Empirics  |  Appendix Theory

Incentive Contracts and the Allocation of Talent Abstract This paper develops a theory of sorting that links ability, pay-performance sensitivity, and pay levels across a wide range of managerial levels. Firms employ managers to improve productivity. Because of limited liability, firms use incentive contracts to elicit unobservable managerial effort; the type of optimal contract depends on a manager’s ability. In equilibrium, individuals are sorted based on ability into production workers, business owners, managers paid an ability-invariant bonus, and managers whose pay varies with ability and firm size. The model predicts that TFP-enhancing technological progress and increased competition in product markets tend to 1) reduce the fraction of business owners while increasing the fraction of pro.t-sharing managers and 2) increase wage inequality between workers and managers and wage inequality among managers. Wage dynamics and employment of the self-employed and various groups of salaried managers in the U.S. are consistent with these predictions.
Economic Journal, 127(607): 2744-2783, December, (2017).
VoxEU   |   LSE Business Review
Online Appendix

Authority, Incentives and Performance: Evidence from a Chinese Newspaper Abstract This paper examines how the allocation of authority within an organization affects workers' incentives and performance, using personnel data from a Chinese newspaper. Relying on an authority change that transferred the right of making editorial decisions from mid-level editors to top editors in four of the eight divisions in the newspaper, I find that the authority change 1) improves reporters' performance, while reducing their activities for private gain; and 2) decreases mid-level editors' journalistic initiative. To reconcile these findings, a synthesis of two theories on authority and incentives - the vertical and the horizontal allocation of authority - is needed.
Review of Economics and Statistics, 99(1): 16-31, March, (2017).
Online Appendix

Why Does China Allow Freer Social Media? Protests vs. Surveillance and Propaganda Abstract This paper documents basic facts regarding public debates about controversial political issues on Chinese social media. Our documentation is based on a dataset of 13.2 billion blog posts published on Sina Weibo—the most prominent Chinese microblogging platform—during the 2009–2013 period. Our primary finding is that a shocking volume of posts on highly sensitive topics were published and circulated on social media. For instance, we find millions of posts discussing protests and even more posts with explicit corruption allegations. We find that a large number of real-world protests can be predicted by social media data one day before their occurrence and that corruption charges of specific individuals can be predicted one year in advance. These findings support the view that social media can be effective surveillance tools for autocratic governments. Finally, we estimate that our data contain 600,000 government-affiliated accounts on Sina Weibo, and the share of government accounts is larger in areas with a higher level of internet censorship and where newspapers have a stronger pro-government bias. Overall, our findings suggest that the Chinese government regulates social media to balance threats to regime stability against the benefits of utilizing bottom-up information.
(with Bei Qin and David Strӧmberg)
Journal of Economic Perspectives, 31(1): 117-40 (2017).
VoxEU  |  VoxDev
(CEPR DP11778; previously circulated as The Political Economy of Social Media in China.)

Organizational Structure and Product Choice in Knowledge Intensive Firms Abstract This paper formulates a model in which a firm simultaneously chooses its organizational structure and product position. The firm’s production is knowledge-intensive, requiring employees to solve problems. A vertical hierarchy in which workers refer unsolved problems to managers facilitates the acquisition and leveraging of managers’ superior knowledge. I show that a larger span of control is complementary to the provision of high-value products. Moreover, this complementarity is sustained when employees acquire sufficient knowledge and is further strengthened when the firm enhances its capability of communicating knowledge. The model yields testable implications concerning: (1) the fit between a firm’s product position and span of control; 2) the effect of information technology on product innovations and skill-biased organizational changes; and 3) the heterogeneity in hierarchical structure and human resource management in professional service firms.
Management Science, 61(8): 1830-1848 (2015).

Knowledge, Communication, and Organizational Capabilities Abstract This paper attempts to bridge a gap between organizational economics and strategy research through an analysis of knowledge and communication in organizations. We argue that organizations emerge to achieve the intensive use of the knowledge that is acquired to perform specific tasks and to integrate dispersed knowledge that is embodied in different human minds. The attributes of the tasks undertaken determine the optimal acquisition and distribution of knowledge. Depending on the codifiability of knowledge, different communication modes arise as a coordination mechanism to deepen the division of labor, leverage managerial talent, and exploit the increasing returns to knowledge. Organizational processes can be adapted through codes and culture to facilitate coordination; organizational structure can be designed to complement the limitations of human ability. We stress that organizational process and structure construct the core of organizational capital, which generates rent and sustains organizational growth. From the analysis, we draw implications for the strategic management of knowledge and human resources in organizations.
(with Luis Garicano)
Organization Science, 23(5):1382-1397 (2012).
 - Technical and enriched version: A Problem-solving Approach to Organizational Design
 - Reprint in Recent Developments in the Economics of Information, The International Library of Critical Writings in Economics Series, C. Antonelli (ed.), Edward Elgar Publishing Limited (Cheltenham, U.K.), 2018.