Publications

• The Devil in the Air: Air Pollution and Dementia
  with Meng Sun and Emily Zheng
  Journal of Environmental Economics and Management (JEEM), 2024. [abstract] | [manuscript] | [online] |
  We study the causal relationship between air pollution and dementia prevalence. Leveraging the strict air pollution regulations implemented during the 2008 Beijing Olympics and employing a Synthetic Difference-in-Differences (SDID) approach, we find that a 1 µg/m³ decrease in annual PM10 levels corresponds to a 0.82 percentage point reduction in dementia prevalence (equivalent to 2.39% of the mean). Analyses across demographics show a more pronounced impact on vulnerable groups. Moreover, an economic assessment suggests that a 10 µg/m³ reduction in China's air pollution in 2010 could generate up to 2.36 billion US dollars in benefits due to a lower dementia prevalence. These results highlight the potential public health gains achievable through air pollution regulations.
• Subjective Belief on Housing Return and Its Implications on the Asset Allocations of Chinese Households
  with Yuting Wang, Bing Xu, and Mawell Pak
  Economic Research Journal (ERJ), 2023. [abstract] |
  Housing plays a dominant role in Chinese households' asset allocation, while the share of risky financial assets, particularly stocks, remains disproportionately low. Existing literature has largely documented the stylized fact of Chinese households' "heavy real estate, light risky financial assets" allocation pattern, but whether this allocation structure is efficient has not been thoroughly studied. Using data from the 2019 China Household Finance Survey and historical asset performance data, this paper first documents that the mean-variance of Chinese households' actual portfolio returns is inconsistent with the mean-variance of efficient portfolios predicted by classical portfolio theory, and proposes that households' biased subjective beliefs about housing return distributions may be a primary driver of this phenomenon. We further incorporate this hypothesis into a standard mean-variance portfolio theory framework, constructing a structural model of asset allocation with subjective beliefs, empirically testing this hypothesis and quantitatively examining its impact on Chinese households' asset allocation. Structural estimation results show that Chinese households generally overestimate the expected return on housing by an average of 2.8 percentage points. More importantly, households' cognitive biases regarding expected housing returns vary at the city level: households in first-tier cities underestimate expected housing returns, while households in non-first-tier cities overestimate expected housing returns by up to 4.9 percentage points. Counterfactual analysis shows that overestimating expected housing returns reduces Chinese households' wealth by an average equivalent to 5.2 percentage points of annual household income, and by up to 8.9 percentage points in non-first-tier cities.
• Uncertainty in Procurement Contracting with Time Incentives
  with Wenzheng Gao and Daiqiang Zhang
  International Economic Review (IER), 2021. [abstract] | [manuscript] | [online] |
  This article studies cost-plus-time (A+B) procurement contracting with time incentives in the highway construction industry. In the presence of construction uncertainty, the contractor's actual completion time may deviate from the bid completion time, and the A+B contract design is not ex post efficient. Using data from highway procurement contracts in California, we show that an ex post efficient lane rental contract would reduce the social cost by $41.39 million (43.11%) on average. Moreover, the average commuter cost would decrease by $62.06 million (78.96%), suggesting a substantial reduction in the construction externality to commuters from lane rental contracts.
• Effects of Minimum Wage on Workers’ On-the-Job Effort and Labor Market Outcomes
  with Meng Sun
  Economic Modelling, 2021. [abstract] | [online] |
  Can higher minimum wages motivate workers to work harder? If so, what are the effects of workers' on-the-job effort responses on the labor market outcomes? To answer these questions, we apply a model with directed on-the-job search and dynamic incentive contracts in a frictional labor market. The steady-state comparison of the calibrated model shows that a higher minimum wage increases workers' on-the-job effort. It also reduces the average hiring and layoff rates. Since the reduction in the hiring rate is higher than the reduction in the layoff rate, the unemployment rate increases, and hence lowers the aggregate output. Moreover, we find that the higher minimum wage has a spillover effect on higher-income workers. It suggests that agents' incentive decisions can provide a new explanation of the spillover effect of the minimum wage. Lastly, shutting down the effort channel leads to greater labor market impacts. These results suggest that workers' on-the-job effort responses have moderate offsetting effects on the cost of the higher minimum wage.
• Income Inequality, Liquidity Constraints and China’s Household Savings Rate
  with Li Gan and Yongzhi Sun
  Economic Research Journal (ERJ), 2018. [abstract] |
  Using three independent nationally representative micro-level datasets — the China Household Finance Survey (CHFS), the China Family Panel Studies (CFPS), and the Chinese Household Income Project (CHIP) — this paper examines China's high household savings rate from the perspective of the interaction between income distribution and liquidity constraints. The findings show that: (1) high-income households have significantly higher savings rates than low-income households; (2) although all households may face liquidity constraints, low-income households face a much higher probability of being liquidity constrained than high-income households, and the presence of liquidity constraints significantly raises household savings rates; (3) widening income inequality and tightening liquidity constraints lead to higher aggregate household savings rates; (4) low-income households have a much higher marginal propensity to consume than high-income households. These empirical findings are not only consistent with the predictions of heterogeneous-agent intertemporal savings models, but also provide new insights for government policies aimed at boosting consumption: reducing income inequality is not only a viable policy for lowering China's household savings rate, but also a policy that promotes economic growth and structural transformation.

Selected Working Papers

• Identification and Estimation of Binary Choice Models with Social Interactions and Unknown Group Structures
  with Yonghong An and Wenzheng Gao
  under review. [abstract] | [manuscript] |
  This paper studies identification and estimation of a game-theoretic binary choice model with social interactions where the group structure is latent. By exploiting two proxies for the group characteristics, we show that peer effects can be identified even though the group structures are unobservable. Based on the identification method, a semiparametric nonlinear least square estimator is proposed. Monte Carlo experiments demonstrate that the semiparametric estimator has good finite-sample performance. In the empirical application of our method, we find positive and significant peer effects among students in their choices regarding private schoolwork tutoring, and show that ignoring the latent group structures can lead to significant bias in the estimation of peer effects.
• Empathy, Social Networks, and Academic Achievement: Experimental Evidence on Cross-Productivity in Skill Formation
  with Flávio Cunha, Qinyou Hu, and Yiming Xia
  under review. [abstract] | [NBER WP35370] |
  We evaluate a parent-directed empathy development program for middle school students in China using a randomized trial with 2,246 students. The four-month mobile app intervention produced short-run gains in empathy and reduced bullying but no immediate academic effects. Over the longer term, treatment students were 3.2 percentage points more likely to enter elite high schools and 2.5 percentage points less likely to miss the entrance exam. We trace these gains to improved self-beliefs and a restructuring of classroom friendship networks that reduced academic homophily, revealing a social channel of cross-productivity: non-cognitive investments can reshape peer relationships to improve academic outcomes.
• Earning Effects of Extending Compulsory Education: Continuation Values and the Credential Mismatch
  with Shiyu Xie and Bing Xu
  under review. [abstract] |
  Extending compulsory education has been officially placed on the agenda of China's 15th Five-Year Plan. As this policy remains in the deliberation stage, its potential earnings effects can neither rely on ex-post evaluation nor be verified through randomized controlled trials in the short run. Ex-ante policy evaluation based on structural models, however, can predict how families adjust educational choices under different institutional designs and the resulting impacts on labor market earnings by characterizing heterogeneity in individual ability and family background, thereby providing quantitative evidence for policy deliberation. Using micro-level data from the CFPS, this paper constructs a generalized Roy model encompassing sequential educational choices and earnings determination, incorporating both cognitive and non-cognitive abilities into a unified framework, and first estimates the returns at each educational choice node under the current system. Unlike traditional Mincerian returns, the educational choice returns measured in this paper encompass two dimensions: direct credential returns and continuation values. Estimation results show that the direct credential return differential between academic high school and vocational high school is not statistically significant, but choosing academic high school yields an 8.9% continuation value — additional expected returns obtained through opening pathways for further education. Meanwhile, approximately 17.3% of high school students would have higher expected returns from choosing vocational education, indicating that optimal educational pathways vary by individual endowment. Building on this, the paper simulates the earnings effects of two reform scenarios: extending compulsory education to 12 years while retaining the academic-vocational track, and extending to 12 years while eliminating the academic-vocational track. Results show that both scenarios significantly increase overall social income, but the former raises per capita annual income by 4.73%, significantly outperforming the latter's 1.9%. Mechanism analysis reveals that income improvements primarily stem from correcting educational mismatches among groups misallocated under the current system: extending compulsory education effectively eliminates vertical mismatches caused by family background — reducing premature exit from education among those with potential — while retaining the academic-vocational track avoids worsening horizontal mismatches by providing individuals without academic comparative advantages a more suitable skill accumulation pathway. Together, these achieve more efficient human capital allocation.
• Reducing Bullying: Evidence from a Parental Involvement Program on Empathy Education
  with Flávio Cunha, Qinyou Hu, and Yiming Xia
[abstract] | [NBER WP30827] | [New draft coming soon] |
  According to UNESCO, one-third of the world's youths are victims of bullying, which deteriorates academic performance and mental health, and increases suicide ideation and the risk of committing suicide. This paper analyzes a four-month parent-directed intervention designed to foster empathy in middle schoolers in China. Our implementation and evaluation study enrolled 2,246 seventh and eighth graders and their parents, whom we assigned, at the classroom level, to the control or intervention condition randomly. We measured, before and after the intervention, parental investments, children's empathy, and self-reported bullying perpetration and victimization incidents. Our analyses show that the intervention increased investments and empathy and reduced bullying incidents. In addition, we measured costs and found that it costs $12.50 for our intervention to reduce one bullying incident. Our study offers a scalable and low-cost strategy that can inform public policy on bullying prevention in other similar settings.

© Naibao Zhao 2026. Last updated: August 2026