My Profile Photo

Junlong Aaron Zhou

Data Scinece, Causal Inference, Experimentation, Social Science, Ph.D. from NYU



  • Private Returns to Public Investment: Political Career Incentives and Infrastructure Investment in China (With Zhenhuan Lei). Journal of Politics, Vol 84, No. 1 (2022), 455–469 [Publisher] [SSRN]

    [Abstract] Why do politicians who have short tenure expectations have incentives to invest in long-term infrastructure projects? This mismatch between politicians’ short tenures and the long-term needed for infrastructure projects to come to fruition is generally expected to result in underinvestment in critical infrastructure. However, recent data show that China makes massive investments in large-scale, long-term transportation projects. By proposing a political exchange model, we demonstrate a fundamental synergy between the incentives of short-term mayors and of provincial leaders that is realized as a result of subway projects. With both a difference-in-differences design and a fuzzy regression discontinuity design, we show that subway projects significantly increase the promotion chances of city mayors. Additional tests also confirm the mechanism of our theory. Mayors who obtain subway projects deliver economic benefits to provincial leaders. The provincial politicians’ prospects of promotion are significantly improved thanks to these economic returns.

Working Papers

  • Generalizing Trimming Bounds for Endogenously Missing Outcome Data Using Random Forests (With Cyrus Samii, Ye Wang)[Arxiv]
    • Presented at APSA 2022, MPSA 2022, Asian Polmeth 2022, PolMeth 2021
    [Abstract] When units drop out or their outcomes are missing for endogenous reasons in experiments or quasi-experiments, random or conditionally-random treatment assignment is insufficient to point identify treatment effects. Non-parametric partial identification bounds are a way to address this problem without having to make ad hoc or disputable parametric assumptions. The problem is that basic approaches to constructing bounds often yield identification sets that are very wide and therefore minimally informative. We present methods for narrowing non-parametric bounds on treatment effects by adjusting for potentially large numbers of covariates, working with generalized random forests. Our approach maintains the agnosticism about the data generating process that makes bounds appealing, and it also allows for honest inference. We use a simulation study and two replication exercises to demonstrate the benefits of our approach.
  • Regression Discontinuity Designs for High-frequency Time-Series Cross-Sectional Data (With Ye Wang, Yiqing Xu)
    • Presented at MPSA 2022, Asian Polmeth 2022, MPSA 2021, APSA 2021, UCSD, NYU
  • Estimating Heterogeneous Treatment Effect on Clustered Data with Application on Get-Out-The- Vote Experiments
    • Presented at PolMeth 2020, NYU
  • Political Hierarchy and Political Pressure: The Regulation Power of Local Governments in Chinese Context.
    • Presented at MPSA 2019