What Makes a Government Respond? A Data-Driven Approach

• In Event: Power Politics in Authoritarian Regimes

Abstract

With access to all petitions submitted online for a major city in China in 2015, this paper aims to directly address the question of what makes the government respond. Specifically, this paper intends to answer the following two questions: (a) from the theory discovery perspective, what contents in the letters make the government more responsive, and (b) from the theory testing and causal inference perspective, conditioned on the discovered theory, how much of the discovered element contained in the letter would increase the government’s probability of responding. Existing observational evidence shows that insiders are more likely to complain to the government (Tsai and Xu, 2017), while experimental evidence defies the hypothesis that insiders get higher rate of response (Chen et al 2016). This paper tries to reconcile the paradox by finding that indeed that the local government in China is more likely to respond if the petitioner sends a credible signal to the government that she is an insider with a data-driven theory-discovery approach. It further tests this proposed hypothesis with an active-labeling-enhanced supervised learning algorithm. Results show that the probability of getting a response from the government is 12-23 percentage points higher if the petitioner sends the insider signal.

Authors

• Shiyao Liu, Peking University

• Tianguang Meng, Tsinghua University