What does it take to elicit a local government response in China? Professor Meng Tianguang of the department of political science at Tsinghua University wanted to know the answer, and on March 17, 2016, he shared stories of his process and results with members of the NexGen Global Forum. Given that many local officials are not directly elected by the populace they serve, response rates to complaints to local officials had to be motivated by something else.
To identify which factors played a roll, Professor Meng designed an experiment to mine a huge amount of data contained on the Local Leadership Message Board, a nationwide online political forum where citizens submit complaints and requests to their local officials. He hypothesized that socioeconomic factors of the petitioners, as well as certain characteristics of the petitions themselves, would predict local officials’ response rates.
How to get a response from your local government:
1. Be a local. Petitions submitted by registered residents, those with a hukou (a local residence registration) in the province they were petitioning, were more likely to receive a response. Generally, provincial officials are more interested in ensuring their registered citizens receive services before turning attention to non-registered residents or migrant workers.
2. Petition collectively. Petitions submitted by a group of citizens have a higher response rate than those submitted by individuals. Although association is not an unlimited right in China, collective petitions are able to indicate to officials that a given issue is affecting multiple people, and thus warrants remedy.
3. Focus your request. The latter two predictors deal with the content of the petition rather than the identity of the petitioner. When a petition covers multiple issues or issue areas, it is less likely to receive a response. Local officials are more likely to respond to petitions when they cover only a single issue as the remedy is likely more easily achievable, or within a single department's jurisdiction.
4. Request remedy for an economic issue. Petitions to remedy issues in certain areas simply receive more responses than others; for instance, asking for solutions to problems in agriculture, employment, law and corruption, and business services received substantially higher numbers of responses than more overtly political topics. This focus is emblematic of the wider post-1979 pattern of increasing economic freedom in China uncoupled from change in the political realm.
Big data and social science
The methods Professor Meng employed to acquire these findings are as interesting as the findings themselves. Rather than using 'field experiments' whereby researchers submit manufactured petitions to local government officials and record response rates, he used data mining, computerized data analysis of a large database, to benefit from more 'natural' petitioning conditions while retaining statistical significance. Whereas a field experiment methodology might result in local government officials 'gaming' the system, using 'big data' analysis lets researchers view patterns that emerge from the wealth of naturally occurring petitions and do so over a long time frame.
The use of 'big data' analysis techniques has been on the rise in the social sciences as information is produced at an ever more rapid pace. They promise to hold up a truer mirror of our activities and behaviors than we could otherwise see from a smaller sample, but the vast quantities of data require careful handling to return useful, accurate results. Professor Meng acquired his data from the Local Leadership Message Board (LLMB), a national level platform for citizens to direct petitions to local government. This allowed him to avoid parochial data bias; as a national database, local governments cannot alter its records.
With a database of over 200,000 records spanning a period of six years, the data are obviously too numerous to analyze individually. Rather, Professor Meng and his research assistants applied an automated text analysis program to the database, which analyzes and gathers the results in a practical time frame. The key to this step is properly instructing the program and calibrating its analysis. To do this, research assistants code a random selection of the petitions by hand to provide samples, and then test random petitions against the program's output to make sure it is consistent. A wide variety of variables had to be controlled against as well, to ensure that the petitions measured only differences in social identity or policy issue. This careful attention to the details yielded the useful guidelines given above, but a number of interesting questions remain. Which petitions received workable solutions, and not just responses, to citizen petitions? What happens at the county and city, not just provincial, level? Is it possible to gain a more detailed picture of who petitions, and for what? Big data analytics has promise, with careful application, for finding answers to all these questions and more.