Bayesian Reasoning for Qualitative Research

On April 21, 2022, the Southeast Asian Studies research group of the Institute for International and Area Studies (IIAS) of Tsinghua University hosted an online lecture entitled “Bayesian Reasoning for Qualitative Research”. The lecture was delivered by Tasha Fairfield, Associate Professor of International Development at the London School of Economics and Political Science (LSE) and presided over by Dr. Li Yuqing, an Associate Professor at the IIAS. Faculty members and students from Tsinghua university, the University of Michigan, and other universities and research institutes at home and abroad attended the lecture.

Professor Fairfield firstly pointed out that Bayesian reasoning is a research method based on statistics and probability calculation, which can calculate and compare different hypotheses’ likelihood of occurrence based on the prior probability of such hypotheses under the given limited information conditions. Bayesian reasoning is more cutting-edge than other research tools in political science. It treats all correlated quantities in statistical inference as random quantities that can be analyzed when they are not fully understood, thus reducing the risk of prediction in political science research to a certain extent, and enjoying a broader application prospect.

Based on Bayes formula, Professor Fairfield then explained the hypotheses, prior probability, prior information, likelihood function, posterior probability and other elements in the formula one by one. In terms of the hypotheses, she stressed that the hypotheses she advocates in Bayesian analysis need to be mutually exclusive. In terms of prior probability, Bayesian analysis treats prior probability as a random quantity. She took “why developing countries are keen on signing bilateral investment agreements” as an example to illustrate. As for the prior information and likelihood function, the research designers need to carefully evaluate them, considering the authenticity of the prior information and its logical correspondence with the hypotheses. The large amount of prior information does not mean that the hypotheses are more likely to be true.

In the conclusion, Professor Fairfield argued that Bayesian analysis, as a research tool, is of great value for qualitative research. It can guide researchers to consider different mutually exclusive hypotheses for the same event, so as to facilitate more effective data collection. It can also urge researchers to further improve theories as well as to measure and think about the hypotheses and the prior evidence in theories more rigorously. Bayesian analysis does not seek to prove the absolute truth of a certain hypothesis in the research, but expounds the possibility of different hypotheses through evaluation and calculation, so as to improve the scientific nature of the research from the perspective of statistics.

After the lecture, Professor Fairfield discussed with the audience on the application of Bayesian analysis in the research of mixed qualitative and quantitative methods, the comparison of Bayesian reasoning with process tracking method and KKV’s quantitative methods (King, Keohane and Verba, Designing Social Inquiry, 1994) in political science research, etc.

Tasha Fairfield is an associate professor of International Development at the LSE. She holds a Ph.D in political science from the University of California, Berkeley. Her research interests include the political economy of inequality, the political process of policy making, and government-enterprise relations in Latin America. Bayesian reasoning for qualitative research has been extensively used in her research. She has published the monograph Private Wealth and Public Revenue in Latin America: Business Power and Tax Politics.

Edited by: Song Tianyun
Proofread by: Southeast Asian Studies research group
Typeset by: Cheng Yao