Natural science has a significant impact on social science. One of its manifestations is that more and more experimental methods are used in social science research. In March, Cambridge University Press published "advances in experimental political science".

The two chief editors of this book are James n. Druckman and Donald P. green. James n. Druckman is a professor of political science at Northwestern University in the United States. His co-author's book who governments presidents, public opinion, and regulation was published in the University of Chicago Press in 2015. Donald P. green is a professor of political science at Columbia University. He has published four books (including field experiments: design, analysis and interpretation) and more than 100 papers. In 2010, he helped establish the experimental research branch of the American Society of political science and served as the first president of the branch.

This book holds that experimental politics has undergone many changes. In the past 20 years, it has evolved from an emerging research method to an industry recognized method, and also into the main method of this discipline. At present, the challenge is how to ensure that the experimenters can design a more reliable research scheme and implement it in a better way, so as to clarify the causality. In addition, ethical boundaries should be respected, research results should be interpreted in a transparent manner, and relevant data and research materials must be shared so that others can continue to explore on the existing basis.

This book discusses the application of new subject design; The introduction of new data sources, measurement methods and statistical methods are introduced; It shows the situation of experiments in more substantive areas; At the disciplinary level, this paper discusses the robustness, scalability and ethics of experiments in political science. On the one hand, the book discusses new opportunities, on the other hand, it focuses on new challenges accompanied by new opportunities, so it helps scholars and experimenters to carry out high-quality experiments and make important knowledge contributions.

The chapters of this book include the new era of experimental politics, joint investigation experiment, audit research in political science, field experiment of investigation results, convenient sampling in political science experiment, experiment using social media data, improving experimental processing of political science, spillover effect of experimental data, experimental reliability and scalability, transparency of experimental research, etc.

In the process of compiling this book, we should try our best to integrate it with practice. For example, the second chapter, joint investigation experiment, provides a user-friendly overview on how to design and implement joint investigation experiments and how to analyze the experimental results, and gives an example, that is, how the personal characteristics of those presidential candidates running for president trump in 2020 affect the support of survey respondents.

Because it is written by many scholars, the style and quality of each chapter are different.

Some chapters are quite wonderful. For example, Ms. ROC í o titiunik, the author of Chapter 6 "natural experiment" and a professor of politics at Princeton University, told us that the usage of the word "natural experiment" has not been agreed. Some people think that as long as the experimental treatment group is randomly assigned by people other than researchers, such an experiment is a natural experiment. Others believe that in order to meet the definition of natural experiment, there should be no controlled random distribution, but the experimental treatment group should be allocated by an external factor, which is roughly similar to random experiment. The third understanding is that as long as a non randomized study compares the treatment group with the control group, it is a natural experiment, and there are no specific requirements on how the treatment group is allocated.

The author puts forward the fourth definition of "natural experiment", which tries to clarify the indispensable characteristics of natural experiment and distinguish natural experiment from random controlled experiment. This definition is that if the processing allocation mechanism of a study meets the following conditions, the study is a natural experiment: (1) the processing allocation mechanism is not designed and implemented by researchers; (2) Researchers do not know what the processing allocation mechanism is; (3) The processing allocation mechanism is probabilistic because it depends on external factors.

The essence of this definition is that the difference between randomized controlled experiment and natural experiment is not in degree, but in essence. Therefore, the author believes that understanding natural experiment as a research design similar to random experiment is neither rigorous nor can it provide useful guidance for experimental analysis.

This chapter also discusses the difference between natural experiment and traditional observation research, and puts forward some practical suggestions for researchers who want to use natural experiment to study causal effects.

The author originally thought that he had a good understanding of natural experiment, but this chapter taught me a good lesson.