Look for the randomness in these chaotic logics, and then use the knife of randomness to cut off other factors that interfere with our judgment and find out the causality we hope to really observe.
On the afternoon of October 11, the Royal Swedish Academy of Sciences announced the 2021 Nobel Prize in Economics . Among them, economist David Card (David Card) won half of the award for his "empirical contribution to labor economics", and the other two economists Joshua Angrist and Guido Imbens shared the other half of the award for his "methodological contribution to causality analysis".
The Royal Swedish Academy of Sciences pointed out that this year's three winners provided new insights into the labor market and showed that the conclusions about causality can be obtained from natural experiments. Their research methods have been extended to other fields and completely changed the empirical research , covering many aspects of our lives, such as medical insurance, pensions, labor compensation, etc. Let's talk about it...
Natural Experiment
In the public’s impression,Experimenting is equivalent to scientific research. The same is true in the field of economics. If we don’t do experiments and compare them only through the observed data, we will often draw some dumbfounding conclusions.
Many major issues in social science involve causality, such as the impact of different local government subsidies on enterprises, and how longer education affects a person’s future income. But these causality questions have always been difficult to answer because it is difficult to conduct comparative studies on them.
For example, some companies have been plagued with debts recently, so if we want to study the impact of government subsidies on corporate debt, what should we do? It is not possible to directly compare different companies in different regions, because companies often observe various preferential treatments offered by the local government to choose whether to invest and build factories there. Because of this, there are too many fields that economics focuses on that cannot be experimented with, but they are also very important. For example, the reform of medical insurance and pension systems: Before the policy is introduced, there can be many arguments, but if it happened, it happened. It is impossible for us to return to the state where it did not happen. Change another choice and observe it again.
But this year’s winners confirmed that one thing is that natural experiments can be used to answer such difficult social questions. The key is to use accidents or policy changes to cause the population to be treated differently, similar to the way of medical clinical trials.
Example: Analyze the impact of increasing the minimum hourly wage from US$4.25 to US$5.05 in New Jersey, the United States.
The traditional view of economists is that an increase in the minimum wage will lead to a decrease in employment because it will increase the wage costs of enterprises-but the evidence for this causal relationship is not entirely convincing.But just studying what happened to New Jersey after this rise does not show any impact, so Carders used a control group with no change in salary but all other factors being the same: they chose a fast food restaurant—a salary Industries where low minimum wages are important.
Conclusion: Two economists compared fast food restaurants in New Jersey with Pennsylvania border fast food restaurants in the opposite economic study. Yes, raising the minimum wage has no effect on the number of employees.
Double difference and breakpoint regression
is the natural method used in the immigration case and the above-mentioned analysis of the minimum wage in the case of David, education, and the above analysis The double difference used in the impact of the labor market is very simple:
The first step is to subtract the labor supply at the time when the minimum wage has not changed in New Jersey and Pennsylvania. The result is two adjacent ones. Regional differences;
The second step, after the New Jersey Act is implemented, the labor supply in the two regions will be reduced by one time, and the new regional differences will be obtained;
The third step ,Subtracting the new regional differences from the inherent differences obtained in the first step, we get the policy effect of raising the minimum wage.
But we will find that in some cases, the control group cannot be found, what should we do? Then use "Breakpoint Regression".
For example, to study the impact of scholarships on future studies, whether students receive scholarships depends on the test scores. Because the exam will inevitably have a problem of performance, that is, a narrow range between the score line and the score line, the scores just meet the award criteria and the students who almost meet are comparable. If the test score is greater than the award-winning standard score, enter the processing group; if the test score is less than the award-winning standard score, enter the control group. Therefore, the processing variables form a breakpoint at the award-winning standard score. The main idea of the research design is to use samples close to this breakpoint to effectively estimate the processing effect.
Angrist used a system of the Israeli education system to perform breakpoint regression when studying the impact of class size on performance; this system limits the upper limit of class size to 40 students, and once more than 40 students (such as 41 Students), the class is divided into two.
The randomness of policies and the randomness of personal play can be cleverly constructed and become natural experiments to study some deeper problems under the surface. In the following 30 years, this method did not attract the attention of academia. Until 1990, breakpoint regression design began to be applied in various fields, and in recent years it has become the most important research method in the field of causal analysis and policy evaluation.
Finally,The award speech to the three said: They provided new insights on the labor market and showed that conclusions about causality can be drawn from natural experiments. Their methods have been extended to other fields and revolutionized empirical research.
.
