James heckman's Academic Contribution

Heckman and Daniel McFadden, another winner of the 2000 Nobel Prize in Economics, have made great contributions to the establishment and development of individual econometrics and made outstanding contributions to the theory and methods of microeconomics. They have designed analytical methods to study people's lifestyle decisions, which have been widely used in the statistical analysis of individuals, families and enterprises in economics and other social disciplines. These theories and methods are of great practical significance to the study of social and economic problems such as education and training plan, urban transportation system and housing for the elderly. To understand their contribution, we must first understand the individual econometric research. The so-called individual econometrics refers to the quantitative study of the behavior of economic individuals such as families and manufacturers. The research object is very extensive, mainly including labor economics topics: labor supply, wage determination, education choice, unemployment period, immigration, career choice, reproductive choice, gender discrimination, racial discrimination and so on. The theme of public economics: the influence of tax policy and social welfare: the research theme of consumer behavior: commodity demand, brand choice; Theme of urban and transportation economics: housing rental and purchase choice, location choice and transportation choice; Special topics of industrial economics: selection of production forms, demand of production factors, evaluation of production efficiency, etc.

It is worth mentioning that an important reason for the rapid development of econometric research on economic individuals in the past few decades is that a number of large-scale "individual data" databases have appeared during this period. The so-called "individual data" is the data collected by economic individuals such as families and manufacturers, generally in the form of "cross section" (that is, investigating multiple economic individuals at a certain time point), but in recent years, "tracking data" (that is, investigating the same batch of economic individuals at multiple time points) has become more and more popular, and tracking data began at the end of University of Michigan 1960.

The introduction of these large-scale individual databases not only helps to verify the existing economic theories more accurately and rigorously, but also leads to many new econometric topics, mainly focusing on the characteristics of individual data itself. In the process of analyzing and discussing these new topics, econometric methods have also developed by leaps and bounds, and the research on individual econometrics has also grown sturdily with the popularization of individual databases. In addition, due to the popularization of computers and the rapid improvement of computing power, it is feasible to process a large number of individual data, which is also an important reason for the progress of individual econometrics research. Simply put, microeconomics is a marginal discipline between economics and statistics, including economic theories and statistical methods used to analyze micro data. That is, the microscopic data reflecting individuals, families and enterprises are analyzed through economic theory and statistical methods, from which more essential economic information reflecting society can be obtained. Microscopic data generally have two forms of expression: one is called cross-sectional data, which is a collection of different situations at the same time; The other is called longitude data, that is, a collection of continuous situations of the same observation unit at different points (such as years). In the past 30 years, the field of microeconomics has been rapidly expanded due to the emergence of large databases including micro-data. At this time, the micro-data is more and more easy to obtain and the computing power is stronger and stronger, which provides brand-new convenience and possibility for the empirical analysis and test of microeconomic theory; On this basis, researchers have the ability to examine and analyze many new problems about "individuals". With the development of econometrics, the application of micro-data has also produced many new statistical problems. Especially because of the inherent limitations of non-experimental data, researchers can usually only observe specific individuals of certain variables, so this non-random sampling sample cannot be representative in the whole population. Even if the sample is representative, it is impossible to observe some characteristics that affect individual behavior, which makes it difficult or even impossible to explain some differences between individuals. James heckman's theory fills these gaps, and he puts forward some solutions to some basic statistical problems related to this kind of micro-data analysis, which promotes the development of econometric theories and methods. The collection of individual data is mostly carried out under the condition of incomplete random sampling, and the reason why sampling is not random is because the observed values of individual data come from economic individuals such as families and manufacturers, and these economic individuals themselves (or other economic individuals around them) have the ability to choose and judge, so it is likely that some actions will be taken to affect sampling, which will make sampling lose randomness and the collected samples cannot accurately represent the matrix. For example, we can only get information about working hours and wages from people who have jobs, but there is always a large proportion of people who have no jobs in the total population. It is impossible for any database to include these people's working hours or wages. No matter how objective and random the sampling process is, the working hours or wages obtained are not strictly representative. If traditional econometric methods are used to analyze such data, any inference can only represent the behavior of workers, but not the behavior description of the whole population. If we still interpret the empirical results as universally applicable findings, we will definitely make a mistake of generalizing, which is the so-called "sample selection error".

Heckman not only made great contributions to the theory of individual econometrics, but also conducted many in-depth empirical studies. He has obtained a wealth of empirical research results, and provided many original opinions on topics such as labor supply, wage determination, unemployment period, benefit evaluation of labor market consultation plan, fertility and gender discrimination. The so-called "duration" refers to the duration of an event. The application of econometric analysis of duration in economics includes unemployment duration, strike time, business cycle, consumer shopping time and many demographic topics, such as marriage, childbirth, longevity and migration. Heckman also made considerable contributions to the study of duration, and he paid special attention to the problem of "hidden differences" in duration data. Now take the analysis of unemployment period as an example to illustrate the influence of hidden differences: among the unemployed, the unemployed with better quality are more likely to find new jobs, so their unemployment period is shorter, while the unemployed with relatively poor quality will definitely have a longer unemployment period. Therefore, the difference between "long-term unemployment sample group" and "short-term unemployment sample group" may not be completely random, but belongs to the difference between two different groups with fundamental differences in quality. What is the difference between the two groups is usually not fully confirmed, so these unrecognizable quality differences are called hidden differences. In other words, the difference in unemployment period is probably caused by the hidden difference that cannot be confirmed. If there are too many,

In this discussion, we should see that the influence of hidden differences on duration analysis is similar to that of sample selection, and the processing of sample selection has always been heckman's interest. In order to solve the problem of hidden differences, heckman put forward some measurement methods without mother number, which were widely used by empirical researchers in continuous period. Heckman himself has done a lot of empirical research on the topic of unemployment and fertility. Sampling is the basic problem in econometrics, and sampling deviation and self-selection are the most basic problems in microeconomics research. If a sample can't randomly represent its population, sampling deviation may occur. Generally speaking, a sample is either the result of data collection rules or the result of economic man's own behavior. The latter is a process of self-selection. Heckman's breakthrough in self-selection mainly occurred in the mid-1970s. These theoretical breakthroughs are closely related to his painstaking research on individual decisions of labor participation and working hours. When only the working hours of those individuals who choose to work are investigated, they may encounter samples with self-selection problems. In an article published in 1974, heckman designed an econometric method to solve the problem of self-selection when studying the labor supply of married women. This research has become an excellent example of the combination of microeconomic theory and microeconomic method.

Heckman put forward another method to solve the problem of self-selection in his subsequent research work, namely the famous heckman correction method, also known as the two-stage method or heckman method. This method is extremely convenient to apply and has far-reaching influence. Heckman correction method is divided into two steps: first, the researcher designs a model to calculate the individual working probability according to economic theory, and the statistical estimation results of the model can be used to predict the probability of each individual; In the second step, the researchers combined these predicted individual probabilities into an additional explanatory variable to correct the self-selection problem together with variables such as education and age. In this way, the estimated wage relationship is statistically appropriate.