In 2021, over 10.78 million Chinese high school students took the college entrance exam (Gaokao): a historical high. While the Gaokao is the most important talent selection channel in China, many questions about the Gaokao system remain unanswered. Does the exam measure ability? What type of ability? Does the exam score make a difference for a young person’s future career? Are higher-score individuals more likely to work in government, become entrepreneurs, or earn higher wages?

It is impossible to provide comprehensive answers to all of these questions in a single study, but here, we provide partial answers to some of these questions. Our results come from a recent study, motivated by a theoretical literature on talent allocation (e.g., Baumol 1990; Murphy, Shleifer and Vishny 1991; Acemoglu 1995). This literature has long noted that talent is general and can be used in both the entrepreneurial and non-entrepreneurial sectors, and that its allocation depends on the reward structure of a society. While this important theory of talent allocation has been developed for three decades, few empirical studies have directly tested it due to data challenges.

In Bai, Jia, Li and Wang (2021), we study whether talented Chinese are more or less likely to become entrepreneurs. Empirically, we link the universe of college admission records in 1999–2003 with the universe of Chinese firms and their owners, and then use a random sample of 20% of the linked data to examine who have become entrepreneurs and how successful their firms are. In total, this yields a sample of 1.8 million college graduates who created approximately 170,000 firms by 2015. We supplement this linked data with a large survey of Chinese college graduates that we conducted during 2010–2015 to study waged jobs. We use students’ Gaokao score as a proxy for talent and validate our measure with data.

Research design: within-college comparison

We focus on within-college analyses -- comparing individuals in their mid-30s with others in their cohort who graduated from the same college -- for conceptual and empirical reasons. Conceptually, the within-college comparison helps to control for the confounding factors of college reputation and network. Empirically, we find that most of the variation in firm creation comes from within colleges. For instance, although college fixed effects can explain up to 19% of the variation in wages of paid jobs (in our survey data), the effects can account for only 1.2% of the variation in firm creation. In addition, we find that the college fixed effects can explain only 46% of the variation in exam scores, leaving the majority of the variation to occur within colleges. The wide variation in scores within a college is driven by considerable uncertainty and political economy factors (particularly provincial quotas) in the college admission process. In addition, because Gaokao scores are comparable only for students from the same province and year, as well as the same academic track (social science vs. natural science), we isolate province-year-track fixed effects in our analyses.

We control for a set of individuals’ personal characteristics, which include gender, Hukou (rural vs. urban), high school quality, birth county’s GDP per capita, and age fixed effects. Although we do not have a specific measure of socioeconomic status (such as parental income), it is reasonable to assume that those from better high schools and wealthier counties have higher socioeconomic status. Within colleges, higher socioeconomic status is positively associated with exam scores.