【正文】 and the productivity of all goods and per capital real ine increase. The purpose of the paper is to explain several important phenomena of urbanization, including the urban landrent escalation, decreases in the relative per capita consumption of land in the urban and rural area, increases in the population size of the urban area pared to the rural area, and the absolute increase of diversity of occupations in the urban area as well as relative to that in the rural area, as different aspects of evolution in the division of labor. The intimate relationship between cities and the division of labor has indeed long been recognized by Xenopnon (Gordon 1975), William Petty (1682), Alfred Marshall (1890, ), Mills (1972), Scott (1988) and others. However, as noted by Stigler (1976, pp. 12091210), there was no formal theory of the division of labor and specialization in the mainstream economics: “The last of Smith?s regrettable failures is one for which he is overwhelmingly famous – the division of labor…… . (A)lost no one used or now uses the theory of division of labor, for the excellent reason that there is scarcely such a theory. … Smith gave the division of labor an immensely convincing presentation – it seems to me as persuasive a case for the power of specialization today as it appeared to Smith. Yet there is no evidence, so far as I know, of any serious advance in the theory of the subject si。 the urban landrent increases absolutely as well as relative to that in the rural area, the relative per capita lot size of residence in the urban and rural areas decreases。ckerman and Ilmakunnas, 2021). In the remainder of the paper, we will refer to this as the fixed effects ordinal logit. Using Verbeek and Nijman?s (1992) test, we found some evidence of attrition bias in the simple logit However, once fixed effects are taken into account。 B246。 Blundell et al, 2021。 and, reported symptoms experienced in the four weeks preceding the survey (fever, headache, rash, diarrhea, joint pain, heart problems or others). Table 2 shows the means of these more objective health indicators. Except for risk factors for chronic conditions, such as obesity and hypertension, most of the ill health indicators have a very low prevalence rate. Therefore, we create a binary variable that equals one if the respondent reported to suffer from at least one of the physical impairments. Also we use a binary variable to indicate whether the respondent reported any of the symptoms in the four weeks preceding the survey. Table 3 confirms that SAH is correlated with each of the more objective measures of health. The first four columns show marginal effects from probity models explaining the probability of reporting fair or poor health. All of the more objective indicators of illhealth are significantly related with an increased probability of reporting fair or poor health, indicating that the latter binary measure captures at least some of the information contained in these more specific measures. The last three columns of Table 3 show marginal effects on the probability of dying by the subsequent wave. These results show that reporting fair or poor health at time t is predictive of mortality by the subsequent wave (column 5), increasing the baseline probability of dying by about one third, and that this predictive power remains after controlling for the set of more objective health indicators (column 6). This demonstrates that not only is the reporting poor or fair health strongly correlated with the other health indicators。 measured hypertension。 obesity (Body Mass Index(BMI)30)。 the time trend in the index indeed reveals increasing urbanization。 Wooldridge, 2021). The main health oute used in the paper is selfassessed health (SAH), reported on a fourpoint scale from excellent to poor. This general measure of adult health has repeatedly been shown to be highly predictive of mortality, even conditional on physiological measures of health (Idler and Benyamin, 1997). We show that SAH predicts mortality in the CHNS and demonstrate that it is highly correlated with more specific health outes such as obesity, hypertension, physical impairments and symptoms of illness. We also estimate the impact of urbanization on these narrower, but more objective, measures of health status. To our knowledge, this is the first paper to estimate the causal effect of urbanization on health from longitudinal data on both individuals and munities. These data allow us to identify the effect of urbanization by paring the health transitions of individuals living in areas that experience rapid transformations to an urban environment with those living in areas that remain rural. We find important, and robust, negative effects of urbanization on health. Urbanization increases the probability of reporting fair or poor health by 5 to 15 percentage points, with a greater degree of urbanization having larger health effects. While people in more urbanized areas are, on average, in better health than their rural counterparts, the process of urbanization is damaging to health. Urbanization raises the probability of suffering from physical impairments, disease symptoms and hypertension, but there is no significant impact on obesity or undernutrition. In the remainder, we first present the CHNS data, and explain construction of the urban city index. This is followed by an explanation of our identification strategy, estimation methods and the various definitions of urbanization used. In the fourth section, we first present the main results for the impact of urbanization on SAH, and then check their robustness, before examining the impact on other health outes. The concluding section provides an interpretation of the implications of the study and acknowledges its limitations. We use the China Health and Nutrition Survey (CHNS)