Human Capital Externalities in China

单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,*,*,Human Capital Externalities in China,Edward L. GLAESER and Ming LU,（,Harvard University, NBER;,Fudan,University,）,I. INTRODUCTION,Are human capital externalities (HCE) important in swiftly-growing economies like China?,HCE generate the increasing returns that enable long-run economic growth,(Lucas (1988).,Strong correlation across American metropolitan areas, between area-level human capital and individual earnings,(Rauch (1993) and,Moretti,(2004a) ),Chinas average years of schooling have increased from 3.7 to 7.5,(,Barro,and Lee, 2011),Chinas per capita GDP has gone up by more than 1200%.,Challenges in Estimation of HCE,Significant problems with interpreting the OLS coefficients (,Acemoglu,and,Angrist,(2001) and others).,Areas may have attracted more educated workers with greater economic opportunity - upward bias.,If more educated people are attracted to areas because of consumption amenities, then the extra labor elicited by these amenities should depress wages. downward bias.,If workers differ in unobserved ways, and if workers with more unobserved ability tend to sort with workers that have more observed ability- upward bias.,IV estimation of HCE,IV:,Acemoglu,and,Angrist,(2001) :,Using area level policies, like compulsory schooling laws.,Little evidence for human capital spillovers across U.S. States.,Moretti,(2004a):,The location of land grant colleges in the U.S. prior to 1940,Eevidence,of large HCE, similar in magnitude to the OLS estimates.,Land grant colleges operate by increasing the upper end of the human capital distribution, while compulsory schooling laws operate by raising the lower end of the human capital distribution.,If the spillovers came from learning from the more skilled, as in,Glaeser,(1999), then increasing the most skilled creates spillovers while increasing the least skilled does not.,One or both of the instruments are contaminated by correlation with omitted variables, either at the state-cohort level (in the case of,Acemoglu,and,Angrist, 2001) or that the metropolitan area level (,Moretti, 2004a).,II. A MODEL OF SKILLS AND LOCATION,(Skipped),III. DATA DESCRIPTION AND ORDINARY LEAST SQUARES RESULTS,Individual-level data from the 2002 and 2007 Chinese Household Income Project Surveys (CHIP2002, CHIP2007) for urban households.,The 2002 survey covers 70 cities and county towns from 10 provinces, 6,835 households and 20,632 individuals,The 2007 sample covers 19 cities and county towns from seven provinces, 5,000 households and 14,699 individuals.,The data of city-level per capita schooling are from population census data in 2000.,Other city level characteristics are from,China City Statistical Yearbook,.,Model,ln,w,ij,is the logarithm of individual-level hourly wage or monthly wage.,educity,j,is the average years of schooling at city level.,edu,ij,is individual-level years of schooling.,X,ij,is a vector of individual characteristics, including:,(1) Experience, defined as the difference between age and years of schooling minus 6;,(2) Gender dummy, male=1;,(3) Marital status, denoted by dummies of being married, being divorced, being widowed and other;,(4) Ethnic group dummy, minority=1;,(5) Dummy variables of occupation, sector, and ownership types of their working units.,City,j,:,City level characteristics.,(1) Public goods.,averoad2000, road area per capita in 2000.,avebus2000, the number of buses per capita in 2000;,(2),university, city-level number of universities.,(3) City size.,medium,and,big,;,(4) Dummy variables of provincial capitals and municipalities,Human Capital Externality (N=11,556 ),(1),(2),(3),(4),loghrsal,logsal,loghrsal,logsal,educity,0.197*,0.184*,0.119*,0.100*,(0.0459),(0.0467),(0.0466),(0.0446),gender,0.169*,0.212*,0.167*,0.210*,(0.0166),(0.0169),(0.0156),(0.0161),edu,0.0469*,0.0403*,0.0494*,0.0430*,(0.00427),(0.00360),(0.00403),(0.00330),exp,0.00218*,0.000528,0.00319*,0.00174,(0.00130),(0.00142),(0.00132),(0.00137),minority,-0.0483,-0.0322,-0.0236,-0.00978,(0.0470),(0.0458),(0.0403),(0.0399),City level var.,N,N,Y,Y,R-squared,0.782,0.498,0.790,0.520,Heterogeneity of Human Capital Externality by Education Group,(1),(2),(3),(4),(5),(6),Dep. Var.,Log hourly salary,Log monthly salary,edu,12,9,edu,12,edu,12,9,edu,12,edu,200,Working hour,7,Salary,200,Working hour,7,educity,0.119,*,0.112,*,0.100,*,0.0986,*,(0.0466),(0.0480),(0.0446),(0.0448),Observations,11,556,11,156,11,556,11,156,R-squared,0.790,0.802,0.520,0.521,IV. UNIVERSITY RELOCATION AND SHIFTING EDUCATION LEVELS,University Relocation,Motivation:,First, having learned the economic system from the Soviet Union, the Chinese leaders also wanted to follow their university system which is highly specialized to serve the economic development.,Second, to widespread the communism ideology, the Party was strongly motivated to remove the influence of the education system in the period of the Republic of China.,Moving across cities,Among 502 departments moved out, 282 were across cities, while 333 among 623 departments moved in were from a different city.,For 314 top scientists who experienced the movement of the relocation of university departments,232 among them, 74% of the 314, were relocated to other universities, colleges or institutions.,43 out of 158 top scientists who changed their working units within university system were moved across cities during the relocation of university departments.,38 out of 74 top scientists who were moved out of universities to other institutions migrated to other cities, while for those 17 from other units to universities, 10 of them left their living cities.,For balance? Not significantly.,A simple correlation between the city-level numbers of departments moved in and moved out.,However, the correlation coefficient is actually 0.44.,The Quantity of Universities in the 1950s,The Quantity of University Departments Moved In,The Quantity of University Departments Moved Out,The Net Number of University Departments Moved In,University Relocation and Regional Characteristics,Departments in,Departments out,Net departments in,No. of Universities,1.170,*,0.748,*,0.421,(0.242),(0.214),(0.334),Population in 1953,0.0267,*,-0.00201,0.0287,(in 10,000),(0.0125),(0.0111),(0.0173),Northeast,-3.097,-2.466,-0.631,(3.403),(3.009),(4.694),North,-6.480,*,-4.381,-2.099,(3.148),(2.784),(4.343),East,-4.920,*,-1.244,-3.676,(2.700),(2.387),(3.724),Southwest,-3.279,1.621,-4.899,(3.249),(2.873),(4.481),Northwest,-6.158,*,-5.781,*,-0.378,(3.599),(3.182),(4.964),Observations,53,53,53,R-squared,0.676,0.404,0.278,F-value,13.40,4.35,2.48,Other channels?,Social network?,A case:,Fudan,vs. Zhejiang U.,Through investment in the 1950s and 1960s?,Formally tested.,The Determinants of Per Capita Fixed Asset Investment in the 1950s and 1960s,(1),(2),(3),(4),(5),(6),department_net,0.013,0.002,0.017,0.013,department_out,0.020,0.002,0.020,0.016,department_in,0.027,0.004,0.017,0.014,fix_4952,0.408,*,0.408,*,0.403,*,0.077,0.077,0.080,Constant,-15.036,*,-15.144,*,-15.183,*,-8.069,*,-8.081,*,-8.180,*,0.136,0.175,0.161,1.312,1.347,1.397,Observations,48,48,48,48,48,48,R-squared,0.013,0.021,0.055,0.395,0.395,0.395,The Determinants of Per Capita Infrastructure Investment in the 1950s and 1960s,(1),(2),(3),(4),(5),(6),department_net,0.015,0.004,0.017,0.013,department_out,0.012,-0.007,0.022,0.017,department_in,0.022,-0.000,0.016,0.014,infra_4952,0.387,*,0.397,*,0.391,*,0.075,0.075,0.078,Constant,-15.075,*,-15.134,*,-15.192,*,-8.444,*,-8.226,*,-8.372,*,0.136,0.176,0.160,1.281,1.314,1.363,Observations,45,45,45,45,45,45,R-squared,0.019,0.008,0.042,0.403,0.404,0.402,V. HUMAN CAPITAL EXTERNALITIES BASED ON UNIVERSITY RELOCATION,IV Estimation for Human Capital Externality,First stage,Second stage,Dep. Var.,educity,loghrsal,logsal,department_in,0.0342,*,educity,0.219,*,0.162,(0.00680),(0.123),(0.119),department_out,-.0270,*,edu,0.0477,*,0.0420,*,(.00909),(0.00374),(0.00337),exp,0.00283,*,0.00152,(0.00137),(0.00143),gender,0.166,*,0.209,*,(0.0153),(0.0159),F test,20.119,Observations,11,556,11,556,R-squared,0.787,0.516,IV Estimation for Heterogeneity of Human Capital Externality by Education Group,(1),(2),(3),(4),(5),(6),Dep. Var.,Log hourly salary,Log monthly salary,edu,12,9,edu,12,edu,12,9,edu,12,edu,9,educity,0.224,*,0.208,0.245,*,0.196,0.140,0.177,*,(0.132),(0.146),(0.104),(0.124),(0.150),(0.0851),Obs,.,3,533,4,520,3,503,3,533,4,520,3,503,R-squared,0.778,0.782,0.769,0.472,0.484,0.492,IV Estimation for Heterogeneity of Human Capital Externality by Industry,(1),(2),(3),(4),(5),(6),Dep. Var.,Log hourly salary,Log monthly salary,abstract,manual,Manuf,.,abstract,manual,Manuf,.,educity,0.231,*,0.250,*,0.214,*,0.197,*,0.157,0.194,*,(0.123),(0.145),(0.0908),(0.119),(0.152),(0.0887),edu,0.0295,*,0.00662,0.0247,0.0307,*,-0.00536,0.00994,(0.00786),(0.0345),(0.0159),(0.00725),(0.0310),(0.0140),exp,0.00755,*,0.000737,0.00293,0.00673,*,-8.88e-05,-0.000166,(0.00222),(0.00383),(0.00282),(0.00204),(0.00321),(0.00270),Obs,.,2,479,1,798,1,157,2,479,1,798,1,157,R-squared,0.814,0.771,0.794,0.528,0.491,0.469,VI. CONCLUSION,The OLS regressions show that 1 year more in the city-level education leads to 11.9% increase in hourly wage.,After using IV, the human capital externality is raised to 21.9%, almost twice the OLS estimates.,The change in the estimate of human capital externality is greater for the most skilled workers,The great human capital externality amplifies the returns to education, and well explains why China has grown so rapidly.,The policy implication:,free mobility of labor,In China, where urbanization and agglomeration are still constrained by the,hukou,system, growth can be fostered through human capital externality in cities, especially when knowledge and information sharing are becoming increasingly important.,Thanks!,-,Comments welcome,