xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200409Ka 4500
controlfield tag 001 002069487
007 cr bnu|||uuuuu
008 100422s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003279
Cook, Finnie B.
Globalization, migration and the U.S. labor market for physicians :
b the impact of immigration on local wages
h [electronic resource] /
by Finnie B. Cook.
[Tampa, Fla] :
University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 123 pages.
Dissertation (Ph.D.)--University of South Florida, 2009.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
ABSTRACT: The healthcare labor market has experienced some significant changes in the last half century, including the establishment of Medicare and Medicaid in 1965, the emergence of managed care in the 1980s, and the worldwide mobility of labor encouraged by globalization. Currently, more than 25% of physicians working in the U.S. are foreign-born. The existing body of literature related to the impact of immigration on local wages has to date found conflicting results. The purpose of this research is to evaluate the impact of immigration of foreign physicians on local physician wages. This study employs physician survey data from the AMA Physician Masterfile for the years 1997 through 2007 combined with wage data published by the Bureau of Labor Statistics and data from other government sources.Several econometric models are employed to analyze the wage impacts of immigration, including ordinary least squares, fixed effects, two-stage least squares and a first-difference approach to control for endogenous location choice. The results of this study provide evidence that in the short-run, the impacts of immigration of physicians on area wages is small but positive. In the long run, however, wages adjust and the impact becomes negative and statistically significant, although the magnitude of the impact of a one percentage point increase in the share of immigrant physicians in an area is less than 0.2%. The negative wage effects of immigration tend to be larger for foreign-born physicians educated in the U.S. compared with foreign-born international medical graduates. The study also finds evidence that the negative effects of immigration tend to be offset by outflows of the lowest paid native physicians.Furthermore, physicians tend to locate in areas where wages are already higher, and foreign-born physicians are more likely than their native counterparts to work in larger cities as opposed to rural areas. The research has important policy implications in the presence of current debate over immigration law and healthcare reform and in an era of increasing mobility of labor due to globalization.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Advisor: Gabriel Picone, Ph.D.
Ordinary least squares
Two-stage least squares
t USF Electronic Theses and Dissertations.
Globalization, Migration and the U.S. Labor Market for Physicians: The Impact of Immigration on Local Wages by Finnie B. Cook A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Economics College of Business University of South Florida Major Professor: Gabriel Picone Ph.D. Donald Bellante, Ph.D. Mark Herander, Ph.D. M urat Munkin, Ph.D. Date of Approval: November 5, 2009 Keywords: ordinary least squares, fixed effects, instrumental variable, two stage least squares, first difference, medical care Copyright 2009 Finni e B. Cook
Dedication I would like to dedicate this dissertation to my husband, Joel, and to my parents, Karl and Carolyn Riedl. To my husband Joel, for being there for me through out the entirety of my graduate studies, for sympathizing with me during times o f stress and always understanding when the demands of my academic life took priority over those at home I owe you a great deal of thanks. You never doubted that I would succeed in this endeavor and the experience was made so much more enjoyable with your love and encouragement throughout. I also thank my parents, who always encourage d me to reach as high as possible, not only with regard to education but also in life. Without your guidance, I would not have had the maturity and dedication required to ac hieve this goal. You always remind me that I am capable of anything I wish to accomplish as long as I put my mind to it, and always believed in my ability to succeed. I am blessed to have such a wonderful family support system.
i Table of Contents List of Tables iii Abstract vi Preface viii Chapter 1: Introduction 1 Chapter 2: Globalization and Immigration 2 2.1: Migration in the Contemporary Period 2 2.2: History of Immigration and Immigration Law in the U.S. 2 2.3: Changes in Immigration in the Contemporary Period 3 Chapter 3: Literature Review 6 3.1: The Effect of Immigration on Native/Local Wages 6 3.2: Area Analysis Approach 7 3.3: Factor Proportions An alysis 11 3.4: Other Research 13 3.5: Immigration of Physicians 15 Chapter 4: Justification for Study 19 4.1: Academic Value 19 4.2: Immigration and Healthcare Policy 20 Chapter 5: Immigrant Physicians: Visa Requ irements 23 Chapter 6: Theoretical Foundations 26 6.1: Physician Labor Demand and Supply 26 6.2: Physician Immigration: Decision to Emigrate 28 6.3: Exceptions to the Rule in the Healthcare Industry 29 6.4: Hypoth eses of this Research 30 Chapter 7: Research Methodology 32 7.1: The Basic Ordinary Least Squares Model 32 7.2: Extension of the Basic Model: Additional Variables 34 7.3: Extension of the Basic Model: Instrumental Variable A pproach 35 7.4: Extension of the Basic Model: First Difference Approach 37
ii Chapter 8: Data Sources and Building the Database 39 8.1: The AMA Physician Masterfile 39 8.2: Wage Data from the Occupational Employment Statistics Survey 41 8.3: Medicare Enrollment Data 42 8.4: Data from the U.S. Census 43 Chapter 9: Descriptive Statistics of the Data 45 9.1: Country of Physician Birth 45 9.2: Country of Medical School Attended 47 9.3: Physician Practice Area 48 9.4: Physician Demographic Characteristics 50 9.5: Wages of Physicians: Summary Statistics 55 Chapter 10: Regression Results 65 10.1: Variable Summary 65 10.2: Ordinary Least Squares and Two Stage Least Squares Results 67 10.3: OLS with Area Fixed Effects and First Difference Results 76 10.4: Two Stage Least Squares First Difference Results 82 10.5: Analyses with Specialty Controls 84 Chapter 11: Concl usions of the Study based on Regression Analyses 9 3 Chapter 12: Healthcare and Immigration Policy Implications 96 References Cited 99 Bibliography 103 Appendices 104 About the Author End Page
iii List of Tables Table 1: Summary of Existing Empirical Literature: Labor Market Effects o f Immigration 17 Table 2: Physician Sample: U.S. Born and Foreign Born by Year 45 Table 3: Summary Statistics of th e Variable I = Immigrant Share 46 Table 4: Foreign Born Physicians Practicing in the U.S. by Country of Birth 46 Table 5: Location of Medical School Attended by Place of Birth 47 Table 6: Summary Statistics of the Birthplace / Education Variables 48 Table 7: U.S. and Foreign Born Physicians by Practice Area Size 48 Table 8: Foreign Born Physicians by MSA 50 Table 9: Physicians by Age and Size of Practic e Area 51 Table 10: Se x of Physic ian by Place of Birth and Year 52 Table 11: Physician Specialty by Sex 53 Table 12: Physician Specialty by U.S. / Foreign Birth and Education 54 Table 13: A verage Physician Wages by Year 55 Table 14: Average Physician Wages by Sex 56 Table 15: Comparison of Means Test: Wages by Sex 56 Table 16: Average Physician Wag es: U.S. Born and Foreign Born 57 Table 17: Comparison of M eans Test: Wages by Birthplace 58 Table 18: Average Physician Wages: by Location of Medical Schoo l Attended 59
iv Table 19: Comparison of Means Test: Wages by Location of Medical School Attended 59 Table 20: Comparison of Means Test: Wages by Birthplace and Location o f Medical School Attended 61 Table 21: Average Wages by Phys ician Specialty 62 Table 22: Average Wages by Physician Specialty and Birthplace 63 Table 23: Average Wages by Physician Specialty and Location of Med School Attended 63 Ta ble 24: Summary of Variables 66 Table 25: Pairwise Correlations between Independent Variables 6 7 Table 26: OLS and 2SLS Estimates of the Relationship between Immigration and Physician Wages 68 Table 27: Fixed Effects and First Difference Estimates of the Relationship between Immigration and Physician Wages 78 Table 28: 2SLS First Difference Estimates of the Relationship between Immigration a nd Physician Wages by Specialty 83 Table 29: OLS and 2SLS Estimates of the Relationship between Immigration and Physician Wages with Specialty Controls 85 Table 30: Fixe d Effects and First Difference Estimates of the Relationship between Immigration and Physicia n Wages with Specialty Controls 88 Table 31: 2SLS First Difference Estimates of the Relationship between Immigration and Physician Wages with Specialty Cont rols 91 Table A1: Physicians by Country of Birth 105 Table A2: Foreign Born Physicians by MSA 111 Table A3: Sex of Physicians by Place of Birth and Yea r 119
v Table A4: 2SLS Estimates of the Relationship between Immigration and Physician Wages: F irst Stage Regression Results 120 Table A5: 2SLS First Difference Estimates of the Relationship between Immigration and Physician Wages: First Stage Regression Results 121 Table A6: 2SLS Estimates of the Relationship between Immigration and Physi cian Wages with Specialty Controls: First Stage Regression Results 122 Table A7: 2SLS First Difference Estimates of the Relationship between Immigration and Physician Wages with Specialty Controls: First Stage Regression Results 123
vi Globalization, Migration and the U.S. Labor Market for Physicians: The Impact of Immigration on Local Wages Finnie B. Cook ABSTRACT The healthcare labor market has experienced some significant changes in the last half century including the establ ishment of Medicare and Medicaid in 1965, the emergence of managed care in the 1980s, and the worldwide mobility of labor encouraged by globalization. Currently, more than 25 of physicians working in the U.S. are foreign born. The existing body of litera ture related to the impact of immigration on local wages has to date found conflicting results. The purpose of this research is to evaluate the impact of immigration of foreign physicians on local physician wages. This study employs physician survey dat a from the AMA Physician Masterfile for the years 1997 through 2007 combined with wage data published by the Bureau of Labor Statistics and data from other government sources. Several econometric models are employed to analyze the wage impacts of immigra tion, including ordinary least squares, fixed effects, two stage least squares and a first difference approach to control for endogenous location choice. The results of this study provide evidence that in the short run, the impacts of immigration of phys icians on area wages is small but positive. In the long run, however, wa ges adjust and the impact becomes negative and statistically significant, although the magnitude of the impact of a one percentage point increase in the share of immigrant
vii physicians in an area is less than 0.2. The negative wage effects of immigration tend to be larger for foreign born physicians educated in the U.S. compared with foreign born international medical graduates. The study also finds evidence that t he negative effects o f immigration tend to be offset by outflows of the lowest paid native physicians. Furthermore, physicians tend to locate in areas where wages are already higher, and foreign born physicians are more likely than their native counterparts to work in larger cities as opposed to rural areas. The research has important policy implications in the presence of current debate over immigration law and healthcare reform and in an era of increasing mobility of labor due to globalization.
viii Preface I would like to express my gratitude to everyone who helped make this dissertation possible. I would like to thank my dissertation committee members: Donald Bellante, Ph.D., Mark Herander, Ph.D., and Murat Munkin, Ph.D. for taking the time to serve on my committee, review my research and provide their expert advice. I would especially like to thank my major professor, Gabriel Picone, Ph.D., for spending countless hours of his busy schedule assisting me in this research. I would not have be en able to complete this study without his expert knowledge and guidance. I would like to thank the Gaiennie Endowment Fund at the USF College of Business for funding the purchase of the AMA Physician Masterfile Data. This data was absolutely essential t o completion of this research. Thanks to Rena Berens at Medical Marketing Association, Inc. for arranging the specific data purchase I needed within my specified budget, and to Joe MacDougald for taking the time to assist me in prepar ing the data for anal ysis and for helping me with database management and STATA. I am especially grateful to my employers and colleagues at Deiter, Stephens, and Durham: to Dr. Deiter and Dr. Stephens for encouraging me to enter the Ph.D. program in the first place and for being so flexible in allowing me to pursue my academic studies on a full time basis, and to Dr. Durham for his help with everything from homework assignments to proofreading the final version of my dissertation.
ix I would like to thank all the professors, both during my undergraduate studies at the University of Florida and my graduate career at the University of South Florida, that taught me the economics required to succeed both as a student as well as in my future career as an e conomist. I would especi ally like to recognize David Denslow, Ph.D. at UF whose undergraduate macroeconomics course was the reason I chose to study economics, and who has been a great mentor to me. Also, special thanks to Michael Loewy at USF, for being such a patient advisor a nd always responding to my every question with a wealth of information. Lastly, I acknowledge all the graduate students at USF with whom I spent hours solving economic homework problems, preparing assignments, studying for exams and simply sharing ideas and knowledge. I thank you all and wish you all the best.
1 Chapter 1: Introduction The market for labor in the medical care industry has experienced some significant changes in the last half century. The establishment of Medicare and Medicaid in 1965, followed by the emergence of managed care in the 1980s, has dramatically changed the way physicians operate their practices and receive payment. Furthermore, globalization has encouraged worldwide mobility of labor and the U.S. labor market for medical professionals no longer consists only of native workers; on t he contrary, data published in the 2009 Statistical Abstract of the United States indicate that as of 2006, more than 25 of physicians in the U.S. were educated at foreign medical schools. In addition, many graduates of U.S. medical schools are foreign s tudents who often stay in the U.S. after achieving their degree. The following research analyzes immigrant doctors: who are they demographically, where do they come from, and where do they locate? Furthermore, this paper investigates the impact of physic ian immigration on the average wages of doctors in the U.S.
2 Chapter 2: Globalization and Immigration 2.1: Migration in the Contemporary Period One important feature of modern globalization involves migration. The world has seen recent growth in i mmigration to developed countries. The number of working age individuals born in one country and living in another country increased from 42 million in 1990 to 59 million in 2000, or 1.7 million per year, on average (Docquier 2005). According to Massey a nd Taylor (2004), contemporary international migration unfolds in a context of the globalization of markets. Castles and Miller (1998) posit, while movements of people across borders have shaped states and societies since time immemorial, what is disti nctive in recent years is their global scope, their centrality to domestic and international politics and their enormous economic and social consequences. The United States is by far the worlds largest immigrant destination. Although the relative inten sity of immigration into the U.S. was greater in the 19 th century than in the contemporary period of globalization, there has been a change in the kind of migrant in the contemporary period. Changes in U.S. immigration law since 1952 have resulted in an i ncrease in immigration of high skilled workers as well as persons from Asia and Latin America. 2.2: History of Immigration and Immigration Law in the U.S. In 1952, the Immigration and Nationality Act (INA), the first legislation emphasizing labor qualific ation as a preference for entry into the United States, became
3 law. Amendments to the Immigration and Nationality Act in 1965 opened the doors to immigrants from non European countries including Asians and Latin Americans. The 1990 Immigration Act was de signed to increase the number of immigrants admitted on the basis of skill level, but also maintained the importance of family reunion and refugee acceptance (Gardner and Bouvier 1990, Castles and Miller 1998). Current immigration law provides preferentia l treatment to family members of U.S. citizens, and otherwise allows for immigration of three types of immigrants: employment based immigrants with their spouses and children, refugees and asylum seekers, and diversity immigrants (immigrants from countries which have been underrepresented in recent immigration) (Massey 2004). 2.3: Changes in Immigration in the Contemporary Period Prior to World War II, most U.S. immigrants were low skilled workers employed in menial jobs in the public services and dirty j obs in the manufacturing sector (Held 1999). As a result of the changes in legislation since 1952, increasing numbers of migrants are employed in private and domestic service industries. Furthermore, a steady movement of highly skilled, highly trained p rofessionals (elite migrants) has been occurring (Held 1999). This phenomenon has been referred to in the development economics literature as the brain drain from developing to developed countries. In addition to the changes in legislation, Castles a nd Miller (1998) argue that an increase in the international mobility of highly qualified personnel, in both temporary and permanent flows is also attributable to globalization. The Immigration Act of 1990 further emphasized selection of high skilled wor kers, employing a system of quotas favoring
4 candidates with academic degrees or specific professional skills (Docquier 2005). Accor ding to the 2005 07 American Community Survey s 19.1 of foreign born U.S. citizens had a bache lors degree, compared with 17 .4 of native U.S. citizens (only 12.9 of foreign born non U.S. citizens living in the United States had a bachelors degree). Furthermore, 12 .6 of foreign born U.S. citizens and 9.3 of foreign born non U.S. citizens living in the U.S. had advanced deg rees beyond the bachelors, compared with 9.7 of native U.S. citizens. The immigrants possessing bachelors degrees or higher overwhelmingly come from Asia. The largest migration of Asians in recent history occurred after the passage of the 1965 amendment s to immigration law. The number of migrants from Asia increased from 17,000 in 1965 to more than 250,000 on average annually in the 1980s, to over 350,000 per year in the early 1990s. In addition, the number of Asian students studying in the U.S. increa sed from 82,709 in 1965 to 453,787 in 1995. This statistic is important, because it is not uncommon for students to come to America to study and eventually stay in the country; in the late 1960s, 90 of students from Taiwan and Korea who came to the U.S. for training never returned home. Since 1980, more than half of all U.S. doctorates awarded in the field of engineering have gone to foreigners, predominantly Asian. According to the 1990 Census, over 60 of immigrants over age 25 from Taiwan and India h ad college degrees, and over 45 of the immigrants from these countries hold management or professional positions (Castles 1998, Massey 2004). Waldinger and Gilbertson (1994) report that in the 1980s, a substantial portion of the new immigrants is far m ore highly skilled than native whites of native parentage. They found that Asian and Iranian immigrants, in particular, had substantially higher
5 levels of education, on average, than their native white counterparts. However, of non European immigrants, only the Japanese had higher average per capita income than the native white population. According to Madhavan (1985), the occupational composition of Indian immigrants in the United States was much more professionally and technically oriented than the U .S. labor force as a whole. Over 80 percent of the Indian immigrant labor force in 1981 was engaged in professional, technical, and management categories of employment as against 27 percent for the nation as a whole.
6 Chapter 3: Literat ure Review 3.1: The Effect of Immigration on Native/Local Wages Imm igration law is currently a topic of national debate in the United States. The U.S. Census Bureau estimates that from July 1, 200 7 to July 1, 2008 8 88,825 people immigrated legally to the United States This represents 3 2 of the total change in population, after considering births and deaths of the native population. Much of the current debate involves illegal immigration, and the impact that both legal and illegal immigration have on th e wages and employment of native citizens. Although numerous studies have focused on the effect of immigration on wages and employment, particularly with regard to low skilled workers, no consensus on the issue has yet been reached. Basic economic theor y posits that an increase in the supply of low skilled (high skilled) workers should result in a decrease in the wage of low skilled (high skilled) workers, all else being equal. If a shift in the supply of labor decreases wages and thus increases employm ent, the country experiences a gain in national income, or immigration surplus Borjas (1995) illustrates that the immigration surplus is proportional to the demand elasticity of factor price for labor. The greater the impact of immigration on wages, the higher the gain to the nation as a whole. Borjas reports that studies summarized in Hamermesh (1993) indicate that the demand elasticity of factor price is greater for skilled workers than unskilled workers; this would suggest that immigration of high sk illed workers would have a greater impact on wages of native high skilled workers than immigration of low skilled workers would have on wages of
7 native low skilled workers. Thus, the immigration surplus would be larger when immigrants are more skilled. F urthermore, skilled labor is likely to have production complementarities with capital, which adds to the potential immigration surplus. However, these results rely on the result that immigration shifts the labor supply curve and decreases native wages. T his result has not been proven in the empirical literature; there is conflicting evidence in the literature as to whether immigration has any effect on native wages at all. Some studies have found that immigration either lowers the wages of low skilled wo rkers or widens the income gap between the rich and poor, although the magnitudes of these effects are relatively small. Furthermore, even if immigration does reduce wages and thus produces an immigration surplus, this surplus generally involves a transfe r of wealth from labor to capital, which may not be in agreement with policy objectives. The impact of immigration of physicians on wages and employment of their native counterparts is even less clear, as it is complicated by licensing requirements and pa yment mechanisms involving third party payors. There are two generally accepted methods of testing the effects of immigration on wages and employment: area analyses and factor proportions analyses. 3.2: Area Analysis Approach Area analyses exploit the f act that immigration is usually geographically concentrated, and contrast the level (or change) in immigration by area with the level (or change) in earnings of nonimmigrant workers (Borjas 1996). They usually calculate a spatial correlation between wages in a particular metropolitan statistical area (MSA) and the ratio of immigrants to native population in that area. These studies have found that,
8 in general, immigration has a minimal effect on native wages. Butcher and Card (1991) employed the area ana lysis approach and analyzed changes in the distribution of wages in 24 major cities during the 1980s, particularly focusing on the lower tail of the wage distribution. They found that the labor market consequences of an immigration influx tended to be r elatively small, in part because immigrant inflows were often offset by outflow migration of natives. They also found that, although there was no significant decline in wages at the lower end of the earnings distribution, higher levels of immigration were associated with more rapid increases in the 90 th percentile of wages, contributing to a widening gap between the rich and the poor. Borjas, Freeman and Katz (1996) use data from the 1980 and 1990 Census to estimate the cross sectional effect of immigrat ion, measured as the ratio of immigrants to natives in the relevant area, on the log of weekly earnings of natives in that area. Their model takes a differences in differences approach to control for changes in local labor market conditions from 1980 to 1 990 as well as educational achievement of natives and immigrants, which allows them to measure the impact of immigration on earnings after controlli ng for other factors that affect wages. They find that, for a small geographical area, such as an M SA, the effect of immigration on wages is nearly zero, but positive. However, widening the geographic focus to the state or region level turns the effect of immigration on wages negative, although the magnitudes of the coefficients are still relatively small. Or renius and Zavodny (2007) use data from the Immigration and Naturalization Service (INS) and the Current Population Surveys (CPS) to study the relationship between immigration and native wages by broad occupational group, over the time period
9 from 1994 to 2000. Their method regresses the average earnings of natives in a particular occupation group (professional, service related, or manual labor) on the fraction of workers in that group who are immigrants and other control variables. The ordinary least squ ares regression results indicate that higher immigrant shares are associated with positive wage effects on more skilled natives and negative effects on less skilled natives. Because immigration into an area may not be independent of local wages, they also control for endogeneity by using immigrants who are admitted to the U.S. in a given year as the spouse of a U.S. citizen as an instrumental variable The results of the two stage least squares regression indicate essentially no effect of immigration on w ages of professionals, a negative but insignificant effect on wages of service workers, and a statistically significant negative effect on wages of manual laborers. In addition, they find that all of the adverse wage impacts come from immigrants who are n ot new arrivals, presumably as it takes some time for immigrants to assimilate into the labor force. Area analyses have been criticized because, by taking a snapshot in time approach to studying the effect of immigration on wages in a particular metropo litan area, the approach implicitly assumes that local labor markets are closed. This assumption obviously does not concur with reality. If, in fact, native workers respond to an influx in immigration by migrating to other U.S. cities, this would reduce the impact of immigration in a particular MSA. This hypothesis is consistent with the findings of Butcher and Card (1991). In addition, Frey (1994) found that during the 1980s, as foreigners from Mexico, China, and other countries immigrated to the citie s of Los Angeles and New York, large numbers of native whites moved away. Specifically, competition for low skilled jobs and the growing population encouraged low and middle
10 income native whites to migrate elsewhere, but generally did not affect migration of high income natives. Furthermore, if immigrants intentionally choose to locate in areas where wages are high, then choice of location is endogenous and the correlation between immigration and wages in that particular area would be artificially positiv e. Card (2001) addresses the two main criticisms of area analyses in a study of the effect of immigration on employment and wages by broad occupational group. His analysis uses data from the late 1980s and focuses mainly on the impact of immigration o n less s killed occupational groups. In contrast to earlier studies that used the fraction of immigrants in a city as a measure of immigrant competition, Card breaks the immigrant share down by occupational grouping. H is analysis of the data indicates tha t, at least in the short term, if anything, immigration encourages native migration to rather than away from the immigrant heavy city. To control for the possibility that unobserved city and occupation specific factors (productivity shocks) attract immig rants and are therefore correlated with both wages and immigration flows, Card constructs an instrumental variable based on the fact that newly arriving immigrants tend to move to cities where earlier immigrants from their country are already established. In particular, he estimates an exogenous supply push component, the actual inflow of immigrants from a given source country moving to a given city based on total inflows from the country and the fraction of earlier immigrants from that country who live in the city, multiplied by the fraction of immigrants from that country who fall into a particular occupation group. After controlling for the two criticisms of area analysis studies, Cards empirical analysis finds a systematically negative effect of imm igration on both employment and wages of native workers.
11 3.3: Factor Proportions Analysis Factor proportions analyses take a general equilibrium perspective, treating immigrants as a source of an increased national supply of workers of the relevant skill, and applying an elasticity of substitution to estimate the effects of changes in the labor supply on native wages (Borjas Freeman and Katz 1996). Borjas, Freeman and Katz (1992) employ a factor proportions approach to determine the impact of trade and i mmigration on wages and employment of native workers, particularly those with a high school diploma or less, using data from the 1970s and 1980s. They identify that in the 1980s, the wages and employment rate of less skilled workers fell in relation to th ose of more skilled workers, and suggest that this change could be caused both by changes in the trade deficit as well as immigration. Their analysis concludes that the immigration flow did not significantly alter the nations relative supplies of high sc hool and college graduates during the 1980s and therefore is unlikely to have had much effect on relative wages of these groups of workers. However, both trade and immigration greatly expanded the supply of high school dropouts relative to other workers a nd had a negative impact on the wages of these workers during this time period. The authors confirmed in their 1996 study (Borjas, Freeman and Katz 1996) that immigration contributed moderately to a rise in wage inequality in the 1980s. Ottaviano and Pe ri (2005) introduce a production function in which capital accumulates endogenously and make a critical assumption that U.S. and foreign born workers with similar levels of education and experience are imperfect substitutes for one another. The imperfect substitutability occurs because of differences in training, job choice, or other unobservable differences between U.S. and foreign born workers. The
12 authors analysis estimates the total effect of immigration on U.S. aggregate wages, both through own an d cross elasticities. They use country level data from the Integrated Public Use Microdata Sample for 1970, 1980, 1990, and 2000, and divide the groups by education and experience level. Their analysis yields an elasticity of substitution of foreign gr aduates to natives of about 4 for college graduates, 7 for high school dropouts and 10 for high school graduates and college dropouts. This finding suggests that, of all education groups, it is hardest to substitute foreign born college graduates for nati ve college graduates. Using these elasticity of substitution estimates, the authors estimate the impact of immigration on native wages to be large and positive in the aggregate: an 8 increase in foreign born workers increases the average U.S. wage by 2.2 The top three education groups gain by about 2.4 each, while the high school dropouts lose; their wages decrease by about 2.4. Furthermore, the authors simulate several counterfactual immigration policy situations, in which different types of worker s were precluded from immigrating. They find that the most harmful scenario for the U.S., in terms of aggregate native wages, would be to replace immigration of high school dropouts with immigration of college graduates: the effect of immigration on aggre gate wages would then drop to approximately zero. Borjas, Freeman & Katz (1996) compare the area analysis and factor proportions approaches, and find that the estimated effect of immigration on native workers is very sensitive to the empirical experiment e mployed. They find very unstable results for both methods of analysis, suggesting that results should be interpreted cautiously. Overall, they conclude that immigration has played a role in reducing the pay of high school dropouts, while having a smaller negative effect on earnings of high school graduates.
13 3.4: Other Research Johnson (1980) and Topel (1994) studied the effect of immigration on wages of low skilled workers and found that, indeed, immigration has a negative effect on native wages. Johns ons analysis is concerned mainly with illegal immigration of low skilled workers and concludes that, although immigration decreases native wages of low skilled workers, it has little effect on employment, and actually increases the wages of high skilled w orkers and holders of capital. Thus, immigration tends to widen the income gap between the rich and the poor. Topels analysis finds that, in particular, increased immigration of less skilled Hispanic and Asian workers in the western United States has ha d a negative impact on the wages of natives, thus causing a greater increase in income inequality than anywhere else in the country. Less research has been done to determine the effect of immigration of high skilled workers on the wages of hi gh skilled n ative workers in the United States. Borjas (2003) analyzed the effect of immigration on native wages and employment by education and years of experience, using data from the 1960, 1970, 1980, and 1990 Census as well as the 1999, 2000, and 2001 Current Pop ulation Surveys. His model assumes that similarly educated workers with different levels of experience participate in a national labor market and are not perfect substitutes for each other. His analysis finds a clear negative impact of immigration on nat ive wages; an influx of immigration was found to decrease the wages of native high school dropouts by 8.9, those of high school graduates decreased 2.6, and those of college graduates decrease by 4.9. Overall, a 10 increase in the supply of workers th rough immigration decreases wages by 3 to 4.
14 Borjas (2005) and Borjas (2006) evaluate the impact of immigration of foreign students on the earnings of doctorates. Using data from the National Science Foundations Survey of Earned Doctorates (SED) and Survey of Doctoral Recipients (SDR), Borjas employs a two stage least squares regression analysis and finds, in agreement with Borjas (2003), that a 10 increase in the supply of doctorates due to immigration results in a 3 to 4 decrease in wages of doct orates in the same field. Camarota (2007) found a negative impact of immigration on the employment opportunities for native born workers. He notes that most of the increase in immigration he studied came not from the very bottom of the labor pool; about 5 0 of the growth in immigrant employment from 2000 to 2004 came from immigrants with at least a high school education. Chiswick (2005) addresses the increased demand in the late 20 th and early 21 st centuries for high skilled technology workers in OECD co untries, largely due to the Computer Revolution, globalization, and immigration of low skilled workers. The United States Immigration Act of 1990 shifted the focus from family reunification (which largely involved low skilled workers) to an increased role of high skilled worker immigration. Chiswick posits that while tending to lower the average wage of high skilled workers, high skilled worker immigration actually increases the average wage of the low skilled workers who complement the productivity of sa id high skilled immigrants. This, in turn, reduces poverty and income inequality, and alleviates the burden of the government to provide welfare programs for the poor. He therefore argues that overall, immigration of high skilled workers to developed cou ntries has a positive effect on the economy. In contrast, he opines that immigration of low skilled workers
15 tends to lower the average wage of low skilled workers and worsens the income gap, increases poverty and results in an increased need for governmen t spending. Low skilled workers tend to draw more government benefits than they pay for in taxes, which can contribute to already fiscally strained social welfare programs in developed nations, including the United States. 3.5: Immigration of Physicians Weiss (2000) made use of what was essentially a natural experiment that took place in Israel: the sudden and large influx of highly skilled migrants from the former U.S.S.R. Specifically, 12,200 medical doctors migrated from the former U.S.S.R. to Israel between 1989 and 1993, compared with 15,600 doctors living in Israel in 1989. Weiss analysis finds that highly skilled immigrants initially accepted lower wages and worked in lower skilled occupations, slowly climbing the occupational ladder and findin g jobs that suited their occupation, skill level, and earning capacity. The study concludes that even a significantly large influx of high skilled immigrants had negligible effects on native wages and employment. Svorny (1991) uses a factor proportions type analysis to estimate the effects of liberalization of restrictions on physician migration in a time of increased demand in the U.S. Specifically, she uses aggregate data for the time period 1966 71, just after the passing of the Immigration Act of 1 965 and shortly after the creation of Medicare and Medicaid. Svorny argues that over such a short time period, the native supply of physicians should be relatively inelastic, even in the face of an unexpected increase in demand. Thus, she argues that any change in consumer welfare during the time period is
16 attributed solely to the change in immigration law. Although the inelasticity of supply assumption is likely to hold in the short run, it seems that her results apply only in the face of an increase in demand. In other words, although she is able to identify the effect of immigration under specific conditions, the results may not apply to a labor market in the absence of such a demand shock. In addition, the study makes assumptions about the price ela sticity of demand for physician services and attempts to calculate the number of physicians and residents that would have been present in the U.S. in a given year under the counterfactual assumption that the law had not changed. The study finds that the dollar value of the benefits to consumers from the 1965 liberalization of immigration restrictions reached 2.9 billion dollars by 1971 and that if the additional migration had not been permitted and if prices had been free to adjust to equate supply and demand, physician earnings would have been at least 11 higher in 1971. Svorny uses physician earnings as a proxy for price. Although this may have been an effective measure in the 1960s, today it is not necessarily the case that physician earnings and prices for medical care are so highly correlated. A reduction in the price charged for medical care may or may not translate into a change in the price of care that patients actually pay for (in the form of insurance premiums, co pays, or actually paying directly for the service) and similarly may or may not affect physician earnings. Table 1 summarizes the results of the existing literature.
17 Table 1 Summary of Existing Empirical Literature: Labor Market Effects of Immigration Author (Year) Result s of Analysis Butcher & Card (1991) No significant effect on wages of low skilled workers, wages of high skilled workers grew more rapidly, increased income gap. (1980s data) Svorny (1991) Consumer gains and decrease in physician wages after the Immigrat ion Act of 1965. (1966 1971 data) Borjas, Freeman and Katz (1992) Negative impact on wages and employment of high school dropouts. (1980s data) Topel (1994) Increased income gap in the West. (1972 1990) Borjas, Freeman & Katz (1996) Small positive effec t of immigration on wages in an MSA, small negative effect in broader geographic area. Rise in wage inequality. (1970s 1980s data) Weiss (2000) Negligible effects of high skilled immigration on wages and employment of high skilled workers. (Israel, 1989) Card (2001) Significant negative impact on wages and employment. (late 1980s) Borjas (2003) Negative impact on wages for all education groups. (1960 2001 data) Ottaviano & Peri (2005) Large positive aggregate effect on wages. High school dropouts los e, more highly educated workers gain. (1970 2000 data) Borjas (2005), Borjas (2006) Negative impact of foreign doctoral student program on native wages. (1968 2000 data) Orrenius & Zavodny (2007) No effect on wages of professionals, insignificant effect on service workers, negative effect on manual laborers. (1994 2000 data) Camarota (2007) Negative impact on employment for all education groups It is evident that, although the literature tends to support a negative impact of immigration on wages of l ow skilled workers, even the most recent studies conflict as to the effect on wages of high skilled workers. Most notably, Ottaviano and Peri (2005) find that immigration increases wages of native high skilled workers, while Borjas (2003, 2005, 2006) find just the opposite: that immigration decreases the wages of native
18 high skilled workers. In addition, research focused on specific occupations is almost nonexistent. The study by Svorny (1991) is an interesting one with respect to physicians; however, ch anges in the medical care industry since 1965 certainly merit further, more current research in this area.
19 Chapte r 4: Justification for Study 4.1: Academic Value As discussed in the previous chapter, earlier studies of the impact o f immigration on local wages find conflicting results. Th is r esearch makes use of an excellent data source, the AMA Physician Masterfile, which has not previously been employed in studies of this type. For the first time, the effect of immigration on a p articular occupation will be studied using detailed data that isolates the effects being studied from the noise which is present in other studies that use aggregate data to analyze broader occupational groups or skill categories. According to Card (2001 ), because of the large differences in skill levels among immigrants in different areas, studying the overall fraction of immigrants in a city is simply too crude an index of immigrant competition for any particular subgroup of natives. Card argues for studying the impacts of immigration at the occupational level. The particular choice of occupation in this study the physician, extends the application of the data and results beyond simply the academic question of whether immig ration decreases native wa ges to policy issues pertaining to healthcare and immigration policy in the U.S. Furthermore this study has applications in many fields of economics; including, but not limited to health, international, development, and labor economics.
20 4.2 : Immigra tion and Healthcare Policy Since the early 1900s, there has been an ongoing debate as to whether the U.S. has too many or too few physicians, and whether the government should be involved in controlling the physician supply. The Flexner report published i n 1910 concluded that there was an oversupply of largely under trained medical professionals in the U.S. The relative supply of physicians subsequently fell due to the closing of a number of medical schools of lesser quality. The general consensus that t he U.S. had an oversupply of physicians continued until the publication of the Bane report in 1959, which predicted a shortage of physicians in the U.S. by 1975. This was followed by government subsidies encouraging the expansion of the number of medical schools as well as the number of students admitted. In 1981, the Graduate Medical Education National Advisory Committee published a report indicating that by 1990, the U.S. would again face a surplus of physicians and recommended that the number of studen ts admitted to U.S. medical schools as well as the number of foreign physicians allowed to immigrate be restricted. Subsequent reports by the Council on Graduate Medical Education (COGME) in the early 1990s, the Bureau of Health Professions, and Tarlov (1 986) and Weiner (1994) confirmed that a surplus of physicians would occur by the year 2000. Thus, policymakers in the 1980s and 90s generally took steps to limit first, the number of students admitted to U.S. medical schools and, later, the number of fore ign physicians admitted to the country as immigrants. However, just as it seemed a consensus had been reached that the U.S. was, in fact, facing a surplus of physicians, the research of Cooper (1998) projected just the opposite: that by 2010 the U.S. woul d be facing a shortage of physicians, especially specialists (Blumenthal 2004). Reports in the recent medical
21 literature have brought to light this pending physician shortage (Moore 2003, Cross 2007, Arvantes 2007). Based on a recent study by the COGME, the America n Medical Association adopted a resolution at their 2008 Annual House of Delegates Meeting recognizing that there is currently a shortage of physicians (Resolution 309). The shortage of physicians is generally thought to be even larger in rur al and inner city areas. Physician supply becomes an even more relevant topic due to the current political climate with regard to healthcare reform. According to an April 2009 article in the New York Times, Obama administration officials, alarmed at doc tor shortages, are looking for ways to increase the supply of physicians to meet the needs of an aging population and millions of uninsured people who would gain coverage under legislat ion championed by the president (Pear 2009). Whether the U.S. is truly facing a shortage of physicians in the economic sense of the word and, if so, whether it is appropriate for the AMA and/or the government to attempt to control the supply are matters for debate. However, given that for the last century these organizati ons have attempted to do just that, the supply of physicians can be affected through a change in the number of students admitted to U.S. medical schools and/or a change in immigration policy with respect to international medical graduates. Current immigra tion policy provides preferential treatment to more highly educated immigrants, and also favors immigrant physicians willing to work in underserved areas. Although many studies have attempted to estimate physician demand and supply to determine whether a shortage or surplus exists, no work has been done to evaluate how changes in the supply due to immigration would affect the wages and employment of the local labor market. Given that there is likely to be a growing demand for medical care in
22 the future du e to the aging population of baby boomers as well as potentially newly insured patients, a study of the effect of an increase in the supply of foreign physicians is certainly merited. Thus, the study has important policy implications in the presence of cur rent debate over immigration law and healthcare reform and in an era of increasing mobility of labor due to globalization. From the point of view of the United States, whether foreign physicians are filling positions in areas that have a physician shortag e or whether they compete directly with domestic doctors for jobs in already populated areas should be considered when determining immigration policy. Current proposals for future immigration law include implementing a point system, under which potential immigrants would receive additional points based on higher levels of education. Thus, the proposed legislation would make it easier, all else equal, for highly educated foreigners to immigrate to the United States. Borjas (1995) concludes that the benefi ts of immigration can be increased by pursuing policies that attract high skilled workers. Chiswick (2005 ) also seems to imply that policies favoring high skilled immigrants are preferable. However, Ottaviano and Peri (2005) find just the opposite; that, in fact, any net welfare gains from immigration could be eliminated by establishing policies that resulted in an increase in high skilled immigration at the expense of low skilled immigration. Thus, before drawing such a conclusion the effect of such imm igration on their native counterparts should be better understood.
23 Chapter 5: Immigrant Physicians: Visa Requirements This study analyzes immigration of both physicians educated in the U.S. as well as in a foreign country. Graduates of foreign med ical schools are typically referred to in the literature as IMGs (international medical graduates). In order to practice medicine in the U.S., IMGs are required to pass federal and state licensing examinations pertaining both to medical knowledge and Engl ish language before they are eligible to apply for a visa. Graduates of international medical schools may enter the United States under one of several visa programs: the temporary work visa, H 1B, is the most common. This visa is restricted to highly sk ill ed individuals with at least a b achelors degree who are obtaining work in a high skilled occupation. This visa requires an employer, such as a hospital, to sponsor the visa of the immigrating physician. Unless the physician obtains another type of vi sa later or obtains citizenship through marriage at some point (which is likely), the H 1B visa is for a term of three to six years. Some IMGs may enter the U.S. under the O visa program. This visa is restricted to persons with extraordinary abilities in their field. Thus, the physician must be considered an expert in their field or have achieved a significant amount of exemplary research experience in order to qualify for such a visa. This visa has less stringent restrictions with regard to employer sponsorship. In addition, although the O visa is typically approved for three years, it may be extended indefinitely and thus there is no limit set on the amount of a time a physician with an O visa can remain in the U.S.
24 Finally, most immigrants who mov e to the U.S. to study at a U.S. medical school do so under the J 1 training visa This visa program was enacted to encourage exchange of knowledge between the U.S. and other countries. The J 1 visa normally requires immigrants to return to their home co untry for a minimum of two years after completing their residency in the U.S. before returning under another visa. However, the U.S. government makes an exception to this requirement in three cases: persecution, hardship, and government recommendation. A persecution waiver may be obtained if the physician is seeking asylum in the U.S. from persecution in their home country. A hardship waiver may be available if, due to immediate family living in the U.S., the physician and/or his family would suffer exc eption al hardship if the physician were forced to return to his or her home country for two years. Finally, and most common, if the physician agree s to provide primary care services in federally designated health professional shortage areas (HPSAs or MUA s) fo r a minimum of three years, a waiver of the two year return home requirement may be granted Foreign physicians who graduate from U.S. medical schools are normally subject to the same licensing examinations as their native born counterparts. All immigrant physicians may eventually be able to apply for permanent residence in the United States under several alternative circumstances. Asylum seekers, physicians who marry a U.S. citizen, and those who start substantial successful business es in the U .S that employ at least ten U.S. workers are eligible to apply for permanent residency. Physicians in the U.S. on an O visa, as previously stated, may continuously extend their visas. Finally, almost any foreign physician who agrees to work in primary fi elds in HPSAs (Health Professional Shortage Areas) or MUAs
25 (Medically Underserved Areas) for a minimum of five years may apply for permanent residency. Thus, there are many circumstances in which an immigrant physician who moves to the U.S. for education purposes may end up living and practicing in the U.S. for many years (Ester 2004, Klasko 1999, Mautino 2002).
26 Chapter 6 : Theoretical Foundations 6.1: Physician Labor Demand and Supply The basic theoretical model underlying this study is a simple model of labor demand and supply. In economics, the study of the demand for healthcare has been framed in terms of the demand for health itself. An individuals demand for health influences their demand for health services, which are produ ced, like other goods, using both labor and capital. The production function for health services can therefore be written: Q = f (L,K). The demand for each input will depend, in part, on the price of the final good (or service, as th e case may be). A s the price of the health service increases, so does the marginal revenue product of labor and therefore, so should the wage rate. The wage rate also depends on productivity; if new technology allows a physician to become more productiv e, the marginal revenue product increases and so should the physician s wage rate. The demand for each input also depends on the degree of substitutability between the factors as well as their relative prices (which, in the case of labor, is the wage rat e). The degree of substitution between capital and labor w ill depend on the particular service desired. For example, new advances in technology have resulted in machines which can dispense pills just as a pharmacist could. In this case, hospitals may so mewhat easily substitute between labor and capital. However, some surgeries may require specific skills of a surgeon that no machine is capable of performing; in this case, the degree of
27 substitutability is very small. In the last few decades, and partic ularly with the increasing pressure from managed care companies to control costs, firms in the healthcare industry have not only attempted to substitute capital and labor but also to substitute less skilled, cheaper labor for high priced physicians. Many services that formerly were performed by physicians are now performed by nurse practitioners or physician assistants. Furthermore, advances in technology now require more technicians to operate machinery and few er specialized physicians to actually perfor m complex tests. The supply curve of labor is upward sloping, such that more doctors will be willing to work for a higher wage rate. Higher wages may encourage physicians to work more hours and/or see more patients. In addition, physicians working in o ther areas, both in the U.S. and abroad, are encouraged to move to areas where wages are higher. The demand curve of labor is downward sloping, such that the lower the price of labor, the greater is the amount of labor demanded. The equilibrium market wa ge rate is determined where the supply and demand curves intersect. The equilibrium physician wage will change in response to a shift in either the demand or supply curve. The supply curve can be shifted outward by an increase in the number of physician s in an area, either due to an influx of new medical school graduates or physicians moving in from other areas, whether from within the U.S. or due to immigration. In the absence of any demand shocks, basic economic theory posits that in the presence of such a supply shift, the wage of physicians will decrease in the short run. In the long run, native physicians may respond by moving out of the area, or fewer students will apply to medical school as the decrease in wages reduces the expected lifetime ret urns of becoming a physician.
28 On the other hand, the equilibrium wage rate could be increased due to an increase in the demand for medical care. It would take several years for U.S. residents who are not already physicians to respond to the increase in wa ges by attending and graduating from medical school and residency programs. Foreign medical graduates practicing outside the U.S. can respond more quickly to such an increase in expected wages. Thus, the physician labor supply is more elastic due to immi gration of foreign physicians (Folland 2007) 6.2: Physician Immigration : Decision to Emigrate Migration theory segregates the reasons for worker migration into two categories: supply push factors and demand pull factors. Supply push factors are gener ally negative factors in the home country that encourage workers to leave and seek work elsewhere, such as unfavorable employment conditions, poor healthcare, and lower wages. Supply push factors are normally more important in migration decisions of worke rs from developing countries than from developed nations. Demand pull factors are those factors in the potential destination countries that attract foreign workers, including better living and working conditions and higher wages. Networks also play an im portant role in migration decisions: potential immigrants often acquire information about living conditions, wages and employment in a potential destination country from people they know who already live there. Furthermore, networks help immigrants in the assimilation process once in the new country (Clark 2006). When the decision of a foreign medical doctor to migrate is primarily based on expected earnings (which is often the case ) the decision will be determined by relative
29 wages in the U.S. and the fo reign country in which the doctor currently lives, as well as migration costs and non wage factors that affect a physicians choice to move. Specifically, a physician will wish to migrate if: Z C W W F US > where W US and W F are the wage rates in the U.S. and the physician s home country (foreign), C is the cost of migrating, and Z represents the compensating differential in favor of staying in the home country which is generally positive and reflects the fact that, all else equal, most potentia l migrants would prefer not to move Thus, the physic ian will wish to migrate if the increase in his wage, less the cost of moving, is greater than the value of his desire to stay in his home country (Vujicic 2004). 6.3 Exceptions to the Rule in the Heal thcare Industry The model predicts that an increase in the supply of physicians through immigration will, consistent with basic economic theory, cause wages to fall. However, this result is complicated by many factors in the medical care industry. First it assumes that a foreign physician is a perfect substitute for a native one. This assumption is likely to be violated, as patients may (rightly or wrongly) believe that a physician educated outside of the U.S. is of inferior quality. Studies have gene rally found no difference in the quality of care provided by foreign medical graduates compared to U.S. medical graduates (Folland 2007) P rejudices may exist even against foreign physicians educated in the U.S. Patients have imperfect information regard ing quality of care from any physician, native or foreign.
30 Furthermore, because they receive payment from third party payors including managed care companies, Medicare, and Medicaid, physicians are limited in the amount they can charge in some cases, and their compensation structure may not behave in the way theory would predict. Specifically, physician wages are more likely to be sticky than wages in many other occupations. On the other hand, there are at least two reasons that allow the effects of immigration on wages of physicians to be more easily identified than most other occupations: first, because physicians are subject to strict licensing requirements by states, there are unlikely to be many foreign physicians working illegally in the U.S. T herefore, it will not be necessary to attempt to estimate a measure of illegal immigration, for which there is little reliable data, a problem that has plagued many of the studies surveyed earlier. Second, immigration can be considered as an importation o f labor. To the extent that production is shipped overseas or jobs are outsourced, policy decisions either limiting or encouraging immigration might be rendered meaningless. Although some outsourcing of medical care, and even physician care, does exist ( media reports of U.S. citizens traveling to foreign countries for medical procedures which would otherwise be unaffordable to them come to mind), it is certainly very minimal compared to other occupations. Physician care unlike many labor services, is di fficult to outsource. 6.4 Hypotheses of this Research This research study hypothesizes that immigration of foreign physicians has a negative impact on area wages, and that this negative impact comes not only from the increase in physician supply caused by immigration but also from the fact that immigrant
31 physicians will be willing to accept lower wages than their native counterparts. This research expects, however, that once country of education is controlled for, the results will show that immigration of foreign educated doctors has a more negative impact on area wages than immigration of U.S. educated foreign born doctors This hypothesis is based on the idea that foreign born doctors educated in the U.S. are likely to be more similar to their U.S. co unterparts. The researcher hypothesizes that an influx of U.S. born, foreign educated doctors, on the other hand, will have a negative impact on area wages. These physicians may be viewed as being of lower quality by potential employers. It is expecte d that wages will be higher in areas with more male physicians. Furthermore, wages are expected to increase with age, but at a decreasing rate. An increase in the physician stock is anticipated to have a negative impact on area wages, consistent with bas ic economic theory of labor supply and demand. Finally, this researcher hypothesizes that, ceteris paribus, areas with a larger percentage of the population covered by Medicare B will have higher wages. Medicare B patients are either disabled and/or ov er the age of 65; therefore, areas with more Medicare B patients are likely to have higher demand for physician services.
32 Chapter 7 : Research Methodology 7.1 : The Basic Ordinary Least Squares Model The intent of this research study is to test the hypothesis that immigration of foreign physicians, by increasing the supply of workers in the field, lowers the average wage of physicians in the United States. The approach employed mimics the area analyses discussed earlier, while correcting for some cr iticisms of that approach, as will be described below. Incorporating methods from Borjas (2005, 2006) and Orrenius and Zavodny (2007), the basic model takes the following form: t a t a t a t a t a t a MedB T A X I w , 4 3 2 1 0 ln e b b b b g b + + + + + + = where ln w represents the natu ral log of the average hourly wage of physicians indexed by MSA ( a ) and year ( t ). D ata have been sampled for every other year from 1997 to 2007 rather than every year, due to budget constraints I represents immigrant share and is measured as the numbe r of foreign physicians in a given MSA in a given year relative to the total number of physicians: t a t a cians TotalPhysi sicians ForeignPhy I , = g is the parameter of interest; it measures the percent change in area wages for a one percentage point i ncrease in the immigrant share and is expected to be negative. This method of calculating the percentage of immigrants (whether it be in general, or b y specific occupation or education level and also at either the MSA, state, or national level) and employing it as the measure of immigrat ion has been previously
33 performed in studies by Altonji and Card (1991), Orr enius and Zavodny (2007) Borjas (2003) a nd Borjas (2005, 2006 ). X is a vector that controls for the demographic characteristics of physicians in the MSA during the given year, including physician sex and age. A and T represent area and year fixed effects, respectively, and control for unobse rved determinants of wages within a particular MSA or year. The OLS regressions are performed both with and without the area fixed effects. M edB measures the percentage of the population in the physicians service area that is covered by Medicare Part B This controls for the influence of large buyers, namely government programs and managed care companies, over physician charges. Managed care penetration rates have been shown to affect wages of medical professionals. Hadley and Mitchell (1999) find ev idence that higher HMO market penetration is associated with lower physician earnings. However, their study is limited to a cross section of data from 1990 and therefore does not address changes over time in physician earnings. Buerhaus and Staiger (199 6) use data from the Current Population Surveys for 1983 through 1994 to analyze the impact of managed care on employment and earnings of nurses. Their research finds that managed care has been associated with a shift in employment away from hospitals to other settings (home health care, nursing homes, physician offices), but has had minimal effects on nurses wages. Simon, Dranove and White (1997) analyze data pu blished in the American Medical Associations Socioeconomic Monitoring System survey from 198 5 to 1993 and find that states with the fastest growth in managed care penetration experienced the highest growth rates in earnings of primary care physicians and the lowest growth rates in earnings of
34 radiologists, anesthesiologists, and pathologists. Du e to difficulties in obtaining managed care penetration rate data, Medicare penetration rate data have been employed in this study. Medicare is the largest single purchaser of healthcare in the U.S. Furthermore, many managed care companies structure thei r pricing using a mark up based on the Medicare Resource Based Relative Value Scale (RBRVS) Reimbursement System. 7.2: Extension of the Basic Model : Additional Variables The first extension to the basic model is to control for changes in physician suppl y. The variable S calculated as t a t a Population ns ofPhysicia S , # = is added to the regression. Note that this variable measures the actual number of physicians divided by the actual population in a given area and year; this variable is not a sample statistic. Thi s variable measures the stock of physicians in a given MSA during a given year, and allows the effect of immigration on wages, holding the number of physicians constant to be identified. By adding the variable S the study can determine whether immigrati on affects wages of physicians only because it increases the supply of physicians, or whether the effects are compounded because there is something specifically different about foreign doctors. Now the parameter of interest, g measures specifically the e ffect that an increase in the proportion of foreign physicians has on wages in a given area. If g is negative, this suggests that foreign physicians are less productive than domestic ones or are willing to accept lower paying jobs. Furthermore, if g is m ore negative than in the analysis without the supply control, this would provide evidence that local physicians move out of the area in response to an influx of immigrant physicians.
35 The next extension to the model is to control for whether the physician o btained their medical degree at a domestic or foreign medical school (for purposes of this analysis, Canadian medical schools are grouped with U.S. medical schools, consistent with the categorization of the American Medical As sociation). Simply including a variable that controls for foreign medical education would not be appropriate, as the effect of foreign medical education is likely to be different for foreign born doctors and U.S. born doctors. Technically, there are four different categories of docto rs to be considered: U.S. born and educated (USBUSE), U.S. born and foreign educated (USBFE), foreign born and foreign educated (FBFE) and foreign born educated in the U.S. (FBUSE). Thus, three additional variables are added to the model, with U.S. born a nd educated physicians being the base category for comparison The variable FBFE_Share is calculated as the total number of foreign born, foreign educated physicians in the sample in a given area and year divided by the total number of physicians sampled in that area and year. The variables FBUSE_Share and USBFE_share are similarly defined. Adding the se variable s allows the determination of whether effects on wages are explained by country of birth or where the physician went to medical school, or both. 7.3: Extension of the Basic Model: Instrumental Variable Approach The next specification of the model involves the incorporation of an instrumental variable to control for endogenous location choice. As discussed earlier, results are potentially bia sed if physicians choose to move to areas where wages are already higher than others. Results could also be biased if occupation specific local productivity shocks
36 raise wages and encourage migration to a particular area. In the presence of such endogene ity the parameter estimates will be inconsistent and causation cannot be assumed (i.e., does immigration actually cause higher wages, or do higher wages attract immigrants?). Prior studies including those of Orrenius and Zavodny (2007) and Friedberg and Hunt (1995) have employed an instrumental variable to control for endogenous location choice. An appropriate instrument is one that is correlated with immigration but not directly correlated with wages or the error term in the wage equation Following Al tonji and Card (1991) and Card (2001) the fraction of immigrants in a given MSA in 1990 can be used as an instrumental variable. Bartel (1989) finds that the number of immigrants already living in a given area acts as a pull factor for future immigrants; thus, this variable should be correlated with im migrant inflows from 1997 2007 the years which the study will analyze. Furthermore, this variable does not directly re late to future wage increases. The instrumental variable applied in this analysis is r eferred to as FB90 and calculated as: 1990 1990 90 a a totalpop npop foreignbor FB = If immigrant share is likely to be endogenous, then the variable FBFE _Share is likely to be endogenous as well. A second instrument FE_IV (the instrument for foreign education) constructed a s FB90 multiplied by the share of physicians sampled in a given area and year who were educated abroad, is added to the model in the specifications that include this physician education variable. This instrument is loosely based on Card (2001), where an i nstrumental variable is constructed using immigrant inflows multiplied by the fraction of immigrants in certain occupational groups (in the case of this study, the category of concern is educational, not occupational, group).
37 7.4: Extension of the Bas ic Model: First Difference Approach Finally, a first difference approach, with and without the instrumental variable, is employed to control for area specific factors that effect wages, endogenous location choice and the possibility that local physicians migrate away from cities when foreign doctors move in. T he first difference approach examines changes in physician wages as a function of changes in the share of immigrant physicians within an MSA This method abstract[s] from differences across cities that might bias a simpler cross sectional analysis the first differenced analysis eliminates any bias introduced by city specific fixed effects that are correlated with the fraction of immigrants in a city and the labor market outcomes of natives (Alton ji and Card 1991). Furthermore, i f immigrants move to a certain MSA because wages are high, but not because they expect them to rise more in the future, regressing the change in wage on the change in immigrant share will correctly identify the parameter o f interest (Friedberg & Hunt 1995). In this specification, the model to be estimated is as follows: As in earlier regressions, the variables FBFE_Share, FBUSE_Share, USBFE_Share and S, as well as the instrumental variable s FB90 and FE_IV are added as a dditional specifications. The first difference approach alleviates some of the potential bias that could be caused by area specific location choice factors; however, it does not completely solve the endogeneity problem. Therefore, following Altonji and C ard (1991), the instrumental variable approach can be combined with the first difference approach to control for local economic conditions that attract immigrants. According to Friedberg and Hunt (1995) an appropriate instrument that is correlated with changes in t a t a t t a t a t a MedB T X I w , 3 2 1 , lne b b b gD + D + D + D + D = D
38 the immigrant share but does not directly influence changes in wages can be used to remove the bias due to immigrant choice of regions with improving outcomes. Altonji and Card (1991) use the existing immigrant population as an instrumental var iable; as previously explained, this analysis employs the instrument FB90 the existing foreign born population in an area according to the 1990 Census.
39 Chapter 8 : Data Sources and Building the Database The data for this project come p rimarily from 4 sources: (1) the American Medical Association (AMA) Physician Masterfile, (2) the Occupational Employment Statistics (OES) Survey published by the Bureau of Labor Statistics (BLS), (3) the Fee for Service Data published by the U.S. Departme nt of Health and Human Services, Centers for Medicare and Medicaid Services and (4) data from the U.S. Census Bureau. 8.1: The AMA Physician Masterfile The Physician Masterfile, compiled annually by the American Medical Association (AMA) contains curr ent and historical data on the over 940,000 physicians practicing in the U.S. and its territories, including more than 243,000 graduates of foreign medical schools. The AMA maintains a high standard of accuracy in data collection and reporting. Relevan t variables included in the Physician Masterfile data are: type of practice (hospital based, office based), state, county, PMSA/MSA and zip code of practice location, size of practice area (population), gender, birth date, birth country licensure informat ion, residency training information, medical school information including country and name of school attended, medical school graduation year and physician specialty. The data on country of physician birth and country of medical school attended are of par ticular importance to this research study. Data on country of birth al low the calculation of the share of immigrant physicians in a given MSA. Data on country of
40 medical school allow further specifications of the model as described above, in order to det ermine whether there is a significant difference in the earnings of immigrant physicians e ducated abroad as compared to the United States. Demographic variables, including gender and age are obtained from this data set. The AMA data are proprietary and can only be purchased subject to strict licensing agreemen ts as to the use of said data. A random sample of 184,563 observations from the AMA Physician Masterfile Survey data was purchased from Medical Marketing Service, Inc. for purposes of this researc h. Specifically, a random five year sample of 30,021 observations from the 1997 AMA survey data, 30,505 observations from the 1999 survey, 30,708 from the 2001 survey, 31,201 from the 2003 survey, 30,975 from the 2005 survey, and 31,153 from the 2007 surv ey were purchased. The number of observations purchased and the number of years of data purchased were limited by budget constraints. The greatest challenge involving the AMA data was the manipulation required on the birth country variable. This varia ble did not contain a unique identifier for the country of origin of each physician Instead, it often contained a city and state of birth, a city and country, a country with a name misspelled, a country abbreviation, a city name only, or, in some cases, an identifier of unknown. As whether the physician is foreign born or not is the most critical variable to the analysis, observations for which the birth country could not be identified as at least either foreign or U.S. had to be dropped. After droppi ng these observations, 170,858 observations remained in the dataset. For the remaining observations, the data had to be cleaned so that the countries of birth had a uniform description.
41 8.2: Wage Data from the Occupational Employment Statistics Survey The physician observations from the AMA Masterfile were assigned a wage using the average wage for their specialty, practice area and survey year as obtained from the Occupational Employment Statistics (OES) Survey. Published by the U.S. Department of Lab or, Bureau of Labor Statistics, the OES Survey reports average wage data for over 800 occupations for 375 MSAs, 34 metropolitan divisions, and over 170 nonmetropolitan areas. This study employs wage data for 1997 through 2007 from the OES Survey. Although the AMA data assign each physician a detailed specialty code, with more than 200 possible specialties, the wage data for physicians for 1999 through 2007 are reported by broad specialty code, including: anesthesiologists, family practitioners, internists, OB/GYN, pediatricians, psychiatrists, and surgeons. Therefore, each specialty designation in the AMA database had to be assigned a broader specialty code to correspond with the OES wage data. Physicians with specialties which clearly fell into one of th e seven categories listed above were assigned wages based on that specialty. Physicians with other specialties, such as cardiology or urology, which did not clearly fit into one of the broad categories, were classified as Other Specialty and assigned th e average wage of all physician specialties in their area for the given year. For physicians who practice within a primary metropolitan statistical area (PMSA), the OES data by MSA was employed. However, for those physicians who do not practice within a PMSA or nonmetropolitan area for which wage data was collected (i.e. they practice in rural areas) the average wage from the OES state wage data for their given specialty and year were used (the OES does not publish county level occupational wage data).
42 The OES data for 1997 report only the average wage by area for physicians and surgeons; the data are not broken down more specifically by specialty. Therefore, for each area (MSA or state level, where appropriate) the average wage of all physician spe cialties for 1999 was calculated (as a second option, when data for a particular specialty and area were not available in 1999, the data for 2001 were employed). Then, for each specialty, the average wage by area relative to the average of all specialties was calculated. This ratio was then applied to the average wage in 1997 to conform to the data for later years. For example, if in a particular MSA anesthesiologists earned 20 more than the average physician in 1999 (or 2001), it is assumed that anesth esiologists also earned 20 more than the average physician in that area in 1997. 8.3: Medicare Enrollment Data Medicare enrollment rates by county are reported in the Fee for Service Data published by the Department of Health and Human Ser vices, Centers for Medicare and Medicaid Services for 1998 through 2006. Medicare Part B is the arm of Medicare that covers physician services and is therefore the relevant variable to this study. In order to calculate a Medicare Part B penetration rate, however, the percentage of the population covered by Medicare Part B, not the number of people, is required. Thus, the Medicare data were merged with annual county population data published by the U.S. Census Bureau. Medicare penetration rates for each county and year were calculated. In order to properly merge the Medicare penetration rate data with the AMA data and wage data, Medicare penetration rates for each physicians practice area were required. Since most physicians in the database practice wi thin an MSA, the Medicare penetration rate data
43 needed to be recategorized from the county level to the PMSA level. A list of counties within each PMSA was obtained from a geocode file available through The Ohio State Universitys Center for Human Resourc e Research. For PMSAs consisting of more than one county, a weighted average Medicare penetration rate was calculated. Medicare data were then merged with the AMA and wage data by PMSA or county, depending on whether the physician practiced within a PMS A or not. Since Medicare penetration rate data were not available for 1997 or 2007 at the time this analysis was performed physicians surveyed during 1997 were assigned the Medicare penetration rates for 1998 and physicians surveyed during 2007 were assi gned Medicare penetration rates for 2006. 8.4: Data from the U.S. Census In order to calculate the variable t a t a Population ns ofPhysicia S , # = for the first extension of the model, population data by MSA were downloaded from the U.S. Census Bureau. Population by MSA for the fifty U.S. states is available by year from 2000 through 2007. The population for earlier years was estimated using the population from the 1990 and 2000 Census and assuming a constant linear growth rate. P opulation data for the U.S. Virgin I slands and Guam are available only from Census 2000; therefore, this analysis is forced to assume that population in these territories did not change significantly between 1997 and 2007. Population data for Puerto Ricos municipalities is available from 2 000 through 2007; Puerto Ricos population is assumed to be unchanged from 1997 through 2000. The number of physicians employed by PMSA and year are reported in the OES data referenced above. As data on the number of physicians employed by county are
44 una vailable, the use of this stock variable S will be limited to regression analysis of the PMSA areas only, and cannot be employed with regressions which include rural (non metropolitan) areas. As will be discussed later, this limitation turns out to not be as problematic as it may seem. Finally, the model required a measure of the immigrant population, as a fraction of total, by MSA (or county, when the physician did not practice within an MSA) in 1990 to be used as an instrumental variable. The foreign b orn population and native born populations, by MSA and county, are recorded in the 1990 Census were obtained from the U.S. Census Bureau online.
45 Chapter 9 : Descriptive Statistics of the Data 9.1: Country of Physician Birth Of the 170,85 8 physicians sampled from the AMA survey for which country of birth was available, 26 (44,503) were born outside the U.S. (foreign born) and 74 (126,355) were born in the U.S. The proportion of foreign born physicians in each year sampled is roughly the same: Table 2 Physician Sample: U.S. Born and Foreign Born by Year Survey Year # Foreign Born Foreign Born # U.S. Born U.S. Born Total 1997 7,421 26 21,497 74 28,918 1999 7,734 27 21,337 73 29,071 2001 7,658 27 21,161 73 28,819 2003 7,521 26 21,280 74 28,801 2005 7,181 26 20,748 74 27,929 2007 6,988 26 20,332 74 27,320 Total 44,503 26 126,355 74 170,858 Summary statistics of the key independ ent variable in this study, I, the share of immigrant physicians in a particular area, are reported by year in Table 3.
46 Table 3 Summary Statistics of the Variable I = Immigrant Share Year Mean Standard Deviation 1997 0.2566222 0.141 7947 1999 0.2660383 0.1392615 2001 0.2657275 0.1390750 2003 0.2611368 0.1368208 2005 0.2571163 0.1332611 2007 0.2557833 0.1336291 The top twenty countries that most of the foreign born physicians in the sample c ome from are reported in Table 4 For a complete listing, see Table A1 in the Appendix. Table 4 Foreign Born Physicians Practicing in the U.S. by Country of Birth Birth Country # of Physicians of Foreign Born Physicians India 8,352 18.8 The Phili ppines 3,905 8.8 Canada 2,105 4.7 Korea (South) 1,705 3.8 Pakistan 1,523 3.4 Iran 1,486 3.3 China 1,343 3.0 Cuba 1,252 2.8 Germany 1,139 2.6 United Kingdom 1,123 2.5 Vietnam 999 2.2 T aiwan 908 2.0 Egypt 823 1.8 Mexico 639 1.4 Syria 601 1.4 Poland 593 1.3 Colombia 537 1.2 Israel 516 1.2 Argentina 484 1.1 Nigeria 473 1.1
47 The largest single sou rce country of immigrant physicians, by far, is India, followed by the Philippines. This is not surprising, given the earlier review of literature that discussed the predominance of highly educated Asian immigrants. Of the 44,503 foreign born physicians in the sample, only 11,353 (26) come from advanced economies as defined by the International Monetary Fund. The remaining 33,150 (74) come from emerging and developing economies. 9.2 Country of Medical School Attended Of the 170,858 physicians in the sample, 21.9 (37,390) attended a medical school outside of the United States (foreign medical school) and 78.1 (133,468) attended medical school in the U.S. Although, as expected, foreign born doctors are much more likely to have attended a foreign med ical school, a significant percentage of them did, in fact, attend medical school in the U.S. Of all U.S. born physicians sampled, only 4 attended medical school outside the U.S. Meanwhile, 72 of foreign born physicians were educated at foreign medical schools while 28 of them were actually educated here in the U.S. Thes e values are reported in Table 5 Table 5 Location of Medical School Attended by Place of Birth Place of Birth: Location of School: United States Foreign Uni ted States 120,890 96 12,578 28 Foreign 5,465 4 31,925 72 126,355 100 44,503 100 Note: Percentages may not add to 100 due to rounding.
48 Summary statistics of the key variables measuring physician education, FBFE_Share FBUSE_Share USBFE_Share, as well as the base category, USBUSE_Share are reported in Table 6 Table 6 Summary Statistics of the Birthplace/Education Variables USBUSE_Share USBFE_Share FBFE_Share FBUSE_Share Year Mean S t. Dev. Mean St. Dev. Mean St. Dev. Mean St. Dev. 1997 .7092 .1570 .0336 .0587 .1944 .1271 .0620 .0475 1999 .7014 .1521 .0322 .0537 .1955 .1248 .0704 .0528 2001 .7006 .1536 .0337 .0580 .1926 .1235 .0730 .0529 2003 .7073 .1516 .0315 .0563 .1868 .1189 .0743 .0512 2005 .7118 .1475 .0309 .0569 .1785 .1145 .0785 .0535 2007 .7143 .1469 .0298 .0530 .1716 .1128 .0841 .0612 9.3 : Physician Practice Area Although the majority of all doctors sampled work in large MSAs, this is especially t rue of immigrant physicians. Foreign physicians in the sample are more likely than U.S. doctors to practice in big cities and less likely than their native counterparts to work in small cities or rural areas. U.S. born physicians are more likely than the ir foreign born counterparts to work in medium and small cities and rural areas. Table 7 U.S. and Foreign Born Physicians by Practice Area Size MSA Population # of U.S. Born Physicians of U.S. Born Physicians # of Foreign Born Physi cians of Foreign Born Physicians A 1,000,000+ 82,567 65 33,661 76 B 250,000 999,999 20,319 16 5,107 11 C 100,000 249, 999 9,163 7 2,203 5 D <100,000 1,091 1 207 0 N/A Rural 13,215 10 3,325 7 Note: Percentages may not a dd up to 100 due to rounding.
49 This would seem to contradict the idea that foreign physicians serve in rural parts of America where doctors are scarce, and suggests that policies encouraging physicians to locate in these areas are unsuccessful. However, the result that foreign physicians are more likely to work in big cities does not invalidate the theory that they are serving in shortage areas per se; the Health Professional Shortage Areas and Medically Underserved Areas as defined by the U.S. Department of Health and Human Services are comprised of rural areas as well as inner city areas in large cities, or anywhere that there is a shortage of primary medical care, dental, or mental health providers. Table 8 reports the twenty MSAs with the largest number of foreign born physicians sampled. The number of foreign born physicians in the sample in each MSA, for all MSAs in the sample, can be found in Table A2 in the Appendix
50 Table 8 Foreign Born Physicians by MSA MS A # of Foreign Born Physicians of Total Foreign Born Physicians New York, NY 3,669 8.2 Los Angeles Long Beach, CA 2,380 5.3 Chicago, IL 2,139 4.8 Boston, MA 1,317 3.0 Washington, DC MD VA 1,315 3.0 Detroit, MI 1,064 2.4 Nassau Suffolk, NY 1,056 2.4 Philadelphia, PA NJ 1,047 2.4 Miami, FL 985 2.2 Houston, TX 941 2.1 Orange County, CA 777 1.7 Baltimore, MD 679 1.5 Cleveland, OH 641 1.4 Newark, NJ 512 1.2 St. Louis, MO IL 511 1.1 Riverside San Bernardino, CA 506 1.1 Bergen Passaic, NJ 505 1.1 Atlanta, GA 504 1.1 San Francisco, CA 490 1.1 Dallas Plano Irving, TX 483 1.1 9.4 : Physician Demographic Characteristics Foreign born doctors in t he sample tended to be slightly older: the mean age of foreign born doctors in the sample is 48.69 (standard deviation: 12.83) and the mean age of U.S. born doctors in the sample is 47.16 (standard deviation: 13.21). For both groups of doctors, the median occurs between ages 51 and 60. Table 9 reports the percent of U.S. born and foreign born physicians working in each size MSA by age group.
51 Table 9 Physicians by Age and Size of Practice Area U.S. Born Physicians MSA Size Age A B C D Rural Total 20 30 75 15 6 0 4 100 31 40 68 16 7 1 9 100 41 50 63 17 8 1 12 100 51 60 63 16 8 1 12 100 61+ 65 15 7 1 12 100 Foreign Born Physicians MSA Size Age A B C D Rural Total 20 30 85 9 4 0 2 100 31 40 76 11 5 1 7 100 41 50 74 13 5 0 7 100 51 60 74 12 5 0 8 100 61+ 76 11 4 0 8 100 Note: Percentages may not add up to 100 due to rounding. The previ ously discovered phenomenon that foreign born doctors are more likely to practice in large cities than their native born counterparts holds across all age groups. The same is true for the conclusion that U.S. born physicians are more likely than their for eign born counterparts to work in rural areas. Younger doctors, no matter where they were born, are most likely to live in the largest MSAs, while older physicians are more likely than younger physicians to practice in rural areas. Table 10 reports the sex of physicians sampled by birthplace and year. There are notably more men than women in the sample: 119,309 men compared with only 42,526 women. Foreign born physicians are more likely to be female. This is especially the case in earlier years of th e survey; in 1997, 27 of foreign born physicians sampled were
52 female compared with only 22 of U.S. born physicians. This difference tends to decrease over time as the gap between native born male and female doctors has been narrowing since 1997. In fac t, for both foreign born and U.S. born doctors as well as for all doctors combined, the proportion of doctors who are female has been increasing since 1997. The nu mber and percentage of physicians born in the U.S. and abroad, by sex and by year, are repor ted in the A ppendix in Table A3. Table 10 Sex of Physician by Place of Birth and Year U.S Born Physicians 1997 1999 2001 2003 2005 2007 Female 22 23 24 26 28 29 Male 78 77 76 74 72 71 Foreign Born Phys icians 1997 1999 2001 2003 2005 2007 Female 27 29 29 30 31 30 Male 73 71 71 70 69 70 Total (All Physicians) 1997 1999 2001 2003 2005 2007 Female 23 24 25 27 29 29 Male 77 76 75 73 71 71 Note: Pe rcentages may not add to 100 due to rounding. For purposes of this analysis, physicians were grouped into the seven categories of specialties reported in the OES wage data: anesthesiology, family/general practice, internal medicine, OB/GYN, p ediatrics, psychiatry, and surgery. Those physicians whose specialties did not fall into one of these major categories were classified as other. The number of physicians by specialty and sex are reported in Table 11
53 Table 11 Physician Specialty by Sex Specialty All Physicians Male Female Anesthesiology 8,939 5 7,030 6 1,909 4 Family/General Practice 23,212 14 16,767 13 6,445 14 Internal Medicine 24,089 14 16,892 13 7,197 16 OBGYN 10,337 6 6,320 5 4,017 9 Pediatrics 14,760 9 7,258 6 7,502 17 Psychiatry 10,900 6 7,332 6 3,568 8 Surgery 27,232 16 24,096 19 3,136 7 Other Specialty 51,389 30 40,429 32 10,960 25 Note: Percentages may not add to 100 due to round ing. As can be seen in the table, men are much more likely than women to specialize in surgery and somewhat more likely to specialize in anesthesiology. Women, on the other hand, are more likely to specialize in psychiatry, and are far more likely than men to specialize in OB/GYN. These specialty choices become important later in this paper in explaining differences in the wages between men and women. The specialties of foreign born and native physicians as well as those educated at U.S. med ical schools and abroad are reported in Table 12
54 Table 12 Physician Specialty by U.S. / Foreign Birth and Education Specialty U.S. Born Foreign Born U.S. Med School Foreign Med School Anesthesiology 6,171 5 2,768 6 6,550 5 2,389 6 Family/General Practice 18,515 15 4,697 11 19,003 14 4,209 11 Internal Medicine 16,170 13 7,919 18 17,311 13 6,778 18 OBGYN 7,815 6 2,522 6 8,289 6 2,048 5 Pediatrics 10,725 8 4,035 9 11,242 8 3,518 9 Psy chiatry 7,664 6 3,236 7 7,750 6 3,150 8 Surgery 21,858 17 5,374 12 23,373 18 3,859 10 Other Specialty 37,437 30 13,952 31 39,950 30 11,439 31 Note: Percentages may not add to 100 due to rounding. As family/ge neral practice and internal medicine are very similar specialties, grouping them together for comparative purposes, there are approximately the same proportion of general practitioners in the sample of U.S. born and foreign born doctors. In fact, the perc entages of physicians in each specialty are almost exactly the same across birthplace with the notable exception of surgery: U.S. born physicians are more likely to be surgeons than their foreign born counterparts. A similar conclusion can be drawn when c omparing physicians educated at medical schools located within and outside the U.S.: physicians educated in the U.S. are much more likely to become surgeons, but there are no other remarkable differences in specialty choice.
55 9.5 : Wages of Physicians: Summary Statistics Although the AMA sample produced 170,858 observations for which physician country of birth was available, there are a few cases where the OES database did not report a wage for a particular physician specialty or locality for a given ye ar. Thus, after merging the AMA physician survey data with the OES wage data, wages were available and assigned to 161,835 physicians. In the discussion that follows, it is important to keep in mind that physician wages are not necessarily the actual w ages of the physician sampled; they are the average wage for the physicians specialty in the physicians practice area in the relevant year. The mean hourly wage of all the wages assigned to the physicians in the sample is $65.87893 with a standard devi ation of $15.14222. Physician wages ranged from a low of $10.20 per hour to a high of $103.00. The average wages by year are reported in Table 1 3 Table 13 Average Physician Wages by Year Year Mean Wage Standard Deviation Min Max 1 997 49.37990 7.693108 10.2 0 95.16 1999 55.56695 8.231586 16.1 0 70.00 2001 58.51647 7.523859 14.32 69.92 2003 75.34824 14.27231 17.00 101.28 2005 74.52094 11.43191 21.15 94.00 2007 78.43643 11.46004 24.00 103.00 The wages a ssigned to women in the sample were, on average, lower than those assigned to men. The mean wage of women is estimated to be $65.378 per hour
56 (standard deviation: $14.92697) while the mean wage of men is estimated to be $66.05748 per hour (standard deviat ion: $ 15.21429). The averages of the wages assigned to male and female physicians by year are reported in Table 1 4 Table 14 Average Physician Wages by Sex Men Women Year Mean Wage Standard D eviation Mean Wage Standard Deviation 1997 49.73773 7.568217 48.14859 7.987421 1999 55.82967 8.229977 54.74475 8.182890 2001 58.87995 7.492362 57.44544 7.514875 2003 76.09952 14.31077 73.32031 13.96856 2005 75.21550 11.30691 72.78799 11.5581 2 2007 79.01703 11.32920 77.02602 11.65265 A comparison of means test rejects the null hypothesis that there is no significant difference in the average wage assigned to men compared to women and supports the alternative hypothesis that women on average, are more likely to live in areas where wages are lower and/or specialize in lower paying fields: Table 15 Comparison of Means Test: Wages by Sex Sex Observations Mean Wage Standard Error 95 Confidence Interval F 42,526 65.37800 .0723843 65.23613 65.51988 M 119,309 66.05748 .0440469 65.97115 66.14381 Total 161,835 65.87893 .0376403 65.80516 65.95271 Difference .6794 786 .08550 25 .8470616 .5118956 Differe nce = mean(F) mean(M) t = 7.9469 H o : Difference = 0 degrees of freedom = 161,833 H a : diff < 0 H a : diff = 0 H a : diff > 0 Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000
57 Th e data further suggest that foreign born doctors are more likely to earn slightly less than U.S. born physicians. The mean wage of native born physicians in the sample is estimated to be $65.96572 per hour (standard deviation: $15.29789) while the mean wa ge of foreign born doctors is estimated to be $65.62861 per hour (standard deviation: $ 14.68132). The averages of the wages assigned to U.S. born and foreign born physicians by year, based on specialty and practic e area, are reported in Table 16 Table 16 Average Physician Wages: U.S. Born and Foreign Born U.S. Born Foreign Born Year Mean Wage Standard Deviation Mean Wage Standard Deviation 1997 49.39742 7.822257 49.31997 7.234232 1999 55.73508 8.26 0620 55.11172 8.135574 2001 58.57167 7.564645 58.36383 7.408262 2003 75.53049 14.40141 74.83422 13.88978 2005 74.71374 11.58896 73.96442 10.94773 2007 78.63872 11.68048 77.84919 10.77394 A comparison of means test rejects the null h ypothesis that there is no significant difference in the average wage assigned to immigrant physicians compared to native born doctors and supports the alternative hypothesis that foreign born doctors, on average, are more likely to live in areas where wag es are lower and/or sp ecialize in lower paying fields.
58 Table 17 Comparison of Means Test: Wages by Birthplace Birthplace Observations Mean Wage Standard Error 95 Confidence Interval U.S. 120,171 65.96572 0441298 65.87923 66.05222 Foreign 41,664 65.62861 .0719258 65.48763 65.76958 Total 161,835 65.87893 .0376403 65.80516 65.95271 Difference .3371175 .0860848 .1683932 .5058418 Difference = mean(U.S.) mean(Foreign) t = 3.9161 H o : Difference = 0 degrees of freedom = 161,833 H a : diff < 0 H a : diff = 0 H a : diff > 0 Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.0001 Pr(T > t) = 0.0000 Physicians educated at foreign m edical schools seem to earn considerably less than those educated at medical schools in the U.S.; the difference in wages appears to be more drastic than the difference in wages based on place of birth. The mean wage of U.S. educated physicians in the sam ple is estimated to be $66.245 per hour (standard deviation: $15.10868) while the mean wage of foreign educated doctors is estimated to be $64.55275 per hour (standard deviation: $ 15.18964). The averages of the wages assigned to physicians by location of medical school and year, based on specialty and practice area, ar e reported in Table 18
59 Table 18 Average Physician Wages: by Location of Medical School Attended U.S. Educated Foreign Educated Year Mean Wag e Standard Deviation Mean Wage Standard Deviation 1997 49.57645 7.427113 48.62338 8.599974 1999 55.84859 8.082030 54.63304 8.644367 2001 58.66989 7.399834 57.99196 7.911378 2003 75.79231 14.03253 73.76393 14.99110 2005 74.80303 11.28907 73.4 5767 11.89602 2007 78.83199 11.34080 76.87175 11.79168 A comparison of means test rejects the null hypothesis that there is no significant difference in the average wage assigned to physicians educated abroad compared to those educated in th e U.S. and supports the alternative hypothesis that doctors who attended foreign medical schools, on average, are more likely to live in areas where wages are lower and/or specialize in lower paying fields: Table 19 Comparison of Means Test: Wages by Loca tion of Medical School Attended Med School Observations Mean Wage Standard Error 95 Confidence Interval U.S. 126,827 66.24500 .0424249 66.16185 66.32815 Foreign 35,008 64.55275 .0811827 64.39363 64.71187 Total 161,835 65.87893 .0376403 65.80516 65.95271 Diffe rence 1.692255 .0913224 1.513265 1.871245 Difference = me an(U.S. ) mean(Foreign) t = 18.5305 H o : Difference = 0 degrees of freedom = 161,833 H a : diff < 0 H a : diff = 0 H a : diff > 0 Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 0.0000
60 These results suggest that location of medical school may be a more important factor in determining wages than birthpla ce. A comparison of wages by both birthplace and location of medical school reveals an interesting result: the estimated wage of U.S. born physicians educated abroad is the lowest of all physicians This could be explained if U.S. born physicians who are educated abroad are doing so out of necessity; perhaps they are of lower quality in terms of scholastic achievement and cannot get accepted to medical school in the U.S. The comparison of means tests of estimated wages of U.S. and foreign born physicians by location of medical school are reported in Table 20
61 T able 20 Comparison of Means Test: Wages by Birthplace and Location of Medical School Attended U.S. Born Physicians Med School Observations Mean Wage Standard Error 95 Con fidence Interval U.S. 115,017 66.17106 .0445914 66.08366 66.25846 Foreign 5,154 61.38346 .2534297 60.88663 61.88029 Total 120,171 65.96572 .0376403 65.87923 66.05222 Difference 4.787601 .2173727 4.361554 5.213648 Difference = mean(U.S.) mean(Foreign) t = 22.0249 H o : Difference = 0 degrees of freedom = 120,169 H a : diff < 0 H a : diff = 0 H a : diff > 0 Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.00 00 Pr(T > t) = 0.0000 Foreign Born Physicians Med School Observations Mean Wage Standard Error 95 Confidence Interval U.S. 11,810 66.96512 .1375882 66.69542 67.23482 Foreign 29,854 65.09989 .0841467 64.93496 65.26482 Total 41,664 65.62861 .0719258 65.48763 65.76958 Differe nce 1.865227 .1593351 1.552927 2.177527 Difference = mean(U.S.) mean(Foreign) t = 11.7063 H o : Difference = 0 degrees of freedom = 41,662 H a : diff < 0 H a : diff = 0 H a : diff > 0 Pr(T < t) = 1.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 0.0000 For both U.S. born and immigrant physicians, the null hypothesis that there is no significant difference in average wages of physicians educated in U.S. or foreign medical schools is rejected in favor of the alternative, that physicians educated in the U.S. are more likely to live in areas and/or specialize in fields for which wages ar e higher. As previously explained, wages were assigned to physicians based on practice area (PMSA, MSA or state, where applicable) and specialty. For all years and areas in
62 the sample, surgeons are the highest paid of all specialists with an average hour ly wage of $74.05, while family/general practitioners and psychiatrists tend to be the lowest paid with average wages of $59.95 and $60.07 per hour, respectively. The average wages by specialty for all years and areas in the sample, as well as estimated w ages by place of birth and educ ation, are reported in Tables 21 through 23 Table 21 Average Wages by Physician Specialty Specialty Mean Wage Standard Deviation Anesthesiology 69.86168 18.35551 Family/General Practitioners 59.94589 1 3.87846 Internal Medicine 66.40683 14.91470 OB/GYN 70.66007 16.42572 Pediatrics 60.99174 13.40880 Psychiatry 60.06922 13.41130 Surgeons 74.05098 16.18843 Other 64.97257 12.49072
63 Table 22 Average Wages by Physici an Specialty and Birthplace U.S. Born Foreign Born Specialty Mean Wage Standard Deviation Mean Wage Standard Deviation Anesthesiology 69.83370 18.38523 69.92542 18.29101 Family/General Practitioners 60.11647 13.85956 59.26293 13.93 452 Internal Medicine 66.18644 15.11365 66.86565 14.48198 OB/GYN 70.59882 16.49930 70.85211 16.19482 Pediatrics 61.18670 13.53934 60.46149 13.03417 Psychiatry 59.90264 13.56181 60.47003 13.03556 Surgeons 74.17250 16.32209 73.54924 15.61601 Other 64.99336 12.64371 64.91615 12.06612 Table 23 Average Wages by Physician S pecialty and Location of Med School Attended U.S. Educated Foreign Educated Specialty Mean Wage Standard Deviation Mean Wage Standard De viation Anesthesiology 70.14947 18.32521 69.05622 18.42028 Family/General Practitioners 60.48336 13.51183 57.50910 15.19694 Internal Medicine 66.40421 14.92556 66.41366 14.88749 OB/GYN 70.90430 16.32767 69.65452 16.79034 Pediatrics 61.38920 13.35084 59.68996 13.51736 Psychiatry 60.09691 13.46288 60.00015 13.28386 Surgeons 74.29921 16.18393 72.52526 16.13460 Other 65.16821 12.43361 64.28210 12.66670 As reported earlier in Table 11 women in the sample were more likely to be general practitioners or to specialize in OB/GYN, pediatrics or psychiatry. Wages for these specialties, with the exception of OB/GYN, tend to be lower than other specialties. Men, on the other hand, are more likely to specialize in anesthesiology and s urgery,
64 which are two of the highest paying fields. This provides evidence that differences in pay by sex may be attributable to specialty choice, and not necessarily based on discrimination. As reported in Table 12 U.S. born physicians were more likely to specialize in surgery than their immigrant counterparts. Again, as surgery is the highest paid specialty, differences in average physician wages by birth country might be explained by specialty choice. Similarly, graduates of foreign medical schools were more likely to be general practitioners and much less likely to be surgeons; therefore, the fact that graduates of foreign medical schools are estimated to have lower wages than doctors educated in the U.S. might also be at least partially explained b y specialty. In this case, specialty could refer to specialty choice, or lack thereof, if foreign medical graduates specialize in other fields but must resort to practicing general medicine in the U.S. Interestingly, when comparing the estimated wages of U.S. born and foreign born physicians by specialty, overall the wages tend to be quite similar. However, when comparing estimated wages of physicians educated in the U.S. and abroad, the wage estimates for physicians educated at foreign medical schools ar e noticeably lower than physicians educated in the U.S. This implies that lower average wages of physicians educated outside the U.S. may result from specialty choice as well as location of practice area, with foreign educated physicians locating in areas where wages are already lower or where migration of foreign educated physicians actually causes wages to be lower in a specific area.
65 Chapter 10 : Regression Results 10.1: Variable Summary A summary of the variables employed in this study is reported in Table 24. It is important to note that in the following regression analyses, the dependent variable, the natural log of wages ln w is not the actual wage of the individual physician sampled from the survey. Instead, average physician wages as reporte d in the Occupational Employment Statistics survey were assigned to the physicians sampled by specialty, MSA or state where applicable, and year.
66 Table 24 Summary of Variables Variable Description L n w a,t Dependent variable natural log of the average hourly wage rate of physicians by specialty, indexed by area and year I a,t Immigration share percentage of physicians in a given area and year who are foreign born Male Dummy variable = 1 if physician is male, = 0 if female Age Physician age A g e 2 Square of physician age MedB a,t Percentage of population in physicians service area in a given year who are covered by Medicare Part B FBFE_ Share a,t Foreign born foreign educated share percentage of immigrant physicians in a given area a nd year who attended a foreign medical school FBUSE_ Share a,t Foreign born U.S. educated share, percentage of immigrant physicians in a given area and year who attended a U.S. medical school USBFE_ Share a,t U.S. born foreign educated share, percentage of native born physicians in a given area and year who attended a foreign medical school S a,t Physician Stock, number of physicians as a percentage of the area population in a given y ear FB90 Instrumental Variable: percentage of area population that was foreign born as of the 1990 Census
67 The correlations between the independent variables employed in this analysis are reported in Table 2 5 Table 25 Pairwise Correlations between Independent Variables Variable: I FBFE Share FBUSE Share USBFE Share Male Age Age 2 MedB S I 1.000 FBFE 1.000 FBUSE 0.109 1.000 USBFE 0.082 0.043 1.000 Male 0.028 0.011 0.048 0.055 1.000 Age 0.030 0.040 0.013 0.040 0.239 1.000 Age 2 0.036 0.044 0.007 0.040 0.229 0.987 1.000 MedB 0.144 0.001 0.370 0.051 0.065 0.047 0.039 1.000 S 0.033 0.014 0.060 0.048 0.019 0.011 0.011 0.104 1.000 10.2 : Ordinary Least Squares and Two Stage Least Squares Results The OLS and 2SLS estimates of the impact of immigration on are a wages are reported in Table 26 Estimation of these regressions is performed at the individual level, where the immigrant share in an area and year are regressed on the natural log of the wage assigned to the individual physicians sampled. All regressions include time fixed effects and robust standard errors are reported in parentheses.
68 Table 26 OLS and 2SL S Estimates of the Relationship between Immigration and Physician Wages Dependent Variable: Natural Log of Physician Wages: ln w Independent Variable (1) OLS w/ Time Fixed Effects (a) (b) (c) (2) 2SLS w/ Time Fixed Effects (a) (b) (c) I = Immigrant Share 0.0131 *** 0.0720 *** 0.1688 *** 0.1723 *** (.0031) (.0044) ( .0055 ) (.0058) FBFE_Share 0.0564 *** 0.0696 *** (.0057) (.0065) FBUSE_Share 0 .1621 *** 0.4770 *** (.0143) (.0253) USBFE_Share 1.2630 *** (.0254) Male 0.0306 *** 0.0310 *** 0.0297 *** 0.0289 ** 0.0295 *** 0.0295 *** (.0010) (.0011) (.0011) ( .0009 ) (.0010) (.0010) Age 0.0011 *** 0.0010 *** 0.0011 *** 0.0006 *** 0.0008 *** 0.0007 *** (.0002) (.0002) (.0002) ( .0002 ) (.0002) (.0002) Age 2 0.0000 *** 0.0000 *** 0. 0000 *** 0 .0000 *** 0.0000 *** 0.0000 *** (.0000) (.0000) (.0000) ( .0000 ) (.0000) (.0000) MedB 0.4714 *** 0.5449 *** 0.7278 *** 0 .3673 ** 0.5627 *** 0.3526 *** (.0117) (.0168) (.0185) ( .0113 ) (.0152) (.0212) S = Physicia n Stock 0.4487 *** 0.4994 *** 0.4643 *** 0.4696 *** (.0102) (.0105) (.0102) (.0103) First Stage F Test of Instrument 68,737 96,185 23,569 29,492 # of obs 161,811 129,375 129,375 159,448 1 28,074 128,074 R Squared 0.4946 0.5076 0.5491 0.5483 0.5513 0.5508 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: Parts (1) (b), (1)(c ), (2) (b) and (2) (c) have fewer observations d ue to lack of data on the physician stock variable. 2SLS regressions have fewer variables due to lack of data on the instrumental variable.
69 T he first specification of the model (1) employs an ordinary least squares analysis, with ti me dummy variables to control for time fixed effects, to test for the impact of immigration on local area wages The joint test of significance reveals that, for all three specifications (1)(a), (1)(b) and (1)(c), the included independent variables are significant in explainin g variation in area wages at the 1 level. The R squared coefficients of determination indica te that between 49.46 and 54.91 of the variability in wages can be explained by the regression models above. The first regression run, (1)(a), indicates a hi ghly significant, although practically small, impact of immigration on local wage s. Specifically, areas with a one percentage point larger immigrant share (where the immigrant share is the percentage of doctors in the area who are foreign born), ceteris p aribus, have 0.0131 lower wages, on average. Adding the variable S (1)(b ), which measures the stock of physicians, the percentage of physicians in a given area and year relative to total population, controls for the increase in supply that occurs when a n immigrant physician moves to an area and the possibility that, in response, other physicians move away from the area. The coefficient on this variable is negative, as expected, large and highly signi ficant. All else constant, areas with one percentage point more physicians per capita have 0.4487 lower average wages. This result is consistent with the basic supply and demand theory that an increase in supply, holding demand constant, should result in a lower wage After including this supply control, the negative effects of immigration become larger, and remain significant at t he 1 level. In this case, holdin g physician supply constant, a one percentage point increase in the number of foreign physicians in a given area is predicted to result in .0720 lower area wages, on average. The result that the effect is smaller when supply
70 is not held constant suggests that local physicians may respond to the lower wages caused by an influx of immigrant physicians by moving away from the area. The third specif ication of the OLS regression model (1)(c) adds controls for country of education. Specifically, the three variables FBFE_Share FBUSE_Share and USBFE_Share are added to the model, with U.S. born and educated physicians omitted as the base group for comp arison purposes. The coefficients on all three of these birth/education variables are significant at the 1 level. Areas with a one percentage point larger share of physicians who were born and educated abroad are predicted to have 0.0564 higher wages, on average. Areas with a one percentage point larger share of physicians born in a foreign country but educated in the U.S. are predicted to have 0.1621 lower wages, on average, and areas with a one percentage point larger share of physicians born in the U.S. but educated abroad are predicted to have 1.263 lower average wages. These results provide evidence that the negative impacts of immigration come not from foreign born and educated doctors, but from immigrant physicians who are educated in the U.S At first this result may seem counterintuitive, but there is a probable explanation for it. Foreign born, U.S. educated physicians could be one of two types: those who migrated to the U.S. as small children and grew up in America and those who came to the United States as adults for purposes of higher education. Both groups of physicians may have benefited from university and medical school admissions policies that favor foreign students. Furthermore, the latter group likely is less restricted by visa issues as the longer they have been in the U.S., the more likely that they have achieved resident alien or citizen status. Foreign born and educated physicians, on the other hand,
71 immigrated at some point in their lives after achieving their medical degr ee. They are more likely than their U.S. educated counterparts to have entered the United States through the H 1B visa program by a sponsored employer or through the O visa program as an extraordinary p hysician in their field. In either of these cases, t he foreign born foreign educated physician is likely to be either the cream of the crop in their specialty or one who is extremely driven and dedicated to achieve his or her goals. This theory is consistent with the findings of Chiswick (1999) that migran ts tend to be favorably self selected. Thus, it is certainly plausible that these physicians do actually earn higher wages than their foreign born, U.S. educated counterparts. The physicians who seem to have the most negative effect on wages are those b orn in the U.S. but educated abroad. This result is likely explained because U.S. born physicians who attend a foreign medical school do so out of necessity; they may not have been accepted to a U.S. medical school. Therefore, they may be of lower qualit y which would result in them being more likely to practice as general practitioners than high paid specialists, potentially having less selection as far as jobs are concerned and thus being forced to work in less desirable, lower paying areas, and being le ss productive and thus paid a lower wage on average. As previously discussed, the results of the OLS regression s likely suffer from endogeneity bias because immigrant location is not randomly selected but, rather, chosen by the immigrant. The results of t he 2SLS analysis are previously reported in Table 26. In the first two specifications (2)(a) and (2)(b), where the measure of immigration is represented by I the immigrant share, the instrument FB90 the foreign born share of the areas population in 199 0, has been employed. The results of the first stage regressions
72 reveal that the instrument is highly significant in predicting I after controlling for the other independent regressors. Specifically, there is a strong positive relationship between the fo reign born population share in an area in 1990 and the current share of immigrant physicians. The F Test of the significance of the instrument is extremely significant and shows that the instrument passes the first requirement, that it be highly correlate d with I and significant in the first stage regressions. The second assumption, that FB90 is not correlated with the error term and is therefore not directly related to current wages, cannot be tested. Following the earlier research of Altonji and Card ( 1991) and Card (2001), this study assumes that the second condition is satisfied. The results of the second stage regressions show a highly significant negative effect of immigration on wages. In this case, the negative effect is even larger than reported in the OLS regres sions. A one percentage point increase in the share of immigrant physicians in an area is associated with 0.1688 to 0.1723 lower average physician wages. The result that the negative impacts of immigration as measured by the 2SLS appr oach which controls for endogenous location choice are more negative than the OLS results provides evidence that immigrants are in fact drawn to areas where wages are higher prior to their arrival. A comparison of the results from the OLS regression (1)( b) and the 2SLS regressions (2)(b) provides evidence that local physicians respond to the decrease in wages caused by immigration by moving out of the area, and that the lowest paid physicians in the area are more likely to move out than the higher paid ph ysicians. The se two specifications of the model include the physician supply control variable S and measure the impact of immigration on wages holding physician supply constant
73 However, this does NOT mean that these analyses do not allow local physicia ns to move out. Specifically, holding physician supply constant requires that for each physician moving into the area, one physician must move out. The results of the 2SLS analysis are more negative than those of the OLS regression; adding the control fo r endogeneity indicates that the true impact of immigration is more negative than is observable using the OLS approach. If immigrant physicians move into an area and drive out the lowest paid physicians, the decrease in average physician wages in the area would be offset by the flight of these low paid doctors, and the average wage observed in the area would be higher than it would have been if these lowest paid physicians had remained in the area. Performing the 2SLS analysis on the impact of immigration by country of medical school on area wages requires the inclusion of not only the instrument FB90 but also the FE_IV the calculation of which was described earlier in Chapter 7.3. The results of the first stage regressions show a strong and highly statis tically significant correlation between both instruments and the two endogenous regressors, FBFE_Share and FBUSE_Share. The F tests for joint significance of the instruments in the regression of FBFE_Share and FBUSE_Share on the instruments and other exog enous regressors produce a value of 29,492 and 23,569, respectively, and show that the two instruments are highly significant in predicting immigrant share by country of education. Breaking down the results of the 2SLS analysis by location of medical sch ool attended in part (2)(c) also produces a uniformly more negative effect of immigration on wages. The coefficient on the FBFE_Share variable (foreign born, foreign educated share of physicians in an area) changes signs from positive to negative and is s tatistically significant at the 1 level. Specifically, after controlling for endogenous location choice,
74 a one percentage point increase in the foreign born, foreign educated share of physicians in an area is associated with a 0.0696 decrease in average wages. A one percentage point increase in the foreign born, U.S. educated share of physicians in an area is associated with a 0.477 decrease in average area wages. This effect is also statistically significant at the 1 level. The negative impacts of immigration remain larger for the foreign born, U.S. educated physicians, the explanation for which was previously discussed. The change in sign of the FBFE_Share variable and the increase in the magnitude of the negative coefficient on the FBUSE_Share va riable both support the previous hypotheses that immigrant physicians, regardless of where they are educated, are drawn to areas where wages are higher, and that local physicians, particular those from the low end of the pay scale, move out of the area in response to an influx of immigration, diffusing the potential negative wage impacts of the immigrant physicians. The 2SLS analysis controlling for both country of birth and education had to be performed without the inclusion of the USBFE_Share variable. T his variable is likely endogenous as well, as U.S. born, foreign educated physicians are just as likely to be drawn to higher paying locations as foreign born physicians are. However, without a valid third instrument it is not possible to include this var iable in the analysis. Thus, the base category that the wage impacts of FBFE and USBFE physicians is to be compared to is all U.S. born physicians, regardless of where they were educated. As the focus of this study is to determine the impact of immigrati on of foreign physicians on wages, this limitation does not substantially hinder the research. The signs of all of the demographic characteristic variables in part s (1) and (2) are as expected and all are significant at the 1 level. M ale doctors are e xpected to be paid
75 about 3 higher wages, on average, than females. Physicians wages increase with age but at a decreasing rate. The coefficient on MedB is large, positive, and highly significant. Areas where one percentage point more of the populati on are covered by Medicare Part B are associated with 0.3526 to 0.7278 higher physician wages, on average, depending on the specification of the model employed. T he strong positive impact is likely due to areas with a higher percentage of patients cover ed by Medicare Part B having higher demand for physician services and, perhaps, also because physicians are more likely to actually receive payment when a patient is covered by Medicare compared to self pay or uninsured patients. In order to be covered by Medicare Part B, p atients must be over the age of 65 or disabled; these populations are large consumers of medical care. In all specifications o f the model reported in Table 26 the impact of an increase in the number of area physicians per capita is uni formly negative and highly significant. Areas with one percentage point more physicians per capita have 0.4643 to 0.4994 lower physician wages, on average. This result is consistent with the basic economic theory of labor supply and demand. A larger s upply of physicians, ceteris paribus, should be associated with lower equilibrium wages.
76 10. 3 : OLS with Area Fixed Effects and First Difference Results The followi ng regression analyses employ the ordinary least squares method with area fixed effec ts and the first differencing approa ch Both of these methods control for unobserved area specific factors that affect wages and isolate the change in wages due to changes in the immigrant share. The coefficients in each of these regressions are interpre ted as the marginal effect of a change in immigration on area wages. Following Orrenius and Zavodny (2007) and Altonji and Card (1991), for both of these approaches the data are collapsed and analyzed at the area level. The number of observations, theref ore, drops from 161,811 physicians to 599 areas (comprised of MSAs and counties) for the OLS regressions with area fixed effects, and even further to 487 areas for the first difference analyses (areas which did not have physicians sampled in every year wer e dropped so as to create a balanced panel). The dummy variable, male is not los t in these regressions as when it is collapsed at the area level it is no longer a dummy variable and instead becomes the proportion of area physicians who are men in a given year. It is important to note that the first difference approach is more likely than the cross sectional (OLS and 2SLS) approach to capture short run effects of immigration, in which local labor markets have not had time to adjust. Furthermore, the rel ative magnitude of the short run and long run effects on wages depend on whether there are barriers to wage adjustments in the short run (Altonji and Card 1991). The results of the OLS regressions with area fixed effects and the first difference regres sions are reported in Table 27. All coefficient estimates are reported with robust standard errors in parentheses. For presentation purposes the independent variables are listed on the left, and it should be noted that in the case of the first difference model they
77 are not in level form but instead represent the change in the variable from one time period to the next (for example, male in the first male, but for ease of presentation is simply listed as male )
78 Table 27 Fixed Effects and First-Difference Estimates of the Relationship between Immigration and Physician Wages Dependent Variable: ln w Independent Variable (3) OLS w/ Time and Area Fixed Effects (a) (b) (c) (4) First-Difference (a) (b) (c) I = Immigrant Share -0.0931 0.0058 0.1859 ** .0982 (.1111) (.1038) (.0749) (.0738) FBFE_Share 0.0435 0.1776 (.1262) (.0930) FBUSE_Share -0.1157 0.0181 (.1849) (.1677) USBFE_Share -0.2701 -0.8188 ** (.2478) (.3920) Male 0.0479 0.0757 0.0758 0.2057 ** 0.2226 ** 0.1859 (.1023) (.1052) (.1036) (.0950) (.0956) (.0992) Age -0.0151 -0.0160 -0.0168 0.0700 *** 0.0220 0.0328 (.0230) (.0226) (.0227) (.0243) (.0203) (.0206) Age 2 0.0001 0.0001 0.0001 -0.0008 *** -0.0003 -0.0004 (.0002) (.0002) (.0002) (.0003) (.0002) (.0002) MedB 0.1521 0.3886 0.3856 0.8642 *** 0.3661 0.5601 *** (.6359) (.6373) (.2475) (.2130) (.2106) S = Physician Stock -0.2295 -0.2329 -0.5954 *** -0.6873 *** (.1814) (.1832) (.1288) (.1461) # of obs 599 533 533 487 398 398 R-Squared .8833 .8976 .8980 .1149 .0814 .1421 Adj. R-Squared .8511 .8662 .8661 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: Parts (3)(b), (3)(c), (4)(b) and (4)(c) have fewer observations due to lack of data on the physician stock variable. First-Difference Regressions represent a balanced panel and therefore have fewer observations than OLS Regressions.
79 The first set of regression results presented in Table 27 are those of the fixed effects estimation. The results of these regressions are somewhat unsatisfactory in that none of the individual regressors is statistically significant. A loss of efficiency compared to other econometrical methods is generally associated with fixed effe cts estimation, and in this case the sample size compared to the previous OLS and 2SLS regressions has been greatly reduced. So far all coefficient estimates, even those of the OLS and 2SLS models, have been relatively small. Therefore, if there is even less variation of wages within areas than there is between areas it will be difficult to obtain statistically significant results. Furthermore, any sampling or measurement error that exists in the calculation of the I, FBFE_Share, FBUSE_Share, and USBFE_S hare variables, as well as the assignment of wages to physicians, is likely to be more problematic in the fixed effects model than in other models. Aydemir and Borjas (2006) posit that it is precisely this measurement error that causes attenuation bias in studies of the impact of immigration on wages. Specifically, they report that the inclusion of these fixed effects implies that there is very little identifying variation left in the variable that captures the immigrant supply shift, permitting the samp ling error in the immigrant share to play a disproportionately large role. As a result, even very small amounts of sampling error get magnified and easily dominate the remaining variation in the immigrant share. Indeed, the three fixed effects specifica tions of the model (3)(a), (3)(b) and (3)(c) provide no evidence that changes in the share of immigrant physicians have any significant effect on average physician wages, once time invariant area specific facto rs that effect wages are accounted for.
80 The second set of results reported in Table 27 is those of the first difference regressions. Because unobservable factors that affect wages are likely to be correlated over time, serial correlation of the error term is likely to be present ( this has been acco unted for by using robust standard errors). In the presence of serial correlation of the error term, the first difference model is more efficient than the fixed effects model. Thus, it is likely the better model for measuring the impact of immigration on wages within an area in this case. The first specification which allows per capita phys ician supply to vary indicates that, in the short run immigration of foreign physicians actually has a positive impact on average area wages Specifically, a one pe rcentage point increase in the share of immigrant physicians in an area is associated with a 0.1859 increase in the average physician wage in the area. This result can be explained as follows: as previously discussed, wages of physicians, like all wages, are likely to be sticky, and even more so than wages in other professions due to payment mechanisms involving third par ty payors. Suppose the supply of physicians in an area increases due to immigration. To the extent that physicians are paid based on a fee for service mechanism, physician salaries will not be affected until third party payors such as Medicare or insurance companies realize that the physician supply has shifted and that they can lower reimbursement rates without jeopardizing their cont racts with the providers. In the meantime, if the immigrant physicians are not the lowest paid doctors in the area, there presence could actually increase the average physician wage in the area in the short run. In the long run, wages will adjust to the supply shift a s expected and be reduced.
81 The results of part (4)(b) include the physician supply control and show that the short run positive impacts of immigration on wages would be lower if supply were held constant, although this effect is not statis tically significant. There is thus some inconclusive evidence that, even in the short run, the wage impacts of immigration are diluted by local physician labor movement. The results of part (4)(c), which break down the impact of immigration by location o f medical school, are in agreement with the results of (4)(a) and (4)(b). The short run impact of changes in the share of foreign born immigrants in an area on average local wages is predicted to be positive, whether the immigrant physician was educated a broad or in the U.S. The positive wage impact is stronger for the foreign born and foreign educated, and is st atistically significant at the 10 level, while the small positive effect of foreign born, U.S. educated physicians is not statistically signific ant. Specifically, a one percentage point increase in the share of foreign born, foreign educated physicians in an area is expected to increase average physician wages in t he area by 0.1776 in the short run. The result that foreign born, foreign educate d physicians have a larger positive impact on wages than their U.S. educated immigrant physician counterparts can be explained in the same way as the earlier 2SLS results where these physicians had a less negative long run impact on wages than their U.S. e ducated counterparts. As was the case in all prior regressions, migration of U.S. born physicians has the most negative effect on average area wages. It seems that, whether in the short run or long run, these physicians are paid the lowest of any of the four categories of physicians.
82 The signs and magnitudes of all of the other control variables included in the first difference regressions are as expected and consistent with the results of the OLS and 2SLS regressions reported in Table 26. 10.4: Two Sta ge Least Squares First Difference Results The final specification of the model involves incorporating the instrumental variables described earlier to control for endogeneity that is not completely accounted for by the first difference approach. The resul ts of this specification of the model are reported in Table 28.
83 Table 28 2SLS First-Difference Estimates of the Relationship between Immigration and Physician Wages Dependent Variable: Change in Natural Log of Physician Wages: ln w Independent Variable (5)(a) (5)(b) (5)(c) Immigration Share 0.1590 0.0653 (.1086) (.1134) FE_Share 0.0751 (.1456) Share 0.3992 (.4028) 0.2593 *** 0.2406 ** 0.2410 ** (.0965) (.0999) (.0982) 0.0483 ** 0.0154 0.0154 (.0218) (.0192) (.0192) 2 -0.0006 ** -0.0002 -0.0002 (.0002) (.0002) (.0002) 1.0149 *** 0.5438 *** 0.5259 (.2281) (.1790) (.3057) -0.5500 *** -0.5517 *** (.1192) (.1211) First-Stage F-Test of Instrument 74.29 59.99 51.52 25.65 # of obs 482 395 395 R-Squared 0.1292 0.1107 0.1113 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: All regressions have fewer observations than Part (4) above due to lack of data on the IV. Part (5)(a) has more observations than (5)(b) and (5)(c) due to lack of data on the S variable. The results of the first stage regression analyses indicate that the instruments are highly significant in predicting immigrant share after controlling for the other exogenous regressors. The F test of joint significance further indicates that the instruments are highly correlated with the measures of immigration employed in this analysis and thus meet the first criterion of an acceptable IV. These first stage resul ts are reported in the appendix in Table A5. Following Altonji and Card (1991) and Friedberg and Hunt (1995), the instruments are not likely to be correlated with future changes in wages.
84 The estimated short run marginal effects of changes in the share o f immigrant physicians on average area wages, after controlling for endogenous location choice, are similar to the first difference estimates without the instrumental variable but are smaller in magnitude. This provides additional evidence that immigrant physicians do in fact choose to locate in areas with higher wages. However, none of these estimates are statistically significant. There is evidence that had the immigrant physician not chosen to move to the highest paying area, there would have been no significant short run effect on wages at all. 10.5: Analyses with Specialty Controls It is possible that the short run positive and long run negative effects of physician immigration could be attributable to specialty choice. In particular, if foreign physicians are more prevalently represented in lower paying specialties, this would explain their long run negative impact on average area wages. The following analyses incorporate controls for specialty choice for each of the models previously presented. The results of the OLS regressions with time fixed effects and the 2SLS regressions are presented in Table 29. The base specialty category for comparison purposes is Other S pecialty which, as defined in Chapter 8.2, is the average wage of all physicia ns in a given area. These results compare to those previously reported without specialty controls in Table 26.
85 Table 29 OLS and 2SLS Estimates of the Relationship between Immigration and Physician Wages with Specialty Controls Dependent Variable: Natural Log of Physician Wages: ln w Independent Variable (6) OLS w/ Time Fixed Effects (a) (b) (c) (7) 2SLS w/ Time Fixed Effects (a) (b) (c) I = Immigrant Share -0.0232 *** -0.0769 *** -0.1728 *** -0.1759 *** (.0028) (.0041) (.0052) (.0054) FBFE_Share 0.0517 *** -0.0557 *** (.0054) (.0062) FBUSE_Share -0.1811 *** -0.5319 *** (.0133) (.0227) USBFE_Share -1.230 *** (.0250) Male 0.0050 *** 0.0048 *** 0.0041 *** 0.0043 *** 0.0039 *** 0.0039 *** (.0010) (.0010) (.0010) (.0009) (.0009) (.0009) Age 0.0008 *** 0.0008 *** 0.0010 *** 0.0003 0.0006 *** 0.0006 *** (.0002) (.0002) (.0002) (.0002) (.0002) (.0002) Age 2 -0.0000 *** -0.0000 *** -0.0000 *** -0.0000 ** -0.0000 *** -0.0000 *** (.0000) (.0000) (.0000) (.0000) (.0000) (.0000) MedB 0.5083 *** 0.5256 *** 0.6933 *** 0.4051 *** 0.5419 *** 0.2964 *** (.0101) (.0156) (.0174) (.0104) (.0139) (.0196) S = Physician Stock -0.4569 *** -0.5062 *** -0.4722 *** -0.4784 *** (.0080) (.0084) (.0081) (.0082) Anesthesiology 0.0507 *** 0.0503 *** 0.0500 *** 0.0479 *** 0.0487 *** 0.0492 *** (.0020) (.0022) (.0022) (.0019) (.0021) (.0021) Family/General -0.0834 *** -0.0859 *** -0.0848 *** -0.0796 *** -0.0812 *** -0.0810 *** (.0014) (.0016) (.0015) (.0012) (.0014) (.0014) OBGYN 0.0760 *** 0.0738 *** 0.0751 *** 0.0775 *** 0.0760 *** 0.0760 *** (.0017) (.0019) (.0017) (.0015) (.0016) (.0016) Pediatrics -0.0792 *** -0.0852 *** 0.0824 *** -0.0761 *** -0.0826 *** -0.0828 *** (.0017) (.0018) (.0017) (.0015) (.0016) (.0016) Psychiatry -0.0814 *** -0.0860 *** -0.0847 *** -0.0820 *** -0.0879 *** -0.0874 *** (.0019) (.0020) (.0020) (.0017) (.0019) (.0019) Surgery 0.1148 *** 0.1135 *** 0.1120 *** 0.1116 *** 0.1119 *** 0.1123 *** (.0010) (.0011) (.0010) (.0009) (.0009) (.0010) Internal Medicine 0.0118 *** 0.0133 *** 0.0144 *** 0.0141 *** 0.0140 *** 0.0138 *** (.0013) (.0013) (.0013) (.0011) (.0012) (.0012) First-Stage F-Test of Instrument 68,151 95,898 29,240 23,472 # of obs 161,811 129,375 129,375 159,448 128,074 128,074 R-Squared 0.5726 0.5875 0.6268 0.6311 0.6373 0.6365 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note Parts (6)(b), (6)(c), (7)(b) and (7)(c) have fewer observations due to lack of data on the physician stock variable. Part (7) has fewer observations than Part (6) due to lack of data on the instrumental variable.
86 The results reported in Table 29 as to the impact of physician immigration on local wages are almost identical to those reported earlier in Table 26 without the specialty controls. The small negative impacts of immigration on local wages remain highly significant even when specialty is controlled for. Thus, the negative impact of an increase in the share of immigrant physicians on local area wages cannot be attributed to specialty choice. T hese results provide evidence that immigration of physicians lowers the average physician wage in an area, most likely because they are willing to accept a lower wage than physicians already practicing in the area. Specifically, the OLS results in part (6 )(a) average wage of physicians in areas with a one percentage point larger share of immigrant physicians is expected to be 0.0232 lower on average, and this lower wage is not attributable to foreign physicians specializing in lower paying fields. This e ffect is significant at the 1 level. Because the effect becomes even more negative when physician supply is held constant, one can further infer that foreign born physicians must be paid less, on average, than native born physicians. If the immigrant sha re increases but per capita physician supply is held constant, the share of native born physicians in an area must decrease. The larger negative coefficients in both specifications of the 2SLS model, (7)(a) and (7)(b), continue to provide evidence that al though foreign born physicians are willing to accept less than the average area wage, they are still not the lowest paid of all physicians in the area, on average. The 2SLS results suggest that the negative impacts of immigration would have been even larg er on the level of about 0.17 for a one percentage point increase in immigrant share, if the lowest paid physicians in the area had not responded to the influx of immigration by moving away. There is evidence provided throughout this
87 study that the lowe st paid of all physicians are the U.S. born, foreign educated. The results of the 2SLS regressions also continue to provide evidence that foreign born physicians are drawn to areas where wages are already higher, on average. The regression results control ling for location of education in parts (6)(c) and (7)(c) are also consistent with those reported in Table 26. Even after controlling for specialty, the impact of immigration of foreign born, foreign educated physicians initially appears positive, but, af ter controlling for endogenous location choice, is small and negative but highly significant. The effects of immigration of foreign born, U.S. educated physicians remain more negative than the effects of the foreign born, foreign educated physicians and t he magnitudes of the coefficients are similar to those reported without specialty controls. The negative effects of immigration, and the larger negative effects of an increase in the share of foreign born, U.S. educated physicians in an area, are therefor e not attributable to specialty choice. After adding controls for specialty, the positive coefficients on the male dummy variable, although small to begin with, become even smaller in magnitude, and remain highly significant. Once specialty choice is co n trolled for, men earn about 0. 4 higher wages than women, on a verage, as compared to about 3 .0 higher wages reported in Table 26 without specialty controls. Thus, much of the wage inequality between men and women can be attributed to the fact that women are more likely to specialize in lower paying fields. The results of the OLS regressions with area fixed effects and the first difference analyses, with specialty controls included, are reported in Table 30.
88 Table 30 Fixed Effects and First-Difference Estimates of the Relationship between Immigration and Physician Wages with Specialty Controls Dependent Variable: ln w Independent Variable (8) OLS w/ Time and Area Fixed Effects (a) (b) (c) (9) First-Difference (a) (b) (c) I = Immigrant Share -0.0758 00419 0.1279 0.0583 (.1140) (.1034) (.0687) (.0700) FBFE_Share 0.1004 0.1308 (.1245) (.0921) FBUSE_Share -0.1100 -0.0007 (.1824) (.1641) USBFE_Share -0.1913 -0.7953 ** (.2527) (.3796) Male 0.0154 00819 0.0870 -0.0045 0.0970 0.1118 (.1089) (.1129) (.1115) (.1122) (.1137) (.1039) Age -0.0120 -0.0152 -0.0170 0.0682 *** 0.0152 0.227 (.0237) (.0223) (.0224) (.0243) (.0222) (.0221) Age 2 0.0001 00001 0.0001 -0.0008 *** -0.0002 -0.0003 (.0002) (.0002) (.0002) (.0002) (.0002) (.0002) MedB 0.2036 03999 0.3979 0.7608 *** 0.3071 0.5081 *** (.6482) (.6323) (.6358) (.2259) (.2112) (.1969) S = Physician Stock -0.2403 -0.2424 -0.6512 *** -0.7423 (.1789) (.1810) (.1251) (.1413) Anesthesiology -0.0911 -0.1289 -0.1230 0.1281 0.0493 0.0282 (.1900) (.1780) (.1777) (.2042) (.1974) (.2066) Family/General -0.0996 -0.1078 -0.0971 -0.1271 -0.0622 -0.0859 (.1680) (.1429) (.1436) (.1226) (.1159) (.1158) OBGYN 0.1863 0.2557 0.2620 0.0036 0.3091 0.4060 ** (.1947) (.1911) (.1912) (.2126) (.2146) (.1973) Pediatrics -0.1284 -0.0226 -0.0157 -0.4201 ** -0.0765 0.0334 (.1653) (.1651) (.1638) (.1971) (.1729) (.1593) Psychiatry -0.3023 -0.1087 -0.1009 -0.7930 *** -0.6100 ** -0.5408 *** (.1984) (.2107) (.2119) (.2523) (.2414) (.2092) Surgery 0.1167 0.2595 0.2622 ** 0.2315 0.2631 0.1947 (.1321) (.1219) (.1232) (. 1480 ) (.1405) (.1324) Internal Medicine -0.1412 -0.1102 -0.1131 -0.0010 0.0546 0.0859 (.1656) (.1768) (.1760) (.1830) (.1645) (.1610) # of obs 599 533 533 487 398 398 R-Squared 0.8850 0.9004 0.9007 0.1667 0.1288 0.1831 Adj. R-Squared 0.8512 .08675 0.8673 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note Parts (8)(b), (8)(c), (9)(b) and (9)(c) have fewer observations due to lack of data on the physician stock variable. First-Difference Regressions represent a balanced panel and therefore have fewer observations than OLS Regressions.
89 With regard to both the fixed effects and first difference regressions, it is important to note that, as before with the male dummy variable, collapsing the data at the area and year le vel transforms the dummy variables for each specialty code into the proportion of area physicians in each year that specialize in the given field. Therefore, these variables are not dropped from the regression analyses. The results of the regressions wit h area fixed effects are problematic, as before. The probable explanation for the insignificant results and attenuation bias is discussed in Chapter 10.3. The signs of the coefficients on the immigrant share variables in the first difference analyses (9) (a) and (9)(b) are the same as in the first difference regressions (4)(a) and (4)(b) without specialty controls, but the magnitudes of the coefficients are smaller. This suggests that some of the short run positive effects on average area wages resulting from a change in the immigrant physician share within an area can be explained by specialty: the foreign born physicians moving into the area tend to specialize in higher paying fields. However, the positive effect without controlling for physician supply is still significant at the 10 level: a one percentage point increase in the share of immigrant physicians in an area is associated with a 0.1279 increase in average physician wages in that area, even after controlling for the effect of physician specia lty. The effect becomes even smaller and insignificant one changes in physician supply are controlled for, evidence that, even in the short run, some of the lowest paid physicians may be leaving the area in response to the immigrant inflow. After contr olling for both physician specialty and foreign medical education, the only group of physicians whose migration has a significant impact on area wages in the short run is the U.S. born, foreign educated group. A one percentage point increase in
90 the share of area physicians who are U.S. born and educated abroad is predicted to reduce the average area wage by 0.7953. This result is significant at the 5 level and the magnitude is similar to the first difference result without specialty controls. Thus, one may conclude that this group of physicians tends to be paid the least, regardless of their specialty. The results of the first difference regressions including specialty controls and the instrumental variables to address the potential endogeneity probl ems are reported in Table 31. The results of the first stage regressions are reported in the appendix in Table A7. Consistent with earlier results, even after including specialty controls the instruments are highly significant in predicting changes in th e immigrant stock.
91 Table 31 2SLS First-Difference Estimates of the Relationship between Immigration and Physician Wages with Specialty Controls Dependent Variable: Change in Natural Log of Physician Wages: w Independent Variable (10)(a) (10)(b) (10)(c) 0.1673 0.0793 (.1134) (.1150) 0.1024 (.1510) 0.0201 (.4073) 0.1114 0.1807 0.1818 (.1049) (.1062) (.1047) 0.0450 ** 0.0053 0.0053 (.0220) (.0205) (.0205) 2 -0.0005 ** -0.0001 -0.0001 (.0002) (.0002) (.0002) 0.9718 *** 0.5205 *** 0.4789 (.1987) (.1670) (.2944) -0.5946 *** -0.5978 (.1181) (.1197) 0.0202 -0.0250 -0.0231 (.1932) (.1893) (.1903) -0.0272 0.0225 0.0246 (.1106) (.0982) (.0998) 0.1835 0.4746 *** 0.4759 (.1825) (.1726) (.1727) -0.3098 0.0240 0.0201 (.1837) (.1536) (.1533) -0.6129 *** -0.4015 ** -0.3921 (.2212) (.1953) (.1982) 0.1077 0.1637 0.1666 (.1307) (.1217) (.1222) -0.1180 -0.0252 -0.0254 (.1730) (.1487) (.1489) First-Stage F-Test of Instrument 65.31 56.67 50.58 26.99 # of obs 482 395 395 R-Squared 0.1670 0.1529 0.1535 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: All regressions have fewer observations than Part (9) above due to lack of data on the IV. Part (10)(a) has more observations than (10)(b) and (10)(c) due to lack of data on the S variable.
92 The results of the 2SLS first difference analyses with specialty controls are similar to those without specialty controls reported earlier in Table 28. None of the estimates of the impact of changes in immigration on changes in ave rage area wages are statistically significant. The positive effect of immigration that remained even in the presence of the specialty controls becomes insignificant once the instrument is included to control for endogenous location choice. These results provide further evidence that had the immigrant physician not chosen to move to the highest paying area, there would have been no significant short run effect on wages at all.
93 11 : Conclusions of Study based on Regression Analyses This stu dy of immigrant and native physicians has been able to provide evidence that in the short run, the wage impacts of physician immigration tend to be positive, because physician wages are sticky, because immigrants tend not to be the lowest paid physicians in an area and because they tend to settle in areas where wages are higher, on average. The short run positive effects of immigration on wages are somewhat, but not completely, explained by specialty choice, with immigrant physicians tending to specialize in somewhat higher paying fields. In the long run, however, immigration tends to have a negative impact on area wages. This study has been able to provide evidence that immigrant physicians are drawn to areas where wages tend to be higher. There is al so evidence that the negative effect of immigration on wages is mitigated by outward migration of lower paid local physicians in response to the lower wages caused by the influx of foreign physicians. The long run negative effects of immigration are not e xplained by specialty choice. Therefore, immigrants seem to be willing to accept a lower than average wage. The negative effect of immigration on average wages tend s to be relatively small, with a one percentage point increase in the share of foreign bor n physicians having a less than 0.2 effect on average area wages by all measures. In addition, analyzing the wage impacts of physician immigration reveals that foreign educated immigrant physicians tend to have a more positive short run effect on wages and a less negative long run effect on wages, while the opposite is true for foreign
94 born, U.S. educated physicians. Because foreign born, U.S. educated immigrants are more likely to have lived in the United States longer and less likely to have to overc ome difficult immigration hurdles, such as visa requirements and migration costs associated with moving an already established family to a new country, likely at an older age, they are less likely to be favorably selected. Simply analyzing the data in t erms of descriptive statistics also reveals that immigrant physicians are more likely than their native counterparts to work in larger cities as opposed to rural areas. Younger doctors are even more likely to live in larger MSAs, while older physicians ar e more likely than younger physicians to practice in rural areas. The results of all regression analyses performed in this study show that male physicians are likely to earn m ore than female physicians, although the difference becomes minimal after contr olling for specialty choice. The number of women entering the medical profes sion has been increasing every year. The regressions also show that physician wages increase with age but at a decreasing rate in agreement with labor economic theory. T he analy sis consistently predicts that physicians who practice in areas where a larger percentage of patients are covered by Medicare Part B are likely to earn higher wages This is likely due to area specific factors such as age of population and disability stat us that are likely to increase demand. This result could also be explained by a higher probability that physicians will actually be paid when patients are covered by Medicare compared with self pay and uninsured patients.
95 The conclusions about the impac t of immigration on wages drawn from this study are consistent with many of the earlier studies on the topic. In fact, the majority of research in this area has concluded that immigration tends to have a very small negative effect on wages. Even studies that specifically address the issue of high skilled immigration have, in general, found a negative impact of immigration on wages. Borjas (2003) concluded that immigration has a negative effect on wages of college graduates. Borjas (2005) and Borjas (200 6) found a small negative impact of immigration of foreign students on earnings of doctorates in the science and engineering fields; the magnitude of the impact found in his studies is somewhat larger than that found in this analysis. Earlier studies have also found evidence of outward migration of natives in response to immigration. Specifically, Butcher and Card (1991) concluded that the potential impacts of immigration are diminished due to outflow migration of natives while Frey (1994) found that duri ng the 1980s, large numbers of native white s moved away from immigrant heavy destinations such as Los Angeles and New York. The results of the first difference approach are qualitatively in agreement with those of Borjas, Freeman and Katz (1996), who foun d a small positive (almost zero) impact of changes in immigration on wages at the MSA level. Card (2001 ) employed an instrumental variable approach similar to that used in this study and found a systematically negative impact of immigration on wages. H is study however, focused primarily on low skilled workers. As is the case in this study, Orrenius and Zavodny (2007) found a systematically more negative effect of immigration on wages when employing the two stage least squares method as opposed to the ordinary least squares analysis.
96 Cha pter 12 : Healthcare and Immigration Policy Implications The results of this analysis have different policy implications depending on the goals of policy makers. If the goal of policy makers is to address an allege d shortage of physicians, this can be achieved by either increasing the number of U.S. medical school graduates by subsidizing the cost of constructing new medical school as well as the cost of atte ndance for prospective students or by encouraging immigration of foreign born physicians through less restrictive visa requirements and supporting programs that make the immigration process easier for foreign born physicians by reducing costs and easing th e adjustment process. This study has shown that immigrant physicians tend to lower the average wage of physicians; this may or may not be in agreement with policymakers objectives. To the extent that lower physician wages trickle down to consumers in th e form of lower charges for physician services, this could be beneficial to society. However, native physicians will likely suffer a reduction in their wage in the long run due not only to the increase in supply caused by immigration but because the immig rants are will ing to work for lower wages, on average, than their native born counterparts. The results of this study reveal that the negative impacts of immigration are actually largest for foreign born, U.S. educated physicians. Therefore, if reducing physician wages, and potentially the cost of physician care, is the goal of policymakers then this is more likely to be accomplished by implementing programs that encourage foreign born students to attend U.S. medical schools. However, if the goal is to i ncrease the supply of physicians while maintaining the current physician wages, it would be advisable to instead
97 encourage immigration of foreign physicians who already have their degrees by relaxing current visa restrictions. If the goal of policy maker s is specifically to increase the supply of primary care physicians, this study has shown that foreign educated physicians and women are more likely to specialize in general practice than native born male physicians. Thus, policy makers may wish to implem ent programs to encourage U.S. women to enter the medical profession and/or make it easier for foreign medical graduates to practice in the U.S. As previously discussed, the impact of immigration of foreign medical graduates whether born in the U.S. or a foreign country, on local area wages is negative; again, this may or may not be in agreement with policy makers objectives. In addition to the potential reduction in the cost of physician services, a negative wage impact, theoretically, causes an immigra tion surplus by increasing area employment. As to increasing the number of physicians, whether primary care or specialist, in the Health Professional Shortage Areas and Medically Underserved Areas, there is not enough evidence at this point to say whethe r this can be achieved through immigration; more information is needed for further study. This study has shown that immigrant physicians are less likely than native physicians to work in rural areas, which are traditionally thought of as being medically u nderserved. However, there are many areas in MSAs, for example inner city neighborhoods, that are actually classified as HPSAs and MUAs. Immigrants tend to locate in areas where wages are higher, on average. It seems reasonable to assume that wages in H PSAs and MUAs are lower than in more desirable locales. However, the study is unable to determine whether wages are higher or lower, on average, in these areas or whether immigrant physicians who practice in large
98 MSAs are practicing in the Health Profess ional Shortage Areas or Medically Underserved Areas. The AMA Physician Masterfile contains data on physician city, county, state, and even zip code. The Health Professional Shortage Areas and Medically Underserved Areas, however, are defined by census tr act. There is no readily available database, to this researchers knowledge, that allows census tracts to be cross referenced with zip code such that physician s who are practicing in an underserved area can be identified. This task could likely be accomp lished using geocoding data that maps zip codes and census tracts. Such a study would prove extremely useful and is an important area for future research. Finally, it is important to note that the wage impacts of any of any policy that affects the supply of physicians through immigration will not immediately take effect. The results of the first difference analyses show that an increase in immigration of foreign physicians is likely to have a small positive impact on average physician wages in the short r un ; however, as wages adjust to the supply shock over time, the effect on wages becomes negative.
99 References Cited Altonji, Joseph and David Card. 1991. The Effects of Immigration on the Labor Market O utcomes of Less skilled Natives, in Immigration, Trade and the Labor Market Abowd and Freeman eds Chicago: The University of Chicago Press. A rvantes, James. March 22, 2007. Primary Care Physician Shortage Creates Medically Disenfranchised Population. AAFP News Now Washington, D.C. Aydemir, Abdurrahman and George J. Borjas. November 2006. Attenuation Bias in Measuring the Wage Impact of Im migration. Working Paper. Bartel, Ann. 1989. Where Do the New U.S. Immigrants Live? Journal of Labor Economics 7 Borjas, George J., Richard B. Freeman, and Lawrence F. Katz. 1992. Labor Market Effects of Immigration and Trade in Immigration and the Workforce Borjas and Freeman eds Chicago and London: The University of Chicago Press. Borjas, George J. 1995. "The Economic Benefits from Immigration." The Journal of Economic Perspectives 9: 2 Borjas, George J., Richard B. Freeman, and Lawrence F. Katz. 1996. "Searching for the Effect of Immigration on the Labor Market." The American Economic Review 86:2. Borjas, George J. 2003. "The Labor Demand Curve is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market." The Quarte rly Journal of Economics 118:4. Borjas, George J. 2005. The Labor Market Impact of High Skill Immigration. The American Economic Review 95: 2. Borjas, George J. 2006. Immigration in High Skill Labor Markets: The Impact of Foreign Students on the Earn ings of Doctorates. NBER Working Paper 12085 Buerhaus, P.I. and D.O. Staiger. 1996. Managed Care and the Nurse Workforce. The Jo urnal of the American Medical Association 276: 18. Butcher, Kristin F. and David Card. 1991. "Immigration and Wages: Evide nce from the 1980's." The American Economic Review 81:2.
100 Camarota, Steven A. 2007. "Immigrant Employment Gains and Native Losses, 2000 2004 in Debating Immigration Carol M. Swain ed New York: Cambridge University Press Card, David. 2001. Immigran t Inflows, Native Outflows, and the local Market Impacts of Higher Immigration. Journal of Labor Economics 19:1. Castles, Stephen and Mark J. Miller. 1998. The Age of Migration: International Population Movements in the Modern World Second Edition Lo ndon: Macmillan Press Ltd Chiswick, Barry R. 1999. Are Immigrants Favorably Self Selected? The American Economic Review 89:2. Chiswick, Barry R. 2005. "High Skilled Immigration in the International Arena." Forschungsinstitut zur Zukunft der Arbeit/ Institute for the Study of Labor Discussion Paper Series IZA DP No 1782 Clark, Paul F., James B. Stewart and Darlene A. Clark. 2006. The globalization of the Labour Market for Health Care Professionals. International Labour Review 145:1 2. Country Composition of WEO Groups. April 2009 International Monetary Fund, World Economic and Financial Surveys, World Economic Outlook Database WEO Groups and Aggregates Information. Cross, Margaretann. June 2007. What the Primary Care Physician Shortage Mea ns for Health Plans. Managed Care Magazine Docquier, Frederic and Abdeslam Marfouk. 2005. "International Migration by Education Attainment, 1990 2000, in International Migration, Remittances, and the Brain Drain M. Schiff and C. Ozden eds World Ban k Ester, Karma A. 2004. Immigration: Foreign Physicians and the J 1 Visa Waiver Program. CRS Report for Congress, Congressional Research Service, The Library of Congress. Frey, William H. 1994. The New White Flight. American Demographics 16 Frie dberg, Rachel and Jennifer Hunt. 1995. The Impact of Immigrants on Host Country Wages, Employment and Growth. The Journal of Economic Perspectives 9: 2. Gardner, Robert W. and Leon F. Bouvier. 1990. "The United States in Handbook on International Mi gration William J. Serow, et al eds New York: Greenwood Press.
101 Hadley, Jack and Jean M. Mitchell. 1999. HMO Penetration and Physicians Earnings. Medical Care 37:11. Hamermesh, Daniel S. 1993. Labor Demand Princeton, NJ: Princeton University Press. Held, David, Anthony McGrew, David Goldblatt, and Jonathan Perraton. 1999. Global Transformations: Politics, Economics and Culture Stanford, CA: Stanford University Press. Johnson, George E. 1980. "The Labor Market Effects of Immigration." Industrial and Labor Relations Review 33: 3. Klasko, H. Ronald. 1999. Foreign Physicians Face Tough Immigration Hurdles. www.klaskolaw.com Madhavan, M. C. 1985. "Indian Emigrants: Numbers, Characteristics and Economic Impact." Population and Development Review 11: 3 Massey, Douglas S. and J. Edward Taylor, eds. 2004. International Migration: Prospects and Policies in a Global Marke t New York: Oxford University Press. Mautino, Kathrin. 2002. Physicians and Immigration. Journal of Immigrant Health 4:4. Moor e, Gordon and Jonathan Showstack. 2003. Primary Care Medicine in Crisis: Toward Reconstruction and Renewal. Annals of Internal Medicine 138: 3. Orrenius, Pia M. and Madeline Zavodny. 2007. Does Immigration Affect Wages? A Look at Occupation Level Evi dence. Labour Economics 14 Ottaviano, Gianmarco I.P. and Giovanni Peri. September 2005. Rethinking the Gains from Immigration: Theory and Evidence from the U.S. National Bureau of Economic Research Working Paper 11672 Pear, Robert. Shortage of D octors an Obstacle to Obama Goals. April 27, 2009 The New York Times S0501: Selected Characteristics of the Native and Foreign Born Populations. 2005 07 American Community Survey, U.S. Department of Commerce, Bureau of the Census. Simon, Carol J ., David Dranove, and William D. White. 1997. The Impact of Managed Care on the Physician Marketplace . Public Health Reports 112. Statistical Abstract of the United States 2009 U.S. Department of Commerce, Bureau of the Census.
102 Svorny, Shirley. 199 1. Consumer Gains from Physician Immigration to the US: 1966 1971. Applied Economics 23. Table 5: Estimates of the Components of Resident Population Change for the United States, Regions, States, and Puerto Rico: July 1, 2007 to July 1, 2008. Pop ula tion Division, U.S. Department of Commerce, Bureau of the Census. Topel, Robert H. 1994. "Regional Labor Markets and the Determinants of Wage Inequality." The American Economic Review 84: 2 Vujicic, Marko, Pascal Zurn, Khassoum Diallo, Orvill Adams and Mario R. Dal Poz. 2004. The Role of Wages in the Migration of Health Care Professionals from Developing Countries. Human Resources for Health 2:3. Waldinger, Roger and Greta Gilbertson. 1994. "Immigrants' Progress: Ethnic and Gender Differences among U.S. Immigrants in the 1980s." Sociological Perspectives 37: 3 Weiss, Yorum. 2000. High Skill Immigration: Some Lessons from Israel. Swedish Economic Policy Review 7:2
103 Bibliography Cameron, A. Colin and Pravin K. Trivedi. 2005. Microeconometrics: Methods and Applications New York: Cambridge University Press. Folland, Sherman, Allen C. Goodman and Miron Stano. 2007. The Economics of Health and Health Care, 5 th E dition. New Jersey: Pearson Prentice Hall. Greene, William H. 2003. Econometric Analysis, 5 th Edition. New Jersey: Pearson Prentice Hall. Wooldridge, Jeffrey M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, Massachusetts: The M IT Press. Wooldridge, Jeffrey M. 2006. Introductory Econometrics: A Modern Approach, 3 rd Edition. Ohio: Thomson South Western.
105 Table A1 Physicians by Country of Birth BIRTH COUNTRY FREQUENCY OF FOREIGN BORN PHYSICIANS UNITED STATES 126,355 INDIA 8,352 18.8 PHILIPPINES 3,905 8.8 CANADA 2,105 4.7 SOUTH KOREA 1,703 3.8 PAKISTAN 1,523 3.4 IRAN 1,486 3.3 CHINA 1,343 3.0 CUBA 1,252 2.8 GERMANY 1,139 2.6 UNITED KINGDOM 1,123 2.5 VIETNAM 999 2.2 TAIWAN 908 2.0 EGYPT 823 1.8 MEXICO 639 1.4 SYRIA 601 1.4 POLAND 593 1.3 COLOMB IA 537 1.2 ISRAEL 516 1.2 ARGENTINA 484 1.1 NIGERIA 473 1.1 HONG KONG 471 1.1 JAPAN 459 1.0 SOUTH AFRICA 449 1.0 ROMANIA 438 1.0 ITALY 430 1.0 LEBANON 416 0.9 PERU 403 0.9 JAMAICA 376 0.8 HAITI 373 0.8 TH AILAND 360 0.8 DOMINICAN REPUBLIC 354 0.8 RUSSIA 329 0.7 GREECE 323 0.7
106 Table A1 (Continued) Physicians by Country of Birth BIRTH COUNTRY FREQUENCY OF FOREIGN BORN PHYSICIANS SPAIN 293 0.7 UKRAINE 285 0.6 TURKEY 281 0.6 IRELAND 278 0.6 HUNGARY 254 0.6 BRAZIL 241 0.5 BURMA 232 0.5 FRANCE 232 0.5 SRI LANKA 214 0.5 USSR 204 0.5 LATVIA 202 0.5 BANGLADESH 192 0.4 JORDAN 191 0.4 IRAQ 182 0.4 AUSTRALIA 180 0.4 VENEZUELA 164 0.4 GHANA 163 0.4 TRINIDAD AND TOBAGO 160 0.4 CZECH REPUBLIC 156 0.4 KENYA 154 0.3 PANAMA 147 0.3 NETHERLANDS 146 0.3 GUYANA 143 0.3 CHILE 142 0.3 AUSTRIA 137 0.3 ETHIOPIA 133 0.3 MALAYSIA 131 0.3 ECUADOR 119 0.3 SWITZERLAND 119 0.3 INDONESIA 117 0.3 NICARAGUA 114 0.3 BOLIVIA 109 0.2 GUATEMALA 103 0.2 SWEDEN 100 0.2 BELGIUM 99 0.2
107 Table A1 (Continued) Physicians by Country of Birth BIRTH COUNTRY FREQUENCY OF FOREIGN BORN PHY SICIANS BULGARIA 90 0.2 EL SALVADOR 83 0.2 UGANDA 82 0.2 SINGAPORE 79 0.2 TANZANIA 75 0.2 AFGHANISTAN 65 0.1 NEW ZEALAND 63 0.1 YUGOSLAVIA 61 0.1 SERBIA 60 0.1 CROATIA 57 0.1 HONDURAS 51 0.1 DENMARK 50 0.1 PORTUGAL 49 0.1 LITHUANIA 48 0.1 COSTA RICA 47 0.1 BARBADOS 44 0.1 BELARUS 43 0.1 URUGUAY 42 0.1 PARAGUAY 41 0.1 KUWAIT 39 0.1 UZBEKISTAN 38 0.1 NORWAY 36 0.1 SAUDI ARABIA 36 0.1 ZIMBABWE 36 0.1 CYPRUS 35 0. 1 FINLAND 35 0.1 ZAMBIA 34 0.1 MOROCCO 33 0.1 CAMBODIA 31 0.1 BAHAMAS 30 0.1 BOSNIA AND HERZEGOVINA 30 0.1 NEPAL 30 0.1 ARMENIA 28 0.1 SUDAN 27 0.1 CZECHOSKLOVAKIA 25 0.1
108 Table A1 (Continued) Physicians by Cou ntry of Birth BIRTH COUNTRY FREQUENCY OF FOREIGN BORN PHYSICIANS GRENADA 25 0.1 ICELAND 25 0.1 CAMEROON 24 0.1 LIBYA 24 0.1 COLUMBIA 23 0.1 LIBERIA 22 0.0 MOLDOVA 22 0.0 TUNISIA 20 0.0 CZECHOSLOVAKIA 19 0.0 GUAM 19 0.0 AZERBAIJAN 18 0.0 BERMUDA 18 0.0 SLOVAKIA 18 0.0 ESTONIA 17 0.0 ZAIRE 17 0.0 ALGERIA 16 0.0 FIJI 15 0.0 GEORGIA 15 0.0 LAOS 15 0.0 SIERRA LEONE 15 0.0 MACEDONIA 14 0.0 ST KITT 14 0.0 UNITED ARAB EM IRATES 13 0.0 MALAWI 12 0.0 ALBANIA 11 0.0 BELIZE 11 0.0 MALTA 11 0.0 YUGO 11 0.0 ARUBA 10 0.0 MAURITIUS 9 0.0 SLOVENIA 9 0.0 SOMALIA 9 0.0 ANTIGUA AND BARBUDA 9 0.0 BAHRAIN 8 0.0 ST LUCIA 8 0.0
109 Table A1 (Continued) Physicians by Country of Birth BIRTH COUNTRY FREQUENCY OF FOREIGN BORN PHYSICIANS SURINAME 8 0.0 BRUNEI 7 0.0 KAZAKHSTAN 7 0.0 KYRGYZSTAN 6 0.0 NORTH KOREA 6 0.0 MOZAMBIQUE 5 0.0 VINCENT AND THE GRENADINES 5 0 .0 CAPE VERDE 4 0.0 CONGO 4 0.0 MACAU 4 0.0 NETHERLANDS ANTILLES 4 0.0 NIGER 4 0.0 RWANDA 4 0.0 INDIES 3 0.0 LUXEMBOURG 3 0.0 MARSHALL ISLANDS 3 0.0 QATAR 3 0.0 SENEGAL 3 0.0 YEMEN 3 0.0 ANGOLA 2 0.0 BURKINA FASO 2 0.0 GUINEA 2 0.0 MONTSERRAT 2 0.0 OMAN 2 0.0 REPUBLIC OF KIRIBATI 2 0.0 SAMOA 2 0.0 ZANZIBAR 2 0.0 ANGUILLA 1 0.0 BOTSWANA 1 0.0 CO TE D'IVOIRE 1 0.0 DJIBOUTI 1 0.0 EAST TIMOR 1 0.0 GABON 1 0.0 LESOTHO 1 0.0
110 Table A1 (Continued) Physicians by Country of Birth BIRTH COUNTRY FREQUENCY OF FOREIGN BORN PHYSICIANS MADAGASCAR 1 0.0 MALDIVES 1 0.0 MONTENEGRO 1 0.0 NAMIBIA 1 0.0 PALAU 1 0.0 PAPUA NEW GUINEA 1 0. 0 REUNION 1 0.0 SWAZILAND 1 0.0 TAJIKISTAN 1 0.0 TOGO 1 0.0 TONGA 1 0.0 TURKS AND CAICOS 1 0.0 OTHER FOREIGN 381 0.8 TOTAL 170,858
111 Table A2 Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BOR N PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS New York, NY 3,669 8.2 Los Angeles/Long Beach, CA 2,380 5.3 Chicago Naperville Joliet, IL Metropolitan Division 2,139 4.8 Boston Cambridge Quincy, MA NECTA Division 1,317 3.0 Washington, DC -MD -VA M SA 1,315 3.0 Detroit Livonia Dearborn, MI Metropolitan Division 1,064 2.4 Nassau Suffolk, NY MSA 1,056 2.4 Philadelphia, PA NJ PMSA 1,047 2.4 Miami, FL 985 2.2 Houston, TX 941 2.1 Orange County, CA 777 1.7 Baltimore, MD MSA 679 1.5 Clevelan d Elyria Mentor, OH 641 1.4 Newark, NJ 512 1.2 St. Louis, MO -IL MSA 511 1.1 Riverside San Bernardino, CA PMSA 506 1.1 Bergen Passaic, NJ MSA 505 1.1 Atlanta, GA MSA 504 1.1 San Francisco, CA PMSA 490 1.1 Dallas Plano Irving, TX Metropolitan Division 483 1.1 Pittsburgh, PA MSA 480 1.1 San Jose, CA PMSA 465 1.0 Tampa -St. Petersburg -Clearwater, FL MSA 441 1.0 San Diego, CA MSA 404 0.9 Phoenix, AZ MSA 401 0.9 Oakland, CA PMSA 398 0.9 Seattle Bellevue Everett, WA PMSA 367 0.8 Mid dlesex Somerset Hunterdon, NJ MSA 344 0.8 Fort Lauderdale Pompano Beach Deerfield Beach, FL 336 0.8 Buffalo Niagara Falls, NY 278 0.6 Orlando, FL MSA 274 0.6 Minneapolis -St. Paul, MN WI MSA 273 0.6 Raleigh Durham Chapel Hill, NC 260 0.6 Milwau kee/Waukesha, WI 256 0.6 Sacramento, CA MSA 252 0.6 Hartford, CT 251 0.6 Monmouth Ocean, NJ MSA 248 0.6 Jacksonville, FL MSA 237 0.5 San Antonio, TX MSA 235 0.5 Cincinnati, OH KY IN PMSA 226 0.5 New Orleans, LA MSA 225 0.5 Las Vegas, NV MSA 217 0.5
112 Table A2 (Continued) Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS Indianapolis, IN MSA 216 0.5 West Palm Beach -Boca Raton -Delray Beach, FL MSA 213 0.5 Portland Vancouv er, OR WA PMSA 211 0.5 Norfolk -Virginia Beach -Newport News, VA MSA 199 0.4 Columbus, OH MSA 185 0.4 Rochester, NY MSA 182 0.4 San Juan Bayamon, PR MSA 179 0.4 Ann Arbor, MI 178 0.4 Providence Fall River Warwick, RI MA MSA 177 0.4 Kansas Cit y, MO -KS MSA 174 0.4 Denver Aurora, CO 172 0.4 Bridgeport, CT 171 0.4 New Haven -Meriden, CT MSA 171 0.4 Dayton -Springfield, OH MSA 169 0.4 Richmond -Petersburg, VA MSA 167 0.4 Honolulu, HI MSA 166 0.4 Rochester, MN 166 0.4 Nashville, TN MSA 162 0.4 Louisville, KY -IN MSA 157 0.4 Albany -Schenectady -Troy, NY MSA 155 0.3 Syracuse, NY MSA 152 0.3 Fresno, CA MSA 147 0.3 Jersey City, NJ MSA 143 0.3 Oklahoma City, OK MSA 143 0.3 Fort Worth Arlington, TX Metropolitan Division 135 0.3 Memphis, TN -AR -MS MSA 135 0.3 Birmingham, AL MSA 134 0.3 Gary, IN 122 0.3 Tucson, AZ MSA 122 0.3 Austin, TX MSA 115 0.3 Toledo, OH MSA 111 0.2 Wilmington Newark, DE MD 105 0.2 Trenton, NJ MSA 104 0.2 Youngstown -Warren, OH MSA 104 0 .2 Charlotte -Gastonia -Rock Hill, NC -SC MSA 102 0.2 Flint, MI MSA 99 0.2 Bakersfield, CA MSA 97 0.2 Albuquerque, NM MSA 95 0.2 Harrisburg -Lebanon -Carlisle, PA MSA 93 0.2 Akron, OH 90 0.2 Salt Lake City -Ogden, UT MSA 88 0.2 Greensboro -Winston Salem -High Point, NC MSA 85 0.2
113 Table A2 (Continued) Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS Lexington Fayette, KY MSA 85 0.2 El Paso, TX MSA 84 0.2 Sarasota, FL MS A 84 0.2 Allentown -Bethlehem -Easton, PA -NJ MSA 83 0.2 Mcallen -Edinburg -Mission, TX MSA 82 0.2 Omaha, NE -IA MSA 81 0.2 Ventura, CA MSA 81 0.2 Iowa City, IA MSA 80 0.2 Gainesville, FL MSA 79 0.2 Galveston/Texas City, TX 78 0.2 Scranton Wilkes Barre, PA MSA 75 0.2 Worcester, MA MSA 75 0.2 Tacoma, WA PMSA 73 0.2 Little Rock -North Little Rock, AR MSA 71 0.2 Newburgh, NY/PA 71 0.2 Dutchess County, NY MSA 69 0.2 Grand Rapids, MI MSA 68 0.2 Stockton, CA MSA 68 0.2 Augusta, GA -SC MSA 67 0.2 Lansing -East Lansing, MI MSA 67 0.2 Melbourne -Titusville -Palm Bay, FL MSA 67 0.2 Saginaw -Bay City -Midland, MI MSA 66 0.1 Shreveport, LA MSA 65 0.1 Charlottesville, VA MSA 64 0.1 Corpus Christi, TX MSA 64 0.1 Charleston, S C MSA 63 0.1 Daytona Beach, FL MSA 63 0.1 Madison, WI MSA 63 0.1 Tulsa, OK MSA 61 0.1 Wichita, KS MSA 60 0.1 Vallejo Fairfield Napa, CA MSA 59 0.1 Beaumont -Port Arthur, TX MSA 58 0.1 Columbia, MO MSA 58 0.1 Lakeland -Winter Haven, FL MSA 5 7 0.1 Binghamton, NY MSA 56 0.1 Jackson, MS MSA 56 0.1 Fort Myers -Cape Coral, FL MSA 52 0.1 Lubbock, TX MSA 52 0.1 Santa Barbara -Santa Maria -Lompoc, CA MSA 52 0.1 Springfield, IL MSA 52 0.1 Atlantic City, NJ MSA 51 0.1 Canton, OH MSA 50 0.1
114 Table A2 (Continued) Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS Charleston, WV MSA 50 0.1 Utica -Rome, NY MSA 49 0.1 Knoxville, TN MSA 48 0.1 Peoria, IL MSA 48 0.1 Mode sto, CA MSA 47 0.1 Tallahassee, FL MSA 47 0.1 Kalamazoo, MI MSA 46 0.1 Greenville -Spartanburg, SC MSA 45 0.1 Huntington -Ashland, WV -KY -OH MSA 45 0.1 Columbia, SC MSA 44 0.1 Johnson City -Kingsport -Bristol, TN -VA MSA 44 0.1 Punta Gorda, FL MSA 44 0.1 Mobile, AL MSA 43 0.1 Naples, FL MSA 42 0.1 Odessa, TX 42 0.1 Brownsville -Harlingen, TX MSA 41 0.1 Champaign Urbana, IL 41 0.1 Fayetteville, NC MSA 41 0.1 Baton Rouge, LA MSA 40 0.1 Chattanooga, TN -GA MSA 40 0.1 Evansville IN -KY MSA 39 0.1 Fort Pierce, FL MSA 39 0.1 Fort Wayne, IN MSA 39 0.1 Greenville, NC MSA 38 0.1 Rockford, IL MSA 38 0.1 Macon -Warner Robins, GA MSA 37 0.1 Santa Rosa, CA MSA 37 0.1 Erie, PA MSA 36 0.1 Brazoria, TX MSA 35 0.1 Davenport -Rock Island -Moline, IA -IL MSA 35 0.1 Lowell, MA NH MSA 35 0.1 Reading, PA MSA 35 0.1 Ocala, FL MSA 34 0.1 Spokane, WA MSA 34 0.1 Killeen -Temple, TX MSA 33 0.1 Pensacola, FL MSA 33 0.1 Johnstown, PA 32 0.1 Portland, ME MSA 32 0.1 Amar illo, TX MSA 31 0.1 Appleton -Oshkosh -Neenah, WI MSA 31 0.1 Des Moines, IA MSA 31 0.1 Montgomery, AL MSA 31 0.1
115 Table A2 (Continued) Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIAN S Roanoke, VA MSA 31 0.1 Stamford Norwalk, CT MSA 31 0.1 Visalia -Tulare -Porterville, CA MSA 31 0.1 South Bend -Mishawaka, IN MSA 30 0.1 Columbus, GA -AL MSA 29 0.1 Wheeling, WV -OH MSA 29 0.1 Reno, NV MSA 28 0.1 San Luis Obispo Atascadero Paso Robles, CA MSA 28 0.1 Savannah, GA MSA 28 0.1 Yolo, CA MSA 28 0.1 Huntsville, AL MSA 27 0.1 Burlington, VT MSA 26 0.1 Monroe, LA MSA 26 0.1 Ponce, PR MSA 25 0.1 Salinas -Seaside -Monterey, CA MSA 24 0.1 Topeka, KS MSA 24 0.1 Tyler, T X MSA 24 0.1 Wichita Falls, TX MSA 24 0.1 York, PA MSA 24 0.1 Alexandria, LA MSA 22 0.0 Altoona, PA MSA 22 0.0 Chico, CA MSA 22 0.0 Springfield, MO MSA 22 0.0 Springfield, MA MSA 22 0.0 Terre Haute, IN MSA 22 0.0 Bangor, ME MSA 21 0.0 C larksville -Hopkinsville, TN -KY MSA 21 0.0 Hagerstown, MD MSA 21 0.0 Lancaster, PA MSA 21 0.0 Olympia, WA MSA 21 0.0 Fargo -Moorhead, ND -MN MSA 20 0.0 Lafayette, LA MSA 20 0.0 Lima, OH MSA 20 0.0 Mayaguez, PR 20 0.0 Merced, CA MSA 20 0.0 Wilmington, NC MSA 20 0.0 Colorado Springs, CO MSA 19 0.0 Laredo, TX MSA 19 0.0 Mansfield, OH MSA 19 0.0 Parkersburg -Marietta, WV -OH MSA 19 0.0 Sioux Falls, SD MSA 19 0.0 Vineland Millville Bridgeton, NJ MSA 19 0.0
116 Table A2 (Continued) F oreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS Albany, GA MSA 18 0.0 Biloxi -Gulfport, MS MSA 18 0.0 Boise City, ID MSA 18 0.0 Lawton, OK 18 0.0 Redding, CA MSA 18 0.0 Richland K ennewick Pasco, WA MSA 18 0.0 Tuscaloosa, AL MSA 18 0.0 Victoria, TX MSA 18 0.0 Waco, TX MSA 18 0.0 Yuma, AZ MSA 18 0.0 Cumberland, MD -WV MSA 17 0.0 Hamilton/Middletown, OH 17 0.0 Kankakee, IL MSA 17 0.0 Portsmouth Rochester, NH ME PMSA 17 0.0 Salem, OR MSA 17 0.0 Dover, DE MSA 16 0.0 Eugene -Springfield, OR MSA 16 0.0 Lafayette -West Lafayette, IN MSA 16 0.0 Panama City, FL MSA 16 0.0 Pittsfield, MA MSA 16 0.0 Santa Cruz Watsonville, CA MSA 16 0.0 Anchorage, AK MSA 15 0.0 Bismarck, ND MSA 15 0.0 Bloomington -Normal, IL MSA 15 0.0 Boulder/Longmont, CO 15 0.0 Fort Smith, AR -OK MSA 15 0.0 Racine, WI PMSA 15 0.0 Sharon, PA 15 0.0 Barnstable Yarmouth, MA MSA 14 0.0 Bellingham, WA MSA 14 0.0 Grand Forks, ND/MN 1 4 0.0 Green Bay, WI MSA 14 0.0 Hickory -Morganton, NC MSA 14 0.0 Lake Charles, LA MSA 14 0.0 Steubenville -Weirton, OH -WV MSA 14 0.0 Yuba City, CA MSA 14 0.0 Arecibo, PR MSA 13 0.0 Danville, VA MSA 13 0.0 Duluth, MN -WI MSA 13 0.0 Eau Cl aire, WI MSA 13 0.0 Elmira, NY 13 0.0 Jackson, MI 13 0.0
117 Table A2 (Continued) Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS Janesville -Beloit, WI MSA 13 0.0 Lincoln, NE MSA 13 0.0 Waterloo -Cedar Falls, IA MSA 13 0.0 Abilene, TX MSA 12 0.0 Benton Harbor, MI MSA 12 0.0 Caguas, PR MSA 12 0.0 Fort Walton Beach, FL MSA 12 0.0 Jackson, TN MSA 12 0.0 Asheville, NC MSA 11 0.0 Athens, GA MSA 11 0.0 Bremerton, WA PMSA 1 1 0.0 Bryan -College Station, TX MSA 11 0.0 Gadsden, AL 11 0.0 Goldsboro, NC MSA 11 0.0 Hattiesburg, MS 11 0.0 Kenosha, WI PMSA 11 0.0 Auburn Opelika, AL MSA 10 0.0 Billings, MT MSA 10 0.0 Cedar Rapids, IA MSA 10 0.0 Florence, SC MSA 10 0 .0 Fort Collins -Loveland, CO MSA 10 0.0 Glens Falls, NY 10 0.0 Pueblo, CO MSA 10 0.0 Santa Fe, NM MSA 10 0.0 Yakima, WA MSA 10 0.0 Dothan, AL MSA 9 0.0 Jacksonville, NC MSA 9 0.0 Kokomo, IN MSA 9 0.0 La Crosse, WI/MN 9 0.0 Lewiston -Au burn, ME MSA 9 0.0 Manchester, NH MSA 9 0.0 Muncie, IN MSA 9 0.0 New London -Norwich, CT -RI MSA 9 0.0 St. Joseph, MO MSA 9 0.0 Sherman Denison, TX MSA 9 0.0 Sioux City, IA -NE MSA 9 0.0 Elkhart -Goshen, IN MSA 8 0.0 Jamestown, NY 8 0.0 L as Cruces, NM MSA 8 0.0 Myrtle Beach, SC MSA 8 0.0 Owensboro, KY MSA 8 0.0 Rocky Mount, NC MSA 8 0.0
118 Table A2 (Continued) Foreign Born Physicians by MSA MSA DESCRIPTION # OF FOREIGN BORN PHYSICIANS OF TOTAL FOREIGN BORN PHYSICIANS Williamsp ort, PA MSA 8 0.0 Anniston, AL MSA 7 0.0 Dubuque, IA 7 0.0 Florence, AL MSA 7 0.0 Houma -Thibodaux, LA MSA 7 0.0 Missoula, MT 7 0.0 Provo -Orem, UT MSA 7 0.0 Decatur, IL 6 0.0 Fayetteville -Springdale, AR MSA 6 0.0 Greeley, CO MSA 6 0.0 Longview -Marshall, TX MSA 6 0.0 Lynchburg, VA MSA 6 0.0 State College, PA MSA 6 0.0 Wausau, WI MSA 6 0.0 Aguadilla, PR MSA 5 0.0 Bloomington, IN MSA 5 0.0 Joplin, MO MSA 5 0.0 Medford, OR MSA 5 0.0 Rapid City, SD MSA 5 0.0 St. Cloud, MN MSA 5 0.0 San Angelo, TX 5 0.0 Sheboygan, WI MSA 5 0.0 Texarkana, TX Texarkana, AR MSA 5 0.0 Corvallis, OR MSA 4 0.0 Flagstaff, AZ/UT 4 0.0 Jonesboro, AR 4 0.0 Cheyenne, WY MSA 3 0.0 Decatur, AL MSA 3 0.0 Pine Bluff, AR MSA 3 0.0 Sumte r, SC MSA 3 0.0 Enid, OK 2 0.0 Great Falls, MT 2 0.0 Casper, WY 1 0.0 Lawrence, KS MSA 1 0.0 Pocatello, ID MSA 1 0.0 Rural/No MSA 3325 7.5
119 Table A3 Sex of Physicians by Place of Birth and Year U.S. Born Physicians 1997 1999 2001 2003 2005 2007 F 4,712 22 4,823 23 5,073 24 5,523 26 5,751 28 5,859 29 M 16,785 78 16,514 77 16,088 76 15,757 74 14,995 72 14,473 71 21,497 21,337 21,161 21,280 20,746 20,332 Foreign Born Physicians 1997 1999 2001 2003 2005 2007 F 1,981 27 2,220 29 2,229 29 2,231 30 2,230 31 2,102 30 M 5,440 73 5,51 4 71 5,429 71 5,290 70 4,951 69 4,886 70 7,421 7,734 7,658 7,521 7,181 6,988 Total (All Physicians) 1997 1999 2001 2003 2005 2007 F 6,693 23 7,043 2 4 7,302 25 7,754 27 7,981 29 8,061 29 M 22,225 77 22,028 76 21,517 75 21,047 73 19,946 71 19,359 71 28,918 29,071 28,819 28,801 27,927 27,320
120 Table A4 2SLS Estimates o f the Relationship between Immigration and Physician Wages: First Stage Regression Results Dependent Variable: I = Immigration Share FB FE Share FB USE Share Independent Variable ( 2 )(a) (2 )(b) (2 )(c) Male 0.0045 *** 0.0042 *** 0.0039 *** 0.0009 *** (.0006) (.0005) (.0004) (.0002) Age 0.0010 *** 0.0008 *** 0.0003 *** 0.0003 *** (.0001) (.0001) (.0001) (.0000) Age 2 0.0000 *** 0.0000 *** 0.0000 *** 0.0000 *** (.0000) (.0000) (.0000) (.0000) MedB 0.5634 *** 0.8965 *** 0.5921 *** 0.1881 *** (.0151) (.0 101) (.0090) (.0042) S = Physician Stock 0.0857 *** 0.1214 *** 0.0010 (.0025) (.0020) (.0006) FB90 1.012 *** 1.0126 *** 0.4970 *** 0.4694 *** (.0039) (.0033) (.0071) (.0033) FE_IV 3.3796 *** 0.5766 *** (. 0248) (.0078) F Test for Instruments 68,737 *** 96,185 *** 29,492 *** 23,569 *** # of obs 159,448 128,074 128,074 128,074 Prob>F 0.00 0.00 0.00 0.00 R Squared .2934 .4800 .5755 .5292 Adj. R Square d .2933 .4799 .5755 .5292 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: All regressions have fewer observations than Part (1) due to lack of data on the FB90 variable. Part (2 )(a) has more observati ons than (2)(b) and (2 )(c) due to lack of data on the physician stock variable.
121 Table A5 2SLS First-Difference Estimates of the Relationship between Immigration and Physician Wages: First Stage Regression Results Dependent Variable: n Share Independent Variable (5)(a) (5)(b) (5)(c) -0.3114 *** -0.3874 *** -0.3296 *** -0.0880 *** (.0750) (.0834) (.0684) (.0316) -0.0438 *** -0.0553 *** -0.0328 ** -0.0176 *** (.0156) (.0177) (.0145) (.0062) 2 0.0007 *** 0.0008 *** 0.0005 *** 0.0002 *** (.0002) (.0002) (.0002) (.0001) 0.3473 *** 0.2736 0.5891 *** -0.4105 *** (.1321) (.1520) (.1265) (.0584) 0.0703 0.1247 ** -0.0344 (.1050) (.0609) (.0520) FB90 0.7365 *** 0.6924 *** -0.2951 *** 0.3472 *** (.0854) (.0894) (.1069) (.0572) FE_IV 2.3406 *** -0.4554 *** (.2860) (.1392) F-Test for Instruments 74.29 *** 59.99 *** 51.52 *** 25.65 *** # of obs 482 395 395 395 Prob>F 0.00 0.00 0.00 0.00 R-Squared 0.4412 0.4415 0.4796 0.3918 Adj. R-Squared 0.4353 0.4329 0.4702 0.3808 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: All regressions have fewer observations than Part (4) due to lack of data on the FB90 variable. Part (5)(a) has more observations than (5)(b) and (5)(c) due to lack of data on the physician stock variable.
122 Table A6 2SLS Estimates of the Relationship between Immigration and Physician Wages with Specialty Cont rols: First Stage Regression Results Dependent Variable: I = Immigration Share FB F E_ Share FBUSE_ Share Independent Variable (7 )(a) (7 )(b) (7 )(c) Male 0.0041 *** 0.0039 *** 0.0038 *** 0.0010 *** (.0006) (.0005) (.0004) (.0002) Age 0.0011 *** 0.0009 *** 0.0003 *** 0.0003 *** (.0001) (.0001) (.0001) (.0000) Age 2 0.0000 *** 0.0000 *** 0.0000 *** 0.0000 *** (.0000) (.0000) (.0000) (.0000) MedB 0.5733 *** 0.8949 *** 0.5907 *** 0.1883 *** (.0150) (.0102) (.0090) (.0042) S = Physician Stock 0.0849 *** 0.1208 *** 0.0009 (.0025) (.0020) (.0006) Anesthesiology 0.0031 ** 0.0032 *** 0.0005 0.0003 (.0012) (.0010) (.0008) (.0004) Family/General 0.0187 *** 0.0074 *** 0.0056 *** 0.0015 *** (.0010 ) (.0007) (.0006) (.0003) OBGYN 0.0025 ** 0.0019 0.0004 0.0005 (.0013) (.0010) (.0008) (.0003) Pediatrics 0.0003 0.0011 0.0003 0.0006 ** (.0010) (.0008) (.0006) (.0003) Psychiatry 0.0024 ** 0.001 7 0.0006 0.0003 (.0011) (.0009) (.0007) (.0003) Surgery 0.0069 *** 0.0052 *** 0.0024 *** 0.0007 *** (.0009) (.0007) (.0005) (.0002) Internal Medicine 0.0060 *** 0.0023 *** 0.0032 *** 0.0000 (.0009) (.0007) (.0005 ) (.0002) FB90 1.0076 *** 1.0109 *** 0.4985 *** 0.4691 *** (.0039) (.0033) (.0071) (.0033) FE_IV 3.3795 *** 0.5766 *** (.0247) (.0078) F Test for Instruments 68,151 *** 95,898 *** 29,240 *** 23,472 *** # of obs 159,448 128,074 128,074 128,074 R Squared 0.2963 0.4807 0.5763 0.5294 Adj. R Squared 0.2962 0.4807 0.5762 0.5293 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: All r egressions have fewer observations than Part (6 ) due to lack of data on the FB90 variable. Part ( 7 )(a ) has more observations than (7)(b) and (7 )(c) due to lack of data on the physician stock variable.
123 Table A7 2SLS First-Difference Estimates of the Relationship between Immigration and Physician Wages with Specialty Controls: First Stage Regression Results Dependent Variable: Independent Variable (10)(a) (10)(b) (10)(c) -0.3791 *** -0.4615 *** -0.3803 *** -0.1093 *** (.0778) (.0863) (.0714) (.0341) -0.0475 *** -0.0638 *** -0.0378 *** -0.0196 *** (.0153) (.0175) (.0145) (.0064) 2 0.0007 *** 0.0009 *** 0.0006 *** 0.0002 *** (.0002) (.0002) (.0002) (.0001) 0.3151 ** 0.3171 ** 0.5863 *** -0.3876 *** (.1362) (.1572) (.1342) (.0536) 0.0371 0.0864 -0.0328 (.1218) (.0757) (.0536) Anesthesiology 0.3782 ** 0.3801 ** 0.1988 0.1401 ** (.1573) (.1784) (.1438) (.0619) Family/General -0.0327 0.1092 -0.0010 0.0674 (.0868) (.0932) (.0737) (.0387) OBGYN -0.0697 0.0515 0.0786 0.0191 (.1479) (.1651) (.1303) (.0642) Pediatrics 0.0088 0.1078 0.0529 -0.0022 (.1225) (.1319) (.1080) (.0578) Psychiatry -0.5499 *** -0.5261 *** -0.4954 *** -0.0333 (.1210) (.1402) (.1148) (.0586) Surgery -.0460 -0.0071 -0.0276 0.0302 (.1045) (.1155) (.0900) (.0442) Internal Medicine 0.2783 ** 0.3362 *** 0.2384 ** 0.0936 ** (.1153) (.1245) (.0992) (.0459) FB90 0.7065 *** 0.6828 *** -0.3124 *** 0.3532 *** (.0874) (.0907) (.1063) (.0572) FE_IV 2.3527 *** -0.4734 *** (.2862) (.1425) F-Test for Instruments 65.31 *** 56.67 *** 50.58 *** 26.99 *** # of obs 482 395 395 395 R-Squared 0.4808 0.4804 0.5170 0.4081 Adj. R-Squared 0.4676 0.4627 0.4993 0.3863 *p<.10, **p<.05, ***p<.01 All coefficients reported with robust standard errors. Note: All regressions have fewer observations than Part (9) due to lack of data on the FB90 variable. Part (10)(a) has more observations than (10)(b) and (10)(c) due to lack of data on the physician stock variable.
About the Author A National Merit Scholar, Finnie B. Cook graduated summa cum laude from the University of Florida, earning a B.S.B.A. with a major in International Economics and a minor in German in 2003. She completed an undergradua te thesis on the topic of price changes in Germany after conversion from the Deutschmark to the Euro under the supe rvision of Dr. Denslow. In 2007 she earned a n M.A. in Business Economics from the University of South Florida, was accepted into the USF Co llege of Business Ph.D. program and was awarded a graduate fellowship for her first year of doctoral study. Her fields of study included health economics, international/development economics, econometrics, and international affairs. As an un dergraduate, Ms. Cook was employed as a teaching assistant by the UF Department of Statistics. She has been employed as an economist with the Tampa consu lting firm Deiter, Stephens, & Durham since 2004.