Immigration and Occupational Choice of Natives: The Case of Nurses in the United States Jessica Pany

Patricia Cortés

August 4, 2014

Abstract We analyze the e¤ects of foreign nurse immigration on the occupational choice of natives and the quality of native entrants into the nursing sector. Using an empirical strategy that exploits large geographical di¤erences in the distribution of foreign nurses across US states, we …nd that, in response to foreign nurse immigration, fewer native nurses sit for the nursing licensure examinations in states that are historically more dependent on foreign nurses. Moreover, we …nd that states with larger increases in foreign nurses have fewer young natives choosing to enter nursing, with the decline o¤set by an increase in the supply of young natives to primary school teaching. Using data on the passing rates of native nurses in a state as a proxy for nursing quality, we …nd robust evidence that an increase in the ‡ow of foreign nurses increases the passing rate of natives in more dependent states relative to less dependent states. We suggest that the increase in quality may be induced by native nurses’response to an increase in competition or potentially consistent with the predictions of a Roy (1951) model of occupational selection where native nurses are positively selected into the nursing sector.

y

School of Management, Boston University. Email: [email protected] National University of Singapore. Email: [email protected]

1

1

Introduction

To meet the rising demand for healthcare professionals, the US nursing market has relied extensively on the hiring of foreign-educated nurses. The number of foreign educated nurses working in the US has increased rapidly in the past few decades - in the mid-1980s, 5 percent of nurses passing the licensure examination were foreign-educated and this proportion has increased to close to 17 percent in the mid-2000s. The importation of foreign-educated nurses is a contentious issue. On the one hand, proponents of foreign nurse importation argue that foreign nurses provide critical temporary relief in times of nursing shortages. On the other hand, critics raise concerns that the reliance on foreign-educated nurses will reduce the attractiveness of a nursing career, either by lowering wages or delaying key workplace improvements, thereby deterring natives from choosing nursing (ANA 2008). We attempt to address this issue by examining the extent to which the importation of foreign nurses has an impact on the occupational choice of natives. This analysis builds on our earlier paper (Cortes and Pan 2014) that provides strong evidence that the importation of foreign-born registered nurses (RNs) has a negative impact on the long-run supply of native nurses.1 More speci…cally, we …nd that over a ten-year period, for every foreign nurse that migrates to a city, between 1 to 2 fewer native nurses are employed in the city. We provide some suggestive evidence that part of the displacement e¤ects count be driven by a decline in the perceived quality of the workplace environment. While our previous paper focused mostly on changes in the labor supply of native nurses, in this paper, we seek to better characterize how immigrants to the nursing sector has shaped native occupational choice by examining, more speci…cally, (1) the alternative occupations that displaced nurses are entering and (2) the e¤ect of the importation of foreign-nurses on the quality of natives choosing to enter the nursing profession. Both of these questions are important in understanding the long-term e¤ects of foreign nurse importation on the quality and quantity of natives in the US nursing sector. While there is an extensive literature on the e¤ect of immigration on native employment and wages,2 the literature on the impact of immigration on occupational choice is far more limited. The few existing studies generally focus on the e¤ects of high skill immigration, in particular of Science, Technology, Engineering and Math (STEM) workers, on native occupational and educational choices. Peri and Sparber (2011) use data on the characteristics of occupations to document that native-born workers with graduate degrees appear to respond to an increased presence of highly-educated foreign-born workers by choosing occupations that require more interactive and communication skills (and less quantitative and analytical skills). In a similar vein, there is some 1

Throughout the paper, our use of the term “nurses" refer speci…cally to Registered Nurses (RNs). We use the terms nurses and RNs interchangeably. 2 See for example, Borjas (2003, 2006), Card (2001, 2005) and Wozniak and Murray (2012).

2

evidence to suggest that in response to high-skill immigration, natives may be moving toward certain educational programs in line with their comparative advantage. Orrenius and Zavodny (2013) …nd that as the share of immigrants working in STEM …elds rise, natives are less likely to major in these …elds in college. Groen and Rizzo (2007) also document that although the share of PhDs granted to US citizens in the sciences declines between 1963 and 2000, the propensity for native students to pursue an MBA increased markedly. While they do not link these patterns directly to immigration, this interpretation is consistent with the …ndings by Borjas (2007 and 2009). Borjas (2007) documents a strong negative correlation between increases in the number of foreign students enrolled in a particular university and the number of native men in that graduate program, with the crowd-out e¤ect strongest at the most elite institutions. Borjas (2009) shows that an increase in foreign students in a particular doctoral …eld had a signi…cantly adverse e¤ect on the earnings of doctorates graduating from the same …eld. To our knowledge, we are the …rst to look at the impact of immigration of health care workers, nurses in particular, on native occupational choice. There are a number of reasons why the nursing sector is an important and interesting case study in the immigration context. RNs are the single largest group of healthcare professionals in the United States and their demand is expected to grow steadily over the next ten to …fteen years. In 2012, the Bureau of Labor Statistics (BLS) projected that registered nursing, which requires at least an associate’s degree, will have the largest employment growth of all occupations.3 This represents a signi…cant labor market opportunity for many middle/high-skilled natives. At the same time, the nursing market relies heavily on the importation of foreign nurses, raising the question as to how this reliance on immigrants may have impacted the quantity and quality of native workers choosing a nursing career. Moreover, because the immigration shocks that we consider are localized to particular occupation,4 this setting provides us with an opportunity to examine how natives’ career choice responds to immigration, particularly among middle/high-skilled natives. Finally, understanding how immigration impacts natives’decision to choose a nursing career is also critical in evaluating policies aimed at increasing the long-run supply and quality of nurses. Our main empirical strategy exploits the large geographic variation in immigrant concentration in the nursing sector in the United States to identify how the share of foreign nurses has a¤ected (a) the occupational choice of native workers and (b) the quality of native nurses. To answer the …rst question, we use a previously unexplored dataset on the number of native and foreign nurses sitting for the National Council Licensure Examination (NCLEX) to examine how the supply of native nurses has responded to increases in foreign nurse immigration across states that historically di¤er in their dependency on foreign nurses. This dataset has the advantage that it is annual and the number of nurses passing the examination is a close proxy for the actual supply of new RNs in the 3 4

http://www.bls.gov/news.release/pdf/ecopro.pdf Foreign born nurses typically enter the US through special visa programs such as the H1A.

3

nursing market. We …nd evidence that in response to an increase in immigrant ‡ow at the national level, fewer natives sit for the licensure examination in states that are historically more dependent on foreign nurses. Next, to better characterize the occupational choice of native nurses, we use data from the US Census and American Community Survey to examine the alternative occupations that displaced native nurses potentially enter. We follow the empirical strategy in Cortes and Pan (2014) that utilizes Card’s (2001) spatial correlations approach to relate ten-year changes in the number of natives age 21 to 31 in alternative occupations per capita in a state to ten-year changes in the number of foreign-born nurses per capita. We consider a number of alternative occupations such as other non-nursing health occupations, teachers, managers, lawyers, and doctors and dentists. Consistent with the displacement results in Cortes and Pan (2014), we show that states with a larger increase in foreign RNs have fewer young natives choosing nursing, with the decline o¤set by an increase in the number of young natives entering teaching. We …nd little evidence of an increase in the number of young natives in other alternative occupations. Finally, to examine the impact of foreign nurse importation on native nurse quality, we use annual data on the performance of native nurses sitting for the NCLEX. Interestingly, we …nd that increases in the aggregate (national) ‡ow of foreign nurses are associated with a larger share of natives passing the exam in states that are historically dependent on foreign nurses relative to less dependent states. While we are unable to test the precise mechanism that explains this increase in average quality, we propose a couple of potential explanations for this …nding. These results are consistent with a situation in which native nurses face greater competition in states with more foreign nurses, leading them to potentially respond by improving their quality. Furthermore, this result can also be reconciled with the predictions of a simple Roy model (1951) where natives are positively selected into nursing. A decline in the supply of native nurses to the nursing sector (or a rise in the opportunities in the alternative sector) can increase the quality of natives in the nursing sector as marginal, less skilled, natives switch to the alternative sector. The rest of the paper is organized as follows. Section 2 describes the data and presents descriptive statistics. The empirical strategy and results are discussed in Section 3. We conclude on Section 4.

2

Data and Descriptive Statistics

To examine the e¤ects of foreign nurse importation on the quantity and quality of natives entering the nursing profession, we use annual data on the number of US-educated individuals who sat for the nursing board examinations for RNs (NCLEX) by state from 1986 to 2010.5 Figure 1 shows annual data on the number of native and foreign educated RNs who passed the exam. The time-series 5

The data was obtained from annual publications from the Nursing Board and is disaggregated up to the state level.

4

patterns indicate that a signi…cant share of the variation observed in the share of foreign nurses passing the NCLEX is the result of nurse speci…c migration laws, such as the 1989 Nurse Relief Act, which created a non-immigrant visa category (the H1A) exclusively for nurses. As observed in the …gure, while the law was in e¤ect, the foreign educated share increased signi…cantly. This increase was short-lived with the expiration of the law in 1995. The spike in 2006-2007 was the result of an immigration policy in 2005 that released 50,000 green cards to be allocated exclusively to foreign nurses and their families. Our second dataset is drawn from the 1980 to 2000 US Censuses and the American Community Survey (ACS) three-year aggregate for 2010 (2008-2010). As we are interested in the occupational choice of potential native nurses at the beginning of their career, we focus on native workers age 21 to 31 with at least two years of college who reported their occupation to be that of a Registered Nurse as well as other alternative occupations. To give a sense of how occupational choice of native women has changed over time, Figure 2 shows the share of a cohort choosing nursing, teaching and other skilled occupations.6 As observed, nursing was a common occupation among baby boomers,7 but its popularity steadily declined for the following cohorts. The most recent cohorts, as suggested by NCLEX data, might be reversing the trend, a change attributed to rising wages in the health care sector and to the recession. Figure 2 also shows how teaching followed a di¤erent pattern. The decline in popularity started earlier and cohorts following baby boomers were more likely than baby boomers to choose teaching. Finally, Figure 2 clearly illustrates the expanding opportunities of women in prestigious occupations such as medicine, law and business. Our main empirical strategy, using both the NCLEX and Census/ACS, exploits the large concentration of foreign-born nurses8 in particular areas of the country. Table 1 presents the share of foreign-born nurses in each state in the US for each Census decade from 1980 to 2010.9 In 2010, the top three states include the District of Columbia, California and Nevada where more than a third of RNs were foreign-born. In contrast, in states like Mississippi, Nebraska and West Virginia, less than 2 percent of all RNs were born abroad. The increase in the share of foreign-born nurses has also evolved di¤erently across states - while the average increase in the foreign-born share from 1980 to 2010 is 5 percentage points, the increase ranged from 20 to 30 percentage points in the 6

In this graph, we focus on women as the overwhelming majority of registered nurses are female (more than 92 percent in 2010). 7 Baby boomers are people born during the demographic post-World War II baby boom between the years 1946 and 1964. 8 When using the Census and the ACS, we concentrate on foreign born nurses instead of foreign educated nurses given that these data sets do not include information on the country of education. Although we could potentially use the year of immigration variable to construct a proxy, the measurement error is likely to be large, especially for 1980 and 1990 when this variable is aggregated in …ve year periods. 9 A very similar ranking of states is observed for the number of foreign nurses per capita, the variable we are going to use in our regressions. The correlation between these two measures is 0.96.

5

top-three states (DC, California and Nevada), while the bottom-four states saw declines of -3 to -1 percentage points in the foreign-born share of RNs (Arkansas, Montana, Oregon and Mississippi).

3

Empirical Strategy and Results

3.1

Occupational Choice of Natives

Evidence from the Nursing Licensure Examinations We begin our analysis by looking at the e¤ect of foreign nurse importation on the supply of native nurses. This exercise builds on earlier work by Cortes and Pan (2014). To directly test whether foreign nurse migration a¤ects the number of natives choosing to enter nursing, we use annual data on the number of individuals choosing to take the NCLEX from 1986 to 2010. While we have statelevel information on the numbers of US-educated individuals taking the exam,10 unfortunately, data on foreign-educated exam takers is only available at the national level.11 Therefore, we use a reduced form approach inspired by Kerr and Lincoln (2010). Our empirical strategy tests whether increases in the aggregate (national) ‡ow of foreign nurses (normalized by the country’s population) are associated with fewer natives joining the occupations four years later in states that are historically more dependent on foreign nurses relative to less dependent states. The rationale for using a four year lag is that it takes between two to four years to become a nurse. To measure historical dependency on foreign nurses, we use data from the 1980 Census to construct the share of RNs who are foreign-born at the state level.12 For ease of interpretation, we normalize the dependency measure to have unit standard deviation. Our empirical speci…cation is the following:

N ative T akers = + P opulation st

Dependency1980;s

F oreign Educated P assers P opulation t

+

s + t + rt +Xst +"st ;

4

(1) where s is for state, r is for region,13 and t for year. The regressions include state …xed e¤ects ( 10

s ),

Appendix Table 1 provides the key descriptive statistics for this sample. The number of foreign nurses passing the NCLEX is considered a good proxy for the actual in‡ow of foreign nurses to the US. In order to take the exam, the candidate has to have applied for a nursing license in one of the states. This usually requires having obtained a VisaScreen certi…cate from the Commission on Graduates of Foreign Nursing Schools, who checks that the nurse has a valid license from her country of residence, has passed the TOEFL and has passed a qualifying exam. For most foreign educated nurses the process is sponsored by the potential employer or by a recruiting agency (CGFNS, Nichols, and Davis 2009). 12 Results are almost identical when we use as dependency measure the number of foreign born nurses per 1000 people in the state, constructed using the 1980 Census. 13 We use the 9 region classi…cation de…ned by the US Census. 11

6

year …xed e¤ects ( t ), region x year …xed e¤ects ( (Xst ).14

rt )

and a set of state-level time-varying controls

As discussed in the previous section, a large part of the variation in the number of foreign nurses taking the NCLEX can be attributed to nurse speci…c migration policies. One concern is that if the passing of these laws resulted from heavy lobbying by employers in states that are highly dependent on foreign nurses, the interaction term might be proxying for demand and supply shocks experienced by these states. To address this issue, we present speci…cations in which we omit the …ve states that depended most heavily on foreign nurses in 1980 (California, Florida, Illinois, New Jersey, and New York). To further ensure that our results are not driven by the most dependent states and that the e¤ects are also observed for states at other points in the dependency distribution, we estimate a model that replaces the interaction term by quintile dummies interacted with the aggregate ‡ow measure:

N ative T akers = P opulation st

+

1

I(T op quintile)1980;s

+

2

I(Second quintile)1980;s

+

3

I(T hird quintile)1980;s

+

4

I(F ourth quintile)1980;s

F oreign Educated P assers P opulation t 4 F oreign Educated P assers P opulation t 4 F oreign Educated P assers P opulation t 4 F oreign Educated P assers + P opulation t 4

(2)

s

+

t

+

where states in the bottom quintile of the dependency distribution serve as the reference group. Even if the ‡ow was orthogonal to state speci…c shocks, one might be concerned that states that depend heavily on foreign nurses are di¤erent from less-dependent states in other dimensions that make them subject to di¤erent shocks or to exhibit di¤erent trends. One important confound is the expansion of managed care organizations during the 1990s and the large state variation in the speed of adoption of this new form of health care delivery, with some of the states characterized by high dependency on foreign nurses also being early adopters of managed care, California in particular. As Buerhaus, Staiger, and Auerbach (2009) show, in the …rst half of the 1990s, growth in the employment of RNs was signi…cantly lower in states with high health maintenance organization (HMO) rates. To deal with this issue, we present speci…cations in which we include interactions of 14

Note that we use foreign educated passers as our key explanatory variable, but native takers as our dependent variable. The passing rate of foreign educated nurses is not very high (average across years of about 35%) so many of those who take the exam never end up working as registered nurses in the US. On the other hand, we are interested in the number of natives who graduated from a nursing program (a prerequisite to register for the exam) and not necessarily in the number who passed the exam. Note, however, that the passing rates for natives are extremely high, so results using passers are very similar.

7

rt

+ "st

a dummy for early adopter with year …xed e¤ects. An early adopter is de…ned as being a top 10 state in the percentage of population enrolled in a HMO in 1994.15 In addition, we also control for other variables that might be correlated with historical dependence and that are likely to a¤ect our outcome. Using CPS data we construct the following variables and include a 4 year lag of each in our model: the share of whites in the population, age composition of the population (share aged 0-19, 20-39, 40-59), a cubic in the state’s population size and the share of females in professional occupations. Finally, by including region x year …xed e¤ects in the speci…cation, our analysis compares states that are arguably more similar and more likely to be subject to common shocks. The estimates of equations (1) and (2) are reported in Table 2. Most regressions are weighted by the state’s population and standard errors are clustered at the state level. We also estimate our most complete speci…cation without using weights (column (6)) for robustness and because it gives us the e¤ect of the in‡ow of foreign nurses into a randomly selected state. We focus on results using a four-year lag, and present results using other lags in Appendix Table 3. Our estimate of the reduced form e¤ect ( ) is always negative, statistically signi…cant and robust to the inclusion of a variety of controls and alternative forms of weighting. The magnitude of the coe¢ cient in our preferred speci…cation (Column (4)) suggests that increasing the number of foreign-educated nurses passing the exam per capita by 10 at the national level is associated with approximately 7 fewer natives taking the exam 4 years later for each standard deviation growth in state dependency. As suggested by Column (5), which excludes the top 5 states and by the quintile speci…cations, the e¤ect is not driven by the most dependent states. Reassuringly, all the coe¢ cients in the quintile speci…cation are negative (the reference group is the bottom quintile), with e¤ects generally decreasing in magnitude as we move down in the distribution of dependency. An increase in one foreign educated nurse per capita at the aggregate level reduces the number of native nurses taking the exam in states with the highest dependency by about 2.4 and in states with medium level dependency (quintiles 2 and 3) by about 1.4, relative to the bottom quartile. The coe¢ cient on the interaction of the fourth quintile with the aggregate ‡ow is negative and the magnitude is not small, but we cannot reject that it is equal to zero. We present a series of robustness checks in Appendix Table 3. Column (1) re-estimates our preferred speci…cation using the same sample as that for di¤erent lags and the results are similar to that in Table 2. Since it takes between 2 to 4 years to become a nurse, it is reassuring that there is little correlation between the number of native nurses taking the exam in a given year and the ‡ow of foreign nurses six years before (Column (2)) or in the same year (Column (3)). 15

The percent enrolled in HMO by state was taken from Buerhaus, Staiger, and Auerbach (2009) Table 5-1. Results are robust to changing the de…nition of an early adopter to being a top 5 or a top 17 state (the de…nition used by Buerhaus, Staiger, and Auerbach 2009) in the percentage enrolled in a HMO.

8

Evidence from the Census and ACS Although the NCLEX has the advantage that it provides a good proxy for the actual supply of native nurses at a particular point in time, it does not provide any information on individuals who choose not to enter the nursing sector. Therefore, to identify the e¤ect of foreign nurse importation on the occupational choice of natives, we turn to the Census and ACS that has population-level information on the stock of RNs and natives in other occupations in each decade. We exploit variation across states over time in foreign nurse concentration and relate these to changes on the number of natives choosing nursing and other alternative occupations in a state. This empirical methodology is similar to Cortes and Pan (2014) and Card (2001, 2005). Unlike Cortes and Pan (2014), we use states as the main unit of analysis for comparability with the results using the NCLEX data which are only available at the state level. We also focus on the occupational choice of young natives age 21 to 31 as this corresponds to the point at which natives are most likely to be choosing among alternative careers.16 Our empirical speci…cation is as follows: N atives in Occkijt = P opulation srt

+

F oreign N urses P opulation

+ Xsrt +

rt

+

s

+

t

+

srt

(3)

srt

where s refers to the state, r the region, and t the time period (t =1980, 1990, 2000 and 2010). For our main dependent variables, we focus on the number of natives age 21 to 31 in a given occupation (k) in the state as a fraction of the population in the state. The occupations that we consider include RNs, other health occupations (excluding physicians and dentists), primary school teachers, secondary school teachers, managerial positions, doctors and dentists, and lawyers. The key independent variable is the number of foreign nurses (age 20 to 70) per capita in a state in each census year. Xsrt is a vector of time-varying state level controls, rt is a vector of region x time …xed e¤ects, s is a vector of state …xed e¤ects and t is a vector of time period …xed e¤ects. The vector Xsrt also includes a cubic polynomial in state population and proxies for demand and supply determinants. The demand determinants include the share of the state population over 65, the log of average hourly wages as a proxy for the state’s income level and the number of physicians per 1000 population. Variables to capture the supply side of the native nursing labor market include the share of the state population age 25 to 34, 35 to 44, 45 to 54 and 55 to 64, the share of females in professional occupations,17 the labor force participation of skilled married women, the log average hourly wage of skilled women outside of nursing and the share of whites in the population.18 Standard errors are clustered at the state level and most speci…cations are 16

Older nurses may respond to the increase in foreign-nurses by exiting the labor force rather than to switch occupations. 17 Buerhaus, Staiger, and Auerbach (2009) suggest that the expansion of career opportunities for women in traditionally male-dominated …elds could be one of the main causes of the declining interest in nursing among natives. 18 Whites are typically over-represented in nursing.

9

weighted by the state population. To account for the potential endogeneity of the OLS estimates resulting from the fact that the number of foreign nurses to a state is likely to be correlated with unobserved demand and supply shocks to the native nurse labor market, we adopt an instrumental variables strategy common in the immigration literature that uses the historical distribution of migrants across US cities (Card, 2001). The instrument exploits the tendency of immigrants to settle in cities with large enclaves of immigrants from the same country. This is based on the idea that prospective immigrants choose locations on the basis of the strength of immigrant networks (Munshi, 2003). The instrument uses the 1980 distribution of skilled immigrants from a given country across US states to allocate the new waves of foreign nurses from that country to a given state. For example, if a third of skilled Filipino immigrants in 1980 were living in California and a quarter were living in Hawaii, the instrument allocates a third of all Philippine-born nurses in each time period (1980, 1990, 2000 and 2010) to California and a quarter to Hawaii. Note that we chose speci…cally to use the 1980 distribution of skilled immigrants (de…ned as those with some college or more) as opposed to the historical distribution of immigrant nurses as it is likely to be the case that the distribution of skilled immigrants is more exogenous to persistent shocks to the local nursing labor market than the latter. Formally, the instrument for the number of foreign nurses in state s and decade t can be written as: X Skilled Immigrantscs;1980 F oreign N ursesct; s (4) Skilled Immigrantsc;1980 c Skilled Immigrants

where s denotes state, c country of origin and t time period.19 Skilled Immigrantscs;1980 represents the c;1980 fraction of skilled immigrants from country c who were living in state s, and F oreign N ursesct; s stands for the total number of foreign-born nurses from country c to the United States in decade t, net of the contribution of state s to this total.

Most of the speci…cations include state and region*decade …xed e¤ects. Therefore, the instrument will help in identifying the causal e¤ect of the displacement of native nurses under the following conditions: 1. Unobserved factors that determine that more skilled immigrants decide to locate in state s vs. state s0 in 1980 are uncorrelated with changes in the demand for nurses during the 1980s to 2000s. To ameliorate this concern, we use the distribution of skilled immigrants, excluding immigrants who are nurses. As mentioned before, this distribution is likely to be 19

We restrict the set of countries to those that account for the large majority of nurse immigrants. These countries include Canada, Mexico, Cuba, Haiti, Jamaica, Trinidad and Tobago, England, Ireland, Germany, China, Japan, Korea, Philippines, Thailand and India. Together, these countries account for 70% of all foreign-born nurses in the US from 1980 to 2010.

10

more exogenous to persistent shocks to the local nursing labor market than the historical distribution of immigrant nurses.20 2. The national ‡ow of foreign nurses in a given decade (second term in the interaction) is exogenous to di¤erential shocks to states within a given region. This is particularly relevant for large states, where it is possible that the aggregate national foreign nurse ‡ow at time t may be correlated with local conditions at the state level. To circumvent this concern, we omit the contribution of state s to the national foreign-nurse in‡ow in each time period when constructing the instrument for each state. We report the estimate for the …rst-stage regression of foreign-born nurses per 1000 population in a state on the instrument (predicted number of foreign-born nurses in a state per 1000) in Appendix Table 4. The coe¢ cient on the instrument indicates that as the predicted number of foreign-born nurses in a state increases by 10, this is associated with an in‡ow of 3 to 7 foreign-born nurses to the state. This …rst stage coe¢ cient is generally highly statistically signi…cant, in particular in speci…cations that use population weights, and robust to the introduction of di¤erent sets of timevarying controls (Column 2) and …xed e¤ects for region x year (Column 4). In all three speci…cations in which states are weighted by their population the cluster robust F-statistic is larger than the 20% maximal IV size critical value from Stock-Yogo (2005).21 Columns (1) to (5) in Table 3 report the OLS estimates for the occupational choice regressions while Columns (6) to (10) report the 2SLS estimates. Most regressions are weighted by the state’s population and standard errors are clustered at the state level. The …rst row estimates the impact of foreign-born nurses on the number of native nurses age 21 to 31 in each state. Consistent with the results reported in Cortes and Pan (2014) and the results for the NCLEX in the section above, the 2SLS estimates provide evidence of economically and statistically signi…cant displacement e¤ects for every 10 foreign-born nurses that enter a state, between 3 to 5 native nurses age 21 to 31 are displaced.22 The results are robust to the inclusion of time-varying state-level controls and region x year …xed e¤ects. The OLS estimates are generally smaller than the 2SLS estimates, indicating that 20

We also constructed a similar instrument using the 1970 geographic distribution of foreign-born high skilled workers; results are similar, but noisier and available upon request. Similarly, when we estimated our models using the instrument based on the 1980 Census, but dropping 1980 from the analysis, the …rst-stage and second-stage results not including region x year …xed e¤ects are noisier but still marginally signi…cant and similar in magnitude to those using all years. However, the …rst stage coe¢ cient is no longer statistically signi…cant when region x year …xed e¤ects are added. Finally, it is worth mentioning that results do not change in any signi…cant way when we use the geographic distribution of foreign-born nurses to construct the instrument. 21 The Stock-Yogo weak ID test critical values for K1=1 and L1=1 (as reported by the ivreg2 command in stata) is 6.6 for 20% maximal IV size. 22 For example, Cortes and Pan (2014) found that for every 10 foreign-born nurses that enter a city, between 3 to 5 native nurses age 25 to 34 are displaced.

11

the OLS estimates are downward biased by unobserved positive demand shocks or measurement error. In the next few rows of Table 3, we replace the key dependent variable with the number of natives age 21 to 31 in each of the following occupations (i) other health occupations23 (ii) primary school teachers (iii) secondary school teachers and (iv) managerial positions. For doctors, dentists and lawyers (Panel B), we change the age range to 27 to 37 and use the lag of the number of foreign nurses per capita as explanatory variable. We …nd robust evidence suggesting that states that had a larger in‡ow of foreign-born nurses experienced a signi…cantly larger number of natives reporting being primary school teachers.24 While the coe¢ cients indicate that the number of primary school teachers increased by between 5 to 7 for every 10 foreign-born nurses in a state, we are unable to reject that these estimates are signi…cantly di¤erent from the magnitude of the native nurse displacement at conventional levels. In contrast, we …nd little evidence that the importation of foreign-born nurses is associated with signi…cantly more natives choosing other health occupations or taking on higher-skill occupations such as managers, doctors and dentists or lawyers. Note that foreigners account for a relatively small share of primary school teachers (between 3% to 7% from 1980 to 2010), so that our explanatory variable is unlikely to be picking up the e¤ect of the in‡ow of foreign teachers.25 Additionally, even if the ‡ows were signi…cant, we would expect ‡ows of foreign teachers and nurses to be positively correlated at the state level, biasing us toward …nding a negative e¤ect.

3.2

Quality of Native Nurses

Next, we turn to the question as to how the importation of foreign-born nurses may have impacted the quality of native nurses entering the profession. The advantage of the NCLEX data is that it provides us with a direct measure of native nurse quality across state - the passing rate of US-educated individuals who sat for the nursing board examinations. This measure captures the quality of nursing students who decide to sit for the NCLEX in each state. While the passing rate for US-educated individuals is relatively high (from Appendix Table 1, the average passing rate is 88%), there is variation across states and over time (the standard deviation is 4.2%, minimum is 61% and maximum is 98%). Our empirical speci…cation is identical to (1) and (2) except that we replace the dependent variable 23

Other health occupations include physician assistants, all types of therapists, optometrists, podiatrists, pharmacists and veterinarians. 24 The p-value for weak-instrument-robust inference using the Stock-Wright LM S statistic is smaller than 0.05 for all models of nurses and primary school teachers. 25 See Appendix Table 2 for summary statistics on the share foreign-born in each of the occupations considered over time.

12

P assers 26 We …nd robust evidence that, in with NNative ative Takers st . The results are presented in Table 4. response to an increase in the national ‡ow of foreign nurses, the passing rate of natives is higher in states that are historically more dependent on foreign nurses. The results from our preferred speci…cation in Column (4) are all positive and signi…cant using both the linear and quintile speci…cation. The results are also robust to dropping the top …ve states and controlling for region*year …xed e¤ects. The results indicate that a one standard deviation increase in the number of foreign passers per 1000 population at the national level (the average is about 0.032,with a standard deviation of 0.014) is associated with a 0.25 percentage point (or 6% of the standard deviation of passing rate across states) increase in the share of native nurses passing the exam, for a standard deviation change in state dependency. For states in the top quintile of foreign-nurse dependency, a standard deviation increase in the foreign passers per 1000 population at the national level is associated with a 1.1 percentage point (or 26% of the standard deviation of passing rate across states) increase in the share of native nurses passing the exam.

3.2.1

What Explains the Rise in Nursing Quality?

While the previous result might appear counter intuitive, we propose two reasons as to why we may observe an improvement in the average quality of native nurses in response to an in‡ux of foreign nurses. One possibility is that native nurses may respond to the increased competition in the nursing market, induced by the entry of foreign nurses, by improving their skills. This improvement in average nursing skill would be consistent with native nurses achieving higher scores on the licensure examinations in states that are more a¤ected by foreign nurse importation. Second, the results could also be interpreted through the lens of a simple Roy (1951) model of comparative advantage with positive selection of natives into the nursing sector. To illustrate this point, we sketch a simple Roy (1951) model to provide a framework for thinking about the occupational choice decision of a potential nurse. Our set up is similar to Bacolod (2007) and Borjas (1987). Let 0 denote the non-nursing sector and 1 the nursing sector. An individual considering nursing faces the following earnings (pecuniary plus non-pecuniary) distributions:

ln w0 =

0

+

0

where

0

N (0;

2 0)

ln w1 =

1

+

1

where

1

N (0;

2 1)

and represents the correlation between 0 and 1 . Both 0 and 1 are the values of an individual’s draw of talent or ability in each sector. Assuming that individuals maximize income, the probability 26

The robustness checks using other lags are presented in Appendix Table 3.

13

that an individual will choose the nursing sector is P = P r[ln w1 > ln w0 ] = P r[v > where v =

1

0,

z=

(

0)

1 v

and

(

1

0 )]

=1

(z)

is the standard normal cumulative distribution function.

Let us consider what happens to the average quality of native nurses resulting from a relative decline in the average wages of the nursing sector. In our context, we argue that part of the displacement e¤ect of native nurses is driven by foreign-born nurses reducing the perceived quality of the workplace (Cortes and Pan, 2014), which results in a lower perceived e¤ective wage (pecuniary plus non-pecuniary) of the nursing sector. This model indicates that the quality of natives in the nursing sector may increase or decrease when wages in the nursing sector falls. The e¤ect on average nurse quality depends on the distribution of ability in each sector and the nature of self-selection. If skills are su¢ ciently correlated across sectors and the wage distribution in the nursing sector is more dispersed than the non-nursing sector, average nursing quality could increase with the decline in wages of the nursing sector.27 The intuition is that as 1 falls and shifts the nursing sector distribution downwards, some workers who had chosen nursing before the change see their relative productivity and earnings across sectors change. More speci…cally, the group that gains most from switching to the non-nursing sector would be those individuals at the lower end of the nursing skill distribution, because their skills are now relatively more productive in non-nursing. This would tend to raise the average quality of nurses.28 While we may not expect that the wage distribution in the nursing sector is, in general, more dispersed than the non-nursing sector, our results from Section 3.1 suggests that individuals who are interested in nursing are typically choosing between nursing and an alternative occupation like teaching. It is possible, that relative to teaching, the wage distribution in the nursing sector is more dispersed, given the range of potential employment opportunities that RNs can undertake in the health sector and the rising demand for healthcare services. For example, in 2010 the standard deviation of the hourly wage distribution for female nurses aged 40-44 was 9.2 but only 7 for comparable teachers. Furthermore, the 90th percentile was 27 dollars for nurses and 23 dollars for teachers. This would tend to favor positive selection in the market for nurses. 27

Formally, this occurs when > 01 and 1 > 0 . In contrast, if skills are su¢ cient correlated across sectors and the wage distribution in the nursing sector is less dispersed than the non-nursing sector, average nursing quality could decrease with the decline in wages in the nursing sector. 28

14

4

Conclusion

The reliance on foreign-trained nurses is not particular to the United States, a recent OECD/WHO (2010) policy brief estimates that there are a large number of advanced economies actively recruiting nurses from developing countries to meet their healthcare needs - for example, foreigneducated nurses accounted for approximately 8% of the nursing workforce in the United Kingdom and Canada, 10% in Italy29 , 16% in Australia, 23% in New Zealand and 47% in Ireland. With the increasing reliance on foreign nurses to meet health care needs in many developed countries, understanding how these immigration ‡ows a¤ect the occupational choice of natives is an important issue particularly for policymakers concerned about the long-run supply of nurses. In this paper, we build on the …ndings by Cortes and Pan (2014) and explore in greater detail the impact of foreign nurse importation on native occupational choice and the resulting quality of natives choosing to enter nursing. We show that states the experience an in‡ux of foreign nurses have signi…cantly fewer natives choosing to sit for nursing licensure examinations, have fewer natives in the 21 to 31 age group choosing to enter nursing, and more individuals choosing to enter the teaching profession. This evidence is consistent with the idea that potential native nurses respond to the importation of foreign nurses by entering the teaching profession instead of nursing. Using data on the passing rates of natives sitting for the licensure examinations in each state, we also document that states more a¤ected by foreign nurse immigration have a larger share of natives passing the examination, suggesting that the average quality of native nurses appears to improve in response to the entry of foreign nurses into the nursing sector. We suggest that these results could be due to native nurses improving their quality in response to the increase in competition. We also show that these results are potentially consistent with the predictions of a Roy (1951) model of comparative advantage. If native nurses are positively selected into nursing, then a decline in the attractiveness of the nursing section (due to the entry of foreign nurses) would lead to an exit of marginal, lower quality, nurses, thereby raising the average quality of native nurses. Unfortunately, in this paper, we are not able to distinguish whether the increase in quality is driven by competition or self-selection. Understanding the mechanisms as to why average nursing quality rose in states that experienced a larger in‡ux of foreign nurses is an interesting avenue for future research. Our results also have implications for understanding the impact of immigration on other occupations that rely on foreign-trained personnel. The impact of immigration is likely to be occupation-speci…c and depend, among other things, on the degree of substitution (or complementarity) between natives and immigrants and on the existence of economies of agglomeration in the relevant production function. Previous research has largely focused on the e¤ects of high-skill immigration and its effects on the occupational and educational choices of natives (Peri and Sparber 2011). For example, 29

The number for Italy is the share of foreign-born nurses, not foreign-educated.

15

in the US, the share of college graduates working in STEM occupations doubled from 9% in 1970 to nearly 19% in 2000 (Orrenius and Zavodney 2013). Consistent with the evidence presented in these papers, we …nd that natives do respond to the immigration in‡ux by entering a sector with signi…cantly fewer immigrants (teaching). On the other hand, our …ndings are in contrast to other studies such as Kerr and Lincoln (2010) and Hunt and Gauthier-Loiselle (2010) who …nd limited crowding out e¤ects of very high-skill immigration (H-1B visa holders) on native employment in science and engineering …elds. These …ndings underscore the importance of taking into account occupation-speci…c factors in understanding the potentially heterogenous e¤ects of immigration on occupational choice.

References [1] American Nurses Association. 2008. “Statement for the Committee on the Judiciary - Subcommittee on Immigration, Citizenship, Refugees, Border Security, and International Law On Registered Nurse Immigration." http://www.nursingworld.org/DocumentVault/GOVA/Federal/Testimonies/Testimony061208.pdf. [2] Bacolod, Marigee. 2007. “Do Alternative Opportunities Matter? The Role of Female Labor Markets in the Decline of Teacher Quality." Review of Economics and Statistics 89(4): 737– 751. [3] Borjas, George J. 1987. “Self-Selection and the Earnings of Immigrants." American Economic Review 77(4): 531–553. [4] Borjas, George J. 2003. “The Labor Demand Curve Is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market." Quarterly Journal of Economics 118(4): 1335– 1374. [5] Borjas, George J. 2006. “Native Internal Migration and the Labor Market Impact of Immigration." Journal of Human Resources 41(2): 221–258. [6] Borjas, George J. 2007. “Do Foreign Students Crowd Out Native Students from Graduate Programs." In Science and the University, ed. Paula E. Stephan and Ronald G. Ehrenberg, 134–150. Madison: University of Wisconsin Press. [7] Borjas, George J. 2009. “Immigration in High-Skill Labor Markets: The Impact of Foreign Students on the Earnings of Doctorates." In Science and Engineering Careers in the United States, ed. Richard B. Freeman and Daniel L. Goro¤, 131–161. Chicago: University of Chicago Press.

16

[8] Buerhaus, Peter I., Douglas O. Staiger, and David I. Auerbach. 2009. The Future of the Nursing Workforce in the United States. Sudbury, MA: Jones and Bartlett Publishers. [9] Card, David. 2001. “Immigrant in‡ows, Native Out‡ows and the Local Labor Market Impacts of Higher Immigration." Journal of Labor Economics 19(1): 22–64. [10] Card, David. 2005. “Is the New Immigration Really So Bad?" Economic Journal 115(507): 300–323. [11] CGFNS International, Barbara L. Nichols, and Catherine R. Davis, eds. 2009. What You Need to Know About Nursing and Health Care in the United States. New York: Springer Publishing Company. [12] Cortes, Patricia and Jessica Pan. 2014.“Foreign Nurse Importation and the Supply of Native Nurses." Journal of Health Economics 37: 164-180. [13] Groen, Je¤rey A. and Michael J. Pizzo. 2007. "The Changing Composition of AmericanCitizen PhDs." in Ronald G. Ehrenberg and Paula Stephan, eds., Science and the University, University of Wisconsin Press. [14] Hunt, Jennifer, and Marjolaine Gauthier-Loiselle. 2010. “How Much Does Immigration Boost Innovation?" American Economic Journal: Macroeconomics 2(2): 31–56. [15] Kerr, William R., and Willian F. Lincoln. 2010. “The Supply Side of Innovation: H1-B Visa Reforms and US Ethnic Invention." Journal of Labor Economics 28(3): 473–508. [16] Munshi, Kaivan. 2003. “Networks in the Modern Economy: Mexican Migrants in the U.S. Labor Market." Quarterly Journal of Economics 118(2): 549–99. [17] OECD/WHO. 2010. "International Migration of Health Workers". Policy Brief February. Available at: http://www.who.int/hrh/resources/oecd-who_policy_brief_en.pdf. [18] Orrenius, Pia, and Zavodny Madeline. 2013. "Does Immigration A¤ect whether U.S. Natives Major in a STEM Field." Working Paper. [19] Peri, Giovanni, and Chad Sparber. 2011. "Highly Educated Immigrants and Native Occupational Choice." Industrial Relations 50(3): 385–411. [20] Roy, A. D. 1951. “Some Thoughts on the Distribution of Earnings.“ Oxford Economic Papers 3(2): 135–146. [21] Stock, J.H., and M. Yogo. 2005. "Testing for Weak Instruments in Linear IV Regression." In D.W.K. Andrews and J.H. Stock, eds. Identi…cation and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg. Cambridge: Cambridge U Press, pp. 80-108. 17

[22] Wozniak, Abigail, and Thomas J. Murray. 2012. “Timing is Everything: Short-Run Population Impacts of Immigration in U.S. Cities." Journal of Urban Economics 72(1): 60–78.

18

160000

0.18

140000

0.16 0.14

120000

0.1 80000 0.08 60000

Share Foreign

0.12 100000

0.06 40000

0.04

20000

0.02 0

0 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Number of RNs who passed the exam

Figure 1. Flow of Nurses by Foreign Status - NCLEX Passers

Native

Foreign

Share Foreign

Source: Authors' own calculations based on the share of foreign RNs and number of RNs who passed the exam. Note: The data is from the National Council Licensure Examination (NCLEX) statistics.

19

Figure 2. Occupational Distribution of Skilled Native Women by Cohort 0.4

Share in Occcupation

0.35 0.3 0.25 0.2 0.15 0.1 0.05 1982

1979

1975

1972

1969

1965

1962

1959

1955

1952

1949

1945

1942

1939

1935

1932

1929

1925

1922

1919

1915

0

Cohort RN

Teacher

Other Health

Doctor/Dentist/Lawyer/MBA

Other Manager

Note. The data is from the Census and ACS and the sample is restricted to native women with at least two years of college. The outcomes for cohorts 19X5 to 19X8 are measured when women are 32 to 35 years old, cohorts 19X9 to 19X1 when they are 29 to 31 and cohorts 19X2 to 19X4 when they are 26 to 28 years old. Note that the dips in the graph for doctors/dentists/lawyers/MBAs for cohorts 19X2 and 19X4 are due to the fact that these cohorts are too young to have graduated from medical or law school. Other health category includes physician assistants, all types of therapists, optometrists, podiatrists, pharmacists and veterinarians. Other manager refers to women in managerial occupations with no graduate degree.

20

Table 1. Share of Foreign-Born Nurses in Nursing Workforce by State 1980 1990 2000

2010

National

0.08

0.09

0.13

0.15

California District of Columbia Nevada New York New Jersey Hawaii Florida Maryland Texas Illinois Washington Georgia Connecticut Arizona Virginia Delaware Massachusetts Rhode Island Alaska New Mexico Utah Michigan Pennsylvania North Carolina Colorado Maine Oregon Minnesota New Hampshire Vermont South Carolina Oklahoma Tennessee Ohio Arkansas South Dakota Kansas Idaho Alabama Indiana Missouri Wisconsin Louisiana Montana Kentucky Nebraska North Dakota Iowa Wyoming Mississippi West Virginia

0.17 0.10 0.12 0.18 0.12 0.15 0.10 0.08 0.09 0.12 0.07 0.03 0.05 0.05 0.05 0.03 0.04 0.03 0.06 0.04 0.04 0.07 0.02 0.03 0.04 0.06 0.07 0.03 0.04 0.04 0.03 0.03 0.02 0.02 0.07 0.02 0.03 0.02 0.02 0.02 0.03 0.02 0.03 0.03 0.01 0.02 0.00 0.01 0.02 0.03 0.01

0.23 0.14 0.08 0.20 0.15 0.20 0.14 0.09 0.10 0.12 0.09 0.06 0.06 0.05 0.08 0.06 0.07 0.05 0.07 0.03 0.03 0.06 0.03 0.03 0.04 0.06 0.06 0.03 0.03 0.05 0.01 0.03 0.02 0.02 0.03 0.03 0.01 0.03 0.02 0.02 0.03 0.02 0.04 0.03 0.01 0.01 0.02 0.01 0.04 0.03 0.02

0.31 0.16 0.18 0.27 0.25 0.25 0.21 0.18 0.16 0.15 0.13 0.09 0.12 0.10 0.09 0.08 0.10 0.05 0.13 0.07 0.08 0.09 0.04 0.06 0.07 0.06 0.09 0.04 0.06 0.06 0.05 0.04 0.04 0.03 0.05 0.02 0.04 0.04 0.03 0.03 0.03 0.02 0.04 0.04 0.02 0.02 0.04 0.02 0.02 0.04 0.02

0.39 0.37 0.32 0.30 0.28 0.26 0.24 0.23 0.19 0.18 0.15 0.13 0.13 0.13 0.12 0.12 0.11 0.10 0.09 0.09 0.08 0.07 0.07 0.06 0.06 0.06 0.06 0.06 0.06 0.06 0.05 0.05 0.05 0.04 0.04 0.04 0.04 0.04 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.02 0.02 0.02 0.02 0.02 0.01

Note: Constructed using Census and ACS data. The sample is restricted to individuals aged 20-70 who reported registered nurse as their occupation.

21

Table 2. Reduced Form Effects of Foreign Educated Flow on the Number of New Native Nurses Dependent Variable: Native Exam Takers / Population s,t (2) (3) (4) (5) (6) I. Linear Effects -0.638*** -0.736*** -0.719*** -0.650*** -1.168** -0.978** [0.135] [0.150] [0.172] [0.226] [0.538] [0.460] (1)

Dependency Variables, 1980*(Foreign Passers/Population)t-j

Dummy Top quintile s, 1980*(Foreign Passers/Population)t-j

-3.001*** [0.513]

II. Quintiles Specification -3.125*** -2.980*** -2.398*** [0.617] [0.624] [0.822]

Dummy Second quintile s, 1980*(Foreign Passers/Population)t-j

-2.091*** [0.580]

-2.271*** [0.698]

-2.237*** [0.706]

Dummy Third quintile s, 1980*(Foreign Passers/Population)t-j

-2.138*** [0.750]

-1.927*** [0.693]

Dummy Fourth quintile s, 1980*(Foreign Passers/Population)t-j

-1.146 [0.687]

Lag (j) State FE Year FE State level time-varying controls HMO Early Adopter X year FE Region X Year FE Excludes Weighted by Population Period No. Observations

-2.180** [0.950]

-3.466** [1.595]

-1.450 [0.944]

-0.335 [0.953]

-2.002 [1.784]

-2.050*** [0.708]

-1.527* [0.905]

-0.890 [0.959]

-2.343 [1.811]

-1.117 [0.694]

-1.167 [0.706]

-0.969 [0.842]

-1.048 [1.014]

-3.235* [1.883]

4 years X X

4 years X X X

4 years X X X X

Yes 1990-2010 1,067

Yes 1990-2010 1,067

Yes 1990-2010 1,067

4 years 4 years 4 years X X X X X X X X X X X X X X X Top 5 states Yes Yes No 1990-2010 1990-2010 1990-2010 1,067 962 1,067

Note: The data is from NCLEX statistics from 1986-2010. The dependency variable (share foreign born in 1980) is normalized to have unit standard deviation before interacting. The regions refer to the 9 regions defined by the Census. The state level time-varying controls are: Lag of a cubic polynomial in state population, lag of share of the population aged 0-19, 20-39, 40-59, lag of share of whites in the population, lag of share of females in professional occupations. HMO Early Adopter is defined as being a top 10 state in the percentage of population enrolled in a HMO in 1994: DC, CA, MA, OR, CO, AR, HI, NY, MD, WI. The top 5 states are: California, Florida, Illinois, New Jersey, and New York. Standard Errors are clustered at the state level. * significant at 10% level, ** at 5% level and *** at 1% level.

22

Table 3. The effects of the flow of foreign nurses on the occupational choice of natives: Evidence from Census Data Dependent variable (measured as natives aged 21-31 in occ/state pop):

Panel A. Explanatory Variable: Foreign Nurses / state population OLS IV (3) (4) (5) (6) (7) (8)

(1)

(2)

Registered Nurses

-0.311*** [0.112]

-0.228* [0.129]

-0.135 [0.137]

-0.094 [0.091]

-0.049 [0.112]

Other Health Occ (excl Physicians/Dentists)

-0.072*** [0.020]

-0.023 [0.025]

-0.033 [0.037]

-0.066* [0.035]

-0.047 [0.047]

-0.045 [0.043]

-0.026 [0.051]

Primary School Teachers

0.378*** [0.100]

0.134 [0.147]

0.238 [0.191]

0.236 [0.154]

0.225 [0.143]

0.724*** [0.204]

0.100 [0.061]

-0.025 [0.040]

-0.056 [0.058]

-0.000 [0.066]

-0.028 [0.066]

-0.445*** [0.137]

-0.026 [0.159]

0.156 [0.226]

0.122 [0.172]

0.326 [0.254]

Secondary School Teachers Managerial Positions

Dependent variable (measured as natives aged 27-37 in occ/state pop):

(9)

(10)

-0.316* [0.166]

-0.168 [0.312]

-0.128 [0.120]

-0.074 [0.080]

-0.122 [0.115]

0.484** [0.232]

0.860** [0.348]

0.629** [0.291]

0.705* [0.424]

0.205*** [0.061]

0.049 [0.070]

0.317 [0.247]

0.194 [0.121]

0.308 [0.245]

-0.328 [0.279]

0.157 [0.185]

0.637 [0.391]

0.118 [0.263]

0.099 [0.440]

(9)

(10)

-0.531*** -0.472*** -0.461** [0.169] [0.108] [0.234]

Panel B. Explanatory Variable: Lag of Foreign Nurses / state population OLS IV (3) (4) (5) (6) (7) (8)

(1)

(2)

Doctors and Dentists

-0.044*** [0.014]

-0.007 [0.032]

0.053 [0.063]

0.021 [0.052]

0.074 [0.062]

-0.006 [0.035]

0.089 [0.084]

1.054 [1.688]

0.186 [0.187]

2.261 [13.300]

Lawyers

-0.068** [0.027]

0.063 [0.067]

0.288 [0.226]

0.115 [0.145]

0.256 [0.197]

-0.051 [0.051]

0.181 [0.131]

1.379 [2.031]

0.396 [0.288]

2.850 [16.632]

X X

X X X

X X X

X X X

X X X

Yes

No

X X X X No

X X

Yes

X X X X Yes

Yes

Yes

No

X X X X Yes

X X X X No

Controls State FE Year FE State level time-varying controls Region X Year FE Weighted by Population

Note: The data is from the 1980, 1990, 2000 US Census and 3-year aggregate 2010 American Community Survey. Number of observations is 204 for regressions in Panel A and 153 for regressions in Panel B. For the 2SLS regressions, the foreign-born nurses/population is instrumented using the predicted foreign-born nurses/pop constructed by using the historical distribution of high-skilled immigrants across cities in 1980 to allocate the national flow of nurses to each state (net of the contribution of each state to the national flow). The state level time-varying controls are: population (cubic polynomial), share of state population aged 25-34, 35-44, 45-54, 55-64, 65+, Log(average hourly wages), physicians per 1000 population, share of females in professional occupations, LFP of married skilled women, log(avg hourly wage of skilled women), share of whites in the population. Standard errors are clustered at the state level. * significant at 10% level, ** at 5% level and *** at 1% level.

23

Table 4. Reduced Form Effects of Foreign Educated Flow on the Quality of New Native Nurses Dep. Var: Share of Native Nurses passing the exam s,t (2) (3) (4) (5) (6) I. Linear Effects 17.762** 16.057** 15.476** 17.842** 38.266*** 30.422*** [6.961] [6.569] [6.916] [7.778] [9.986] [9.798] (1)

Dependency Variables, 1980*(Foreign Passers/Population)t-j

Dummy Top quintile s, 1980*(Foreign Passers/Population)t-j

II. Quintiles Specification 79.419*** 76.827*** 76.247*** 79.838*** 77.498*** 96.324*** [19.009] [18.643] [17.885] [17.301] [14.410] [22.189]

Dummy Second quintile s, 1980*(Foreign Passers/Population)t-j

48.126*** 56.002*** [14.304] [13.815]

55.755*** [11.186]

43.416** [18.605]

Dummy Third quintile s, 1980*(Foreign Passers/Population)t-j

58.274*** 56.899*** [21.374] [19.261]

61.253*** [17.353]

64.240*** 56.047*** 47.509** [16.890] [14.607] [20.388]

39.091** [15.208]

44.280** [20.178]

Dummy Fourth quintile s, 1980*(Foreign Passers/Population)t-j

25.421 [19.581]

24.608 [17.424]

30.708* [16.742]

41.615** [17.860]

Lag (j) State FE Year FE State level time-varying controls HMO Early Adopter X year FE Region X Year FE Excludes Weighted by Population Period No. Observations

4 years X X

4 years X X X

4 years X X X X

4 years 4 years 4 years X X X X X X X X X X X X X X X Top 5 states Yes Yes No 1990-2010 1990-2010 1990-2010 1,067 962 1,067

Yes Yes 1990-2010 1990-2010 1,067 1,067

Yes 1990-2010 1,067

28.884 [18.402]

34.222* [18.300]

Note: The data is from NCLEX statistics from 1986-2010. The dependency variable (share foreign born in 1980) is normalized to have unit standard deviation before interacting. The regions refer to the 9 regions defined by the Census. The state level time-varying controls are: Lag of a cubic polynomial in state population, lag of share of the population aged 0-19, 20-39, 40-59, lag of share of whites in the population, lag of share of females in professional occupations. HMO Early Adopter is defined as being a top 10 state in the percentage of population enrolled in a HMO in 1994: DC, CA, MA, OR, CO, AR, HI, NY, MD, WI. The top 5 states are: California, Florida, Illinois, New Jersey, and New York. Standard errors are clustered at the state level. * significant at 10% level, ** at 5% level and *** at 1% level.

24

Appendix Table 1. Descriptive Statistics NCLEX State-level data: 1990-2010 Mean Std. Dev Min Max Obs. Native Takers

1826.66

1778.80

64

11371

1067

Native Takers *1000 / pop

0.39

0.32

0.03

6.59

1067

Passing Rate Natives

87.84

4.20

60.70

98.26

1067

Note: The passing rate is obtained by dividing the number of US-educated natives who passed the NCLEX exam by the number of natives who sat for the exam in each state. The descriptive statistics are not weighted by population. We also dropped a few state-year observations with obvious coding mistakes.

25

Appendix Table 2. Robustness Checks - Foreign Educated Flows and the Number and Quality of New Native Nurses Dependent Variable: % of Natives passing NCLEX s,t Native Exam Takers / Pop s,t

Dependency Variables,1980*(Foreign Passers/Population)t-j

Lag (j) State FE Year FE State level time-varying controls HMO Early Adopter X Year FE Region X Year FE No. Observations

(1) -0.588*** [0.190]

(2) -0.101 [0.223]

(3) -0.176 [0.222]

(4) 14.022* [8.100]

(5) -2.739 [6.732]

(6) 10.535 [9.171]

4 years

6 years

0 years

4 years

6 years

0 years

X X X X X

X X X X X

X X X X X

X X X X X

X X X X X

X X X X X

965

965

965

965

965

965

Note: The data is from NCLEX statistics from 1986-2010. All specifications are weighted by the state's population. The dependency variable (share foreign born in 1980) is normalized to have unit standard deviation before interacting. The regions refer to the 9 regions defined by the Census. The state level time-varying controls are: Lag of a cubic polynomial in state population, lag of share of the population aged 0-19, 20-39, 40-59, lag of share of whites in the population, lag of share of females in professional occupations. HMO Early Adopter is defined as being a top 10 state in the percentage of population enrolled in a HMO in 1994: DC, CA, MA, OR, CO, AR, HI, NY, MD, WI. Standard errors are clustered at the state level. * significant at 10% level, ** at 5% level and *** at 1% level.

26

Appendix Table 3. First Stage

Instrument

(1) 0.724*** [0.264]

Dep. Variable: Foreign Nurses / Pop. (2) (3) (4) 0.593** 0.224 0.469** [0.231] [0.136] [0.179]

Controls State FE Year FE State level time-varying controls Region X Year FE Weighted by Population

X X

X X X

X X X

Yes

Yes

Cluster-robust F statistic

7.52

6.60

(5) 0.287* [0.151]

No

X X X X Yes

X X X X No

2.71

6.86

3.6

Note: The number of observations is 204. The instrument is the predicted foreign-born nurses/pop constructed by using the historical distribution of high-skilled immigrants across cities in 1980 to allocate the national flow of nurses to each state (net of the contribution of each state to the national flow). See the text for the exact formula. Standard errors are clustered at the state level. * significant at 10% level, ** at 5% level and *** at 1% level.

27

Appendix Table 4: Share Foreign-Born in Different Occupations 1980 1990 2000

2010

RNs

0.08

0.09

0.13

0.15

Other Health Occupations

0.06

0.08

0.12

0.14

Primary School Teachers

0.03

0.04

0.06

0.07

Secondary School Teachers

0.04

0.05

0.06

0.08

Managers

0.07

0.09

0.12

0.15

Doctors and Dentists

0.19

0.19

0.26

0.28

Lawyers

0.03

0.05

0.06

0.08

Note: Constructed using Census and ACS data. The sample is restricted to individuals aged 20-70

28