Skilled Immigration, Innovation and Wages of Native-born American *

Transcription

1 Skilled Immigration, Innovation and Wages of Native-born American * Asadul Islam Monash University Faridul Islam Utah Valley University Chau Nguyen Monash University March 2012 Abstract The paper examines the effects of skilled immigration on wages in the US that are due to innovation. We extend the studies by Hunt and Gauthier-Loiselle (2010), and Hunt (2011) to explore the immigrationinnovation-wages nexus. Using the National Survey of College Graduates (NSCG) and the US Census datasets we find a significant positive effect of immigration on wages that are attributable to immigrants contribution to innovation. Our findings suggest that as the share of immigrants increases in a particular group, the wages of both the natives and the immigrants also get a positive boost. In particular, the effects are more pronounced through immigrants impact on patent granted and patent commercialized, where the results remain robust regardless of the use of different controls. The results also show that the immigrants are more likely to present a paper at a conference or publish in professional journals, primarily because they are more educated or concentrated in the related occupation compared to the natives. Our findings indicate that the immigrants make substantial contribution to the host economy s innovation which is a major driver of economic growth. The results have important implication for immigration policy and economic growth. Key Words: Innovation, immigration, wages JEL codes: J15, J31 * We thank Paul Frijters, Dipanwita Sarkar, seminar participants at University of Queensland, Queensland University of Technology, and Monash University for helpful comments and suggestions. The usual disclaimer applies. Financial support from Australian Research Council and Utah Valley University are acknowledged. 1

2 1. Introduction Does innovation boost wages? What are the impacts of immigration on innovation? In this paper we re-visit the immigration-innovation-wages nexus to address these questions by extending the study by Hunt and Gauthier-Loiselle (2010). Our interest is in examining the wages accruals in the US that are due to immigrants contributions to innovation. In other words, we provide estimates for the contribution to wage increases that can be attributed to immigrantled innovation. In a subsequent paper Hunt (2011) examines how the skilled immigrants perform relative to natives in boosting U.S. productivity. The paper contributes to the literature by providing estimates for wage increases to both the natives and the immigrants which is directly attributable to innovation by the skilled immigrants. The research signifies marked departure from the earlier studies which did not consider this important interaction. Immigrants are defined as persons who were born outside of the US and its territories. Innovation, widely regarded as the driver of technological progress, belongs to the domain of the macroeconomic growth literature, especially in the context of the endogenous growth models. [see Romer (1990); Grossman and Helpman (1991); Aghion and Howitt (1992); Jones (1995b); and Howitt (1999).] The link between innovation and the number of scientists is also central to the idea of production function in these models. Jones (1995a), and Ha and Howitt (2007) have gone to great length to empirically examine the relationship. Although the share of skilled immigrants in the relevant labour market has been increasing over the past few decades, their contributions to innovation and thus to technological progress have received little attention in the literature. Quibria and Islam (2010) have shown that while the short-run impact of immigration on wages generally is negative, the long-run impact may not be so, if innovation effect of immigration is sufficiently strong. Inflow of immigrants who are largely unskilled and bereft of entrepreneurial abilities can have negative impact on native wages. Hunt and Gauthier-Loiselle (2010) point out that immigrants, in addition to directly contributing to research, indirectly contribute to technological progress through positive spillovers on fellow researchers, the achievement of critical mass in specialized research areas, and the provision of complementary skills such as management and entrepreneurship (p.31). However, these positive spillovers can be offset by negative spillovers or due to crowd-out effects, possibly through competition for university admission and research funding. Borjas (2007) finds little evidence of crowd-out effect of foreign students on a typical native student in 2

3 graduate programs, but strong negative correlation between increases in the enrolment of white native students and the enrolment of foreign students, especially at the most elite institutions. Most papers examining the effects of immigration on innovation find positive effect. Results from the state-level analysis by Hunt & Gauthier-Loiselle (2010) show that 1 percentage point increase in the share of immigrant college graduates in the population during the period increased patenting per capita by about 12 per cent. The positive spillovers also account for a significant share of the total impact. Stephan and Levin (2001) use indicators such as, election to sciences and engineering academies, publication and citations in professional journals, and key role in launching biotechnology firms find that immigrants disproportionately contribute to science and engineering sectors in the US. Chellaraj, Maskus, and Mattoo (2008) incorporate skilled immigrants as a component of the idea generation input into the production function concept. Using time series methodology, they find that an increase in the number of foreign graduate students could lead to increases in patent applications, university patent grants and non-university patent grants. Similarly, Stuen, Mobarak and Maskus (2010) find immigrant students add to patenting and publication in the science and engineering journals. Their findings suggest that a 10 per cent decrease in the foreign share of doctoral students reduce production of scientific publications and citations at US universities by around 5-6 per cent. Kerr and Lincoln (2010) examined the effect of fluctuations in H-1B visa entry. They find that by granting more H-1B visas, total invention could be boosted; while would have minimal negative effect, if any, on the natives science and engineering employment and/ or patenting. Peri (2007) found that the impacts of foreign-born PhD degree holders on innovation rate is higher compared to those born in the US. However, this difference could be attributed to the disproportionate employment of immigrants in the R&D sectors; or due to the better quality of scientist and engineer immigrants. Hunt and Gauthier-Loiselle (2010) used individual-level data from the National Survey of College Graduates (NSCG) and extended the state panel. They found that a college graduate immigrant could contribute at least twice as much to patenting as a native counterpart. However, by examining the sample of only scientists and engineers and showing that scientist and engineer immigrants are less likely to patent than their counterparts, they explain the advantage of immigrants in innovating by their higher share among those with education in science and engineering, not because of their better innate ability. In a recent paper He used state panel data, but controlled for state and time fixed effects as well as R&D inputs. 3

4 Hunt (2011) examines the performances of immigrants across different visa types. She found that college graduate immigrants who enter the US on temporary work or student/ trainee visas do better than their native counterparts in wages and patenting; while the reverse is true for immigrants entering as their dependents, or on other temporary visas. In other words, immigrants generally do better, compared to their dependents. As an offshoot of research in labor economics, economists have extensively studied the effects of immigration on the host country s labour market, particularly on natives wage and employment. However, the measured effects vary widely, raising much controversy in research and policy discussions. Most studies using cross-area approach find a small negative or insignificant effect of immigrants on wages (Card, 2001; Card and DiNardo, 2000; Friedberg, 2001; Lewis, 2005). Borjas (2003) criticizes this approach for ignoring the strong currents that tend to equalize economic conditions across cities and regions (p.1336). Using national-level data, some other studies found large and even negative impact of immigrants on the wages of less educated workers (Borjas, 2003; Borjas, Freeman, and Katz, 1997; Borjas and Katz, 2007). Borjas (2003) also examined the variation in immigration supply shifts across educationexperience groups. He found that a 10 per cent increase in such supply reduces average wages by 3-4 per cent. He argues that while there is wide variation in wages across different education groups, the high-school dropouts bear the brunt of the large negative impact. Dustmann, Fabbri, and Preston (2005) find no good evidence which might suggest that immigration affects aggregate employment or wage rates in the U.K.; but the evidence for Germany is mixed. De New and Zimmermann (1994) identify detrimental effects while Pischke and Velling (1997) find no such effect of migration on employment. Islam (2009) finds no displacement of Canadian-born workers caused by immigrants. He finds that the impact of more recent immigrants on the native-born is positive, but older ones are neither substitute nor complement for the natives. Card (2005) in reviewing the recent evidence on U.S. immigration, concludes that there is scant overall evidence to support the claim that immigrants adversely affect natives labour market outcomes. He also finds that immigration-induced changes in the skill composition have little effect on wages. It appears that the empirical evidence of negative effects of immigration is typically generated by aggregate factor proportions analysis or nationwide approaches (Borjas, 2003). Studies based on local labour markets or on natural experiments such as political developments in sending countries (Card, 1990) find no 4

5 significant effect of immigration on the outcomes for natives. His argument is based on the evidence from the Mariel boatlift of 1980 when the instantaneous rise in the fraction of lowskilled workers in the Miami labor force, produced no discernible effect on wages or unemployment rates of less-skilled natives in the city. Card (2001) notes that a 10-percentagepoint increase in the fraction of immigrants reduces native wages by no more than 1 percentage point. Studies that exploit geographic variation using correlation of immigration with changes in native outcomes across cities or regions find less sizable effects (Altonji and Card, 1991) or none at all (Pischke and Velling, 1997). The findings of adverse effects of immigration on wages are consistent with the usual textbook labour supply theory within a static framework where an influx of immigrants (perhaps, less skilled) could lower the relative wages of their counterpart natives. However, it is important to recognize that any increase in the fraction of immigrant population does not necessarily raise the supply of low-skilled labor, rather comes as a mixed bag of high and low skill. Larger the proportion of high skill, more it is likely to trigger an immigration-induced innovation and thus technological progress in the long run. (For theoretical argument, see Quibria and Islam, 2010). Academicians and policymakers have expressed concern over the real or perceived adverse impact of immigration on labor market a central theme in the immigration debate. With the large-scale immigration in recent decades, a sizeable literature has emerged trying to measure these labor market outcomes in the host nations. Extending the national approach of Borjas (2003), Ottaviano and Peri (2008) identify a small but significant degree of imperfect substitutability between native and immigrant workers within the same education-experience group. (p.2). The authors explain this result mostly in terms of imprecise estimates of the elasticity of substitution between workers with at most a high school degree vs. workers without a degree. While the elasticity between workers with some college education or more and those with high school education or less is much smaller (equal to 2), the balanced inflow of immigrants belonging to these two groups implies very small relative wage effects of Levy and Murnane (1992) report that over the past 3 decades, immigration rates into the United States have gone up. Apparently, this outcome has coincides with a fall in the real wages of younger and less-educated workers. He used nested CES production function which allows combining supply shocks affecting workers of different education and experience levels in order to identify wage effects. By extending the so-called national approach. he produced a more plausible estimate of the elasticity of substitution between workers with a high school degree and workers without one which appears quite large, precise and consistent with the labor literature. He identifies a small but significant degree of imperfect substitutability between native and immigrant workers even within the same education-experience group. 5

6 immigration, and a very small negative impact on wages of less educated immigrants. (p.39). The authors argue that Borjas (2003) failure to account for capital adjustment in the short run might have caused such effect. So they grouped workers into highly educated (at least some college education), and less educated (high school education). Card (2001) argues that a growing body of research finds only modest evidence that immigrant competition has hurt the labor market opportunities of low-wage native (p. 22). Butcher and DiNardo (1998) observed that the average immigrant worker is only slightly less skilled than an average native worker. This reflects the fact that much of the immigrants to the US are low skill **, although in many cities, immigrants actually earn higher wages than natives. (See Card, 2001). Borjas (2003) new approach to estimate labor market impact of immigration aims at exploiting this variation in supply shifts across education-experience groups. (p. 1335). He considers workers with similar education but differing levels of experience. They participate in a national labor market and are imperfect substitutes. He writes, Despite the intuitive appeal of this theoretical implication and despite the large number of careful studies in the literature, the existing evidence provides a mixed and confusing set of results. (p. 1335). A classification of workers by education groups based on very high and low skill might have addressed the issue. Chellaraj, Maskus, and Mattoo (2008) argues that skilled immigrants could foster innovation, thereby contribute to future productivity gains and real wages of the natives. Peri in a forthcoming paper finds that immigrants increase total factor productivity growth in the U.S. However, Card (2001), Borjas (2003) or Ottaviano and Peri (2008) did not consider the impact of potential human capital embedded in the skilled immigrants and their contribution in promoting innovation, thus technology and productivity. It is plausible that by ignoring human capital, the authors might have failed to identify the true contribution the immigrants can make. Productivity which measures wages cannot be fully captured unless technological progress is taken in to account. While, historically immigrants, regardless of their background, have contributed to US economy, we argue that some college education (as in Ottaviano and Peri, 2008) is unlikely to impart skill enough to be innovators which is the key to technological progress. If the aim is to quantify contribution of the immigrants to innovation, the appropriate ** Capped at 65,000 H1B visa for science engineering and computer specialists (excludes some university teachers). He uses CES production function method. 6

7 classification of skill should include at least college degree, classified further into science and/or engineering. In our paper we consider the innovative skill, which is in marked departure from the works of the earlier authors. Despite the existence of a sizeable body of empirical literature with focus on the impact of immigration on wages, the foregoing discussion so far, offers mixed evidence. Given the fact that historically the average wages in the US has been rising over the past despite the influx of immigrants in significant numbers, the latter alone is unlikely to be the reason for the negative outcome on wages. One approach might be to examine how the skilled immigrants might have contributed to innovation and thus technological advance. This paper considers immigrants separately based on their skills as would be appropriate to capture innovation capability. The focus is on skilled immigration, defined as those with at least a college degree. We argue that any impact on the wages of native-born workers resulting from these immigrants would come through their contribution to innovation, research and publication. These are the key ingredients of technological progress which help boost in U.S. productivity. It is likely that a supply shift resulting from the influx of unskilled immigration could itself lower the relative wages of their native-born counterpart. It is, however, less likely that a supply shift, resulting from skilled immigration, would crowd-out natives and negatively impact wages; if we consider the complementarities of skilled workers in production in view of their high demand in the U.S. A body of empirical work associated with Borjas and some of his collaborators (e.g., Borjas, George J., Jeffrey Grogger, and Gordon H. Hanson, 2006; Aydemir, Abdurrahman and George J. Borjas, 2007; Borjas, Grogger and Hanson, 2008) implicitly suggest that the native and immigrant workers are perfect substitutes. However, based on a set of recent empirical studies, reviewed by Card (2009), it appears that these two types of workers are far from perfect substitutes, even within the same skill group. This heterogeneity in skill levels is more pronounced at the aggregate level of the economy, something that makes it more realistic to assume imperfect substitutability between native and migrant workers at the economy level. Our empirical strategy relies on the premise that the skilled immigrants contribution to increased wages is attributable to productivity boost from the former s innovation and research. In contrast to the previous research which largely neglected the immigration-innovation-wages nexus, in this paper we provide estimate for wage increases to the natives which can be attributed to innovation by the skilled immigrants. We do this in two stages. In the first-stage, we estimate 7

8 the probability of innovation using share of immigration and other characteristics that support innovation. In the second-stage, we estimate the effects of predicted innovation (obtained from the first-stage) on wages, assuming that skilled immigrants do not impact wages directly, other than through their contribution to innovation and research capacity. Our empirical strategy closely follows Hunt and Gauthier-Loiselle (2010). They show that the advantage of the skilled immigrants lies entirely in the nature of their education, and not in any selection on unobservables, such as ability. Thus, after controlling for skilled immigrants education and other observable characteristics, we expect the unobservables not to directly influence natives wages. We use two different datasets, the national survey of college graduates (NSCG) dataset and the integrated public use microdata samples (IPUMS), from the original sources of the U.S. Decennial Census Our results show significant positive effect of immigrants on wages, channelled through innovation. We find that a 1 percentage-point increase in share of immigration leads to 0.2 to 0.3 per cent average increase in wages and the effects are much more pronounced in samples of post-college graduates. Data also suggest that immigrants are more concentrated in the fields of science and engineering, and they are more likely to publish a book or journal article or present a paper at a professional conference. However, these results might also be due to the fact that immigrants are overrepresented in these areas. As such, it is plausible that the benefits form the immigrants do not flow to the natives working in these specialized areas. The rest of the paper is organized as follows. Section 2 outlines the empirical Strategy adopted in this paper. Section 3 provides a detailed description of the sources of data. Section 4 reports results; section 5 draws conclusion and identifies directions of future research. 2. Empirical Strategy In line with our objective, we concentrate on the relatively educated immigrants as well as their counterpart natives (i.e., holding at least college degrees). We assume that, conditioning on the observable covariates, the impact of skilled immigration on wages is entirely through their impact on innovation and research capacity. In order to take into account the potential spillover effects on fellow researchers, we group immigrants by their education, occupation, experience, 8

9 and region of workplace in the U.S. To estimate the effect of immigrants on wages arising out the former s impact on innovation, we run the following regression: (1) Pr( Innovation ) γ State + e ijklr = α + β IMS jklr + δx + γ1eduij + γ 2Occik + γ 3Expil + 4 ir ijklm The indices, jklr represent individuals with education j (j=1,, 4), occupation k (k=1,.., 6), experience l (l=1,..., 8), and area of workplace (for state of residence) r (r =1,.., 9). Innovation is a dummy variable to capture: (a) whether an individual was granted a patent; (b) whether patent was commercialized; and (c) whether an individual publishes any book or journal article or presented a paper in conference in the last three years prior to X is the demographic variable (age and age squared). IMS is share of immigrants, defined as: M ijklr /(M ijklr +N ijklr ) where M ijklr provides the number of immigrants in cell (j, k, l, r) and N gives the corresponding number for the natives (i is the index for individual). To compute the share of immigrants for purpose of the analysis, we took the share of immigrants in the population in the group of individuals with same highest level of education, work experience, occupation group, and regions of work. Edu is the vector of indicator for the highest degree achieved, Occ is the indicator for occupation, Exp is the vector of dummy for years of experience, State is the workplace of the individual. In the regression, we also added interaction terms for each pair of education, occupation, experience and state to allow for possibility of, for example, the experience profile for a particular labor market and whether it differs across schooling groups (interaction of education and experience). There are four groups of educational attainment: bachelor, master, doctorate and professional. Occupations are categorized into six major groups: computer and mathematical scientists, life and related scientists, physical and related scientists, social and related scientists, engineers, and other occupations. In defining work experience, we follow the effective work experience as in Borjas (2003) in measuring the years of work exposure that are valued in the US labour market. Thus, only individuals who have between 1 and 40 years of experience are included; and have been classified into one of the eight experience groups, each defined in terms of five-year intervals. The principal estimating equation of interest (the 2 nd stage) is specified as follows: (2) log( w ijkl ) = 0 + Γ * Pdtinnoijkl + θx i + λ1edu j + λ2occk + λ3 α Exp + τ l jkl 9

10 Where log(w jklm ) denote log of wages for natives only or both natives and immigrants who have education j, occupation k, experience l. Pdtinno jkl is the predicted value of innovation for individual i, who belongs to education group j, occupation group k, experience group l. We follow the suggestion of Borjas, Freeman and Katz (1997), who argue in favour of estimating wage regression at the national level rather than at the geographic level. As in equation (1), we also try to capture the interactions between each pair of education, occupation, and experience. Our primary interest is in the coefficient Г, which we interpret as measuring the effect of innovation on wages due to immigration. In the actual estimation, we run an OLS regression in both stages using equation (1) as the first-stage, and use the predicted value of innovation from the first-stage to run equation (2). Standard errors are corrected to take into account of the estimated value of innovation in equation (2). We cluster standard error at the occupation, degree and workplace levels to allow arbitrary correlation among different workers. The regressions are weighted by the survey weight provided in the two data sets. Since the distribution of wages of natives and immigrants are very similar we focus on the mean regression (see Figure 1). The distribution also does not differ across different groups, e.g., samples of college graduates, postcollege graduates, and scientists and engineers. 3. Data and Descriptive Statistics Our data come from two sources: the 2003 US National Survey of College Graduates (NSCG), by the National Science Foundation; and the 2000 Integrated Public Use Microdata Series (IPUMS). The NSCG data comprises of a random sample of 100,402 respondents from the 2000 Decennial Census long form. These people were living in the US during the reference week of October 1 st, 2003, had a bachelor s degree or higher, and was under the age of 76. All respondents who had ever worked were asked whether they had applied for/been granted/had commercialized any US patent since October The sample thus does not include patents obtained, if any, by those without a college degree. Our second dataset consists of the 2000 Borjas (2003) argues the unit of analysis should be at the national level to examine the impact of immigration. It is to be noted that in equation (1) we consider the probability of innovation at the local level as by aggregating workers at the national level we lose a lot of variation in labour supply shocks across different regions of residence that could affect innovation. However, we follow Borjas (2003) in equation (2) and allow the possibility that innovation can have impact at the national level. Thus our exclusion restriction is in fact immigration and their workplace (region of residence). We also experiment using probit in the first-stage. The results are reported in appendix Tables A1-A3. As we use a number of fixed effects and their interactions in our regression, probit estimates could not converge when we use interactions of various fixed effects. Therefore, appendix table reports results without the interactions. 10

11 IPUMS data set, created from random sample which is 5 per cent of the population. In this sample we include only those currently living in the US, age 65 years or younger (the youngest is 25) and are participating in the labour force. The regressions of innovation measures and annual and weekly wages are done in the sample of only the individual about whom we had the information: not enrolled in school, have positive annual earnings, weeks worked, and weekly hours; including the self-employed. Innovation measures include three dummy variables to identify whether (a) the individual has any patent granted; (b) the individual has any patent commercialized; and (c) an individual published a book or journal article or presented any paper during the last five years prior to the survey. These information relating to innovation are, however, not available in the IMPUMS data set. We compute innovation for individuals in IPUMS dataset using coefficients from the regressions of innovation in the NSCG data set. Table 1 presents descriptive statistics for the innovation and wage variables computed from both the data sets. Following Hunt and Gauthier-Loiselle (2010), we consider three (not mutually exclusive) skill categories: those with at least a college degree, post-college graduates, and those working in the science and engineering sectors. Average annual and weekly wages show little differences between natives and immigrants in all samples. In both data sets, the share of immigrants in the population is higher in the scientist and engineer sample, compared to the sample of college graduates and sample of post-college graduates. In all three samples in the NSCG data set, immigrants appear to be more likely to innovate than their native counterparts, with immigrants having a higher average chance of getting a patent granted, and a patent commercialized. The gap in innovation probability is the largest in the post-college graduate sample while it is the smallest in the sample of scientists and engineers. Of foreign-born postcollege graduates, 5.9 per cent have a patent granted more than double the 2.2 per cent by their native counterparts. The gap is much smaller in the sample of scientists and engineers, with 8 per cent of immigrants having a patent granted, compared with the 5.9 per cent of the native counterparts. Predicted innovation measures for the IPUMS sample as presented, show immigrants higher innovation probability in all three categories. In terms of proportions, more immigrants are likely to have a paper published, or presented at a conference/seminar during the last five years prior to the survey (Table 1). The gap between immigration and natives is the largest for post-college graduates: 35.4 per cent among immigrants as opposed to 28.8 per cent 11

12 among natives who have post-graduate degree and are publishing or presenting a paper at a conference during the last 5 years prior to As these innovation variables are not available in the census dataset, we computed the predicted innovation coefficient from the NSCG dataset and then applied them to the census dataset. Incidentally the predicted innovation variables for the census dataset are very close to and consistent with those observed in the NSCG dataset. 4. Results We now report the estimated results obtained from regression. The first stage estimates from equation (1) reported in Table 2, show the probability of innovation using samples of both natives and immigrants from the NSCG dataset. The estimates capture the effect of the share of immigrants on both the probability of having a patent, (a) granted; and (b) commercialized. The coefficients corresponding to immigration share variable, IMS, are positive and significant in both the samples of college graduates and sample of post-college graduates. Consider the sample of college graduates for which the coefficient of immigration share is with a standard error of for the whole sample. We interpret this coefficient as the impact of the increase in labor supply due to immigration on innovation probability. The point estimates suggest that a 1 percentage point increase in the influx of college graduate immigrants is expected to increase the probability of patent granted for an individual in that group by percentage-points. For the post-college graduates, the corresponding increase in innovation would be percentagepoints (Table 2, row 2). We also report separate results for the sample of only the native-born using equation (1). We consider the probability of innovation for natives using share of immigrants in each group by taking the highest education, occupation, experience and state of residence. The results, reported in row 3 (Table 2), suggest positive spillover effects of immigrants on natives innovation capability. The coefficients of immigration share are, in general, smaller than those obtained from both the samples of immigrants and natives. This suggests that the effects of immigrants on natives innovation are smaller than those on the immigrants own innovation probability. However, the difference is too small to carry out any statistical significance in the coefficients. Now we use the predicted innovation for each group, obtained from the first-stage, (i.e., regression equation-1) and use those to estimate equation (2). We find that the effects of innovation on individual annual and weekly wage are positive and significant for samples of 12

13 college graduates and post-college graduates (Table 2). A college graduate with a 1 percentage point higher probability to have a patent granted is expected to cause wages increase by about 3-6 per cent. As a result, in the sample of all college graduates, a 1 percentage -point increase in the share of college graduate immigrants translates to about 0.3 per cent increase in both annual and weekly wage for individuals in that group (column 1). Put differently, a one standard deviation rise (a change of percentage-points) in (predicted) patent granted would result in percent increase in wages. Similarly, in the sample of only post-college graduates, a 1 percentage-point increase in the share of immigrants in one group contributes to a 0.3 per cent increase in annual wage. In this case, a one-standard deviation increase in patent granted would lead to percent increase in wages (column 2). The addition of interaction terms leaves the results very similar for the two samples. For example, the results reported in the right-hand bottom panel of Table 2 suggest that a 1 percentage-point influx of immigrants leads to expected increase of annual wages by about 0.14 per cent and 0.17 per cent for college graduates and postcollege graduates, respectively. However, the overall results do not change significantly when interactions of education, occupation and experience are added. The results show that the effects of an increase in the share of immigrants on the probability of having patent granted are positive, but smaller and statistically insignificant for the sample of scientists and engineers. This could be explained, in part, by the relatively smaller sample of scientists and engineers. However, this could also be due to the fact that the share of immigrants in the sample of scientists and engineers is higher than in the whole sample of college graduates; and also a majority of patents are owned by those working in science and engineering sectors. This result is consistent with the findings by Hunt and Gauthier-Loiselle (2010). However, the results using the effect of predicted innovation on wages are positive for both samples, i.e., the natives only, and the combined sample. This suggests that even though the effects of immigration on innovation is modest perhaps, because of higher concentration of immigrants among scientists and engineers, any such effect translates into a positive impact on wages for both natives and immigrants. Ignoring the statistical significance of the first-stage, the coefficient in the wage regression suggests that a 1 per cent influx of immigrants in the fields of science and engineering result in per cent increase in wage of the natives. These results stay same regardless of whether we consider interactions of education, occupation and experience in both stages of regressions. 13

14 We now report the results based on the 2000 census dataset. As noted, the census data do not have any information on patent granted or commercialized. So, applying the NSCG dataset to equation (1) we estimate the coefficients to predict the innovation probability for individuals from the census data. It may be noted that the sample frame for the 2003 NSCG survey was drawn from the 2000 Census. As reflected in Table 1, we also find similar share of immigrants and native-born for the three groups of sample. The predicted innovation derived from the census dataset is also very similar for all the three skill groups that we consider here. In the second stage, we find that the contribution of innovation capability measured either by probability of having a patent granted or commercialized, to wages is positive and highly significant in most cases. The contribution is however, smaller in the sample of scientists and engineers, compared to the whole sample and the sample of post-college graduates. This again demonstrates that part of the positive spillover effect from the immigrants is the result of their disproportionate concentration in science and engineering occupations. The effects of predicted innovation on wages are also higher in the sample of only native college graduates and native post-college graduates, compared to the sample of all college graduates and all post-college graduates, respectively. The reverse is also true when applied to the sample of scientists and engineers. Overall, the effects of innovation that are attributable to immigration are also very similar when we apply the coefficients of predicted innovation from the NSCG sample to the IPUMS sample. For example, for the sample of post-college graduates, a 1 percentage-point increase in share of immigrants is expected to increase weekly wages of the natives alone by 0.10 per cent (column 5). The corresponding estimates from the NSCG sample show an increase of about 0.07 per cent. Overall, we find that the coefficients of predicted innovation in IPUMS dataset are very close to those obtained from the NSCG dataset in terms of both magnitudes and statistical significance. The estimated effects using IPUMS dataset are at times lower and occasionally higher than those found from the NSCG dataset. Visual inspection of the regression coefficients obtained from the two datasets suggests no significant difference. Table 3 reports the results for patent commercialized. The results are very similar to those reported in Table 2. Column 1 of Table 2 indicates that a 1 percentage-point rise in immigrant share leads to percentage-point, and percentage-point increase in the probability of patent commercialized among samples of college graduates and post-college graduates, respectively. The estimates are statistically significant for the sample of scientist and engineers 14

15 when we use interactions of education, occupation, experience and place of work. As for the results of this group, a 1 percentage-point increase in share of immigrants of scientists and engineers leads to 0.04 percentage-point expected increase in patent commercialized. This in turn raises wage by about 0.15 per cent for the NSCG sample; and by about per cent for the IPUMS sample. The results indicate that a one standard deviation (0.0337) of patent commercialized increases the annual average wage by about 0.10 per cent. Comparing the results between native-only sample and the whole sample which also includes immigrants, we find a slightly lower magnitude for the coefficients of immigration share in innovation for native-only sample. However, the resulting effect of innovation on wages is opposite. When using nativeonly sample, we get a higher effects of innovation on wages that are attributable to immigration share in a particular group defined by education, occupation and experience. We now consider the effects of immigration on professional work such as, writing books or papers for publication or presenting at conference and their effect on wages. As reported in the descriptive statistics (Table 1) all three samples college, post-college graduates and scientists and engineers have higher likelihood to publish or present a paper in a conference. Table 4 reports results using the share of immigrants in a given education, occupation and experience group after controlling for other covariates. The results indicate insignificant or sometimes even negative coefficient of immigration share on publishing or presenting. In this case, it is difficult to interpret the coefficient of the effect of innovation on wages as the relevant coefficients are statistically not significant. In appendix Table A4, we report results using various controls. The table shows that increase in immigration share in one group is likely to increase both patent and publication when considered without any control (column 1). However, when we control for education, as in column (3), the positive effect disappears and coefficient becomes insignificant. In fact, for samples of scientists and engineers we find that an increase in share of immigration in that group is likely to lower the publication and presentation. The results remain qualitatively similar when we add controls such as occupation, experience and work place. Overall, the results are not surprising as the people who are publishing and presenting are more educated; and once we control for education, the positive effects of immigrants disappear. The conclusion is that immigrants are publishing more; and the outcome is the result of their higher concentration in the sector. 15

16 We also report results of reduced form coefficient estimates the effects of immigration on wages. Table 5 displays the results of the effects of immigration share on wages obtained from NSCG dataset under different controls regimes. The results in column (2), with just age and age-square as controls, show substantial positive effects of immigration share in any group on wages for both natives as well as and the whole samples. Overall, inclusion of age and agesquare in the regression does not change alter the magnitude of the coefficients much. The results remain robust for samples even for the post-graduates when we add education in column (3). The relevant coefficients however, become insignificant for samples of college graduates when controlled for experience, occupation and work place. The results in column 4 for scientists and engineers suggest that a 1 per cent-point increase in share of immigrants in that group is expected to raise annual wages of the natives by 0.35 per cent, conditioning on all controls. Note that these reduced form results neither contradict nor support the results presented earlier. There is no oneto-one relationship between the reduced form estimates we presented in Table 5 and those in Tables 2-4. *** Results from the IPUMS sample are reported in Table 6. The results in column (1) show that an increase in the share of immigrants among scientists and engineers does not increase the wages of natives when we do not control for any covariate. Thus, it may appear that scientists and engineers have no influence on natives wages. However, when we control for age, we find a large influence of their share in the wages of both natives and immigrants, or the natives alone. Immigrants with similar age can make large positive effects on wages of their counterpart natives. The results remain same if we add education as control, but the magnitude reduces to one-third or less when experience, occupation, and work place are added. The later results remain significant in all but one case. The results for the college and post-college graduates are statistically insignificant when we apply full controls. The results in column (4) for samples of college graduates imply that there is no effect or insignificant, but negative effect of immigration on wages of the natives. Since the sample of college graduates also includes other two groups, the results indicate that the estimates from census data show no positive effect of immigration on wages. The result is in line with the findings of Borjas (2003) when he considers only sample of *** Generally, there need not be a relationship between the significance of the reduced form and the 2SLS estimates. However, we need a strong first-stage to ensure that we are not using a weak instrument. The standard IV/2SLS estimator, (z x) -1 z y, with dependent variable (regressor x), and instrument z, breaks down when z x is near singular but does not if z y approaches zero. 16

17 college graduates. Taken together, the results from both NSCG and IPUMS datasets suggest that, overall there is no significant adverse effect of immigration on wages of native-born once we control for full set of covariates; although there are positive effects when we consider sample of post-college graduates and scientist and engineers. These skilled immigrants play a key role in innovation and thus to technological progress which are the main players in determining productivity, economic growth and thus wages. 5. Robustness of Results: Using Alternative Estimates Our estimates above rely on an important assumption that skilled immigrants do not directly affect wages other than their influences through innovation in terms of patent granted, patent commercialized or publication in academic journal or presentation in conferences. This is indeed the exclusion restriction assumed in the instrumental variable estimation method. Unfortunately, the exclusion restriction is not testable. In our context, the exclusion restriction implies that the coefficient to immigration share in the wage equation (e.g., equation (2)) is equal to zero. Note that the reduced form estimates we provide in Tables 5-6 do not indicate any direct effects of immigration, as the effects of immigration on wages could run through innovation channels. We attempt to test the sensitivity of results using a recently developed method proposed by Conley et al. (2012). Instead of imposing the exclusion restriction, Conley et al. proceed with the assumption that the instrument satisfies the exclusion restriction approximately. That is, the instrument may have non-zero but very small effects in the structural equation. Under the approximate exclusion restriction, the instrument is plausibly exogenous in the terminology of Conley et al. Under this weaker exogeneity condition, they show that one can estimate bounds on the causal effect of the endogenous variable. Their method provides tools to applied researchers working with than perfectly valid instruments. We apply this method in our context to conduct inference while relaxing the exclusion restriction. Following this procedure, we relax the exclusion restriction, assuming that the immigrant share can be correlated with the unobservables influencing the wage of natives. Using the local to zero approximation method developed by Conley et al (2012), we estimate the possible range for the effect of innovation on native wages (the coefficient of interest in the second-stage equation), based on a likely value range of Ω. There is no consensus 17

18 in the literature on the direct effect of change in immigration supply on native wages, varying from a large adverse effect (Borjas (2003)) to a small positive effect (Ottaviano & Peri (2008)). According to Borjas (2003), a 6%-points increase (estimated by averaging the immigrant shares of all experience sub-groups in the college graduate group) in the share of immigrants in the male working population of college graduates during led to a decrease in the logarithm wage of natives in the same group. This implies a value of -0.8 for the parameter Ω. Using this value as a benchmark, we calculated the following local-to-zero estimate for the effect of innovation on native wages in the group of college graduates in the NSCG data set. Overall our results do not change using different values of Ω in the neighbourhood of estimates obtained by Borjas. Obviously, the coefficient estimates of effects of innovation on wages changes for different values of Ω, as expected. The results indicate that our point estimates support the values of Ω that lies between 0 and Conclusion: In this paper, we examine the impact of innovation, measured by patents, as a possible channel through which immigrants could positively affect wage level in the host country. Using two data sets from the US, we find significant contribution of immigrants to innovation propensity, which appear to impact increases in wages for both natives and immigrants. A 1 percentage-point increase in the share of immigrants in each group of individuals with same education level, work experience, occupation, and work place, is expected to raise the probability of having a patent granted for the individual in that group by percentage-point. This increase in innovation probability in turn contributes to increase in annual or weekly wage by about 0.2 to 0.3 per cent. In most cases, the overall impact of immigrants on wages is larger in the sample of post-college graduates only, compared with the whole sample of college graduates. The results are similar when we replace the probability of having a patent granted with the probability of having a patent commercialized. Our results also confirm the earlier findings on immigration and innovation nexus, which found positive impact of immigrants on innovation. The findings suggest that the disproportionate concentration of immigrants in science and engineering occupations is at the heart of the outcome. However, any possible positive effect on innovation in the sample of 18

19 scientists and engineers is helpful for technological progress which translates into improvement in productivity, and to wage levels in the group; and perhaps to others as well. Besides the two innovation measures, we also examined the effects of immigration on wage through the likelihood of publication and conference presentation. However, we failed to find evidence in support of positive effect of immigrants on the probability to publish or present papers. There is a lack of clear relationship between this probability and wage determination. The probability of research and publishing does not alter the outcome in the labor market unless translated into real technological progress and thereby into productivity gain. Future research aimed at identifying research and publication in particular area and how they translate to specific innovation and thus lead to technological development may help establish a link. Although the paper does not aim at finding direct link between innovation and economic growth, the positive relationships found among immigration, innovation, and wages implicitly suggest important contributions immigrants make to the host country s economic performance, given the generally established role of innovation to growth. Indeed, innovation is the logical prior to technological progress. Government policies aimed at addressing the immigration issue thus should give due consideration to the potential contribution immigrants make through innovation when comparing the costs and benefits of immigration into their respective countries. Also, additional measures to encourage positive spill-over effects from immigrants to natives innovation propensity may boost the whole country s innovative capacity. 19