Evidence From California

By:Giovanni Peri

يعمل  بروفسر” جوفيان بيري” في الوقت الحالي استاذاً لعلم الاقتصاد في جامعة كاليفورنيا. عمل محاضراً في نفس المجال في عديدٍ من الجامعات في ايطاليا و الولايات المتحدة الامريكية. كتب و نشر العديد من البحوث و الاوراق فاقت ثمانية و ثلاثين عملاً نشرتها دوريات و فصليات و مجلات علمية مرموقة.
نال بروفسر “بيري” عدداً من الجوائز في

حقل الاقتصاد، و كذلك حصل علي العديد من المنح البحثية في اوربا و الولايات المتحدة الامريكية، مثل جائزة الزمالة العالمية يالمعهد العالمي بجامعة كاليفورنيا في لوس انجلس، و جائزة زمالة “جين مونيه” من الجامعة الاوربية بفلورنس  في ايطاليا، و غيرها.
اشرف بروفسر” بيري” علي جملة من  الرسائل العلمية في الدراسات العليا. هناك عددٌ من طلابه يشغلون مناصب رفيعة في مجالات الاقتصاد و الادارة و في مناطق مختلفة من انحاء العالم.
اتصلت ادارة “سودانيز ايكونومست” ببروفسر “بيري” طالبة منه نشر اعماله علي صفحات الموقع، و كان ان وافق و بحرارة علي التعاون ، بل انه سمح لموقع “سودانيز أيكونومست” بنشر آخر اوراقه و التي كان ان  انهى كتابتها للتو في مايو 2007م.
فباسم موقع “سودانيز ايكونومست”  وباسم قرائه تتقدم الادارة بجزيل الشكر للبروفسر “جوفيان بيري”، و تامل الادارة في ان تتمكن من ترجمة بعضاً من كتاباته الي العربية، وخاصة تلك التي تتعلق يالنشاط الاقتصادي للمهاجرين و دورهم.

I am grateful to Hans Johnson, Jed Kolko, Ethan Lewis, David Neumark, Steven Raphael, Deborah Reed and participants to several seminars for helpful comments. Benjamin Mandel provided extremely competent assistance in editing the paper and very valuable comments. The views expressed herein are those of the author(s) and do not necessarily reflect the views of the National Bureau of Economic Research. © 2007 by Giovanni Peri. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. 

ABSTRACT
As of 2004 California employed almost 30% of all foreign born workers in the U.S. and was the state with the largest percentage of immigrants in the labor force. It received a very large number of uneducated immigrants so that two thirds of workers with no schooling degree in California were foreign-born in 2004. If immigration harms the labor opportunities of natives, especially the least skilled ones, California was the place where these effects should have been particularly strong. But is it possible that immigrants
raised the demand for California’s native workers, rather than harming it? After all immigrants have different skills and tend to work in different occupations then natives and hence they may raise productivity and the demand for complementary production tasks and skills. We consider workers of different education and age as imperfectly substitutable in production and we exploit differences in immigration across
these groups to infer their impact on US natives. In order to isolate the “supply-driven” variation of immigrants across skills and to identify the labor market responses of natives we use a novel instrumental variable strategy. Our estimates use migration by skill group to other U.S. states as instrument for migration to California. Migratory flows to other states, in fact, share the same “push” factors as those to California but clearly are not affected by the California-specific “pull” factors. We find that between 1960 and 2004 immigration did not produce a negative migratory response from natives. To the contrary,
as immigrants were imperfect substitutes for natives with similar education and age we find that they stimulated, rather than harmed, the demand and wages of most U.S. native workers.
Giovanni Peri
Department of Economics
University of California, Davis
One Shields Avenue
Davis, CA 95616
and NBER
gperi@ucdavis.edu 

1 Introduction
In the year 2004, California was home to almost 30% of all foreign-born individuals working in the U.S.; in turn, these foreign-born represented roughly one third of the almost 15 million workers employed in California, two thirds of California workers without a high school degree and almost half of the California workers with a doctoral degree. Many U.S.-born Californians moved out of the state during the nineties and job competition from immigrants has sometimes been regarded as a key factor for this outflow. It stands to reason that if recent inflows of immigrants indeed crowded out the labor market options of U.S. natives, specifically the low skilled ones, then such an effect should have been particularly strong in California. But is it possible that immigrants actually lifted California’s wages, rather than harming them? After all, immigrants have different skills and tend to work in different occupations than natives; they could make natives performing complementary production tasks more productive, thus increasing the demand for those tasks! The present study analyzes the effect of the migratory inflow on the employment, population and wages of U.S. natives in California, using data from the decennial U.S. Censuses and from the American Community Survey spanning the period 1960-2004. Our approach combines elements of a ”general equilibrium” (more structural) approach to immigration and wages, as proposed in recent national studies (Borjas 2003; Aydemir and Borjas 2007; Ottaviano and Peri, 2006) with the study of employment and inter-state migratory response of native workers typical of the so called ”area approach” (exemplified in Card 1990, 2001, Lewis, 2005 and Borjas, 2006). As in Borjas (2003) we consider labor as a differentiated input in production and we model the interactions between workers with different education and age using a nested CES production function. As in Ottaviano and Peri (2006) we allow for imperfect substitution between native and foreign-born workers within an education-age group (due to differences in skills, occupational choices and job opportunities) and we estimate the elasticity of substitution between natives and immigrants. The degree of substitutability between these two groups is a key parameter to determine whether immigrants increase or depress the demand for native workers. For a large degree of substitutability between natives and immigrants, uneducated immigrants mostly depress the demand for uneducated natives and augment the demand for highly educated natives. However, for a smaller degree of substitutability between natives and immigrants, uneducated immigrants have a much smaller depressive effect on (and may even increase) the demand for uneducated natives while still increasing the demand for more educated natives. As pointed out by the critics (e.g. Borjas et al, 1997) of the ”area approach” when focussing on a state economy it is important to account for the fact that any labor market effect of immigration can be ”diffused” to other states by out-migration of native workers. We carefully account for this effect in our empirical analysis. Moreover, by focussing on California over time, we are able to use a new identification strategy that addresses the problems of unobserved demand shocks and measurement errors; these are often deemed responsible for biased estimates of the impact of immigration on the labor market outcomes of natives in a state (e.g. Aydemir and Borjas, 2006; Borjas, 2006). Specifically, we use immigration to the other U.S. states by skill group over decades as a reasonable instrument to proxy the ”push”-driven component of immigration to California by skill group and decade. While sharing the ”push” factors behind international migrations with California, the flows of foreign-born workers to other U.S. states are not affected by California’s specific pull factors (i.e. unobserved demand shocks). The intuition of such identifying assumption is as follows. Suppose that immigrants with a college degree were “pulled” to California in the 90’s by the boom in the high tech sector, which increased the demand for workers with their qualifications. This would be a “pull factor” specific to California. The same boom probably would also have attracted (or reduced the potential outflow of) natives in the same skill group. It could thus create a positive correlation between foreign immigration and natives’ migration and wages–even if foreign migrants compete with natives for the same jobs. Such pull, however, would not be shared by other states and hence the instrument would not capture it. On the other hand take the cases of the increased international mobility of the college-educated Chinese middle-class or the worsened job outlook for young uneducated workers in Mexico during the nineties. Both are “push factors” that could increase immigration of some age-education groups to California (a large receiver of Chinese and Mexican migrants) as well as to other states. Push factors generate more migrants to California as well as to other states and are not related to changes in California’s local demand for labor. Thus, how native employment responded to those immigration changes would correctly estimate how immigrants affect natives’ employment opportunities–for a given local demand. The second purpose served by our instrument is to reduce the measurement error bias. As the measure of immigrants to other states is based on large national samples (excluding California), the instrument is also largely exempt from the measurement error affecting state-level measures of immigration by skill due to the potentially small size of cells1. Hence, we have an instrument for the inflow of immigrants that is potentially uncorrelated (or weakly correlated) with California-specific labor demand shocks and the measurement errors while still correlated with the supply-driven shocks to the composition of immigrants to California. This instrumental variables strategy that we adopt is novel for the ”area” approach and it is an improvement on the ”national” approach (e.g. Borjas, 2003) in that, while we cannot rely on some natural experiment, we have a more sophisticated way of isolating supply-driven variations of immigrants going beyond the simple inclusion of education-age, education-year and age-year specific effects (which we still include). The rest of the paper is organized as follows: Section 2 presents the data and shows some statistics on the skills of foreign-born and recent immigrants to California. Section 3 presents the production function used as framework to estimate the effect of immigration on wages. The skill-structure defined in the production function is used in the empirical estimations. Section 4 presents the identifying strategy and shows the migration and employment responses of California’s native workers to immigration for the period 1960-2004. Section 5 1Recent work by Aydemir and Borjas (2006) analyzes the role of measurement error in generating potential bias in the estimates of the impact of immigrants based on local data. estimates the substitutability between U.S. and foreign born workers within education-age groups. Section 6 uses the estimated parameters to calculate the effects of immigrants on wages of natives (by education) for the 1990-2004 period. Section 7 concludes. 2 Immigration to California: A Look at the Data The data we use are from the integrated public use microdata samples (IPUMS) of the U.S. decennial Census and of the American Community Survey (Ruggles et al, 2005). In particular, we use the general (1%) sample for Census 1960, the 1% state sample, Form 1, for Census 1970, the 5% state sample for the Censuses 1980 and 1990, the 5% Census sample for year 2000 and the 1/239 American Community Survey (ACS) Sample for the year 2004. As those are all weighted samples we use the variable “personal weight” to construct all the average and aggregate statistics relative to California. We consider people aged 17 to 66 not living in group quarters, and we included them among the workers if they worked at least one week in the previous year and earned a positive amount in salary income. When using wage data, we converted the current wages to constant wages (in 2000 U.S. $) using the CPI (Consumer Price Index)-based deflator across years. We define the four schooling groups using the variable that identifies the highest grade attended (called “HIGRADEG” in IPUMS) for census 1960 to 1980 while we use the categorical variable (called “edu99” in IPUMS) for censuses 1990 and 2000 and ACS 2004. Age groups are identified using the variable “AGE”. Finally, yearly wages are based on the variable for salary and income wage (called “INCWAGE” in IPUMS). Weekly wages are obtained dividing that value by the number of weeks worked2. The status of “foreign-born” is given to those workers whose place of birth (variable “BPL”) is not within the USA (or its territories overseas) and did not have U.S. citizenship at birth (variable “CITIZEN”)3 Table 1 illustrates the evolution of the percentage of foreign born in employment and population for the period 1960-2004. Employment is defined as the sum of individuals aged 17 to 66 years old, not residing in group quarters, who worked at least one week in the previous year. Population is defined as the sum of individual between 17 and 66 years of age not residing in group quarters. We report the figures for California as compared to the corresponding figures for the whole U.S.A. First, note that in California the percentages of foreign-born already began growing in the sixties, while in the U.S. as a whole they only began to grow during the seventies. During the seventies, eighties and nineties, California experienced increases in the share of foreign-born workers by about 7% in each decade, with similar rates continuing after the year 2000; in the whole U.S., the increases were far more modest over that period, amounting to 2-3% per year. Second, over the entire period considered,
2For Census 1960 and 1970 only a categorical variable that measures weeks worked exists, called ”WKSWRK2”. Individuals are assigned the middle value of the interval in the variable. 3The variable CITIZEN is not available in Census 1960. For that year we consider all people born outside the U.S. as foreignborn born. the share of foreign-born workers in employment was larger than their share of population, denoting higher
employment/population ratios of immigrants relative to natives; this was due in part to their age distribution.
Finally, notice that the percentage of immigrants in California’s population and employment as of 1980 is similar to the percentage of immigrants in population and employment for the nation in 2004. In terms of percentages, a continuation of the present trend would imply that the future of the nation may look like the last 25 years of California’s experience. Moreover, in addition to their percentages of population and employment, the distribution of foreign-born across education levels in California in 1980 is similar to that for the whole nation in 2004. Table 2 shows the percentage of foreign-born workers in California by education group between 1960 and 2004. Notice the higher concentration of foreign-born among the low (less than high school) and high (college or more) schooling levels and the lower concentration in the intermediate education levels. Notice that, as of 2004, two thirds of high school dropout workers in California were immigrants as well as 42% of Ph.D.’s, while only 20% of workers with some college education but no degree were not U.S. born. The distribution of immigrants, predominantly at the two ends of the schooling spectrum, will be dubbed ”U-shaped”. This U-shape is also a feature of the national data; Figure 1 shows the percentage of foreign-born workers by education level in 2004, comparing California and the whole U.S. One clearly notices the same qualitative U-shaped distribution, but each bar is much higher for California, denoting a higher average percentage of foreign-born. We need to go back to the year 1980 (see Table 2) to find percentages of foreign born across education groups in California similar to the ones for the U.S. today: back then, in California, 32% of high school dropouts, 12% of high school graduates, 10% of college
dropouts4 and 14% of college graduates were foreign-born. The numbers presented above are relative to the ”stock” of foreign-born living in California (or nationally).
The more recent flow of immigrants to California and to the U.S. (1990-2004) also has a similar distribution over schooling. Figure 2 shows the net growth in employment due to immigrants as percentage of initial employment by education group for California (light shaded columns) and for the U.S. as a whole (dark shaded columns). College graduates, Masters and Ph.D.’s flowed in much larger percentages of initial employment in the group
than college dropouts both in California and the whole U.S. At the opposite end of the schooling spectrum, the inflow of high school dropouts was much larger as a percentage of the group than inflow of high school graduates. Overall, aggregating across groups, immigrants to California during the 1990-2004 period equaled 20% of its total employment in 1990, while in the U.S. as a whole they were only 11% of initial employment. These aggregate numbers provide a good sense of how large immigration has been in California. The focus of this paper is the effect of immigration on the labor market outcomes of Californian workers. It is useful, therefore, to start by showing the behavior of native worker real wages during the most recent fourteen 4For brevity, and somewhat improperly, we often use the term ”College Dropouts” to indicate those people who have some College education but not a four-year degree. years of data (1990-2004); these years correspond to the period of largest immigration flows. Figure 3 shows the
percentage change in real wage for native workers, by education group, in California (light shaded columns) and in the U.S. as a whole (dark shaded columns) for the 1990-2004 period. We use real weekly wages calculated as yearly wage and salary income divided by weeks worked, and convert to constant 2000 dollars by dividing for the Consumer Price Index deflator. First of all, we notice that the real wage changes across education groups are very similar in California compared to the whole nation; the difference in real wage growth between California and the rest of the nation was never larger than 4% in any group. In terms of magnitude across education groups, high school dropouts’ wages decreased in real terms over the period by as much as 17%, real wages
of high school graduates were rather stationary, while real wages of college graduates and above experienced a substantial increase, generally above 20%. Aggregating across groups, the average real wage grew by 10.7% in California and by 9.7% in the U.S. as a whole, again denoting a similar performance (less than 0.1% difference in growth per year), with no apparent wage “penalty” at all for the high-immigration state of California.
California’s closeness to the national average in terms of wage dynamics over the last 15 years denotes substantial integration of the Californian labor market with the rest of the U.S. Implied are small costs of moving that arbitrage away large differences in wage changes across states. The poor performance of uneducated worker real wages, contributing to an increase in wage dispersion and income inequality, has certainly been a thorny issue in California as well as in the rest of the nation. The question is whether it was immigration flows that contributed to these real wage changes in California and the U.S. and by how much. 

3 The Framework: Production Function and Imperfect Substitutability
To evaluate the effects of immigrants on the wages and employment of native workers in California we use a framework similar to Ottaviano and Peri (2006). Workers differ by their education and age; different types of workers and physical capital are combined in a production function to produce output. The marginal productivity (wage) of each group depends on skill-specific technology and the supply of each group is affected by immigration. We extend that framework to allow for changes in the labor supply of natives (via migration to/from other states and in/out of employment) in response to immigration, and we estimate these responses and corresponding wage elasticities maintaining the same groupings by skill in the production function. Then, we use the estimated responses and the estimated wage elasticities to calculate the overall effect of immigration on the wages of U.S. natives in California 

3.1 Production Function
Following previous work with Gianmarco Ottaviano (Ottaviano and Peri, 2006) that, in turn, builds on Borjas (2003), we represent output in California as produced by physical capital and different types of labor. Labor types are grouped according to education and age and combined in a constant elasticity of substitution (CES) aggregate; age groups are nested within educational groups that are themselves nested into the labor composite
Lt. U.S.-born and foreign-born workers are allowed a further degree of imperfect substitutability even when they have the same education and age. More specifically, the aggregate production function is given by the
following expression:  

where Yt is aggregate output , At is total factor productivity (TFP), Kt is physical capital, Lt is a CES aggregateof different types of labor (described below), and α ∈ (0, 1) is the income share of labor. All variables reflect data for the state of California in year t. The labor aggregate Lt is defined as: 

where Lkt is an aggregate measure of workers with educational level k in year t; θkt are education-specific productivity levels (standardized so that Pk θkt = 1 and thus any common multiplying factor can be absorbed in the TFP term At). We group educational achievements into four categories: high school dropouts (denoted as HSD) , high school graduates (HSG), college dropouts (COD) and college graduates (COG), so that k = {HSD, HSG, COD, COG}. The parameter δ > 0 measures the elasticity of substitution between workers with different educational achievements. Within each educational group we allow workers with different experience levels to be imperfect substitutes. In particular, following the specification used in Card and Lemieux (2001),
we write: 

where j is an index spanning age intervals of ten years between 17 and 66, so that j = 1 captures workers 17-26 years old , j = 2 captures those who are 27-36 years old, and so on. Within an education group age groups are identical to groups based on years of experience and sometimes we will use the terms ”age” and ”experience” interchangeably. The reason to choose a ten year interval is that, by so doing, we can track ten year cohorts
across censuses and control for their demographic tendencies when evaluating the impact of immigration on employment, revealing the internal migratory response of natives to foreign immigrants. The parameter η > 0 measures the elasticity of substitution between workers in the same education group with different experience levels and θkj are experience-education specific productivity levels (standardized so that Pj θkj = 1 for each kand assumed to be invariant over time, as in Borjas, 2003) . As we expect workers within an education group to be closer substitutes than workers across different education groups, our prior (consistent with other findings in the literature) is that η > δ. Finally, we define Lkjt as a CES aggregate of home-born and foreign-born workers. Denoting the number of workers with education k and experience j who are, respectively, home-born and foreign-born, with Hkjt and Fkjt, and the elasticity of substitution between them by σ > 0, our assumption is that: 

Foreign-born workers are likely to have different abilities pertaining to language, quantitative skills, relational skills and so on. These characteristics, in turn, are likely to affect their comparative advantages in production and hence choices of occupation, therefore foreign-born workers should be differentiated enough to be treated as imperfect substitutes for U.S.-born workers, even within the same education and experience group. While in a more general specification the substitutability between U.S.- and foreign-born workers, σ, may vary across education groups (k), the findings in Ottaviano and Peri (2006) suggest that those differences are not very relevant; hence, herein we maintain a common elasticity. Finally, the terms θHkjt and θF kjt measure the specific productivity levels of foreign- and home-born workers, and they may vary across groups and years (in
the empirical identification we impose a systematic structure on their variations over time) . They are also standardized so that (θHkjt + θF kjt) = 1. 3.2 Effects of Immigration on  

Employment and Wages in a State Economy
Using the production function (1) we can calculate the wage response of each group to total immigration once we know the parameters δ, η and σ, and have the data on immigration flows, wage shares and employment. In particular, assuming a given supply of U.S.-born workers in each skill group, Hkjt, and assuming that physical capital adjusts to immigration so as to keep its real return constant, one can easily show that the effect of total immigrants on the real wages of U.S. natives of education k and experience j (expressed in units of output Yt, which is taken as the numeraire) is given by the following expression:

and the effect of immigration on wages of foreign-born (previous immigrants) of education k and experience j is given by:  

The term is the percentage change of foreign-born in skill group k, j due to immigration; the variable sF kjt is the wage share paid in year t to foreign workers in group k, j, namely sF kjt  Analogously, is the share of wage income in year t paid to all workers in skill group k, j. While appropriate when considering the overall U.S. economy, the assumption of fixed labor supply of natives, Hkjt, may not hold when we analyze the effect of immigrants on a state economy. In response to an inflow of immigrants, ΔFkjt, native workers may move out of or be attracted to California, depending on the effect of a larger supply of immigrants on the demand for natives. We denote as  response the percentage change of native employment of education k and age j in period t in response to total immigration during period t, and we need to account for this response when evaluating the overall effect of immigration on wages. It is easy to show that, in this case, the long-run effect of immigration on the wages of natives and foreign-born would be given by the following two expressions: 

The terms containing  are identical to those in the formulas (5) and (6). The terms containing  ´response account for the wage shift due to the change in native supply of labor in response to immigration. If the induced adjustment of native employment has the opposite sign of the change in foreign-born due to immigration and a relevant magnitude, accounting for it will attenuate the impact of immigration on wages. On the other hand, if the induced adjustment is negligible accounting for it will not significantly change the impact on wages obtained by the simple application of (5) and (6). The next section describes how we estimate the response of native employment to immigrants, response , and the instrumental variable strategy behind those estimates. 

4 The Response of Native Labor Supply
4.1 Specification and Instrumental Variables
While we do not model in detail the mechanism producing the response of native employment to immigration, we estimate the elasticity of Hkjt (i.e. the supply of natives in an education experience group) to changes in Fkjt by running the following regression: 

ΔHkjt is the change in native employment in cell k, j during the decade t. The left hand side of (9) measures the change in native employment as a percentage of overall initial employment in the skill group k, j: Hkjt+Fkjt. The regression controls for education by age (Dkj) and education by year (Dkt) fixed effects. These controls are supposed to capture systematic differences in employment growth across skill groups as well as those that are education-decade specific; for instance, the latter could potentially be due to skill-biased productivity changes. We also control for the predicted change of employment in each cell, denoted  that accounts 10 for demographic trends (cohort size and mortality rates) and national employment/population ratios 5. Any deviation of ΔHkjt from the predicted change in employment is due either to net migration to/from other states or changes in the employment/population ratio of the group6. The coefficient γ captures the elasticity of native employment changes,  to immigration flows, ukjt are zero-mean cellspecific shocks. Once we have an estimate of the elasticity γ, we can obtain the response of native employment to immigration, used above as  In order to obtain an estimate of the coefficient γ that captures the response of native labor supply to immigration in California we adopt the following estimation and identification strategy. First, as already noted, we control for education-specific labor demand shocks (Dkt), that would induce correlation between the residual and the immigrants’ inflow. Second, we perform an instrumental variables estimation using the variable over skill groups and decades calculated for the rest of the U.S. as an instrument for immigration flows to California. The flows of immigrants to any U.S. state by education and experience groups are determined by the interaction of ”push” (supply) factors, relative to the countries of origin of immigrants and “pull” (demand) factors specific to the U.S. states where they move to. By using immigration by skill in the rest of the U.S. as
an instrument for its counterpart in California we are able to isolate the supply-driven variation of immigration (common for flows to California and rest of the U.S.) from the demand-driven variation specific to California. Furthermore, if the dependent variable is measured with error due to the fact that the size of some skill cells in California is small and the 1% samples used may have a non-trivial measurement error, the instrument, based
on national data (excluding California) is likely to have larger size in all cells and to measure immigration much more precisely as a percentage of initial employment. This would drastically reduce any measurement error bias. In general, using migration by cell for the rest of the country as an instrument, we rely on the fact that the education-age distribution of immigrants to California is correlated to the distribution for the rest of the U.S. due to common countries of origin and push-factors. For instance, scant job opportunities for young uneducated workers in Mexico during the nineties was a supply factor inducing a large ΔFkjt for low education, young age groups, both for California and the U.S. as a whole. On the other hand, good employment opportunities for middle aged, highly educated engineers in Silicon Valley would certainly affect ΔFkjt in some education-age groups in California, but would not affect ΔFkjt for the same groups in the rest of the country. These “pull” factors, that also affect ΔHkjt for California, and induce correlation between ΔFkjt and ukjt, would not affect the instrument. To our knowledge, this method isolating ”pull” factors that determine migration to California, is novel to the area approach. At the same time, it is an improvement on the existing aggregate literature that adopts a similar CES production structure (Borjas 2003; Borjas and Katz, 2005; Ottaviano and Peri, 2006) and 5The construction of