CHAPTER 12 EMPIRICAL STUDIES OF LABOR MARKET EFFECTS Following on the theoretical material in the previous chapter, this chapter discusses the recently-burgeoning body of empirical studies of the labor-market effect of immigration. ISSUES IN EMPIRICAL RESEARCH DESIGN It is difficult to determine if immigrants cause unemployment among citizens at large because the job-creating process--which offsets the job-taking process--is so much more indirect and diffuse than is the job-taking process. It is even difficult to establish how many advertised jobs would stand open if immigrants do not come, because after a while employers make other arrangements, either by using machines instead of human labor or by reducing the scale of the enterprise. Assuredly, one learns nothing about the extent of "displacement" by simply observing either that immigrants are "occupying" jobs (see discussion below of Huddle's writing), or that immigrants are working at wages that do not attract sufficient natives to fill the jobs. A cross-national study is not feasible, because few developed nations have experienced substantial quantities of immigration, though we may note that those countries which have had relatively high immigration in recent years -- the U.S., West Germany, Israel, Hong Kong, Canada, and Australia, for example -- certainly have not been marked by unusually high rates of unemployment by natives during the periods of high immigration. One might argue that this is not convincing evidence because the receiving country's unemployment rate influences the immigration rate (Jerome, 1928; Thomas, 1941; Kuznets and Rubin, 1954). But much of the post-World War II immigration into such countries as Hong Kong, West Germany, and the U.S. has been pushed by politics and limited by quotas rather than just pulled by economic conditions, thereby reducing the importance of the pull factor aod making the direction of causation from immigration to unemployment more clearcut. And in the U.S. in recent years the overall quota has always been filled, so the pull factor cannot operate differentially. Nevertheless, on balance, cross-national evidence is not likely to produce a satisfactory answer to the question at hand. Time-series analysis of a single country's experience is not likely to produce a satisfactory answer, either, because of the paucity of observations as well as because of the pull effect mentioned above. (But see the Withers-Pope study of Australia's experience discussed below.) Additionally, in the U.S. in recent decades, the proportion of immigrants of working age to the total population of working age has been sufficiently small so that one would expect difficulty in detecting an effect in the overall unemployment rate even if the effect of an additional immigrant is relatively large. Cross-sectional study of the experiences of American labor markets would seem to overcome some of the difficulties mentioned above. But another pull factor arises in such a study: Immigrants may choose labor markets with low unemployment rather than choosing locations randomly (in the sense that the choice is unconnected to the unemployment level). Also, some natives move from areas of high unemployment to areas of low unemployment. Both of these forces may mask the effect of immigrants upon the unemployment level. Additionally, unemployment is measured in very different ways in different states. With respect to the last-named difficulty, individual- level data permit better measurement of unemployment. But the main body of data available are censuses, and such data are available only every ten years. Current Population Survey data do not offer sufficiently large samples in smaller areas. Several studies of this type are useful, however, and will be described below. Another way to mitigate the difficulties of a cross- section of labor markets is to work with changes in unemployment rather than with absolute levels, which may reduce much of the noise and bias caused by differences in state measurement methods. This device -- used in the Simon-Moore study described below -- also makes it possible in principle to detect an effect on unemployment of immigrants who move into low-unemployment areas. Another possible way to grapple with the latter problem is to independently assess the extent of the unemployment rate upon immigrants' choices of city, and then to adjust for it; others' research (e.g., Bartel, 1983; Rogers, 1969) as well as our own suggests that the effect may be very small, however, as we shall see. The phenomenon of natives moving away from a labor market in response to additional competition for jobs caused by immigrants is more difficult to deal with. It might at first seem that if the pattern was one of immigrants moving to low- unemployment cities, this complication would not exist. But this need not be so; if a native were, for other reasons, just on the margin of moving from one city to another, even from a low-unemployment to a high-unemployment city, immigrants could slightly worsen the employment situation in the original city and thereby cause the move to take place. If this effect is important, it would vitiate an across-city analysis; aggregate (all-U.S.) time series analysis would then be the only conceivable way to assess the effect in which we are interested, and that method is not feasible for reasons given earlier. Though there seems to be no feasible design that will avoid this latter difficulty, in principle it is possible to allow for the bias with the elasticity of domestic migration with respect to unemployment, as estimated by other studies, i.e., the total (unbiased) displacement effect would equal the observed (biased) displacement effect plus the domestic migration effect. The special relationship between low-skill immigrants and the unemployment of low-skill natives is another problem, conceptually as well as empirically. Consider, for example, the following not-atypical news story about the difficulty that employers have in recruiting low-skill workers: Help Wanted: Burger Flippers, Teens Preferred Berlin, N.J.--Betty Hagen stands out among the employees at the Kentucky Fried Chicken restaurant here, not because her uniform is usually covered in biscuit flour, but because the hair tucked under her cap is gray. While nearly all her colleagues are teen-agers, Mrs. Hagen is a 60-year-old grandmother. About 3.5 million people work in fast-food outlets, and nearly three-fourths of the workers are between the ages of 16 and 20. But lately, youngsters have become more choosy about their jobs. Many don't like the fast-food product, pay, or image. The result is giving fast-food managers headaches. More of them are closing the gap by recruiting people like Mrs. Hagen; McDonald's Corp. and Kentucky Fried Chicken Corp. look for help in churches and retirement homes. But generally, fast-food managers are tackling the problem head-on, vigorously going after the nation's teen-agers. This summer, for instance, Wendy's International Inc. will recruit at a national Boy Scout Jamboree. In the Northeast, Marriott Corp.'s Roy Rogers division offers jobs on the spot to regular customers. To retain young workers, Pillsbury Co.'s Burger King offers its workers a bonus of up to $2,000, depending upon length of employment, that goes toward paying for college. (The Wall Street Journal, May 28, 1985, p. 33.) If the flow of low-skill immigrants is sufficiently great that employers do not actively recruit young people and retired persons, should that be counted as employment harm to natives? In some sense, surely yes, even though harm would not appear in the unemployment rate. And some native workers' wages are higher than they would be if there were more immigrants applying for these jobs. On the other hand, some jobs that would be filled if there were more immigrants are instead standing open, with no harm to natives, and some are mechanized. It would contribute to an enlightened understanding if we knew the quantitative relationships among these phenomena -- numbers of immigrants, wage rates, jobs filled by natives, open jobs, unemployed natives, and so on. But even then, to make real sense of the matter it would be necessary to go beyond one- industry short-run considerations, just as with tariffs and other trade barriers. And in a long-run perspective, these short-run effects wash out. Several types of empirical evidence concerning the relationship between immigration and unemployment are discussed in this chapter: 1) Comparison of unemployment rates, and of labor-force participation rates, in areas with different levels of immigration. 2) Comparison of wages in places where there are more and less immigrants. 3) Comparison of wages and employment in industries when there have been more and less workers of the same labor class that characterizes a particular group of immigrants, such as factory workers. 4) Time-series analysis over business cycles. 5) Miscellaneous studies, including studies of the analogous phenomena of the effects of school leavers, and of increasing numbers of women entering the labor force. 6) Questionnaire studies of immigrants' experiences in the labor market. Experiments that remove illegal immigrants from jobs and then make the jobs available to legal residents are discussed in Chapter 14. Because some studies fall into more than one of these categories, however, the order of studies to be described cannot neatly follow this classification. Kuznets on Southern Blacks Kuznets greatly admired immigration's economic effects. Nevertheless, he was concerned about the effects upon sub- groups of the population. He judged that rural Southern blacks benefited from the cessation of mass immigration. It meant, once the worst of the depression of the 1930's was over, that the reduction in the inflow of immigrant labor opened up opportunities for more employment of native labor at similar skill levels, particularly of Negroes from the South. It is hardly an accident that while the proportion of all Negroes in the country residing in the South hovered at about 90 percent from 1770 to 1910, it began to decline with World War I, and by 1970 dropped to 53 percent, with substantial shares of Negro population appearing in the other regions, particularly the North. Conversely, the proportion of foreign born in the total white population of the non-South, at a peak of about 21 percent in 1910, dropped to below 6 percent in 1970. (1977, p. 4). Smith-Newman Study of Wages in Texas Smith and Newman (l977) carried out the first convincing study of the effect of immigration upon wages, analyzing the effects of legal Mexican immigrants in various Texas cities with differing proportions of Mexicans in the population and at different distances from the border. They found that wages (actually, yearly incomes with adjustments for hours and weeks worked, as well as for age, education, occupation and migration history) are indeed much higher--a 20% lower wage near the border, where the Mexican population is proportionally greater than away from the border. But when they adjusted for the cost-of-living differences between these cities, the differential was reduced to 8%. And if they were to adjust for the well-known differentially-higher wage in larger cities-- Houston, in this case, compared to Corpus Christi, Brownsville and Laredo--the differential would be even less. On the other hand, the differential is higher than average for the lower- skilled occupational group. Urban Institute Study of Los Angeles County Comparison of one area with the rest of the nation is fraught with the danger that special conditions of one sort or another, not connected with the relationship in question, may account for the phenomenon to be explained. And this difficulty must be kept in mind when considering the results of the comparison of labor-market conditions in Los Angeles County with the rest of the United States by Muller (1984) and by Muller and Espenshade (1985). Yet these findings are of interest even if they are less conclusive than we would like (as all studies on this topic are less conclusive than some social-scientific studies). This is Muller's description of his results concerning unemployment: To what extent did the influx of immigrants entering Southern California in the 1970's reduce the jobs available to nonimmigrant workers? The answer for the 1970's is little if at all. Although Hispanic workers filled a large proportion of the jobs added during the decade, particularly in manufacturing, there is no indication that work opportunities for nonimmigrants lessened. Despite mass immigration to Southern California, unemployment rates rose less rapidly there than in the remainder of the nation. Furthermore, the labor-force participation rate (the proportion of the population in the labor force) did not seem to be affected. In fact the participation rate for both blacks and whites was higher in Southern California than elsewhere in the state and nation. Moreover, the difference in the participation rate between Southern California and the rest of the country remains essentially unchanged since 1970, indicating that the influx of immigrants did not discourage people from seeking employment (p.13). He continues: Even the job prospects for black teenagers do not appear to be adversely affected by the influx of immigrants. Total teenage unemployment in Southern California is close to the national average, but unemployment among black teenagers is substantially lower than average (p.14). Muller and Espenshade also conducted a cross-sectional study of black unemployment in 247 metropolitan areas in the U.S., and 51 metropolitan areas in California, New Mexico, and Arizona, states with large proportions of Mexican-origin persons. They regressed the rate of unemployment among blacks upon the percentage of Hispanics in the population, holding constant the percentage change in population between 1970 and 1980, the percentage of income from construction and durable goods industries, the percentage of blacks with a high school education, and the rate of unemployment for whites. They found as follows: Black unemployment rates are not increased--if anything, they are lowered--by a rise in the proportion of Mexican immigrants in a local labor market. In the U.S. sample regression, signs of the remaining coefficients are as one would expect. Thus, after accounting for general labor market conditions, most of the variation in black unemployment rates among metropolitan areas can be attributed to differences in black educational attainment, in the rate of population growth, and in the degree of durable goods manufacturing and construction. In the regression based on the Southwest sample, only the level of white unemployment stands out as statistically significant. (1985, pp. 99, 100). The picture with respect to wages is different, however. From the evidence contained in Table 12-1 Muller concludes: There is little doubt that wages in several occupations and industries rose more slowly in Los Angeles than elsewhere as low-skilled immigrants, primarily Hispanics, entered the labor force...most notabl[y] in the manufacturing sector, particularly among production workers in industries where wages have been traditionally low such as in apparel and textile production and in relatively low-wage industries such as restaurants, personal services, and hotels where many Mexicans are employed. The relative declines in low-skill wages are especially noteworthy because wages in general rose nine percent more rapidly in Los Angeles than in the rest of the country between 1972 and 1980. ---------- Table 12-1 ---------- Muller and Espenshade also made a special study of the effect of Hispanic immigration upon blacks, the group which they adjudged to be the Hispanics' closest competition in the labor market. They first examined the rates of labor-force participation and unemployment for the years 1970, 1980 and 1982, covering a period of heavy immigration with Los Angeles County, with these results: Blacks generally, and black teenagers especially, do not appear to have been harmed by immigration in the period from 1970 to 1981. During the 1970s and into the 1980s, adult labor force participation rates increased in Los Angeles metropolitan area and in California, reflecting a national pattern of rising labor force participation. Throughout the period, participation rates in Los Angeles continue to exceed the national average, maintaining a fairly constant lead. Teenage labor force participation rates also increased over the period, and the rates for black teenagers in Los Angeles and in the state showed gains relative to the rate for black teenagers in the nation. By contrast, participation rates for all teenagers in Los Angeles declined relative to the national average, dropping below the national labor force participation rate for teenagers by 1982. An examination of labor force participation data for Los Angeles by sex and race from the 1970 and 1980 censuses indicates that black women had gains that were above the average for them nationwide, while black men experienced a decline that was somewhat lower than the decline for them nationwide. And in 1982, when unemployment in California reached its highest rate in four decades, nonwhite labor force participation rates for both teenagers and adults in the Los Angeles area continued to exceed national rates. Native workers who find their jobs jeopardized by immigrants may experience higher rates of unemployment, if they do not drop out of the labor force altogether... The period from 1970 to 1982 was marked by rising rates of unemployment, both nationwide and in California. For all groups in the United States, unemployment rates more than doubled. The smallest increases were for blacks in Los Angeles--27 percent for adults and 35 percent for teenagers--followed by black teenagers in California. In sum, trends in unemployment rates do not provide evidence of sharp job competition between immigrants and blacks. (1985, pp. 96-97.) Table 12-2 (their Table 16) shows a puzzling lack of effect upon blacks in three of the four lower-pay categories. And this finding seems not to be a statistical artifact; Muller and Espenshade ran cross-sectional regressions similar to those described above for unemployment, but now for wages of blacks, and found an effect that was statistically significant but very small in magnitude -- $85 yearly, if the proportion of Hispanics would be 7.5% rather than 5% in the given area, to be compared to average black family income of $15,818. Muller and Espenshade suggest that Hispanics and blacks are not close competitors in labor markets;1 yet this finding needs more explanation. ------------ Table 12-2 ------------ Muller and Espenshade (1985) also tackled the difficult problem of estimating the number of new jobs created by Mexican immigrants. They compare the ...average proportion of low-skill low-wage workers in the total labor force in the nation's twelve largest metropolitan areas -- areas that, with the exception of Dallas, Houston, and Los Angeles, have relatively few Mexicans in the labor force -- and use this average as a guide to what might be expected in Los Angeles in the absence of Mexican immigration. The actual number of operatives and laborers in Los Angeles in 1980 was 60,000 larger than the number predicted using this procedure. (p. 149) This estimate jibes well with their estimate of the relevant number of Mexicans in various manufacturing jobs in Los Angeles. This "direct" effect may be compared with the estimate of 210,000 recent immigrants from Mexico. Muller and Espenshade then go further to estimate the jobs indirectly caused by the Mexicans through complementary employment, consumer services to Mexicans, and so on, reaching a total that is not incomparable in size with the total Mexican migration. But this procedure is necessarily tenuous, so I will refer the reader to their book for further information. McCarthy and Valdez Study of California Using methods similar to Muller and Espenshade -- analysis of Census and Department of Labor data on employment, unemployment, wages, and population for Los Angeles and California compared to the U.S. as a whole, in 1970 and 1980 -- McCarthy and Valdez arrived at results similar to those of Muller and Espenshade. Immigrants appear to have provided a net benefit to the California economy by supporting industrial and manufacturing growth. Their negative labor market effects have been minor and concentrated among the native-born Latino population. (McCarthy and Valdez, 1985, p. 24.) This lack of effect is particularly strking because the growth in employment was large: While between 1970 and 1980 low- wage employment fell by 5.2% in the U. S. as a whole, it grew by 46.1% in California and by 52.7% in Los Angeles; the corresponding figures for moderate wage industry were increases of 4.3%, 20.6%, and 6.9%, and in high-wage industry were 7.8%, 27.6%, and 11.4%. And the retardation in growth of wages relative to the U.S. as a whole was less than seen in Tables 12-1 and 12- 2, though the data are not strictly comparable; the ratio of LA growth to U.S. growth was 0.80, 0.86, and 0.96 for the three levels of industry, and the California-wide ratios were even closer to the U.S. averages. (1986, p. 40). DeFreitas and Marshall DeFreitas and Marshall related the percentage in 1980 of manual workers in SMSA manufacturing sectors who were born abroad to the average annual rate of change from 1970 to 1978 in the hourly earnings of manufacturing production workers in those SMSA's. They found that . . . higher concentrations of foreign-born manual workers have a statistically significant negative impact on wage growth rates. However, the magnitude of the estimated impact is relatively modest (less than one percentage point slower wage growth with every 10 percent increase in the immigrant share of manufacturing jobs), and it is significant only in SMSAs in which 20 percent or more of manual workers are immigrants (Year ?, p. 155). This effect is not easy to interpret, however. It may be partly caused by education or other characteristics of natives rather than by increased competition from immigrants (and indeed, productivity growth has a much larger coefficient in regressions including immigrants than without them) because the independent variable refers to the stock rather than to the flow of immigrants, and the recent flow must be a rather small proportion of the stock. But flows are related to stocks (see Bartels, 1982), providing some support for the analysis. Grossman In theory, immigrants will lower the wages of natives a) to the extent that there is substitution between capital and labor; if the quantity of labor is relatively large with respect to the quantity of capital, the output per worker is relatively small, and hence wages are relatively lower; and b) to the extent that there is substitution between immigrants and natives; if immigrants and natives can do the same job, in the presence of condition a), the wages of native will be depressed by immigrants. To the extent that the two kinds of labor do not substitute for each other, the effect of the capital constraint upon natives is mitigated. Grossman (1982) estimated the effect of foreign-born workers upon second-generation workers, and upon native workers, in a sample of 19 SMSA's. In addition to the proportions of these groups in the labor force, she introduced a variable for quantity of capital. Her results implied a slight negative effect of immigration upon native wages. Her results must be qualified by the smallness of the number of observation, by the fact that the foreign-born represent all ages and lengths of time in the United States, and by the puzzling fact that the number of natives seems to have a strong negative effect upon their own wages -- puzzling because larger cities are observed to have higher rather than lower wages. DeFreitas DeFreitas (1986), in a study noteworthy for its detail and depth, used a sample from the 1980 census to investigate the effect of Hispanics -- "the majority illegal," in his words -- upon male and female groups of Anglos, Blacks, and Hispanics, separated into native and foreign-born, who immigrated between 1975 and 1980. He found no negative effect upon the wages of any group except black females. "For all low-skilled native men the results indicate that there are no significant negative effects on their wage levels from recent Hispanic immigration. In fact, such migration has a significantly positive influence on the Anglo-male earnings. . . The only persons whose wages appear to have been somewhat adversely affected by illegal migration since the mid-70's are black women," but this effect is "relatively small." (page 23) It must be noted, however, that the immigrants had been in the country up to five years; if the effect is mainly in the first year or so, rather than being more permanent, these data could not be expected to show it. DeFreitas also examined the effect on the number of weeks worked per year, and on the amount of unemployment. He found that "For no racial/ethnic group, male or female, is there a discernable negative effect of illegal immigration on employment. In fact, most of the estimated co-efficients are positive." (p. 24) Borjas (1983) studied the substitutability of black, hispanic, and white workers in the 1976 Survey of Income and Education, and found hispanics to be complements rather than substitutes for blacks, and perhaps for whites as well. No distinction was made between native and immigrant hispanics. Nevertheless, given the proportion of immigrants among the hispanics certainly is sufficiently great, the similarity of labor-market characteristics betweeen immigrants and natives among the 61% of the hispanics who are of Mexican origin, and the precision of the methods Borjas used, this evidence provides strong confirmation for deFreitas's finding. Chiswick-Chiswick-Miller B. Chiswick, C. Chiswick, and Miller (1985) estimated the elasticity of substitution by examining the ratio of skill- adjusted immigrant wages to native wages in five countries where the proportions of the two kinds of labor differ markedly2 -- the U.S. Britain, Canada, Australia, and Israel. (Similar data may be seen in Figures 3-5 and 3-6.) The fewness of the observations causes doubt about the validity of the analysis, but experiments checking the sensitivity to the dropping of particular observations provides some confidence. "The estimated elasticity is high (about 28), but it is significantly less than infinity." That is, they found that immigrants and natives are close (though not perfect) substitutes. But this does not imply that immigrants increase native unemployment, for reasons given in the previous chapter, though it does indicate that absorption requires more adjustment than if immigrants were poor substitutes or complements to natives. Bean-Lowell-Taylor Bean, Lowell, and Taylor (1986) estimated the effect of the number of undocumented Mexican workers (1980) in the labor markets across Southwestern United States upon annual earnings in 1979 of six labor force groups -- a) undocumented, b) legal, and c) native-born Mexican-origin males; d) black males; e) non- Mexican origin white males; and f) females. The (statistically significant) effects of illegal Mexican immigrants are negative upon white males' wages (substitution) and positive (complementarity) upon females' wages. The effects of legal Mexican immigrants are positive upon native-born Mexican males and black males. The magnitudes of the effects are hard to interpret [see letter] but the authors refer to them as "not very sizable. The concern that undocumented immigration may be depressing the earnings of native born workers does not appear to be borne out by these results" (p. 15). Morgan-Gardner Study of the Bracero Program's Effects Morgan and Gardner studied the bracero program of guestworkers from Mexico that operated from 1942 to 1964, in the broad context of the history of Mexican immigration and the economic-institutional context of Southwestern agriculture. (Figure 15-1 shows the year-by-year size of the program.) Crucial for our purposes here, they estimated the number of native workers who lost jobs, and the extent of the fall in wages. Because they used data and techniques that seem as well-adapted to studying this particular topic as are available, I shall review the study in some detail. Morgan and Gardner first develop a simple theoretical model to guide the econometric work and its interpretation. Then they discuss the kinds of variables that are necessary to fit the model to the situation at hand, as well as the kinds of data that are available to do the job, and the difficulties encountered in that quest. The resulting econometric model was as follows: [[[[Insert quote M_G from pp. 385-6 of Chiswick]]]]]] (cut and paste) Morgan and Gardner then estimated the model using three geographic groupings to see how the results are affected thereby, and they tried both ordinary-least-squares and two- stage-least-squares regressions. In the light of their various results, as well as the results of other studies of the market for agricultural labor, they worked up an idealized summary as a way of presenting the gist of their results. Calculating on the basis of an effective 35% increase in the supply of labor (210,000 workers) due to the program (the approximate number between 1953 and 1964) they estimate a fall in wages of 9%, and a reduction of 51,000 jobs filled by natives. At the same time, they estimate a total increase of 120,000 jobs due to the decline in the wage rate. Morgan and Gardner are impressed at the relatively high responsiveness in the number of jobs (a labor-demand elasticity of -3.2) to the changes in wages, and they are also impressed at the relatively small decline in the wage rate induced by an increase in the supply of labor of the magnitude of 35% (pp. 398-399). Despite the excellence of this study, it is not easy is to think through its implications. Gardner and Morgan calculate that the aggregate economic benefits to natives are greater than the costs to native workers in foregone wages. The gains are obtained by the growers and by food consumers, but the authors are not able to compute the proportions of the gains between the two groups. From a perceived-pain point of view, we must notice that whereas the benefits are spread widely, the costs fall on one small group, the agricultural workers, which immediately raises questions about "fairness." Going beyond these immediate calculations, there are several other important considerations. First, though our focus is the effect upon natives, there is also an important economic gain to the Mexican workers and Mexican society, constituting a large part of the workers' total wages. (Morgan and Gardner assume for the sake of argument that Mexican wages would be half the U. S. wages, which leads to a net gain to the Mexican workers of more than the gross loss of wages to U. S. workers.) And when we remember that many of the U. S. workers who were "displaced" found jobs elsewhere, their net loss must have been even smaller. Another consideration is that without the bracero program, there would have been more substitution of mechanization than actually occurred, and hence some job loss anyway. As Morgan and Gardner put it: "This relatively high demand elasticity and job creation effect is consistent with the informal observation that braceros were a substitute for mechanization, notably in High Plains cotton, and that the end of the program substantially accelerated the mechanization of Texas cotton. This is also the period in which the tomato harvester came into widespread use in California." (p.399) To put the matter more generally, the appropriate analysis is not a comparative-statistics comparison between a given number of workers N employed in the industry at wages W if there are braceros or illegals, whereas there would be N-X workers and W-Y wages if there are no braceros or illegals, with the difference in wages as derived from static analysis. Rather, as Morgan and Gardner point out, one response to a regime that allows fewer Mexican temporary workers is a shift to more capital-intensive production, which might in the long run reduce employment more than would a regime which allows Mexican guestworkers. Also, some proportion of displaced workers find jobs in other industries, sometimes under conditions that they come to consider superior to the lefts from which they were "displaced." Hence the welfare loss is not that which is computed under comparative-statistics analysis, but rather something less. Simon-Moore Comparison of Immigration and Unemployment Across Cities in the U.S. Moore and I (1986) studied the relationship between the rates of immigration and of unemployment across cities in the United States. Our samples cover the years 1960-1977 for the various numbers of cities in the United States for which Immigration and Naturalization Service data on immigration are available in various years. (This is the only period during which any data are available.) We first investigate the relationship between the level of unemployment and the level of immigration. Here and elsewhere in the study we also experimented with other variables such as the rate of growth of population, the proportion of the labor force working in manufacturing and in durable-goods manufacturing, and the proportion of insured unemployment to total unemployment, but these control variables do not explain an important proportion of the variance. At first glance there appears to be a positive relationship between the levels. But upon further inspection, the correlation is seen to be almost as strong between unemployment in early years and immigration in later years -- obviously not a causal relationship -- as between years in which there might conceivably be a causal relationship. At the heart of the econometric problem is a very high similarity in immigration patterns across cities from year to year. Unraveling this pattern to determine the underlying relationship defied our efforts. To avoid the serious difficulty of differences in measurement of unemployment among states, we also work with changes in unemployment from one year to another. Comparisons of samples of unemployment levels two years apart (surrounding the year of immigration) do not show any clear pattern of effect of immigration level; when the national unemployment level is rising the relationship seems positive, and when the unemployment level is falling the relationship usually seems negative. When we examine the differences between the farthest- distant years (1960 and 1977) the results seem at first to show a significant relationship, but when we allow for the secular rising national trend in unemployment by examining the farthest- distant pair of years having the same unemployment levels, the apparent relationship is no longer seen. It is not obvious which of these ways of looking at the data is the more appropriate. We also carried out analyses with first differences on both sides of the equation, but these runs showed no relationships of any kinds. Analyses of levels of unemployment and immigration in the 1970 Census of Population also show little or no effect, confirming our other analyses. Ultimately, this sort of study ought to render the statistical findings into economic terms, that is, produce an estimate of the quantitative effect (if any) of an additional immigrant on native unemployment. There are many coefficients in the paper that might be the basis of such an estimate, many of them suggesting no effect at all. But for perspective, let us look at the largest of the coefficients, the one for the summed regression for 1960-1975, and consider it an unlikely upper bound. It suggests that for each immigrant who entered during that period, .093 immigrants were unemployed during each of the fifteen years, or 15x.093=1.395 years of native unemployment for each entering immigrant. That would certainly be a meaningful amount of unemployment to be caused by an immigrant on average. But even this most-unlikely upper bound is far distant from a displacement assessment of one job permanently lost to natives for each immigrant admitted; the coefficient would have to be 20 or 30 times that large for there to be any such suggestion. The correlation in this and other regressions is strikingly low. It must be kept in mind, however, that immigration is small in volume relative to other population movements and components, and therefore it is almost impossible that immigration could explain a large proportion of the differences in unemployment, no matter how close the relationship. Furthermore, noise from errors in variables and other sources tends to depress the correlation. Still it is hard to find much policy importance in a variable that explains so little of the variance in the dependent variable, even assuming it to be statistically significant (which is probably an unwarranted assumption), no matter what the explanation of the lack of explanatory power. Huddle's Writings Huddle and "27 of his students in a labor seminar, juniors and seniors, mostly, conducted several field studies during a two-year period, 1981-1982. Their objectives were to determine what industries hire illegal aliens and to what extent illegals displace U. S. workers or shut them out of the labor market in the booming Houston-Galveston metropolitan area" (Huddle, Corwin, and MacDonald, 1985, introductory note). Huddle concluded that "the social and economic implications of the penetration of the economy by undocumented workers is dramatic. If the sample proportion of illegal worker participation is projected onto city, state, and national construction programs alone, we find that all male youths and minority youths, aged 16-24, could, in principle, have been removed from the rolls of the unemployed as of the time of our study, and that adds up to more than one million U. S. workers who have been displaced" (p. 3, "FEDERAL GOVERNMENT...") Huddle asserts that "Some $18 billion per year are 'being siphoned off into the pockets of greedy employers' of illegal aliens" (3/6/84, p. 1). And he recommends that "Arrest and deportation of illegal alien workers is currently the cheapest and fastest way of securing additional jobs for unemployed U. S. citizens" (RAIDS ON..., p. 1). I mention Huddle's reports here because they have been publicized heavily by the INS and discussed extensively in the press. But upon several inquiries to Professor Huddle, I have not been able to obtain any material other than the above-quoted press releases and a general article by Huddle, Corwin and MacDonald (1985) containing no scientifically-detailed description of what he did. Until the research methods and data are made public, it would seem prudent to completely disregard this work. And it is obvious from the press releases alone that Huddle did not make any study at all of displacement--that is, the effect of illegals upon native workers--but rather simply made an unsubstantiated extrapolation from his estimate of the proportion of illegals working in the construction industry. It should also be noted that the latest (1985) Huddle et al. piece asserts that a "modal estimate for 1983 would be 8 to 9 million" illegals in the U. S., a statement made wholly contrary to (and without mentioning) the Census Bureau's 1981 and current subsequent much lower estimate (the NAS's 1986 estimate is even lower; see Chapter 15 on these estimates.) Easterlin's Study of U.S. Cyclical Unemployment In earlier decades of this century, when immigration to the United States was quite free, and much larger relative to the size of the U.S. labor force than now, the waves of immigration coincided with the waves of business activity in the U.S. Immigration increased when the demand for labor was great, and decreased (or even became net out-migration, as in the l930's) when demand for labor fell and unemployment rose. About immigration in the free-entry period before World War I, based on his own work and that of previous investigations, Easterlin concluded: "[T]he swings in immigration were a response to corre- sponding swings in the demand for labor in the United States. The evidence is as follows: In the United States, turning points of long swings in output growth typically preceded those in the rate of immigration, suggesting that immigration was responding to changed conditions in the United States rather than abroad (Figure l2-1)... During long swings in the U.S., a rising immigration rate was typically preceded by a rising rate of growth in hourly wages and, as far as the limited evidence goes, a declining unemployment rate; a falling immigration rate tended to follow a decline in the groth rate of hourly wages and a rising unemployment rate (Figure l2-1). Since the growth of the U.S. labor force from domestic sources, whether from demographic factors or participation-rate change, showed but slight evidence of long swings before World War I (Figure l2-1), the implication is that immigra- tion waves were one of several symptoms of common origin, namely, alternating tightness and slack in the labor mar- ket associated with swings in the growth of labor demand. The immediate stimulus to migration was probably changes in unemployment conditions. There is a substantial similarity in the timing of out-migration waves from diverse areas of origin-- different parts of Europe, Canada, Latin America, Asia, and even the rural sector within the United States. This observation is consistent with the view that these areas were responding to a common external stimulus such as swings in labor demand at destination (l968, pp. 30-3l). ----------- Figure 12-1 ----------- The linkage between employment conditions and immigration had the beneficial effect upon the native labor force of reducing competition for jobs in bad times, while increasing the demand for labor in both good times and bad. There is no reason to doubt that future immigration would also respond to employment conditions in the U.S., and thereby reduce the severity of unemployment cycles, if it were unconstrained by quotas. Withers and Pope (1985) applied the technique of Granger causality to quarterly Australian unemployment and immigration data from 1948 to 1982, enabling them to distinguish the direction of causation. They found no increase in the rate of unemployment due to immigration. They did find that unemployment influences later immigration, which accounts for the correlation between the series. Summarizing this study as well as several earlier studies of the topic, Chapman, Pope and Withers (1985) concluded that "immigration has not increased unemployment within the range of Australian post-war experience." Jones and Smith on Great Britain During the 196l-1966 period studied, general unemployment was sufficiently low that Jones and Smith did not even attempt to determine whether immigrants caused natives to be unemployed. Though the unemployment among immigrants was also very low, it tended to rise relatively faster than general unemployment during the business cycle. Jones and Smith therefore concluded that the immigrants have the positive effect on natives of buffering the changes in their unemployment rates. "...coloured immigrants bear, in these instances (a rise in general unemployment) an above average amount of the economy's real burden of unemployment" (p.45). Bahral on Israel Bahral (1965) studied the effect on wages of the mass immigration of unskilled North African Jews into Israel from 1948 to 1958. He found that the differential between skilled and unskilled workers widened, as standard theory suggests, even though wage determination in Israel takes place mainly through collective bargaining rather than individual employer-worker agreement. That is, even institutional barriers to the operation of supply and demand did not prevent the wages of the category of workers that became more abundant from falling relative to other workers. It should be noted, however, that the quantity of immigration into Israel was extremely high compared to other countries' experiences; the labor force more than doubled between 1948 and 1984, for example. School-Leavers The phenomenon in which we are interested here is analogous to the large surge of new-labor-force-entrant school leavers in the summer time, entrants not offset by a surge of retirements at the same time. This increment to the labor force from school-leavers is much larger than the yearly increment from immigrants, and it all occurs in a single month or so. Figure 12-2 shows seasonal patterns of unemployment and employment for persons of ages 16-19 and 20+. There is no discernible effect of the school-leavers upon older persons. Unemployment among persons older than 16-19 is not higher in the months after the school-leavers hit the job market than in the months before -- indeed it is lower in the second half of the year than in the first half. And the unemployment rate of the youths shows very little impact of school-leaving after the summer. These data call for systematic analysis, however (analysis which might also improve methods of seasonal adjustment of unemployment and employment statistics). ----------- Figure 12-2 ----------- Effects of Cohort Size The post-war "baby boom" cohort of entrants into the labor force in the United States and elsewhere is much larger than previous cohorts. The "additional" persons may be considered analogous to a cohort of immigrants. The large body of research on the labor-market effects has recently been reviewed and summarized by Bloom and Freeman (1986). There is strong consensus that -- at least for a while -- baby-boomers' unemployment is higher, and wages lower, than they would be if their cohort were smaller. This squares with the higher unemployment for immigrants than for natives seen in Australia (though Harrison's analysis [Chapter 11] does not apply so well in this situation, because the baby boomers were consuming even before entering the labor force). There is less agreement about how long the effect lasts and the extent to which the adjustment to the increased "pressure" on the labor market is through lower wages and how much through higher unemployment; the trade-off apparently differs from country to country. If one's main interest is the effect of immigrants upon the native population, however, one wants to know the effects on native unemployment and wages. And the baby-boom studies do not throw light upon those questions. Furthermore, the fact that the baby-boomers' wages are relatively lower than they otherwise would be suggests that others' wages are relatively higher. Theoretical support for such a conclusion follows from the complementarity of age-groups. The only ways that the baby- boomers could lower the wages of others is if a) total output per person falls due to capital dilution, or b) a greater proportion of total output flows to the owners of physical capital. Possibility (b) seems unlikely, given the data on labor's share in recent years. And I know of no data showing that possibility (a) is an actuality. So we must consider as not implausible that, to the extent that the baby-boom studies throw light on immigration, they show that natives on average receive higher wages due to the immigrants. Baby Boom Plus Increased Female Labor-Force Participation: Grant-Hamermesh Grant and Hamermesh examined the interrelated effects of the entry of the unusually large baby-boom cohort of youth into the labor force between 1960 and 1980, and the concurrent increase in the labor-force participation of women. They found that the wages of both groups were lower than if the other group's entry rate had been lower. But the "competition from adult women did not itself increase the rate of unemployment of youths." (Hamermesh and Rees, 1984, pp. 104-105.) Creation of New Businesses Chapter 4 presents evidence that immigrants have a higher propensity to start new businesses than do natives. Usually these are small businesses, because most businesses start in a small way. But one should not therefore underestimate the job- creation capacity of such small businesses. A study of the new jobs created by firms in different size categories indicates that an astonishing 66% arose in firms with 20 or fewer employees; see Table 12-3 from Birch (1981).4-?? ************* ---------- Table 12-3 ---------- Samuel and Conyers considered several of the avenues through which new immigrants into Canada create new jobs. They state (though without further citation) that in 1983-1984 the 3,940 business immigrants -- entrepreneurs and self-employed persons who have funds to invest in Canadian enterprises -- brought "a total of $l,544 million to Canada" with them, a mean of almost $400,000 each (1985, p. 7). And Norman and Meikle report a mean amount had "available" for transfer in l983-84 to Australia by "business immigrants" of Australia $383,600 (1985, p. 123). There is little or no solid information about the relationship between additional capital injected into an economy as investment and the number of jobs created, though numbers such as $30,000 and $50,000 have been talked about in the United States in past years; and of course the relationship must vary among industries; the small businesses that immigrants tend to start should require lesser amounts of such capital than industry on average. There is little doubt that this amount is more than enough to be consistent with the estimate of six jobs per business-starting immigrant found by the Three Years in Canada Study. Additionally, Samuel and Conyers state that immigrants other than business immigrants brought with them a mean of $l7,841 per immigrant family. How much of this goes into investment and how much into consumption and how much remains in idle savings is anybody's guess. But at least some of it must go to investment, providing additional jobs in that way. And of course, sums spent on producer capital also have all the effects of sums spent on consumption. And Norman and Meikle report a mean amount of funds brought by others than business immigrants of $ [Calculate - not in Chapter 7]. (1985, p. 7). As to the relationship between consumption (and also investment spending) while immigrants are still unemployed, Samuel and Conyers cite Vanderkamp's estimate that for inter- regional migration within Canada, for each five unemployed persons who leave, two more persons become unemployed by way of the drop in consumption. They suggest that it is reasonable to turn this around with respect to immigrants -- that each five immigrants who come, even while they are unemployed, cause two more additional jobs for others. (This may be a reasonable basis for estimating the consumption effect that Harrison uses in his model (discussed in Chapter ll). Samuel and Conyers then make calculations based on the adult-equivalent values of the children and non-working adults in the 1983-84 immigrant cohort and, using the Vanderkamp job ratio, they estimated that "the presence of 1983-84 immigrants should have resulted in at least 58,560 new jobs in Canada (16,400 more jobs than required for 1983-84 immigrants)" (p. 10). Experiments Removing Illegals From Jobs Two experimental forcible removals by the authorities of illegal immigrants from hundreds of jobs in California were followed by checkups on what happened to the jobs that the illegals held. In both cases there was a relatively low rate of substitution of natives for the illegals. This evidence is discussed in Chapter 15 in connection with illegal immigration. DISCUSSION 1. As with trade policy, a large difficulty with immigration policy is that -- inevitably -- any labor-force change causes some groups to suffer some harm in the short run. This expected harm causes opposition to the change. But preventing the change foregoes benefits to other groups, and in the long run usually harms even the groups that fear and oppose the short-run effect. The capacity to cry "unfair" is seemingly unlimited. As of 1986, the physicians in the U.S. assert that there is a surplus of doctors, though their net incomes average over $100,000 yearly. Hence they are currently attempting to limit the admission to practice of those who were trained abroad (The Washington Post, June 19, 1986, p. A25). Recently, there appeared an item in the newspaper that "1,250 Kirghiz have applied for visas to start a new life in Alaska because that climate is the best fit for their temperaments and traditions." (The Wall Street Journal, December 16, 1981, p. 21.) Were this possibility to become a live one, chances are that the Eskimos would object to additional competition for economic resources. (Really.) It would seem prudent, then, to discount the urgings of particular occupational groups when making immigration policy. 2. Perhaps it would be a useful exercise for the reader to reflect on the issue of displacement from the perspective of someone else observing you and asking, with tongue somewhat in cheek: What would be the effect upon the labor market if you the reader -- assume to be employed by a private or public enterprise -- were to suddenly stop working? There would immediately be an additional job opening. After some shorter or longer period of time a person who might otherwise not be working might be hired for your old job, thereby decreasing the unemployment rolls. Or, some person now working might take the job, and either the second person's job or else the job of someone else down the line might be filled by an unemployed person. So from the point of view of unemployment as a welfare measure, your ceasing to work might be seen as socially beneficial in the very short run. (Of course it is understood that the valuable output that you produce in your job would be lost to the public that buys it, and material welfare would thereby be decreased unless the unemployed person who finds a job because of your leaving your job produces output as valuable as yours. But this effect is not germane to the immediate discussion.) Does this conclusion leave you feeling satisfied? Do you consider that it tells enough of the story when it says that social welfare, as measured by unemployment, is increased by your leaving your job, implying that it would be a socially desirable act for you to quit tomorrow? Upon reflection, most of us would probably think the matter through a bit more, to counter this threat to our self-respect as well as to make this conclusion jibe with the belief held by most of us that our working is a good thing for other people. And indeed there are several possibilities that immediately mitigate or controvert this conclusion. One such possibility is that you own and run a private enterprise that employs other persons. If you stop working, some or all of your employees will become unemployed, temporarily or permanently. And though some of those persons might later become employed in firms that take over some of your business, there would almost surely be some slippage and loss of output and employment as a result. That outcome certainly would represent a loss of welfare on balance. And indeed, many immigrants and their children do start new business that employ other people, and they almost surely have a higher propensity to do so than natives (see Chapter 4), both because of the otherwise limited opportunities available to them when they arrive as well as because of immigrants' putative greater drive to succeed and build for their children. (After all, was not every one of the businesses in the United States begun by immigrants or their descendants?) Another thought that might occur to you is that your successor might not be as competent as you -- or why else does your employer continue to keep you on the payroll? And indeed, it seems to be true that one of the main reasons that employers hire immigrants is that they are more competent than available natives at the same wage. (It is a rather embarrassing lament on the part of the labor unions that immigrants -- legal and illegal -- should be kept from competing in the labor force on the grounds that they will work harder, and work longer hours, than will natives.) It also is likely to occur to you -- and this is probably the most important line of thought, though the hardest to perceive in its fullness -- that when you cease working you cease earning, and when you cease earning you cease spending. When you cease spending, the demand for labor falls in the industries that produce and sell the products that you might otherwise buy. Thus, some people who would otherwise work will be unemployed, and wages will fall somewhat in those industries. Such is the fundamental nature of an exchange economy; we all take in each other's washing, and if you can't recompense me, there will be no work for me to do. This explains why we can have an economy with as many jobs as there now are, growing at the rapid rate at which the economy has been growing, despite the large numbers of people newly entering the labor force. This process is difficult to comprehend because there always is a period of adjustment between the additional new competitor for a job entering the labor force, and the process playing itself out to its final benign conclusion. One's understanding is also hindered by the process being indirect and hidden from view. But to ignore the process because it is difficult to perceive, and to focus on the immediate effect on the implicit assumption of a fixity in the number of jobs, and to assume that that immediate effect also will be the final outcome, is either to have a too short a time horizon or a defective view of the economy or both. The simple truth is that you, dear reader, probably are not so expendable from your job as the simplest arguments suggest you to be. And if you stopped working for a while and then returned, your coming back to work would not be an imposition and a burden upon other people but rather (aside from the shortest-run adjustment difficulties) would likely be a blessing to the rest of us, even without considering the taxes you pay, the productivity increases you cause, and your other positive effects described elsewhere in this book. And so it is with immigrants; the complaint of taking jobs from natives, and thereby causing damage to them, is a foreshortened and misleading view. But it is the point of view which underlies the public policy of excluding immigrants, and of preventing foreign students and tourists from working, not only in the U.S. but also in many or most countries. CONCLUSIONS Taken together, the empirical studies reviewed in this chapter suggest that general immigration causes little or no unemployment at large, even in the first year or two; the same is true of low-income Hispanic immigration even among the groups most likely to be "displaced" by them. This finding emerges from studies using a variety of methods, studying various periods, and in a variety of geographical areas. Hispanic immigration into Texas and Los Angeles results in lower relative wages in particular occupations where Hispanics are a large presence. But some (even all) of this gap may be due to a real increase in wages among higher-skill groups rather than a real decrease in low-income groups. The extent to which the immediate effect of immigrants is distributed among unemployment of natives, unemployment of immigrants, wage declines, or even advance expectations of immigrants leading to expansion of business, cannot be sorted out at this time. The pattern probably varies considerably from place to place and time to time, depending upon the nature of the economy and of the economic conditions. Immigrants improve the situation of native workers by smoothing out employment over the business cycle when there is no constraint upon immigration. This is especially true of temporary (including illegal) immigrants, because they go and come in response to the job situation. Immigrants not only take jobs, they make jobs. They open new businesses that employ natives as well as other immigrants and themselves. And they do so in important numbers. The businesses immigrants start are at first small, of course. But surprisingly, small businesses are now the most important source of new jobs. Concerning policy: Where the society responds with restrictive action to the request of every group -- even every relatively poor group -- which is threatened by imports or immigrants or technological change or change in public tastes, there will be no end to the claimants, and the economy will suffer from drag. Not only will the overall size of the pie be smaller than otherwise, but as a result of the jockeying of the claimant groups, few or none of the groups do much better even relatively than if the society adopts a more open policy. Almost everyone would be better off in the long run if no group is protected. But this is cold comfort for a particular group that worries about injury, and few of us are willing to take the broader and longer view when it is our ox being gored. It is my view that we do best by trying to help people move and adjust to new occupations, rather than trying to shelter them from competition in their existing occupations. 86-85a Dispda12 12/11/87 FOOTNOTES 1Between 1970 and 1980, for example, net black employment in Los Angeles increased by 107,000 persons, with 98,000 of this total in white-collar occupations. Conversely, Mexican immigrants in Los Angeles who arrived during the 1970s held 210,000 jobs in 1980, but fewer than 25,000 of these were in white-collar employment. Even though the total number of jobs in the low-skill occupational categories of operative and laborer declined during the decade, the number of Mexican immigrants holding jobs in these occupations soared by 108,000. The rising job status of black women is especially noteworthy. In 1980, seven out of ten black women working in Los Angeles held white-collar jobs, a ratio higher than elsewhere in California or in the nation. There was only one black woman textile machine operator for every fifteen Hispanic women in such jobs in Los Angeles in 1980, and almost all the net additional jobs taken by black women since 1970 were in white-collar occupations. By contrast, only one out of every ten Mexican women who came to California during the l970s was employed in a white- collar occupation. (Muller and Espenshade, p. 101). 2This follows from their equation 3 on standard assumptions about perfect maximization and the implication that relative marginal products are equal to relative factor prices. 4-?? 86-85a Dispda12 1/21/87