3. Evidence from the 2nd Great Surge: the Post-1980 U.S. Experience
As noted above, the recent surge in low-skill immigration has coincided with
a collapse in the wages of native-born dropouts. From 1990 to 2004, these
workers saw a decline in the level of real wages of almost 12 percent, and a
decline of more than 24 percent relative to the average wage (Ottaviano and
Peri, 2006, figure 4 and p. 32). It is hard to imagine, at least for
economists, that these two phenomena are unrelated. Yet, also noted above, the
professional consensus was that the aggregate statistical evidence failed to
show strong support for this straightforward supply-side crowding story
(Friedberg and Hunt, 1995; Smith and Edmonston, 1997; Borjas, 1999). For
example, David Card’s (1990) study of the Mariel boat lift appeared to show
that the sudden large influx of Cuban refugees to Miami had no significant
local wage effects. Other, more conventional “spatial correlation” studies
across metropolitan areas seemed to confirm this result.
George Borjas has been outspoken about what he thinks of this spatial correlation research. As he put it in his 1999 book, Heaven’s Gate, “Put bluntly, the spatial correlations are completely uninformative.” The main problem is that native workers “vote with their feet,” dispersing local wage effects to a larger geographical area (1999, p. 73). But it turns out that the evidence for substantial native outflows in response to immigrant competition is at best mixed. Borjas admits that the literature shows a “confusing set of results” (Borjas, 1999, fn 25). Borjas’ own evidence in Heaven’s Gate (1999) is remarkably unconvincing. One figure actually shows a positive association between the change in immigration and the change in native population across states for 1970-90. Even his preferred figure, which makes the changes relative to the 1960-70 trends, appears to indicate that what he terms a “clearly negative” relationship is largely the result of a single outlier (California).
But even if these correlations were stronger, Card counters that it is not total population, but the size of a specific skill group in the labor force that ought to be examined. When this is done there is no evidence of offsetting native outflows; more low skilled immigrant arrivals just tend to produce a higher share of low skill workers in the local labor market (Card, 2005, p. 9). It should be noted that many of these are his own studies (Card and DiNardo, 2000; Card, 2001; Card, 2005).
Although there remain mechanisms that will dilute a substantial immigrant wage effect (local demand, capital mobility, intercity trade), this evidence pointing to little native migration response to immigrant inflows would seem to put the spatial correlations approach back in play. In an important contribution, Card uses 1990 Census data for 175 cities to explore the effects of new immigrants (1985-1990) on 1989 wages for 6 broad occupations. Card concludes that “immigrant inflows over the late 1980s reduced the relative wages of laborers and less-skilled service workers in high immigrant cities by no more than 3%. The effects in other cities, and for other occupation groups that were less affected by new immigrant arrivals, were probably much smaller” (2001, p. 57). In a more recent study with Census data for 2000 using 325 metropolitan areas, Card finds no relationship between the fraction of immigrant high school dropouts (less than a high school degree) in the local labor market and the relative wages of native dropouts. He does find a small but significant negative effect on native employment rates.
What could explain these weak results? Card concludes, like Borjas a decade earlier, that it’s a “puzzle” (Card and Lewis, 2005, p. 26). For whatever reason, firms seem to adjust to large low-skill supply shocks mainly on the quantity side. They do so, according to Card and Lewis “by changes in skill intensity within narrow industries” (p. 26).
For Borjas, spatial correlation studies are inherently flawed. In the flexible U.S. economy, he seems to argue that empirical strategies should never assume isolated local labor markets, whatever the evidence on native migration flows. In any case, in the early 1990s Borjas turned to “factor proportions” time series analysis to explore immigrant wage effects in work co-authored with Richard Freeman and Larry Katz (1992, 1997). These studies begin with a time series estimation of the effects of changes in the relative supplies of college graduates and high school dropouts to their relative wages. An increase in the relative size of the supply of “dropouts” is found, as expected, to lower their wage relative to college graduates. The next step is to calculate the contribution of immigrants to the rise in the relative supply of dropouts. To the extent that immigrants increased the size of this low skill pool, they can be assigned responsibility for part of the decline in relative earnings. In this way, for 1980-88 Borjas et al. conclude that one quarter of the decline in relative dropout wages was due to immigrants (1992); for 1980-95, this relative wage effect if found to increase to half the overall relative wage decline (1997).
The problem with this empirical approach is that, since their data does not distinguish native- from foreign-born workers, the impacts of supply shifts on relative wages is base on the aggregate relationship, which requires them to assume that the native/foreign composition of dropouts does not matter. But as Friedberg and Hunt (1999, p. 355) point out, this tends to overstate the impact of immigration: “there is good reason to believe… that the impact of immigrant dropouts on natives to be lower than the impact of native dropouts on natives: immigrants and native dropouts are unlikely to be good substitutes for each other because even among high school dropouts immigrants have much less education.” It could be added that there are many other, perhaps more important, reasons to believe there is limited substitutability between native and foreign born dropouts, including differences in experience (see Borjas, 2003) and other sorts of skills (language, to begin with) as well as the fact that wage differences may reflect the fact that native- and foreign-born workers are concentrated in different locations, occupations, and industries.
While recognizing the limits of the factor proportions approach, Borjas (1999) finds it far better than the spatial correlations approach. He gives two reasons. First, “the cross-city comparison of native employment opportunities has failed to reveal with any degree of precision the impact that immigration has on the wage structure;” and second, “any interpretation of statistical correlations… requires a story” and the factor proportions “story” should be favored since it is the standard textbook supply and demand account (Borjas, 1999, p. 84, emphasis in the original). This seems to presume that the simple supply-demand story must be correct and the role of empirical research is simply to verify it. It is hard not to agree with Card (2005, p. 25) that “Underlying this argument is the belief that labor market competition posed by immigration has to affect native opportunities, so if we don’t find an impact, the research design must be flawed.”
In an important new paper, Borjas (2003) acknowledged that “the factor proportions approach is ultimately unsatisfactory,” but continues to maintain that “geographic dispersion in native employment opportunities is not an effective way for measuring the economic impact of immigration” (p. 6). Like Card’s recent work, in this new work he turns to an analysis of “skill groups,” but unlike Card he does so at the national level. Workers are aggregated into education-experience groups for each census year from 1960 to 2000 and immigrant wage effects are estimated across these 160 observations. The key assumption is that foreign and native born workers are reasonably close substitutes for one another within each of these skill groups. While he makes a convincing case that workers are more alike when work experience is used to define the skill groups, this is far from establishing that there is perfect, or even reasonably close, substitutability between native- and foreign-born workers within each of his education-experience groups.
Borjas concludes that his results are consistent with those “suggested by the simplest textbook model of a competitive labor market” (p. 36). Specifically, he finds that between 1980 and 2000 “this immigrant influx reduced the wage of the average native worker by 3.2 percent… with the wage falling by 8.9 percent for high school dropouts, 4.9 percent for college graduates, 2.6 percent for high school graduates, and barely changing for workers with some college” (p. 36).
Although broadly following Borjas’ approach, Ottaviano and Peri (2006, p. 3) point out that Borjas’ time series approach provides only “the partial effect of immigration on wages (as it omits all cross-interactions with other types of workers and with capital) and as such is uninformative on the overall effect of immigrants.” The demand for native- and foreign-born workers may be different within each of these education-experience groups. Like Borjas (2003), Ottaviano and Peri (2006) analyze education-experience groups at the national level, but they do not assume that foreign and native-born workers within each of the “skill” groups are perfect substitutes. They adopt a general equilibrium framework in which a production function “describes how these different types of workers interact with each other and with physical capital to produce output. Then, one can derive the demand for each type of labor, which depends on productivity and employment of the other labor types as well as on physical capital.” (p. 3).
These differences in modeling produce dramatically different empirical results. Ottaviano and Peri (2006, p. 4) conclude that 1) “the group of least educated U.S.-born workers suffers a significantly smaller wage loss than previously calculated” (e.g. by Borjas); and 2) that “all other groups of U.S.-born workers (with at least an high school degree) who account for 90% of the U.S.-born labor force in 2004, gained from immigration;” and 3) “the group whose wage was most negatively affected by immigration is, in our analysis, the group of previous immigrants.”
Interestingly, Manacorda, Manning and Wadsworth (2006) employ a similar methodology for the U.K. and find similar results: negative sizable wage effects are found for earlier cohorts of male immigrants, but not for native-born male workers. The authors conclude that “the impact of increased immigration on native wages is muted by the low substitutability between immigrants and natives” (p. 3).
It should be underlined that, like Ottaviano/Peri, even Card (2001) has found negative wage effects for some workers in at least some cities. For example, he concludes that “The results imply that immigrant inflows over the 1980s reduced wages and employment rates of low-skilled natives in traditional gateway cities like Miami and Los Angeles by 1-3 percentage points” (p. 22).
New York City is a traditional immigrant gateway and in tests across jobs (detailed occupation-industry cells), I also found some evidence that the share of recent immigrants in jobs had negative wage effects (Howell and Mueller, 2000). While Borjas insists on national level time series analysis and Card (and many others) have focused on cross-metropolitan (or state) tests, Howell and Mueller (2000) look for immigrant wage effects on white, black and Hispanic native-born male workers and on new immigrants themselves across detailed jobs in a single large metropolitan area. Admittedly, the native-flows problem may apply here as well: recent immigrants will tend to be attracted to the highest paying jobs and faced with intensifying employment and wage competition, native workers may respond by moving to other jobs and/or locations that are more sheltered from immigrant competition. As in the cross-metropolitan tests, native mobility will tend to bias downward the measured wage effects of the immigrant “supply-shock.” For this reason, a finding of substantial negative effects would be particularly notable.
We found that between 1980 and 1990 foreign-born workers increased from about 30 percent to almost 50 percent of all workers in the secondary (lowest quality) job segment. Employment growth for all three native-born groups – white, black and Hispanic – in both secondary and subordinate primary (middle quality) jobs was inversely related to the growth of recent immigrant employment. In contrast, among independent primary jobs, both native-born black and Hispanic employment growth was fastest in the jobs also experiencing rapid growth of recent immigrants. Controlling for mean education, time worked and job quality measures, the recent immigrant share has a significant negative impact on 1979 mean earnings in most of the tests. The “changes on changes” tests for 1979–89 show negative effects of the change in immigrant share on the change in mean earnings for all three native-born groups, although the findings are strongest for white and black workers in the independent primary and secondary segments – the best and worst jobs, which are located mainly in the service sector. In contrast, the negative effect of the growth in recent immigrant share on native-born Hispanic workers is concentrated in the blue-collar goods-producing jobs – the subordinate-primary segment.
Many other studies could be cited. Given space constraints, I will briefly mention just a few. Several studies published in the late 1990s also found some evidence of negative wage effects using the 1980 and 1990 Public Use Microdata Samples of the Census. For Los Angeles, Catanzarite (1998) found that “native workers suffer significant immigrant-related pay penalties, larger for blacks and Latinos than whites” (p. 147). In the Russell Sage volume Help or Hindrance?, both Cordelia Reimers (1998) and Kristin F. Butcher (1998) find that, across metropolitan areas, greater immigration was associated with lower African American earnings in the 1980s, particularly among the least skilled. And finally, much more recent work by Andrew Sum and his colleagues (Sum et al., 2006) finds that across states the probability of employment for youth was significantly negatively related to the size of the immigrant inflow into the state’s labor market. In the first test, immigrant flows between 2000 and 2003 had the largest negative impact on the probability of employment for 16-24 year old black dropouts. The second test, with more observations, also found strong negative effects from 2000-2004 immigrant inflows on 16-20 year old employment rates in 2004, particularly for black men (p. 5).