2. Evidence from the ‘Age of Mass Migration’
For obvious reasons, most empirical work on the labor market effects of
large-scale immigration has focused on the current (post-1980) surge, but
several studies have addressed America’s earlier “age of mass migration”
(1890-1914). The debate over what the evidence in each of these periods of mass
immigration can tell us about labor market effects is similarly
Two highly influential studies of the turn-of-the-century wage effects appeared in the 1990s, one by Claudia Goldin (1994) and the other by Timothy Hatton and Jeffrey Williamson (1998). Both concluded that there was fairly clear evidence that mass immigration had had substantial downward wage effects. Goldin found that “the impact of immigrants on the wages of laborers are generally negative and often substantial… in general, a one percentage point increase in the population share that was foreign born decreased wages by about 1 to 1.5 percent… In men’s clothing, which contained a large proportion of immigrants, wages were distinctly depressed in cities having an increase in the percentage of their population that was foreign born from 1899 to 1909.” At least as persuasive is the evidence from the polling booth. Goldin found that the lower the occupational wage increases in cities from 1907 to 1915, the more likely House Representatives were to vote to override President Wilson’s veto of proposed immigration restriction legislation (involving a literacy test). On the other hand, the higher the foreign-born share of the local population, the less likely representatives were to vote to override the veto (pp. 22-24). This suggests a sharp divide between native-born constituents concerned about labor market competition and foreign-born voters concerned about the native backlash against immigrants. There was at least a widely accepted popular perception of negative wage effects.
Summarizing their earlier work, Hatton and Williamson found that, for both the U.S. (immigration) and Britain (emigration) in the1870-1910 period, taking into account capital mobility “reduces dramatically the effect of migration on real wages.” Still, their estimates of the wage effects from migration are substantial: “In the absence of immigration, the U.S. real wage would have been about 9 percent higher (in an economy with much less capital), and in the absence of emigration, the British real wage would have been almost 7 percent lower than it actually was” (2006, p. 6). They note that Boyer et al. (1994) found similar results for Ireland as a result of the post-1851 emigration – without the mass out-migration there would have been “a 6 percent fall in both rural and urban real wages” (p. 6).
The Hatton-Williamson (2006, p. 5) results are generated by “a multi-sector competitive general equilibrium open economy model based on three factors (labor, capital, and land).” Given the inherent problems with data quality and the extreme simplicity of the model, the confidence with which they present their results seems rather breathtaking: how sure can we really be that the U.S. real wage would have been 9 percent higher, rather than, say, 8 or 10 percent higher, in the absence of immigration? New immigrants were such an important part of the American economy during this period that it seems hard to even imagine what it would have been like without any immigrants, much less to estimate and then assert such precise wage effects.
Carter and Sutch (1999, 2006) have challenged the Hatton and Williamson conclusions, in some cases by using the same data. As they put it, “The state-level data on immigration and native migrant flows do not support the conclusion that immigration during this period reduced the wages of residents” (2006, p. 12). But this critique, it turns out, rests on whether “wage reduction” means 1) lower absolute levels of real wages, or 2) lower wage levels than would otherwise have prevailed. Carter and Sutch agree that in all likelihood, mass immigration reduced wage growth. They write that “The estimates reported by Goldin and Hatton and Williamson of a wage setback for resident workers are only valid if we interpret them to suggest that resident wages would have risen even faster without immigration” (p. 17-18). But it seems reasonable to call a slowdown in wage growth attributable to a surge in the supply of immigrant labor a negative wage effect (or “a wage setback”) for native-born workers.
The conclusions by Goldin and Hatton/Williamson that new immigrants had downward wage effects on native-born wages during the Age of Mass Migration, even when interpreted as manifested in slower rates of wage growth, are viewed by Carter and Sutch to be “problematic” for two reasons. The first is that slower wage growth does not, in their view, provide a sound basis for explaining political opposition to immigration during the period. This may or may not be the case, but it is not clear why this should make the statistical findings problematic. The second reason is that the results reflect a static analysis that fails to recognize that the slower wage growth caused by the surge in low skill immigration “accelerated the rate of economic growth” (p. 18). If new immigrants promoted economic growth, labor demand may have increased as a result, leading in the longer run to rising wages even for the less-skilled. Indeed, Ottaviano and Peri (2006) show how such a dynamic approach applied to the post-1980 surge in U.S. immigration produces much more positive estimates of the overall labor market effects of immigrants, reflecting less negative wage effects for the least skilled (see below).
In assessing the Carter and Sutch (1999, 2006) proposition that a properly dynamic perspective might actually show positive immigrant effects of the first great immigration surge, even on less skilled native wages, the question is whether these less skilled immigrants really triggered significant productivity growth. Unfortunately, Carter and Sutch fail to address the possibility that a surplus pool of less-skilled workers may have done just the reverse, encouraging employers to adopt “low-road” labor intensive methods of production, which delays the introduction of skill-complementary (usually labor-saving) higher productivity technologies. There is, in fact, some evidence for this low-road effect, at least for recent years. Ethan Lewis (2004) found that for manufacturing plants in the 1980s and 1990s, “the adoption of advanced technologies by individual plants is significantly slowed by the presence of a greater relative supply of unskilled labor in the local labor market” (Card and Lewis, 2005, p. 25).