IS THE NEW IMMIGRATION LESS EDUCATED THAN THE OLD? EDUCATIONAL TRENDS OF IMMIGRANTS INTO THE U.S. Julian L. Simon and Ather Akbari It is commonly stated that the newer immigration from Southeastern Europe is more unskilled than the older immigration from Northwestern Europe... If this long-accepted belief in the higher per- centage of skilled workers among the old immigration is, after all, erroneous, why has it been cherished for so long?... [I]t is the custom of each generation to view the immigrants of its day as inferior to the stock that once came over. Some of this prejudice against the newer races has not been wholly absent from some of the writings upon American immigration problems. (Douglas, 1919, pp. 393, 402-3.) It is commonly stated that the educational level of immigrants - their "quality", as it is commonly referred to - has not been rising, or even falling. Barry Chiswick wrote: "There has been a secular rise in the schooling level of adult native- born men... Yet a similar pattern is not to be found among immigrants to the United States", and he cited mean numbers of years of schooling for immigrants in 1950-59, 1965-69, and 1975- 1980 as 12.2, 11.7, and 12.3 respectively (1986, p. 179). George Borjas has made this same assertion with emphasis, and has been the source of many proposition in many public statements (see citations below). This assertion echoes the view that Paul Douglas confronted in 1919 - that the skill level of immigrants had been falling prior to that date. (See headnote above). Though when Borjas is careful with his language he usually (though not always)<1> is referring to amounts of immigrants' education relative to natives, he often writes or speaks in such fashion that the unwary reader can easily read absolute rather than relative amounts of education - that is, present levels relative to past levels.<2> Scholars have inferred that Borjas is writing about an absolute decline. Bashir Ahmed gave a paper to the Population Association of America on immigrant education alone, without even referring to natives in his data, on the supposition that he was replying to Borjas, as we see in these lines that Ahmed quoted from: Borjas (1990, p. 219)...concluded that, "On average, immigrants in recent waves have relatively less schooling, weaker labor market attachment, higher unemployment rates, lower wage rates, higher poverty rates, and higher rates of welfare participation than immigrants in the waves that arrived in the 1950s and 1960s." (Ahmed, 1993, p. 2) Ruben Rumbaut also read Borjas this way: "Such alarmist public debate...is reinforced by recent scholarly research about the putatively 'declining stock' of the newest arrivals, such as economist George Borjas' Friends or Strangers, which argues that recent immigrants are significantly less skilled and poorer than their predecessors" (1991). Barry Chiswick wrote in The New York Times that "the United States has experienced a large increase in low-skilled immigration in the last two decades" (1992), with no mention of relative levels. And in a review of Borjas's book in Journal of Economic Literature (1991), Chiswick read Borjas just this absolute way.<3> Chiswick wrote: The book has two major themes with varying degrees of statistical support. One is that there has been a decrease over at least the past half century in the skill level of immigrants. The other... Borjas emphasizes declining skill levels. He attributes declining skill levels to the dramatic change in the source countries of immigrants...(p. 627) Throughout the book Borjas emphasizes the declining skill level of immigrants and the great importance of this development for public policy. But if as he insists..." (p. 628). Law-and-economics professor Frank H. Buckley wrote: "George Borjas has suggested that the decline in the quality of U. S. immigrants is a consequence of the abandonment of national origins screening in 1965..." (1996, p. 95). I have included a few extra words at the end of each quote above so that the reader can see I am not quoting out of context.<4> This paper therefore provides evidence on the actual trends in the absolute educational levels of immigrants. It also provides evidence on the trends in immigrants' education relative to natives, and it attempts to make some sense of that rather complex body of data. At first thought it would seem straightforward and easy to assemble data that would serve as the basis for the comparisons one seeks to make. But the reader will see, we think, that the job is far more complex than one expects, and solid conclusions are therefore harder to arrive at than would be hoped for. It should be noted in passing that though immigrants with low education may be less desirable than immigrants with high education, all else equal, this does not imply that low-education immigrants are a negative economic force. Rather, all evidence (for reviews, see Simon, 1989 and Borjas, 1990) suggests that on balance low-skill immigrants have a positive economic effect upon natives, in part because of their complementary labor-force role to higher-income natives. TRENDS IN THE ABSOLUTE AMOUNTS OF EDUCATION OF IMMIGRANTS Absolute amounts of education are of interest in part because many have formed a wrong impression of their trends due to confusion with relative levels. But it is even more important that absolute levels of knowledge are of fundamental economic importance. When you need the plumbing fixed or a new computer designed, the extent of the workperson's own education and skill matter, rather than education and skill compared to others' (though the relative levels might influence a hiring decision). Hence we want to know the present state, and the trends, of such absolute levels. Table 1 shows the relevant data on trends in amounts of education of immigrants. There are several different kinds of data in the table - mean years of education, percentage with 8 years of education or less, and percentage with 16 years of education or more; these measures will be discussed separately. Table 1 Mean Years of Schooling Line 1 in Table 1 shows that the mean number of years of schooling increased for the period for which it is available. By itself this is not of great import, especially since data are unavailable for 1990. More will be said about this measure in the context of discussing the amount of schooling of immigrants relative to that of natives. The Very-Low and Very-High Education Categories 1. The 1990 census provides unusual fine-detail data for the education of immigrants age 25+ in various periods during the 1980s (U.S. Department of Commerce, 1993). The data within the 1980s on lines 3 and 8 of Table 1 contain these data just as reported, and the entry for 1980-1990 is an average of these published data. This series shows a tendency for increasing rather than decreasing education during the decade. The proportion in the low-education category (line 3) declined. And the proportion in the high-education category (line 7) increased. One should not expect to see reliable trends during such a short period of time, and therefore these data are not a solid basis to show a short- term increase, in part because the earliest observation in the 1980s includes data from a heavy wave of Vietnamese refugees. But these data certainly are conclusive in falsifying the claim that the amount of immigrants' education went down during the 1980s. Here it should be noted that the education data Borjas refers to all antedate 1980, even when he is writing after 1992 and the title of the article is "Immigration Research in the 1980s: A Turbulent Decade" (see his Table 3). 2. Also on lines 3 and 8 in Table 1 are data for 1975-1980 for persons 25+ from the 1980 census. These data, in combination with the 1980-1981 data, present the main complication in the overall analysis. The large difference - a reversal in trend - between those two sets of data from adjoining short periods, suggests that both sets must be seen in the context of a longer period. When compared against the average for 1980-1990 entry, the 1977-1980 seem "too high" and the 1980-1981 data seem "too low" even though the data are surely very accurate. Such variation must be the result of unusual population movements of some kind; the large Vietnamese refugee immigration does indeed figure in here but is not large enough to explain the entire reversal; some point between the two observations probably best represents the trend. These data showing the trend toward more education for immigrants since 1980 will be buttressed by the data discussed in paragraphs 3 and 4 just below. 3. Published results (including micro-filmed analyses) of censuses prior to 1980 do not give data by year of entry. Hence we must resort to other methods of estimating the data we seek for comparisons of various immigrant cohorts to each other. One such method is to exploit a single census - that of 1980 - that has data for various age-groups who entered in various earlier periods. In Table 1, lines 5 and 6, and 10 and 11, show respectively the proportions in the low-education (8 years or less) and high- education (4 years or more of college) of cohorts who were 20-34 and 25-34 at the time of arrival. That is, the groups who entered in the years 1975-1979 were 25-34 or and 30-34, respectively at census time, of course. These groups cannot be compared to any native groups, but they are useful to show trends among immigrants that we cannot get at in other ways because of the absence of data. The comparable groups caught in the 1980 census who entered in the various five-year periods further in the past were successively five years older - that is, 30-39, 35- 44, and 40-49, or 35-39, 40-44 and 45-49 - at the time of the census (e.g. 25-40 and 30-40 for those 30-39 and 35-39 at the census). Again, these two overlapping cohorts (25-34 and 30-34, etc.) cannot be compared against natives, but are useful, nevertheless, for indicating trends among immigrants. Both of these overlapping groups are shown because the 25-34 group (for 1980, and the comparable groups for earlier years) includes all of the group we aim to describe but also is fuzzy in also including some we do not aim to describe (those who were 20- 25 at the time of entry if they came in 1975, 21-25 at time of entry if they came in 1976, etc.). On the other hand, the 30-34 group includes only the people we aim to describe, but leaves out some of those we aim to describe (those who were 25 in 1975, 25 or 26 in 1976, etc.) The reason for providing data on both these overlapping sub-groups is that the estimates we aim at must be somewhere between the results for those two groups, and in fact we can average them (there not being a huge difference between them in most cases.) There will be a slight bias with these groups because the earlier cohorts have had more time to get additional education, but that bias runs against the hypothesis that immigrants have been getting to be of better "quality" rather than worse. Death affects matters somewhat, but probably only trivially. Though these 25-30 and 25-34 cohorts cannot be compared against natives, the cohorts who entered in various periods can usefully be compared against each other to see trends in this most-important young age group. And we see that there was a very large improvement in the educational level of immigrants from 1960-1980, as seen in both sets of groups - lines 5 and 6, and 10 and 11. This rapid jump before 1980 accounts in considerable part for the off-trend results seen in lines 3 and 8 for 1980- 1981, and for the mixed results (no improvement at the high- education end, while some improvement in the low-education end) between 1975-1980 and the 1980-1990 taken as a whole, for the 25+ group. These two cohorts of immigrants who arrived at young ages, taken in conjunction with the data for the 25+ groups that are available, supports the long-run conclusion of a general increase in the educational "quality" of immigrants, right through the 1980s. 4. Another indirect approach is to look at all those who were 25+ in earlier years as inferred from the 1980 census in a manner similar to that described in paragraph 4. These data - shown in lines 4 and 9 - are biased in that the very oldest group has had a greater likelihood of dying among the earlier entry cohorts. But since this bias goes in the opposite direction from what may be viewed as the hypothesis of this paper - that educational level has been increasing - it may not cause much of a problem. These data for the periods before 1975 - which may be considered the more reliable the later the entry cohort - lines 4 and 9 - show the same very rapid improvement leading up to 1980 as shown by the young cohorts alone (lines 5,6,10, and 11). And to the extent that the data in lines 4 and 9 are reliable, they (together with the data on lines 3 and 8) show a strong upward trend through the decades, right through the 1980s. One cannot construct a unified index that contains all the trend information in Table 1. But every single separate comparison on every index shows that immigrant cohorts have been better educated in each period than in each preceding period. Across the decades the data show the following: 1) Mean number of years of schooling have increased continuously. 2) The proportion of new immigrants with 8 years of education or less has trended downwards. And 3) the proportion with a college degree or even more education has gone up. Fix and Passell (1994) have distinguished between persons from "refugee" countries and others, and showed that the economic characteristics of the groups are very different. But given that refugees have frequently been an important part of the immigrant stream, and may well be in the future, we shall not try to make separate analyses of non-refugee immigrants. It is quite clear, however, that such analyses would strengthen the conclusions of this section rather than weaken them. THE AMOUNTS OF EDUCATION OF IMMIGRANTS RELATIVE TO NATIVES So far we have discussed the absolute educational level of immigrants. It is also of interest to compare the educational level of immigrants with that of the native labor force. When Borjas's work is read very carefully, one sees that all his statements refer to these comparative levels rather than to the absolute levels that he often seems to be referring to, and that so many readers have inferred is his subject.<5> When we move from discussing absolute levels to discussing relative levels of education the matter becomes more complicated, for two reasons: 1) There is the technical problem of lack of availability of data, and incomparability of many of the published data series - for example, the data for the "25 and over" age-group taken as a whole includes retired people rather than just people that are in the labor force. 2) The bi-modal distribution of the immigrants' educational distribution - proportionally large in both the low-education and the high-education categories, and low in the middle - is open to various interpretations. Table 2 shows the levels of education of natives for those measures and periods for which comparable data on both immigrants and natives are available. And Table 3 computes relative levels. Tables 2 and 3 Mean Years of Education. The only long time-series is average education<6>. Borjas calculated the mean years of education since 1940 for natives and immigrant cohorts, and this series shows a relative decline for immigrants over the period from 1940 to 1980. Because of a change in the way the questions were asked, it is not possible to compute a similar measure for the 1990 census. We do not believe, however, that this series since 1940 should be taken as an overall measure of a long-term trend, for two reasons: 1. For reasons of international conditions and the domestic economy, immigration cohorts at the early end of the period (1940) were very different in nature from the cohorts at the end of the period. The latter period is like present conditions, and also like conditions even earlier in U.S. history. In contrast, many of the immigrants who came between 1930 and 1950 were European war refugees. They tended to be people who had enough wealth and survival skills to be able to flee Europe, and they were also people who had enough education to lead them to believe that they would be able to find a job and make a living in the rough U.S. economic climate of the 1930s. Unskilled people abroad, on the other hand, responded to the depression period by not migrating to the U.S.; indeed, there was more U.S. emigration than immigration in the 1930s for this reason. (In this respect, immigration provided its usual salutary cyclical influence upon the labor force - increase it more when times are good than when times are bad.) Evidence that the data for refugees with high education for 1930 to 1950 are an aberration, just as are the data for 1980- 1981 for refugees with low education, may be found in even longer-run data. For earlier decades, P. J. Hill (1975) calculated a measure of the "labor force quality" of immigrants relative to that of natives, roughly equivalent to a percentage. His estimates are: 1870, 0.97; 1880, 0.99; 1890, 0.95; 1900, 0.97; 1910, 0.95; 1920, 0.93. (Bernard Bailyn's research [1986] on the Colonial-period Registry of Emigrants from Great Britain reveals much the same pattern.) Hill's data are plotted together with Borjas's data in Figure 1. We see there a long-run tendency for the mean labor "quality" of new immigrants to be slightly below the mean of the resident labor force, which makes the different sort of pattern for the 1940 and 1950 censuses seem an unusual and temporary happening rather than a reasonable benchmark with which the later data should be compared. Figure 1 2. Another reason to question the meaning of the mean- education data is that the series embodies (and masks) the very different tendencies at the two ends of the educational spectrum. A bi-modal distribution, where immigrants have a very wide range of educational levels, complements natives at both ends of the distribution; this has a very different economic meaning than does a distribution of immigrants heaped in the middle at similar education levels. (The observed bi-modal pattern suggests that immigrants fill empty niches rather than competing head-on with natives). The bi-modal educational distribution of immigrants also constitutes a plausible explanation for the observed long-term stability in the mean "quality" of immigrants relative to natives. That bi-modal distribution complements the middling- skill group which makes up most of the native labor force at any time. Owners of capital - farms and plantations in earlier decades and centuries, and then factories, and now institutions such as hospitals - demanded relatively unskilled laborers who themselves do not own much capital; that was the cause of the importation of slaves and indentured servants in earlier times. And growing affluent economies such as the U.S. represent better opportunities for highly skilled persons than do static poor economies. Proportions With Very Low and Very High Educations. Figures 2 and 3 graph the trends in Tables 1 and 2 for high and low education for immigrants and natives, which we will now discuss. The ratios of the proportions for immigrants and natives in Table 3 help elucidate the analysis. Figures 2 and 3 At the bottom of the educational spectrum there is a disproportionately large number of immigrants. And though the proportion of immigrants with 8 years or less of education has fallen over the decades, the proportion relative to natives has risen over the decades. To repeat a comment made earlier: The presence of many low- income immigrants does not by itself indicate that something bad is happening; rather, it may well represent a behavioral response by potential immigrants to the earlier-mentioned complementarity with the native labor-force distribution. That is, the absence of low-education natives to pick field crops and wash dishes naturally results in a heightened relative demand for such immigrant workers. If all natives, and all immigrants, were at their mean educational levels, the lower mean of immigrants might (or might not) imply a loss to natives through the tax-and- transfer mechanisms. At the top of the education spectrum, there also are disproportionate numbers of immigrants. All the data show that immigrants enhance the workforce and economy with ever-growing proportions of the highest-skilled labor. In addition to the material in Tables 2 and 3, and in Figure 3, Figure 4 shows the recent increase in the proportions of science and engineering graduate students in U. S. universities, and of PhD doctorates who are from abroad. Figure 4 also shows data for the number of foreign engineers and scientists granted permanent residency in the United States. It "rose dramatically - to more than 22,800 - in 1992, compared with 14,100 in 1991...The annual figure hovered near 11,000 during the 1980s" (Science, 22 July, 1994, p. 477). The number of foreign engineering students at all levels rose from under 10,000 in 1955 to more than 75,000 in the 1980s (National Research Council, 1988, p. 92). The ratio of foreign (persons with temporary visas) engineering doctorates in engineering to U. S. citizens rose from about 1:6 in 1970 to more than 1:1 in 1985 (National Research Council, 1988, p. 12). Perhaps most amazing, in just the ten years from 1975 to 1985, the proportions of engineering faculty, and of assistant professors in engineering, aged 35 or less, both rose from about 10 percent to about 50 percent of the total professors in those categories (National Research Council, 1988, p. 68) The trend is so strong that American-born PhD's have been making public protests about the matter, as seen on the Young Scientists Network on internet (Science, 22 July, 1994, p. 477). And the figure is constrained, not by the propensity of foreign scientists to immigrate but rather by U. S. law, a change in which (in 1990) seems responsible for the large increase since then. These talented scientists and engineers are arguably the most important intellectual assets that any country could have. More generally, the absolute and relative numbers of foreign students in the U.S. have been rising - from 154,580 and 1.5 percent of all college studentsin 1974-1975 to 438,620 and 3.0 percent of all college students in 1992-1993 (Washington Post, October 29, 1994, p. 7). These students represent a pool of potential well-educated immigrants whose entrance is restricted by U. S. law. Figure 4 Economic Evaluation of the Observed Pattern Inquiry into the economic meaning of the bi-modal distribution shows how data on mean or median years of schooling can be obscure and possibly misleading. What is the value of an additional year of elementary schooling compared to an additional year of college? Both enter equally into a calculation of average education. But a year of college is likely to be more valuable to the economy, as measured by both the cost of acquiring that additional year of education as well as by the additional earnings that it produces, than the year of elementary school<7>. Even from a pedagogical-psychological point of view, it would seem that an older student is ready and able to learn more in a year than is a young child. Age Distributions of Immigrants and Natives An important determinant of the observed educational levels of immigrants<8> is the ages at which they tend to come. As in all countries in all eras, migrants tend to be young adults just beginning their work lives. On average, young adults have more education than do their elders in the countries in which they grow up. The recent age distributions of the U. S. population and of new immigrants are shown in Figure 5. The relative youth of immigrants is a constant observed in all eras in all countries. Figure 5 In passing, we might note that the youthfulness of immigrants is also important because young persons pay taxes but use relatively few costly social programs (see Simon, 1989). DISCUSSION 1. Gathering data directly from the Census tapes rather than from published sources obviously would improve the quality of many of the analyses above, and would permit additional analyses. But to start from scratch, it is a very large job and would take several years, and hence would not be available to enter in mid-1990s debate on this subject. Furthermore, no data on date of entry was collected in the Censuses prior to 1980, as noted earlier. 2. Rumbaut's compilation (1994) of the proportions of immigrants whose occupation is given as "professional, executive, or manager" from the years 1967 to 1992 runs somewhat contrary to the trends in the education data presented above. The worldwide percentages are: 1967, 32.4; 1977, 36.0; 1977, 33.0; 1982, 31.9; 1987, 26.5; 1992, 24.0. Rumbaut warns that the sharp drop in 1992 is in good part due to Latin American IRCA legalizations, and especially of persons from Mexico. He also points out the huge range among countries in 1992 - from 63.1 percent from India to 1.7 percent from Mexico. But the general trend over the decades in these data still remains inexplicable to us. 3. As noted earlier, by analysing data on in-migrants from various countries Fix and Passell (1994) have shown that there are great differences in the mean characteristics of the various flows of immigrants - legal immigrants, refugees, undocumented. And DaVanzo et. al. (1994) have found huge differences between El Salvadoran and Filipino immigrants - for example, median family incomes of $11,484 and $47,323 respectively (p. 44) - that are certainly related to education. It is reasonable that consideration of a public policy toward some particular group - especially legal immigration, which is most under systematic governmental control - should consult the data on this group alone rather than the characteristics of the overall flow of in- migrants as discussed here. But the data over time for such sub- groups have not yet been organized sufficiently well to serve this purpose. CONCLUSION The impression held widely about trends in the education levels of recent immigrants, deriving from the writings of Borjas, is entirely incorrect. The amounts of education of immigrant cohorts has continued to rise; all the series show a tendency for increasing education. The average levels of immigrants' educations relative to those of the native labor force have remained much the same throughout history except for the period just before and after World War II. There was some decline in the relative levels of education of immigrants compared to natives in the 1970s and 1980s. This is largely due to the rapid upgrade in the educational level of natives starting after World War II. The data in this paper are for all immigrants taken together. But if the data are being brought to bear on policy decisions about the number of people to be admitted to the country under the legal immigration system, one should focus only on the educational levels of people who are admitted under that system; they tend to have higher educations than all immigrants taken together. page 1/article4 immeduc/May 3, 1996 REFERENCES Ahmed, Bashir, "Changing Characteristics of Recent Immi- grants in the United States: 1965-1991," Paper presented at the 1993 Population Association of America meetings, Cincinnati, Ohio, April 1-3. Bailyn, Bernard, "Does a Free-Born Englishman Have a Right to Emigrate?," American Heritage, February/March, 1986, pp. 24- 31. Borjas, George J., Friends or Strangers - The Impact of Immigrants on the U.S. Economy (New York: Basic Books, Inc., 1990). Borjas, George J., "Immigration Research in the 1980s: A Turbulent Decade", xerox, no date, contains reference to a 1993 publication "in press". Buckley, Frank H. , "The Political Economy of Immigration Policies", in International Review of Law and Economics, Vol 16, March, 1996, pp. 81-101 Chiswick, Barry, "Is the New Immigration Less Skilled Than the Old?", Journal of Labor Economics, 1986, vol. 4, #21, pp. 168-192. Chiswick, Barry, review of George Borjas, Friends or Strangers, Journal of Economic Literature, XXIX, June, 1991, pp. 627-628. Chiswick, Barry, letter to New York Times, June 28, 1992. Douglas, Paul H., "Is The New Immigration More Unskilled Than The Old?", in Journal of the American Statistical Association, June, 1919, pp. 393-403. DaVanzo, Julie, Jennifer Hawes-Dawson, R. Burciaga Valdez, and Georges Vernez, "Surveying Immigrant Communities", Center for Research on Immigration Policy (Rand), 1994 Fix, Michael, and Jeffrey S. Passel, "Immigration and Immigrants: Setting the Record Straight", draft, 1994. Hill, Peter J., "Relative Skill and Income Levels of Native and Foreign-Born Workers in the United States," Explorations in Economic History, 12, 1975, pp. 47-60. Lerner, J., and Rustum Roy, "Numbers, Origins, Economic Value and Quality of Technically Trained Immigrants into the United States", Scientometrics, 6, 1984, pp. 243-259, Moore, Stephen, "The Economic Case for More Immigrants", in Vernon M. Briggs and Stephen Moore, Still an Open Door? (Washington: American University Press, 1994). National Research Council, Foreign and Foreign-Born Engineers in the United States (Washington: National Academy Press, 1988). Rumbaut, Ruben G., "Passages to America: Perspectives on the New Immigration", in Alan Wolfe (ed.), America at Century's End (Berkeley: U. of California Press, 1991), pp. 208-244, 518- 526. Rumbaut, Ruben G., correspondence of April 26, 1994, updating Table 3 in his book Passages to America. Simon, Julian L., The Economic Consequences of Immigration (Boston: Basil Blackwell, 1989). Simon, Rita J. and Susan H. Alexander, The Ambivalent Welcome: Print Media, Public Opinion, and Immigration (Praeger: Westport, Conn., 1993). U.S. Department of Commerce, Economics and Statistics Administration, Bureau of the Census. The Foreign-Born Population in the United States, 1990 Census of Population (Washington: GPO, 1993) U.S. Department of Education, Office of Educational Research and Improvement., National Center for Education Statistics (92- 097) Digest of Education Statistics, 1992 (Washington: GPO, 1992). U.S. Department of Commerce, Bureau of the Census, Satistical Abstract of the United States (Washington: GPO, various years). U.S. Department of Commerce, Bureau of the Census, Historical Statistics of the United States, Colonial Times to 1970 (Washington: GPO, 1976) page 2/article4 immeduc/May 3, 1996 APPENDIX After this paper was completed, Borjas published the data shown in Figure 6. These data seem to show a greater relative decline in immigrants' educational levels compared to those of natives than do the data in Tables 1-3. Figure 6 One cause of the discrepancy is that Borjas's data pertain to persons 25-64 in the labor force, whereas our data are for all persons 25+. His category certainly is more relevant for the question under study here. But data so classified are not available as far back as are our data go [check book], and were not available in published tabulations from the Census Bureau that we used (Borjas did the tabulations himself.) And longer series provide a more solid basis for projections than do shorter series in this case. There also are inconsistencies in Borjas's results for the same cohorts from the different censuses that must qualify any conclusions that would be drawn from them. Both selected emigration and increments to education after immigrants have been in the country lead one to expect that the education levels of the remaining members of entry cohorts will rise with successive censuses. And this pattern is seen in the cohorts that entered from 1950 to 1969. But for the cohort that entered in 1970 to 1974 the pattern goes slightly the other direction from expected theoretically and on the basis of the earlier cohorts, and the pattern goes quite far in the "wrong" direction for the 1975-1979 cohort. For example, the average decrease in the drop-out rate between the first two census reports on the cohort was -8.3 percent for the earlier cohorts, but +6.0 percent for the 1975 cohort. The corresponding numbers for college graduates are +3.7 percent and -5.6 percent. Borjas does not comment on these inconsistencies which may be due at least in part from using 2/100 (1970 census) and 5/100 (1980 and 1990 censuses) Public Use Samples rather than larger samples (Borjas, 1994, p. 1673). Nor does he suggest econometric techniques for smoothing so as to surmount the inconsistencies. Indeed, the fewness of the data, much judgment would be required to make pattern that would seem sensible across censuses, across cohorts, and over time. Therefore we simply include the data here for the reader's comparison, and rely on the results from the entire censuses as shown in our tables. page 3/article4 immeduc/May 3, 1996 ENDNOTES **ENDNOTES** <1>: In his analytic work, the reference seems to be to absolute changes over time. He writes that in his 1985 paper "I argued that perhaps the cross-section data was indicating a change in chohort skills or "quality" across successive immigrant waves" (no date, p. 8). <2>: As this is being written, the front page of the Wall Street Journal quotes Borjas as saying about "new Americans" that "we have an increasing pool of lower-skilled labor" (November 28, 1994, p. 1). <3>: Chiswick did not qualify any of these remarks with respect to their (or Borjas's data) referring to relative levels. He only mentioned relative levels with respect to the low-skilled and high-skilled segments, and his comment was incorrect with respect to the former, as the data in this paper show: "There is no question that there has been a growth over the past half century of low-skilled immigration, in both absolute and relative terms, especially relaive to the increasing skills of the native labor force..." (1991, p. 627). <4>: A rare study pointing out that the opposite is true is that of Moore (in Briggs and Moore, 1994, Chapter 5), who reviewed studies of the 1980s and found agreement that direct and indirect evidence showed increasing education levels of immigrants. <5>: The logic of comparing recent cohorts of immigrants to the native labor force as a whole is not transparent; Borjas (1990) makes such a comparison without comment. The implicit purpose in making a comparison is to assess whether the incoming cohort will raise or lower the average level of the work force, but the basis of the comparison is not obvious. A similar practice is easier to justify in the case of cost-benefit analyses because there is no reason to expect that there will be systematic change over time in that case, whereas older people have less education than younger people, on average, and a younger cohort retains their educational levels as they get older. If one were not to discount future economic events, and if the sizes of native age-cohorts remained the same over time, then as long as the average of a group of immigrants were higher than the average of the younger half of the labor force, then even if the upgrading of the native labor force would continue at the rate of the past, the immigrant cohort would have a higher level than natives for the first half its work life, while being lower to the same extent for the second half of its work life (on some reasonable assumptions about changes among natives). The length of work life depends upon the age of the immigrants at arrival, of course, and the older immigrants have less education than do the younger ones. One can, however, examine the 25-34 age group of high-education immigrants, for which data are given in the table, and one can immediately see that their proportion among immigrants level is comparable to the native proportion for the labor force as a whole several decades back. But this analysis would have to be done in much greater detail if one wished to reach any firm conclusions on the matter. We shall here confine ourselves to presenting the data in form similar to that of Borjas (1990) for comparison with his. <6>: The standard government sources provide median rather than mean data. The median data closely parallel the mean data except for the 1970 census, where the official median figure is 12.2 and the Borjas mean figure is about 11.3. This discrepancy is not likely to be due to error, but rather to the properties of the median. <7>: A chart prepared by the U. S. Department of Commerce for the 1995 Economic Report of the President (p. 32) reports that for men in 1992 on average, less than 12 years of education (say 10 years) produced about $19,000 of annual income, or $1900 per year of education; 12 years of education produced on average about $28,000, or $2333 per year of education; and 16 or more years (say 17) produced about $46,000 of income, or $2700 per year of education. Of course age is not held constant here, and the calculations are sensitive to the representative amounts of education used (perhaps 10 years is a poor choice for the les- than-12 group). Also, the marginal increment from less than 12 years, to 12 years, seems higher per year than the increment from 12 years to 16 or more. Yet these data nevertheless suggest increasing gross return to additional years of income for the three categories. (Net return taking into account college costs and foregone earnings would be another matter.) I am grateful to Tom Moore for pointing out a defect in my earlier calculations. <8>: Another influence is that those persons migrate who consider themselves well-fitted to make a living in the new country. For such people the investment in moving costs and personal difficulties is more attractive than for those who do not expect to find good jobs and earnings in the new country. page 4/article4 immeduc/May 3, 1996