As everyone knows, immigration is dramatically transforming the demographic terrain of New York City. Most people interested in understanding this phenomenon have focused their attention on the foreign born. Today, New York City’s population of about 7.5 million contains several million people born outside the U.S. Were it not for their arrival over the last several decades, the city would be smaller, probably poorer and less vibrant, and certainly less diverse. The growth of immigrant communities, whether the Chinese of Sunset Park, East Elmhurst, and Flushing, the Dominicans of Washington Heights, the West Indians of Flatbush, or the Koreans of Bayside, has captured both popular and scholarly attention. As immigration continues apace, the fate of the first generation in American will continue to be an important concern.
    Over time, however, the full implications of the "new immigration" will be worked out not by the first generation, but by and through the lives of their children, the second generation. As their parents become a larger share of the adult population, young in New York City and surrounding areas are also increasingly more likely to be the second generation. This is all the more likely because native born fertility rates are low and dropping compared to rates among immigrants (though they, too, fall over time). As a result, young people are not only more likely to be black, Latino, or Asian than the city’s population as a whole, but they are more likely to be children of immigrants. This group is truly suspended between two worlds – the immigrant sending societies and traditional cultures of their parents and the vortex of youth culture, highly variable schools, and problematic labor market of the contemporary city. They address this matrix of barriers and opportunities, or chutes and ladders with a varying mix of family assets, neighborhood resources, group positions, and individual and family strategies.

The Study
    For the past several years, funding from a consortium of foundations has enabled Philip Kasinitz, Mary Waters, and myself, with help from an extremely able staff, to conduct a large scale study of the immigrant second generation in New York City. During 1999, we completed 3,424 interviews with men and women aged 18 to 32 who lived in New York City (except Staten Island) or the inner suburban areas of Nassau and Westchester Counties and Northeastern New Jersey. We chose them to represent eight groups: those whose parents were from 1) China (including Taiwan and Hong Kong), 2) the Dominican Republic, 3) the West Indies (including Guyana but excluding Haiti), 4) the South American countries of Colombia, Ecuador, and Peru (designated CEP in the remainder of this paper), and 5) Jewish immigrants from the former Soviet Union. For comparison purposes, we also surveyed those whose parents were 6) native born whites, 7) blacks, and 8) Puerto Ricans. The natives were all born in the mainland U.S. About two-thirds of the immigrant second generation respondents were born in the U.S., mostly in New York City, while one-third were born abroad but arrived in the U.S. by age 12 and had lived here for at least ten years. We asked them about their family background, what neighborhoods they lived in and what schools they had attended, what kinds of jobs they had held, what their experiences were with a wide range of official institutions and programs, what activities they participated in, what languages they spoke, and how they felt about a wide range of issues. These data provide us the best picture yet available on the life situations of a representative cross section of the major racial and ethnic groups in metropolitan New York.
    The purpose of this paper is to give a first look at one of the outcomes to be covered in our study: educational attainment. It is a highly preliminary effort to describe the main contours within and across our study groups, to point out some of the more interesting and significant features of that terrain, and to articulate the questions these conditions prompt. Since our data analysis has only recently begun, it will be some time before we can reach well grounded conclusions. In short, this is an exercise more in raising questions than in answering them.

Outcomes
    The basic distribution of educational outcomes across our groups is given in Table 1. It shoes considerable variation within and across the groups. A moment of closer study, aided by the two shaded bands indicating failure to have graduated from high school and success at having earned a bachelor’s degree, reveals a number of patterns, however. (Note that, at age 18, all of our respondents would have graduated from high school if they had followed the prescribed path.) The most obvious is also the most expected: native born whites have achieved the highest level of education, with half graduating from college and only 7.3 percent having failed to get a high school diploma. Moreover the white immigrant group, Russians, also did well.

Table 1 about here
 
    Several other patterns merit comment, however. The first is that blacks (whether native or second generation) on the whole did better than Latinos (though the Colombian, Ecuadoran, and Peruvian second generation matched their performance). The national stereotype puts blacks at the bottom. While far from the top, blacks in New York City are also some distance from the bottom. While there are some obvious reasons why this might be so, including facility with English, higher family incomes, and the like, this is a matter that clearly requires sustained analysis.
    Second, while the differences are small, there is a persistent tendency for second generation groups to perform slightly better than their native born counterparts. West Indians are three percentage points less likely to be high school dropouts than native blacks; Dominicans are ten percentage points less likely than Puerto Ricans, and the CEP group is 17 points less likely. At the other end of the scale, Dominicans and CEP are 3.9 and 5.7 percentage points more likely than Puerto Ricans to have graduated from college and West Indians are also a bit above native blacks. Why the second generation should show small but consistent performance advantages over their native born minority counterparts also demands careful attention. One could think of many family background and group position characteristics that might account for such differences (and we will shortly examine some of them). But an especially troubling aspect of these two patterns is that they seem to interact to put Puerto Rican natives at a consistent educational disadvantage.
    A third, quite remarkable pattern is the consistently strong performance of the Chinese second generation. Their parents have entered the U. S. with low levels of education, occupational status, and income, yet Chinese youngsters have achieved impressive results. Their high school dropout rate lower than any other group but whites, while nearly a third have graduated from college. (Since, as we shall see later, many whites are young college graduates who moved to New York to advance their careers, this achievement is particularly striking.) That Chinese seem to be a "model minority," with all the burdens and challenges that term implies, is not news. But that they have achieved this status against considerable odds that have stymied youngsters from other backgrounds is clearly worthy of close examination.
    If the bad news is that most youngsters from backgrounds other than native white are facing serious challenges in achieving the middle class norm of graduating from college, the good news is that New York is a city of strivers. As Table 2 shows, high proportions of young people from every group are enrolled in school. Only among the natives, whether white, black, or Puerto Rican, does the enrollment rate fall below one-third. All the second generation groups are involved in education, including more than half the Russians and Chinese, almost half of the West Indians, and two-fifths of the Dominicans and CEP. The low enrollment figure among whites might be explained by high levels of attainment. We cannot account for the relatively low rates among native blacks and Puerto Ricans this way. More about these contours is revealed in Table 3, which controls for age. Enrollment rates are practically universal for younger Russian and Chinese second generation individuals, while rates are comparative high among older West Indians, Dominicans, and CEP. (Our Russian and Chinese sample also tends to be younger because of the recency of immigration among these groups compared to the others.) This would suggest that some second generation groups are early studiers while others are late studiers, with the result that, over a longer period of time, they will close at least some of the observed attainment gaps. Once more, it is troubling that a native born group, Puerto Ricans, have the lowest enrollment rate for any age cohort.

Tables 2 and 3 about here

Confounding Factors
    We must be careful to determine that these results are not just an artifact of the samples we collected. That is to say, we must compare apples with apples. Except for gender distribution in some groups, we believe our sample accurately reflects the underlying populations. But these underlying distributions differ on some characteristics that are important to outcomes. Some groups, for example, are simply younger (and more likely to be 1.5 than second generation) than others because their group migrated to the U.S. more recently than other groups. All other things being equal, they cannot be expected to have completed as much education as an older group. Selective migration into and out of New York may also be a factor. In the case of whites, educational attainment is elevated by the fact that many young white college educated people move to New York to advance their careers. In the case of Puerto Ricans, it may be that many better off families have moved to the far suburbs, beyond our sampling area, leaving only the least mobile and worst off to be found in our sample.
    One important corrective frame is offered in Table 4, showing educational attainment by group in terms of where people were born. It shows that half the native white sample was born in the suburbs or outside the New York metro area, and that they have much higher levels of educational attainment than those whites who were born and grew up in New York City. (For example, only a third of city-born whites completed a BA, while three-quarters of those who moved in from elsewhere had a BA.) If we restricted comparisons only to those who grew up in and around the city, the educational attainment gaps between whites and other groups would be much narrower. The same general pattern holds for native blacks, though at far lower overall levels of educational attainment. (Too few people grew up elsewhere in the U.S. for this to be a factor with any of the other groups.) A similar problem might arise if 1.5 generation individuals were outperforming second generation individuals. Table 4 reveals this is not the case. Only among Russians do 1.5ers outperform the second generation, but the second generation is quite small and disproportionately young in this group.

Table 4 about here
    Given the gender balance differences across our sample groups, the question of whether there are gender differences in educational outcomes is central. Table 5 presents results on this question. Except for native whites, females have higher levels of educational attainment than males across the board. In terms of college graduation rates, females do from 2.5 to 7.3 percentage points better than males, with the largest gap being among Chinese and the smallest among Puerto Ricans (who also have the lowest college graduation rates). Similarly, except for native whites and Puerto Ricans, males have higher rates of dropping out of high school than females across the groups. (This departure from the general pattern contributes to why Puerto Ricans appear to be performing the worst in our sample, since this group both has the lowest female educational performance and the next to highest share of females.) The sources of this apparent gender difference in educational attainment will be another major focus of our analysis. Young males are more likely to have negative encounters with the police and other authorities and may face greater pressures to enter the labor market at an early stage that impedes their educational progress. On the other hand, females face the even more difficult challenge of having the primary responsibility for bearing and (often by male default) rearing children.

Table 5 about here

Family Background
    With the preliminaries of age, place of birth, and gender out of the way, let us turn to the sources of educational attainment. By far the strongest determinant, according to the literature, are father’s and mother’s level of education. In general, this relationship is also strongly borne out in our data. But there are also significant twists to the story. As Table 6, respondent’s educational attainment by father’s educational attainment, shows, with one exception, there is a strong and consistent relationship across the groups: the more the father’s education, the more the respondent’s education. Moreover the relationship is not linear. Whether or not the father has received a high school diploma makes a sharp difference both in the child’s high school dropout rate and college graduation rate, cutting the former in half (except for West Indians, Dominicans, and CEP) and significantly increasing the latter (except for Russians, West Indians, and Dominicans). (Why a father’s presence or lack of a high school diploma should not matter in these groups is a question that must be left to a later investigation).
    A second jump is taken when a father has a college education. High school dropout rates are once more reduced compared to fathers with only a high school education (except for Russians and the troubling case of Puerto Ricans) and college graduation rates are significantly enhanced, doubling in some groups. Here, clearly, is another source of native white advantage: the "conversion rate" of transferring a father’s human capital to the child is much higher, 70.1 percent, than for the other groups, which all fall below 50 percent, sometimes far below, and a far higher share of whites have college graduate fathers than any other group. Among the other groups, the Chinese have the highest rate of conversion, despite the fact that they are a relatively young group. As a white group, the strong Russian performance is not a surprise, with a large number also with "some college," likely currently enrolled (they, too, are relatively young.) Among the native and second generation minority groups, native blacks with college educated fathers show the strongest performance (though it is still only half that of whites), though West Indians seem to have large numbers still coming along in the educational system, as do the Dominicans and CEP.
    The Puerto Rican numbers in this table once more raise the question of anomalously poor outcomes. The high school dropout rate among Puerto Rican respondents whose fathers have graduated from college is three times higher than the next highest group, Dominicans. Moreover, the college graduation rate is among the lowest, with only the second generation Dominicans and CEP having lower rates. (Puerto Ricans also have the fewest college educated fathers.) In other groups, a low college graduation rate among those with college educated fathers is offset by relatively high levels of people with "some college," many of whom may still be enrolled. Not so for the Puerto Ricans.

Table 6 about here
    A similar but somewhat weaker pattern holds between the mother’s educational attainment and that of the child. This is important because many of our respondents grew up in households where their father was absent part or all of the time. (A quarter of our respondents did not know what their father’s educational attainment was, while ten percent did not know their mother’s.) Table 7 shows this set of relationships. As in the other tables, the darker bands show high school dropout rate, while the lighter band shows college graduation rate. Again, more mother’s education is generally associated with more respondent’s education. Again, a high percentage of Puerto Rican respondents whose mothers dropped out of high school also dropped out of high school (and Puerto Ricans had a high share of mothers and fathers who were high school dropouts).
    We might accept this as a durable relationship, but Dominicans, who had similar maternal and paternal educational backgrounds, fared fare better. As in Table 6, Table 7 shows that a low percentage of Puerto Rican respondents with college educated mothers graduated from college, while a high percentage dropped out of high school. Unlike the case of paternal education, native blacks join Puerto Ricans in this table as having relatively high dropout rates and low college graduation rates among those whose mothers have a college degree. The case of the Chinese also cuts against the inevitability of low parental educational attainment leading to low graduation rates among the present generation. A remarkable 29.4 percent of Chinese whose mothers lack a high school diploma had graduated from college, and fully one-third of the Chinese sample fell into this category. And this rate jumps for Chinese mothers with high school diplomas or college degrees. Moreover half the Chinese second generation whose mothers had college degrees had also gotten college degrees, a higher rate than any other group but native whites (a pool, one must remember, drawn from across the country, not just New York City). Clearly, having poorly educated parents is not the barrier for Chinese that it is for Puerto Ricans and Dominicans.
    One reason this might be so is indicated in Table 8, which shows whether one grew up with both biological parents or not across the groups. Across the board, high school dropout rates are higher and college graduation rates lower for those who did not grow up with both biological parents. Moreover the effect of this factor apparently differs across groups, being most potent among Puerto Ricans, also potent among native blacks, West Indians, and Dominicans, less potent among CEP and Chinese, and least potent among native whites and Russians. This suggests a model of cumulative disadvantage, in which the negative effect of having mothers and fathers with low levels of education is compounded by the frequency with which one must rely on only one biological parent, usually the mother. Note that family form growing up differed radically across the groups. Over half of native blacks, almost half of West Indians and Puerto Ricans, and two-fifths of Dominicans grew up without both biological parents, while only a third of CEP, a quarter of native whites, and less than one-fifth of Russians and Chinese faced this disadvantage. The collective and cumulative nature of family human (and other) capital among the Chinese most likely help them overcome low individual levels among parents as compared to the other minority and second generation groups.

Table 8 about here

Respondents’ Lives
    Not everything is determined by parental background, of course. The barriers and opportunities facing young adults, the ladders they must climb and the chutes down which they may fall, also have a large impact on educational outcomes. If one studies hard, avoids the perils of the street, and above all defers childbearing, one is far more likely to climb the ladder. If one gets arrested or has a child, one may fall down a chute from which it is extremely hard, though not always impossible, to climb back up. While many possible experiences that our respondents have while growing up may influence educational outcomes, and all will receive attention as our analysis proceeds, let us focus on just one central experience here: whether or not one has a child. The consequences of this factor for educational outcomes is given in Table 9. Clearly it has major importance.

Table 9 about here
    In general, having a child reduces the chance that one will have graduated from college and increases the chance of dropping out of high school. (Though this table does not control for gender, the impact is largest among females, but is also significant for males, except among native whites and West Indians.) The magnitude of the impact on high school dropout rates varies across groups in an interesting way. Childbearing produces the smallest increases in dropouts among Russians, Chinese, and West Indians (where, quite remarkably, dropout rates are lower and college graduation rates are higher among those with children that those who do not!) It is largest among native blacks and the Latino groups, Puerto Ricans, Dominicans, and CEP. Logically, one might think that this is again related to the degrees to which families and kin networks are willing and able to pool support for children of the current generation of respondents. Similarly, childbearing generally reduces a group’s college graduation rate significantly, but the rates are actually higher among West Indian second generation parents, this time joined by Russians. (Why childbearing among West Indians might have such an anomalous impact is another puzzle which must be unravelled later.) The bottom line is that having children can really slow down the rate of educational attainment for most young New Yorkers, and that this chute is steeper for some groups than others.

Multivariate Analysis
    At this point, we can begin to consider the joint effects of these different factors through multivariate analysis. Here, we present only two models that are highly preliminary and meant only to give a sense of how the subsequent analysis might go. The first model includes the various individual and family background characteristics that are typically associated with educational status attainment models. (The data are weighted to show their relationship to the actual universe surveyed. In other words, Chinese cases do not count more in this model just because we oversampled this group.) The dependent variable is the four-category collapsed indicators of educational attainment, with 1 being high school dropout (including those who have a GED), while 4 is a college graduate. This is not an ordinal variable, yet it gives the general sweep of educational outcomes of interest. The first model begins by controlling for age, since older people by definition have had more time to complete their education. Since our samples differ in terms of gender make-up, and gender itself is an important distinction, it also controls for that factor. It also controls for mother’s and father’s level of education, which the literature holds to be crucial determinants of a child’s educational outcomes. Finally, the model includes some features of the respondent’s experience growing up, including the number of times moved, the self-reported high school grade point average (which may measure both the respondent’s ability and the way in which the school system has processed the individual), whether one has been arrested, and whether one has had a child or lives with a partner. Results are given in Model 1.

Model 1 about here

    The model has a respectable though not large R² of .317, is highly significant overall, and contains terms that are highly significant except for the case of partnering. The signs are all in the expected directions. Moving a lot, being arrested, living with someone, and, especially, having children are negatively associated with educational attainment, while age, parental education, and GPA are positively associated. Age, GPA, and father’s education have the strongest positive impact (with standardized coefficients of .415, .225, and .105), while childbearing has far and away the most negative impact (-.282). Interestingly, net of all these factors, males seem to have a slight advantage, notwithstanding the raw distributions given in Table 5.
    But does group still matter? And if so, what lies behind group differences? We can appreciate the dimensions of this issue in Model 2, which enters dummy variables for each of our comparison groups. (The results are presented relative to the omitted group, native whites.) The addition of these factors modestly boosts the R² to .341 and some of the group variables are sizeable and significant. Specifically, as indicated in the raw data, the Chinese appear to be doing somewhat better than native whites, all other factors taken into account, while native blacks, Puerto Ricans, and to a lesser extent Dominicans are doing worse. (All these coefficients are highly significant in the statistical sense.) Indeed, Model 2 reinforces our previous discussion of the Chinese second generation as a positive outlier and Puerto Rican natives as a negative outlier. It also confirms that West Indians are not lagging whites as definitively as native blacks, CEP and Dominicans are not lagging whites as markedly as Puerto Ricans. It is less clear from Model 2 that blacks are doing better than Latinos relative to whites because, after controlling for all these factors, the second generation minorities seem to be doing better than the native minorities. The inclusion of the group dummies also does not change the size or significance of the coefficients in Model 1, suggesting that other factors than these underlie group differences and that future analysis needs to investigate what they might be.

Model 2 about here

Next Steps
    Our data set contains many relevant variables that we have not yet analyzed. We also conducted 440 in-depth, open-ended interviews with our respondents that contain many probing discussions about their experiences with the education system and how they interact with family background, labor market experiences, current household trajectories, and many other factors. As we review this qualitative data, it will undoubtedly reveal many new insights. This preliminary and initial view of our data, however, suggests that we need to see individuals not as isolated units but as part of family systems which can sometimes provide support to help people past (or out of) the chutes into which some may fall and up the ladders which others may find. Looking to the parental generation, it is clear our respondents come from very different positions of initial advantage. It is also clear that the dissolution of parental relationships, while apparent across all groups, happens in different ways and with different consequences for different groups, and indeed for different people within those groups. Looking to our respondents’ generation, it is also clear that both ability and the way its potential is shaped by individual choice and institutional circumstances make a difference. Some people defer child bearing and focus on making immediate educational progress; this tendency is more pronounced in some groups, particularly the second generation Chinese, for reasons that may be particular to that group and which need to be articulated and tested. Other people bear children and form intimate relationships with their peers, sometimes taking the form of marriage, which evidently slows their progress up the educational ladder (and sometimes constitutes a slide away from the educational system altogether). This once more takes different forms in different groups, owing to a group dynamic that must be uncovered and tested against the data.
    Beyond these family factors, we are also very interested in knowing whether and how the kinds of neighborhoods people grow up in, the ways they participate in the local civic fabric, and especially the kinds of schools they attend, particularly high schools, effect their educational outcomes. We believe that, net of all other family and individual demographic characteristics, such factors do have an independent and strong impact. And there are sure to be many individual attitudinal and behavior factors (for example hours spend on homework versus watching TV) reflecting the nature of youth culture in New York City that will also come into play.
    Though we are at the beginning of an interesting analytic journey, we already know some of the major features that define the contours of educational outcomes among young adult New Yorkers, and we have good ideas about some of their sources. Native white advantage is clear, though if one subtracts those who came to seek New York’s opportunities, that advantage diminishes. Among our other respondents, the apparent finding that blacks may be doing somewhat better than Latinos, may dissolve in the face of the trend that the second generation is doing better than their native minority counterparts. This may both counter and confirm aspects of the "segmented assimilation" hypothesis. Finally, we are left with two durable anomalies, the Chinese and Puerto Ricans. The former group is clearly doing better than the individually measured characteristics of its parental generation might indicate, while the latter is doing worse. Learning exactly how and why this comes about will be a major challenge for our project.

ENDNOTES