Convergence Between Black Immigrants and Black Natives Across ...

Convergence Between Black Immigrants and Black Natives
Across and Within Generations*
Alison Rauh**
This Version: October 28, 2013
Abstract
The number of black immigrants living in the US has increased 13-fold from 1970 to 2010, increasing
their share of the black population from 1% to 10%. Black immigrants’ labor market outcomes surpass
those of native blacks. This paper determines in how far the relative success of black immigrants is passed
on to the second generation. While blacks of the second generation have equal or higher education and
earnings levels than the first generation, the return on their unobservable characteristics is converging
to that of native blacks. Race premia are put into a broader context by comparing them to Hispanics,
Asians, and whites. Blacks are the only group that experiences a decrease in residual earnings when
moving from the first to the second generation. Black immigrants do not only converge to native blacks
across generations but also within a generation. For Asians and Hispanics, residual earnings decrease
monotonically with age of immigration. For blacks, the residual earnings-age of immigration profile is
upward sloping for those immigrating before the age of 15. Convergence across generations is mostly
driven by low-educated second generation blacks that drop out the labor force in greater numbers than
low-educated first generation immigrants do. Similarly, convergence within a generation is mostly driven
by low-educated blacks who immigrate when they are young dropping out of the labor force in greater
numbers than those who immigrate when they are older. A social interactions model with an assimilation
parameter that varies by age of immigration helps explain this phenomenon. When making their labor
force participation decision, immigrant men of all races, but not women, generally place more weight on
the characteristics of natives the earlier they immigrate.
*I am grateful to Erik Hurst and Kerwin Charles for their support at all stages of the project. I also thank Steven Durlauf, Steve
Levitt, Steve Davis, James Heckman, Alessandra Voena, Derek Neal, Dan Black, Keith Henwood, Christopher Rauh, the University
of Chicago Microlunch and Lifecycle Working Group for helpful comments. This material is based upon work supported by the
National Science Foundation Graduate Research Fellowship under Grant No. (DGE-1144082). Any opinion, findings, and conclusions
or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science
Foundation.
** University of Chicago, arauh@uchicago.edu.
1

1 Introduction
The number of black immigrants living in the US has increased 13-fold from 1970 to 2010, increasing their
share of the black population from 1% to 10%. Figure 1 illustrates how black immigration from the West
Indies and Africa exploded over the last few decades. Note that black immigrants also have a higher fertility
rate than native blacks, which will naturally increase the share of second generation blacks. 1
In 2011, black immigrant men earned close to $10,000 or 40% more than black native men. This premium
shrinks to $3,000 or 7% once we condition on being employed. The analysis of employment, education,
incarceration, and marriage outcomes, confirms that we are dealing with two completely different
subsets of the population. Black immigrant men are much more likely than native black men to be employed,
married, and not incarcerated. Schooling patterns point towards much higher college completion percentages
and less high school dropouts for black immigrant men. Similar analysis applies to women though the
variation in outcomes is generally compressed. 2
This paper intends to answer three questions. First, how different are first generation blacks from native
blacks and what are the forces underlying these differences? Second, in how far are observable characteristics
such as high education levels transmitted to the second generation? Third, in how far are less salient
characteristics such as motivation and dedication transmitted to the second generation? In addition to across
generation transmission, this paper will also explore within generation transmission. In other words, how do
immigrants that enter the US when they are young differ from immigrants who enter the country when they
are old, and why do they differ?
Throughout this paper I focus on 21-65 year olds in order to capture the working age population. I will
distinguish between three mutually exclusive groups: immigrants, individuals born in the US with immigrant
parents, and others. I will refer to them as the first generation, second generation, and natives. An individual
belongs to the second generation if one or both of the parents are born abroad. Blacks will be compared
to Hispanics, Asians, and whites, where blacks, Asians, and whites always refer to non-Hispanic blacks,
Asians, and whites. 3
This paper is organized as follows. Section 2 summarizes the previous literature. Section 3 presents
descriptive statistics to highlight the differences between native and immigrant blacks as well as the first
and second generation. Section 4 and 5 discuss the convergence with respect to earnings across and within
generations, respectively. Section 6 extends the convergence analysis to labor force participation patterns.
Section 7 introduces a theoretical model that helps explain both phenomena. Section 8 concludes.
1 In 2011, a black immigrant female above the age of 16 had a 48% higher probability of having a child than a native black female.
While black immigrant women have a larger total fertility rate compared to native black women, their teenage pregnancy rates are
half as high. In 2011, 4.2% of black native teenage girls had a child, while only 2.2% of black immigrant teenage girls did (based on
author’s calculations using ACS data).
2 In previous work I show that excluding black first and second immigrants increases inequality measures such as the black-white
wage, employment, college completion, and incarceration gap by amounts ranging from 5 to 25% (Rauh 2013).
3 For simplicity, I will often refer to Hispanics as a racial group, despite the fact that they would more correctly be defined as an
ethnic group.
2

2 Literature Review
Thomas Sowell (1978) was one of the first authors to compare economic outcomes of immigrant and native
blacks. He argued that black immigrants are a natural comparison group for black natives. Using the 1970
census, he finds a positive earnings gap between West Indian immigrants and native-born blacks and argues
that “color alone, or racism alone, is clearly not a sufficient explanation of the disparities within the black
population or between the black and white populations”. He concludes that “cultural traditions” of nativeborn
blacks stand in the way of economic achievement.
Since Sowell (1978), a few studies used the 1980 and 1990 census to evaluate the relative success of black
immigrants from the Caribbean (Butcher 1994, Kalmijn 1996) and from Africa (Kollehlon and Eule 2003).
Butcher (1994) finds that while male annual earnings of black immigrants overall are slightly below those
of native blacks, the subgroup of West indian men already display higher earnings. Kalmijn (1996) finds
that success of black Caribbean immigrants is limited to British Caribbeans. Butler and Herring (1991) and
Bogan and Darity (2008) evaluate entrepreneurship among black immigrants and find that self-employment
rates are higher for immigrant blacks than native blacks. Hamilton (2012) is one of the only researchers to
use post 2000 data to evaluate labor market differences of black immigrants and black natives.
What about the second generation? Analyzing the 1970 census cross-sectionally suggests that the second
generation earns more than both their parents and their children (Chiswick 1977 and Carliner 1980).
Borjas (1993) points out that these cross-sectional calculations ignore the relationship between the national
origin differentials found among immigrants and the differences found among ethnic groups of subsequent
generations.
In terms of black immigrants, there is limited evidence and the little evidence at hand is mixed. On
the one hand, there is the “second-generation decline” theory by Gans (1992). This hypothesis is closely
linked to the “fade to black” phenomenon discussed in the sociology literature. Waters (1999) and Kasinitz
et al. (2001) study the identity formation of children of West Indian immigrants. They identify three main
responses by the second generation Some choose an immigrant identity, stressing national origins. Others
choose a purely American and therefore African-American identity and “fade to black”. Finally some
choose something in between or what Kasinitz calls “an ethnic response” and carry a Caribbean-American
identity or African-not native-American identity. They point out that this might be unsuccessful if whites
do not recognize the difference. Empirically, Kasinitz et al. (2001) find some evidence that among West
Indians in Florida the grade point averages of the second generation are actually lower than those of first.
Suarez-Orozco and Suarez-Orozco (2001) find that greater length of residence in the US is associated with
lower academic motivation. On the other hand, some studies suggest that earnings of the second generation
are larger than those of first and third generations. Most recently, Park and Myers (2010), using a synthetic
cohort method, find overall intergenerational progress for the new second generation of all races. In summary,
there are two opposing forces at work. First, the parents of second generation blacks are relatively
successful, which would point towards positive outcomes. Second, the economic outcomes of black natives,
3

whom the second generation might identify with, have stagnated over the past few decades, which would
suggest negative assimilation.
Differential labor force attachment across generations is an important mechanism underlying the convergence
phenomenon. First generation immigrants are less likely to be citizens than members of the second
generation. Welfare programs are well-known to affect labor market decisions. Hence it is important to
discuss eligibility requirements for public benefits. One major barrier that is easily verifiable is citizenship.
Details for non-citizens are discussed in Appendix D.
3 Data and Descriptive Statistics
3.1 Data
Throughout this paper I use the current population survey data (CPS) from 1994 to 2012, and decennial
census level data from 1970 through 2000 from the Integrated Public Use Series (IPUMS) and the IPUMS
samples of the 2001 to 2011 American Community Survey (ACS). The post 1994 restriction for CPS data
is due to availability of questions and answers on own birthplace. In order to focus on the working age
population I restrict the data to individuals between the ages of 21 and 65. Census data is used from 1970
onwards, since black immigration in large numbers did not start until 1970. 4
3.2 Descriptive Statistics for the First Generation
Table 1 presents the descriptive statistics for males and females in 1980 and 2011. In each of the four
columns under both 1980 and 2011, the first two compare all blacks and all whites and the following two
split blacks into black natives and black immigrants. 5
In 1980 average male unconditional annual earnings for whites are $16,400. Native blacks earn about
60% of that ($9,460). Immigrant blacks on average earn $9,370, which is less then native blacks, making
the relative success of black immigrants a recent phenomenon. Breaking up unconditional earnings into
employment probabilities and annual earnings and conditional earnings we see that both display significant
black-white differences. 6
In 1980, black immigrant males are more likely to be employed than native blacks (75% versus 71%)
but they are still far away from the white male employment percentage of 84%. Note that as few as 57% of
black African immigrants are employed in 1980. In 1980, black immigrants are more comparable to native
blacks with respect to earnings, employment, and marriage rates than they are to whites. Lower incarceration
4 Data limitations and procedures are further discussed in section B of the Appendix.
5 Note that when using ACS and census data, native blacks include the second generation (very small subsample). This is due data
limitation discussed further in Appendix B.
6 Further separating conditional annual earnings out into wages, weeks and hours worked yields that all three are driving forces
behind the black-white gap. While West Indian and South American black immigrants are earning as much or more than native blacks,
black immigrants from Africa and other countries are earning about 90% of native blacks, thereby pulling the black immigrant average
down.
4

ates (1.26% versus 3.38%) and much higher educational investments, however, separate black immigrants
from native blacks. Relative high investments in human capital foreshadow a narrowing of the immigrant
black-white earnings gap.
The relative success of black immigrant women in the labor market is already apparent in 1980. 7
1980, black immigrant females have the highest probability of being employed (67%) compared to native
black women (58%) and white women (56%). Average female earnings conditional on working for black
immigrants are above those of native blacks and right behind those of whites.
By 2011, black immigrant males display a similar relative labor market success over native black men.
In 2011, average male unconditional annual earnings conditional on working for whites are $48,500. Native
black workers earn about half of that equaling $24,600. Black immigrant workers average unconditional
annual earnings equal $32,900 and therefore now earn about $10,000 or 44% more per year than native
black workers. 8
While the gaps in conditional earnings between black immigrants and natives are rather small, they
are immense in employment, incarceration, marriage, and schooling outcomes. 76% of both white males
and black immigrant males aged 21-65 were employed in 2011. Only 59% of black males aged 21-65 were
employed, a number that drops to 57% when black immigrants are excluded. 59% of native blacks’ education
is a high school degree or less, while only 43% of both whites and immigrant blacks fall into this category.
14% of native blacks obtain a college degree while 32% of whites and 30% of immigrant blacks obtain a
college degree or more. 9 Incarceration probabilities of males display the greatest differences. Incarceration
probabilities for white males and black immigrant males lie in between 1% and 2% while native black males
have an over 8% chance of being imprisoned in 2011. A much higher fraction of black immigrant males and
white males than native black males are married (53% and 57% versus 34%).
The African subgroup of black immigrant males display very interesting characteristics (see Table A.1
in the Appendix for break-down by country of origin). From 1980 to 2011 Africans went from being the
least likely black immigrant group to be employed to being the most likely black immigrant group to be
employed. Except for conditional earnings, which are close to the black immigrant average, black African
immigrants outperform not only black natives but also whites in most outcomes. Black African males have
a much higher chance of employment (80% versus 57%), lower chance of incarceration (1.05% versus
8.35%), and more college graduates (43% versus 14%) than black native males. Starting with Sowell in
1978 and followed by numerous studies in the 1990s, authors examined whether West Indian immigrants are
more successful than native blacks and Model’s 2008 book title asked the pertinent question “West Indian
7 When comparing earnings and employment outcomes of white and black women, differences in marriage propensities and labor
force participation rates are important to be kept in mind.
8 Analyzing conditional annual earnings and employment, we see that the difference between native and immigrant blacks is mostly
driven by employment probabilities. Separating annual earnings as before, we see that the black-white gaps in weeks and hours worked
have narrowed. Hence it is primarily wages and to a lesser degree hours and weeks worked that are driving the black-white conditional
earnings gap. The same is true for the native black-immigrant black conditional earnings gap.
9 While the high school and college degree categories include individuals with the relevant degree, the graduate school category
simply implies that individuals continued school after college (whether they graduated form their program or not.)
In
5

Immigrants: A black success story?” (Model 1991, 1995, 2008, Butcher 1994, Kalmijn 1997, Waters 1999).
These results raise the possibility of a new question: “African Immigrants: The black success story?” 10
Panel B of Table 1 displays the same mean outcomes for women in 2011. At first glance, we can
already see that the variation in outcomes across groups is much less than that of men. Conditional and
unconditional annual earnings of native black women are at 91% and 78% of those of white women, which
points to a much more equal distribution than that of men (64% and 48%). The gap in the fraction employed
between whites and native blacks falls from 20 percentage points to 5 points. Incarceration probabilities for
females are below 1% for all subgroups but native black women are still two and a half times more likely
to be imprisoned than white women. Educational patterns display the same relations as those for men but
the variability is compressed. Differences in marriage probabilities are an exception. White women aged
21-65 are more than twice as likely to be married as native black women (59% and 28%), while the black
immigrant female marriage probability lies in between these two rates (45%). 11
3.3 Descriptive Statistics for the Second Generation
Section 3.2 established that black first generation immigrants look very different than black natives. From a
theoretical perspective it is unclear what will happen to the second generation since there are two opposing
forces at work. First, the parents of second generation blacks are relatively successful, which would point
towards positive outcomes. Second, the economic outcomes of black natives, whom the second generation
might identify with, have stagnated over the past few decades, which would suggest negative assimilation.
Table 2 compares education and earnings outcomes of both generations for working age men. Due
to small sample sizes in the CPS, observations are pooled over the 5-year period from 2008-2012. At
first glance, it seems that earnings and employment outcomes of second generation black men are low and
comparable to those of native blacks (Panel A). Note, however, that the average second generation black is
more than 8 years younger than the first generation immigrant. Since earnings increase steeply until the mid
forties, column 5 uses inverse probability weighting to equalize the age distribution of the first and second
generation. Now sons of black immigrants earn $3000 or 8% more than the average first generation black
immigrant. 12 The fraction of second generation blacks with a college degree is 35% or 4 percentage points
higher than those of the first generation. In terms of marriage patterns, however, the second generation seems
to be converging to native blacks, rather than the first generation. While 60% of the black men in the first
generation are married, only about 43% of both natives and sons of immigrants are are married. This points
to different integration patterns between labor market and marriage market outcomes.
10 There are several reasons to be cautious about these results and they certainly require further investigation. First, many African
immigrants come to the US to attend college or graduate school so they are a very select group. This is not a problem with my hypothesis
per se since self-selectivity of immigrants is part of the reason black-white gaps are underestimated. Second, immigrants that do not
become naturalized are likely to be deported for severe crimes. Hence we would expect immigrants’ incarceration probabilities to be
lower even if their probability of being convicted of a crime that would lead to incarceration were identical.
11 See Charles and Luoh (2011) for a link between increased male incarceration rates and decreased female marriage rates.
12 The second generation premium equals 5% for white and Asian men and a striking 35% for Hispanic men.
6

For women both of these trends seem to be even more pronounced (Panel B). The second generation,
once adjusted to have the same age distribution as the first generation, has an earnings premium of $8,600
over native blacks, $6,700 over first generation, and $3,600 over whites. The fraction of second generation
blacks with a college degree is 6% or 2 percentage points higher than those of the first generation. Only
38% of women in the second generation are married, which is close to the average of native blacks (39%)
and much lower than marriage probabilities of the women of the first generation (57%) and white women
(65%).
Note that 81% of second generation black men are participate in the labor force and 70% are employed.
These numbers lie in between those of native and first generation immigrant blacks. This foreshadows
the convergence in labor force attachment and employment probabilities across generations, which will be
further discussed in the subsequent sections.
These results suggest that in terms of observable characteristics, with the exception of employment
probabilities, the second generation is following the relatively successful path laid out by their parents. They
have similar earnings and even higher human capital investments.
4 Convergence Across Generations
We have established that black immigrants have large earnings premiums over black natives. Furthermore,
we know that high human capital investments of the first generation carry over to the second generation.
Now let us look at more subtle differences between generations. Using CPS data, we want to control for
education, potential experience, state and year fixed effects and analyze the residual premium or penalty for
being a black first generation immigrant versus a member of the second generation.
Table 3 and 4 focus on black men and women, respectively. In column (1) all immigrants are included.
Columns (2) - (6) limit first generation immigrants to those who spent over 10 years in the US. This allows
us to focus only on permanent immigrants; immigrants who are more likely to be parents of the subsequent
second generation. We see that if all immigrants are included, both the coefficient on the first and second
generation are positive and insignificant. When recent immigrants are excluded, the coefficient on the first
generation increases to close to 5% and is significant at the 1% level. Including state fixed effects shrinks
the coefficient to about 3.5% and the significance level decreases to 5%. This suggests that location choice
explains some of the premium of first generation immigrants. The coefficient on the second generation is
smaller (around 2%) and insignificant in all regression. For women, the coefficient on being a first generation
immigrant is significant at the 1% level throughout all regressions. The second generation coefficient is
insignificant once state fixed effects are controlled for. The difference between the first and second generation
premium is about 2 percentage points for men and 3.5 percentage points. This difference, however, is
neither statistically significant for women nor for men.
Column (5) and (6) restrict the sample to labor force participants and employed individuals, respectively.
The convergence to natives disappears for both of these subsamples. This suggests that most of the con-
7

vergence is driven by the higher employment probabilities of the first generation compared to the second
generation. This hypothesis will be further explored in Section 6. For women, the convergence remains
although it falls to 2 percentage points for labor force participants and 1 percentage point for employed
women.
So far we compared first and second generation blacks to native blacks. In order to put the residual race
premium in a broader context, column (2) is now also performed for blacks, as well as Asians, Hispanics,
and whites (Table 5 and Table 6). The results are striking. Blacks are the only group that experience a
decrease in the race premium when moving from the first to the second generation. This is true for both men
and women. The increase in the race premia for other races reaches from one percentage point for white
men to 12 percentage points for Hispanic women.
Up until now, we compared races among themselves. In order to compare these estimates directly,
blacks, Hispanics, and Asians are now compared to a common reference group: white natives. Furthermore,
the subsequent plots will explore how these premia/penalties have evolved over time. Regressions control
for education, potential experience, and state fixed effects. In addition, the regression for first generation
immigrants also controls for time spent in the US. 13 Figures 2 and 3 show the earnings premium or penalty
for being black, Asian, or Hispanic for men and women, respectively. The reference group is white native
men for Figure 2 and white native women for Figure 3. 14 The panels are ordered in terms of time spent in
the US, i.e the first, the second, and finally the third generation and beyond. 15
The downward spike from 2001-2003 is associated with two phenomena. First, severe data anomalies
when switching from the census to ACS from 2000 to 2001 are well established in the literature. 16 Second,
September 11th led to a sharp increase in suspicion and animosity towards Muslims as well as other immigrants.
However, the decrease in labor market outcomes are found to be short-lived. 17 For expositional
clarity, I exclude the years 2001-2003 in Figure 4 and 5. This erases the large downward spike in the race
premia of first generation immigrants.
The evolution of the male race penalties in Figure 4 is striking. While Asian and Hispanic men have
experienced a small increase or even a decrease in their penalties across subsequent generations. the penalty
on being black increases substantially as we move from the first to the second, to the third and greater
generations of blacks. Holding education, experience, and state location constant, a black immigrant male
earns 17% less than a white native male in 2011. This penalty increases to 27% for a black male of the
13 Due to small sample sizes, years 2004-2006, 2007-2009, and 2010-2011 are pooled for the ACS samples. For the CPS samples
1994-1998, 1999-2003, 2004-2008, and 2009-2012 are pooled.
14 Focusing exclusively on employed individuals do not change the main take-aways as Figure A.1 and Figure A.2 demonstrate.
Figures are cut off at 1990 to make the comparison with the second generation easier (1994 and beyond). Figure A.3 and Figure A.4
depict the same race premia from 1970 onwards
15 The 95% confidence intervals on Figures 2 through 12, as well as A.1 through A.5 are very tight, barely visible, and therefore not
depicted.
16 The driving differences seem to come from the unemployed.
17 Kaushal et al. (2007) find that September 11th was associated with a 9-11 percent decline in the real wage and earnings of Arab
and Muslim men. Orrenius and Zavodny (2006) find a decline in employment and earnings among Hispanic immigrants. These are just
a select few among many papers exploring the effect of September 11th on labor market outcomes of different types of immigrants.
8

second generation and reaches an astounding 34% for a black native male. In 2000, the gradient was even
sharper starting with a 9% penalty for black immigrants, passing through 28% for the second generation and
reaching 35% for black natives. The smoothing of the gradient is driven by an increase in the race penalty
for first generation immigrants, rather than an amelioration of the race penalty of native or second generation
blacks. As Figure 5 shows, the same pattern holds for black women. For women, the decrease in the race
effect across generations is certainly unique to blacks. Asian and Hispanic women increase their earnings
premia as they move to the next generation. In summary, comparisons within races and across races yield
that black immigrants are the only racial group to consistently experience a decrease in their race premium
as they move across generations.
There are three potential explanations for the convergence of black immigrants to black natives across
generations. First, immigrants are likely to be favorably selected in terms of motivation, dedication, and other
cognitive and non-cognitive skills not captured in years of education and location choice. These might only
get partially transmitted to the next generation. One would expect Hispanics and Asians to be positively
selected along similar dimensions. However, Hispanic and Asian men and women do not experience a
deterioration of the returns to their unobservable skills across generations. The returns, however, do converge
to those of natives the same way those of blacks do.
Second, black immigrants may not face the same potential discrimination and prejudice that native blacks
do. This could be due to a distaste for native blacks or statistical discrimination. Immigrant blacks are
recognizably different from native blacks in that they are likely to have an accent. It is possible that an
accent is a plus in the labor market for blacks, increasing interview and on the job success. Growing up
in the US, the second generation will not have an accent. While their names might hint at an African or
Caribbean origin, they blend into native blacks more smoothly than do immigrant blacks. Figure 4 and 5
suggest that the elimination of this immigration signal might impose a penalty for the second generation.
Third, the fall in the black race premium from the first to the second, to the third and greater generations
could be explained with the “fade to black” phenomenon discussed in the sociology literature. Outcomes of
black natives, whom the second generation might identify with, have stagnated over the past few decades,
which would suggest negative assimilation for blacks.
Comparing Figure 4 with Figure 5, some interesting patterns emerge. The convergence across generations
that does indeed seem to be happening, proceeds somewhat differently for men and women. Moving
across generations, the penalty for men drops significantly more than it does for women. Women, however,
switch their ranking among racial groups. While black women have the highest return on their unobservables
in the first generation, they have the lowest return as natives.
5 Convergence Within a Generation
If there is convergence across generations, it seems reasonable to think there might be convergence within a
generation. In other words, if first generation immigrants immigrate earlier, they are more likely to assimilate
9

with natives, possibly opening themselves up to discrimination. A confounding factor is the well-established
fact that there is a earnings penalty for immigrants followed by a rise in earnings roughly proportional to the
time spent in the US. In order to avoid dealing with the earnings penalty, the regressions control for time
spent in the US and only focus on immigrants who have been in the US for 10 years or more. 18
Immigrant children who get exposed to second languages during childhood generally achieve higher
second language proficiency than those beginning as adults (Krashen et al. 2011). Since many African and
Caribbean immigrants come from countries where English is one of the official languages, accent and dialect
become more important than language proficiency. Many studies link the beginning of puberty to accent
acquisition; in other words, immigrants who arrive before puberty are much more likely to sound native than
immigrants who arrive after. Since language proficiency and ease of assimilation decline monotonically with
age of immigration, these two forces suggest a negative relation between age of immigration and earnings.
Positive forces such as self-selection and motivation due to hardship forged by life experiences in the source
country are counteracting forces.
Figures 6a and 6b plot the coefficients on age of immigration by race for men and women from the
following regression:
ln(Earn)=l s + h t +
M
Â
m=1
b 1m m + b 2 Pot. Exp it + b 3 Pot Exp 2 it +
4
Â
k=1
b 4k School kit + b 5 YrsUSA it + e ismt (1)
where i denotes the individual, s denotes the state of residence, m denotes the age at arrival in the US,
t is the survey year, b 2 and b 3 capture labor market experience, the b 4k s capture different schooling levels
(high school, some college, college, more than college), and b 5 accounts for the increase in earnings due
to increased time spent in the US. Earnings are adjusted to all be in 2010 dollars. Each regression is run
separately for different races and ethnicities and the control group are natives of the same group.
For Hispanics and Asians, the immigrant premium monotonically declines with age of immigration. For
whites and blacks, however, things look different. The premium actually increases until the age of 15 and
then starts to fall but never to the level of Hispanics and Asians. This is well past the average begin of
puberty for boys (11-12) and girls (10-11). What could explain the difference between the evolution of age
of immigration effects for these groups? Perhaps, a foreign accent for blacks and to a lesser extent for whites
serves as a benefit in the labor market. Black native men have certain stereotypes attached to them that can
be a disadvantage in the labor market. Black men who arrived past the age of 11 or 12 likely sound different
than natives and thus can distinguish themselves from black natives in the labor market. The discrimination
explanation is difficult to distinguish from assimilation. Potentially, blacks who immigrate later assimilate
to a lesser degree with black natives, a group that generally falls behind whites in education and income
18 Due to small sample sizes individuals are pooled in three years bins by age of immigration. For example, individuals at the 15
year mark were between 14 and 16 years old when they arrived in the US. The figures are cut off at age 40 due to subsequent sampling
noise with small bin sizes. The sample is restricted to the 2000 census and the 2001-2011 ACS, since pre-2000, year of immigration
was only recorded in bracketed form.
10

levels. 19
Past the age of 15, the black immigrant premium falls monotonically with age of immigration, suggesting
that the benefit of acquiring country specific capital at an early age outweighs any negative effect from
discrimination or assimilation with natives. 20
Group Differences
During the 1960s and linked to the Civil Rights Act of 1964, many elite school started various affirmative
action measures in order to increase the enrollment of African Americans. The emphasis on diversity not
only increased the number of native blacks but also led to a sharp increase in the representation of blacks
of immigrant origin. This spurred a controversial debate about the purpose and effect of affirmative action
policies. At the 2004 reunion of Harvard’s black alumni, Henry Louis Gates and Lani Guinier pointed out
that while Harvard’s undergraduates were 8% black, the majority of them were West Indian and African
immigrants or their children (Rimer and Arenson 2004). They said that only about one third of African
Americans at Harvard were descendants of slaves with all four grandparents born in the US. 21 If blacks
immigrants were much more likely than Asians or Hispanics to immigrate in order to attend high school
or college, teenagers would be much more positively selected and no discrimination or assimilation story
would be needed to reconcile the data. In order to test this hypothesis, equation 1 is run on the subsample
of relatively low-educated immigrants (high school or less). Figure 7a and 7b show that the difference in
patters of coefficients persists and is even more pronounced.
The pronounced pattern for low-educated men and women suggests that it’s the lower tail of the skill
distribution that is driving the convergence phenomenon. Therefore, it makes sense to break-up the sample
by employment status. Figure 8a and 8b restrict the sample to employed individuals and the pattern persists.
Another alternative explanation is that black immigrants choose their location more carefully. While
Caribbeans do not have along distance to travel, Africans do. This could make their location more deliberate
than many Hispanic immigrants crossing the border into California, Arizona, New Mexico, and Texas. In
order to explore this explanation, blacks are split into those with African and Caribbean roots in Figure
9a and 9b. Subsequently, blacks are split by refugee status in Figure 9c and 9d. Refugee status is an
approximate measure constructed by consulting the Office of Refugee Resettlement (ORR) data and the
World Fact Book. 22
19 The possible confounding factor that parents who immigrate when their children are infants may differ systematically from parents
who immigrate when their children are teenagers is explored in section 6.3.
20 If education is viewed as an endogenous outcome (Neal and Johnson 1996), education dummies should be excluded from the
controls in the earnings regression. Excluding education controls does not change the overall shape of the black residual plot, but erases
the downward slope for Hispanics, Asians, and whites. The residual for blacks and whites remains far above those of Hispanics and
Asians.
21 Vigorous debates about the purpose of affirmative action followed and some say black immigrants are not the appropriate beneficiaries
of affirmative action policies that intended to help those disadvantaged by the legacy of Jim Crow laws. Massey et al. (2007)
find that the representation of immigrant-origin blacks at selective institutions is roughly double their share in the population. For ivy
league schools it is almost four-fold.
22 Details on how refugee status was constructed are in the data appendix.
11

The Earning-Age of Immigration profile is more positively sloped for African and Caribbean blacks
than those of other races. For Africans it is steeply upward sloping for both men and women. Both, African
men and women on average have higher earnings premia than black Caribbeans. Refugees have a even
steeper earnings-age of immigration profile than non-refugees, suggesting that deliberate location choice is
not driving this pattern. Refugee men have exceptionally high return on their unobservables. This is in line
with Cortes’s (2004) finding that from 1980 to 1990 refugees made greater gains in English proficiency,
working hours, and wages, than economic immigrants.
Black immigrants are much more likely than Asians and Hispanics to come from countries where English
is an official language. This could lead to a decreased weight placed on arriving early. To explore this
hypothesis, Figure 9e and Figure 9f split black immigrants into blacks who come from countries where
English is an official language and those that do not. Immigrants from English speaking countries indeed
have a less negatively sloped gradient for men and a more positively sloped gradient for women. Immigrants
from countries where English is not listed as an official language still have a considerably less negatively
sloped earnings-age of immigration profile than Asians and Hispanics.
6 The Importance of Labor Force Participation Differences
6.1 Across Generation Comparison
From Table 2 we already learned that labor force participation decisions differ significantly between blacks
of the first and second generations. While 86% of black first generation males participate in the labor force,
only 81% of age-adjusted second generation males are labor force participants. This difference does not
exist for women where 76% of first generation females and 77% of second generation females participate in
the labor force.
Section 4 established that much of the convergence for women and all of it for men can be explained
by differential labor force attachment propensities across generations. In order to understand more about
the convergence phenomenon, the subsequent regressions focus on labor force participation as an outcome
variable.
Table 7 and 8 focus on black men and women, respectively. In column (1) all immigrants are included.
Columns (2) - (6) limit first generation immigrants to those who spent over 10 years in the US. This allows
us to focus only on permanent immigrants; immigrants who are more likely to be parents of the subsequent
second generation. We see that if all immigrants are included, the decrease in labor force participation is 10
percentage points (.0806 - (-.0196)) for men. When recent immigrants are excluded, the decrease increases
to close to 13 percentage points. Including state fixed effects shrinks the gap to about 11 percentage points.
Column (5) restricts the sample to low-educated individuals (high school degree or less). For low-educated
black men the decrease reaches 18 percentage points. This highlights the importance of differential labor
force attachment for low-educated blacks across generations in explaining the convergence between black
12

immigrants and black natives.
As discussed in Section E of the Appendix, the welfare program eligibility requirements are complex
for first generation immigrants and much more restricted than those for the second generation who generally
are citizens. The simplest and most crude way to deal with these welfare program eligibility requirements,
is to restrict the sample to US citizens (column (6)). US citizens, ceterus paribus, face the same eligibility
requirements and hence the first and second generation can be compared readily. The first generation remains
12 percentage points more likely to be attached to the labor force. This suggests that eligibility requirements
of welfare programs alone cannot possibly explain the convergence to native black labor force attachment
patterns from first to second generation. Women display a similar pattern although decrease in the labor
force participation probability is smaller, hovering around 8 percentage points. Tables A.2 and A.3 in the
Appendix show that conditional on being in the labor force, the black first generation is also more likely to
be employed than the black second generation (about 3 percentage point difference for men and 1 percentage
point difference for women).
Note also that the coefficients on the second generation are negative and mostly statistically significant
for men and women. Hence, the second generation, holding potential experience and education constant, are
generally less likely to participate in the labor force than both, black first generation immigrants, and black
natives.
Now the question arises whether the decrease in the labor force attachment when moving from the
first to the second generation is a distinctly black phenomenon or whether it can be found in all races. In
order to avoid confounding factors related to welfare program eligibly requirements, Table 9 and 10 run the
baseline regression on the citizen sample only. Other races also experience a decrease in the labor force
attachment from the first to the second generation, though much smaller than the decrease experienced by
blacks. Hispanic, Asian, and white men decrease their labor force participation by 6, 4, and 3 percentage
points (0, 1, 2 for women), respectively. Tables A.4 and A.5 in the Appendix show that the decrease in
employment probabilities conditional on labor force participation are generally also larger for blacks than
for other races.
While other races also decrease their labor force attachment as they move from the first to the second
generation, there are two main differences compared to blacks. First, the decreases are much smaller than
they are for blacks. Second, this negative effect on overall earnings is outweighed by second generation
benefits, making the overall premium of belonging to the second generation positive for all groups except
for blacks.
6.2 Within Generation Comparison
Since we find convergence in labor force attachment patterns across generations, we would expect there to be
a similar convergence within generations. In other words, as first generation immigrants immigrate earlier,
their probability of participating in the labor force should decrease, ceterus paribus.
13

Regressions underlying Figure 10a and 10b test this hypothesis by running labor force participation on a
set of controls and age of immigration dummies. To avoid differences in welfare eligibility requirements, the
sample is restricted to US citizens. For men, we see a clear upward-sloping trend in labor force participation
by age of immigration for all races. For black, Hispanic, and Asian men the coefficient starts at about
0, suggesting that these immigrant groups portray the same labor force participation patterns as natives
when they immigrate as newborns. Similar to the across-generation analysis, blacks experience the greatest
gradient in labor force participation patterns in the within generation analysis. All else equal, a male black
immigrating in his late thirties is 20 percentage points more likely to participate in the labor force than a
black immigrant who arrives as a newborn. For females this difference reaches a striking 30 percentage
points.
For log earnings, the convergence between black immigrants and black natives is especially pronounced
for low-educated immigrants. To see whether the same holds true for labor force participation, Figure 11a
and 11b restrict the sample to individuals with a high school degree or less. Now both, black immigrant men
and women who arrive in their late thirties are 30 percentage points more likely to participate in the labor
force than blacks who arrive when they are infants.
Self-selection is more likely to be important as age of immigration increases and interacts with the
assimilation effect. If immigrants in their 30s migrate to the US, they are likely migrating in order to find a
job. Hence we would expect labor force participation behavior of their peers to be less important. To test this
hypothesis Figure 12a and 12b split the black sample into immigrants from countries with many refugees and
from countries without. Refugees have different motives for migrating than economic immigrants. While
refugees are likely fleeing political or religious persecution, economic immigrant’s main motive is likely
work-related. Figure 12a and 12b show that the strong assimilation in labor force participation of blacks is
indeed driven by immigrants from non-refugee countries. There are two reasons to believe that assimilation
is important beyond its interaction with self-selection. First, children under the age of 15 are probably
accompanying their parents and not selecting to leave their home country on their own. The upward-sloping
pattern holds for young children as well as adults. Second, the gradients differ greatly by race, which is
likely to be, at least in part, due to differences in assimilation patterns.
6.3 Robustness Check
Focusing on children alone, the possibility remains that parents who immigrate when their child is an infant
might be very different from parents who immigrate when their child is a teenager. In order to address
this issue, we can look at children who still live with their parents and see whether there are any apparent
differences in parental characteristics across age of immigration of the child.
The new sample is restricted to children and teenagers that are 15 or younger in the 1990 census. This
allows for most individuals to be old enough to hypothetically make it into the 2000-2011 adult sample
without having to work with the coarse 5-year year of immigration bins in earlier censuses. Figures 12a
14

and 12b show the results of regressing log earnings of the household head and family log income on age of
immigration dummies by race. The only control is years spent in the US due to the well-established fact that
immigrant earnings increase significantly with time spent in the US. A child who immigrated as an infant is
likely to have parents who have been in the US longer than a 15-year-old who just immigrated. We can see
that for Hispanics, blacks, and whites, there is no clear up- or downward trend that holds for both earnings
measures. For Asians there is downward trend suggesting that parents who immigrate with infants are more
positively selected in terms of earnings power than Asians who immigrate with an older child.
Figure A.6 depicts the same results without controlling for time in the US and indeed all of earnings/income
- age of immigration gradients are now downward-sloping. If we take this specification to
be correct, the convergence story is actually strengthened. Black parents who immigrate with an infant are
more positively selected in terms of earnings power, and yet, their children earn less and drop out of the
labor force more.
7 Labor Force Participation Choice Model
7.1 Theoretical Model
Given that the labor force participation decision is the strongest force behind the convergence between black
immigrants and black natives, this section builds a model to better understand the decision-making process.
The model builds on the decision-making problem introduced in Blume et al. (2010) among others. For
simplicity, the model presented here is a linear probability model. To avoid differences in welfare eligibility
requirements, the sample again is restricted to US citizens.
The model studies the joint behavior of individuals who are members of a common group g. The population
size of each group g is n g . The goal is to probabilistically describe the individual i’s choice: w i,g . The
set of possible behaviors from which choices are made is W i,g . Using standard notation, w i,g denotes the
choices of others in the same group. Let x i be a vector of observable individual specific characteristics. y g is
a vector of observable group-specific characteristics. µ i e(w i,g) is an unobservable probability measure that
describes the beliefs individual i has about behaviors of other in the group. Let e i be a vector of unobservable
random individual-specific characteristics associated with i and y g be a vector of unobservable random
group-specific characteristics. Individual choices are characterized as maximizations to a payoff function V,
w i,g = argmax l2Wi,g V (l,x i ,y g , µ e i (w i,g ),e i ,y g ) (2)
Much of the empirical literature has focused on a linear in means model of the following form 23
23 This linear in means model can be derived as a Bayes-Nash equilibrium of a game in which each individual’s choice is determined
by a private benefit and a conformity benefit. The utility functions are quadratic and the conformity benefit is linearly decreasing in the
quadratic deviation of an individual’s choice from the average behavior of the group (see Blume et al. (2010) for a derivation).
15

w i,g = k + cx i + dy g + Jm e i,g + e i (3)
where m e i,g denotes the expected average of other people’s decisions in the group,
m e i,g = 1 n  j2gE(w j |F i ). (4)
As in Manski’s original formulation, y g = ¯x g , where ¯x g = 1 n g
 i2g x i denotes the average across i of x i
within a given g. Coefficient vector c and scalar J capture contextual and endogenous effects. 24 Under the
assumption y g = ¯x g , every contextual effect is the average of a corresponding individual characteristic
m g = k +(c + d) y g
1 J
We see that m g is linearly dependent on the other regressors, i.e. the constant and y g . This linear
dependence implies that the identification of the full set of structural parameters fails. For this study, the
parameter of interest is a (to be introduced), and not c or J. Hence, this paper will not further deal with
conditions that could be put in place to restore identification of all parameters. 25
What group g do first generation immigrants belong to? In other words, which individuals affect their
decision to participate in the labor force? Figures 10a and 10b showed that an immigrant’s labor force
participation decision varies considerably by age of immigration. This suggests that immigrants assimilate
with natives to varying degrees.
Consequently, I represent an immigrant’s choice by a mixture of two
choices, the choice of individual i in group 1 and that of individual i in group 2. Group 1 represents all
natives of the same racial group and group 2 represents a different group.
(5)
It is not clear how to cleanly define group 2, and the following two sections will go through two possibilities.
w i,g1,2 = a i w i,g1 +(1 a i )w i,g2 (6)
Case 1: Group 2 Members Are only Affected by own Characteristics
Let’s assume that g 1 evolves as the baseline model in (3) while members of group 2 are only affected by
their own characteristics. In other words,
After some algebra, we get 26
24 Assumptions on the error terms are discussed in Appendix C.
25 See Blume et al. (2010) for a detailed discussion and ways to restore identification.
26 Section C provides an explicit derivation of the model.
w i,g2 = k + cx i + e i (7)
16

w i,g1,2 = p 0 + p 1 x i + a i p 2 ¯x g1 + v i,g1,2 . (8)
where a i measures degree of assimilation. Given previous results, we would expect a to be strongly
negatively correlated with age of immigration.
Econometric Model
From the perspective of econometric evaluation, we see that a i cannot easily be disentangled from p 2 in
equation (8). Hence it makes sense to begin by recovering the p 2 from running the baseline regression on
natives
w i,g1 = p 0 + p 1 x i + p 2 ¯x g1 + v i,g1 (9)
We then estimate the following regression separately by each age of immigration
w i,g1,2 = b 0 + b 1 x i + b 2 ¯x g1 + v i,g1,2 (10)
For each age group we are left with the following condition that can be used for identification:
a j p 2 b 2 j = 0 (11)
where a j is a scalar that varies by age of immigration j. Each regression will be estimated separately
for blacks, Hispanics, Asians, and whites. Members of group 1 will be based on unique metropolitan area ⇥
race interactions. Variables in x i include education (high school, some college, college, more than college),
and a quadratic in potential experience. ¯x g1 includes the group averages of these variables.
Case 2: Group 2 Members Are Affected by Behavior of All Metropolitan Area Residents
Instead of having hypothetical group 2 members be affected by no one else, we could imagine them being
affected by the behavior of all metropolitan area residents. This implies that these members put the same
weight on members of their own race as members of a different race.
Let’s assume that both, g 1 and g 2 , evolve as the baseline model in (3) In other words,
After some algebra, we get 27
27 Section C provides an explicit derivation of the model.
w i,g j
= p 0 + p 1 x i + p 2 ¯x g j
+ v i,g j
, j = 1, 2 (12)
17

w i,g1,2 = p 0 + p 1 x i + a i p 2 ¯x g1 +(1 a i )p 2 ¯x g2 + v i,g1,2 . (13)
where a i again measures degree of assimilation. Given previous results, we would expect a to be
strongly negatively correlated with age of immigration.
Econometric Model
As before, we recover p 2 from running the baseline regression on natives 28
w i,g1 = p 0 + p 1 x i + p 2 ¯x g1 + v i,g1 (14)
We then estimate the following regression separately by each age of immigration
w i,g1,2 = b 0 + b 1 x i + b 2 ¯x g1 + b 3 ¯x g2 + v i,g1,2 (15)
For each age of immigration group we are left with the following two conditions that can be used for
identification:
a j p 2 b 2 j = 0 (16)
(1 a j )p 2 b 3 j = 0 (17)
where a j is a scalar that varies by age of immigration. Each regression will be estimated separately
for blacks, Hispanics, Asians, and whites. The definition of group 1 members remains the same. Members
of group 2 will be based solely on metropolitan area. Each immigrant will belong to one of both groups
and the weights on both groups depend on age of immigration. As before, variables in x i include education
(high school, some college, college, more than college), and a quadratic in potential experience. ¯x g1 and ¯x g2
includes the group averages of these variables.
7.2 Estimation
Case 1: Group 2 Members Are only Affected by own Characteristics
From (11) we see that a j is overidentified as p 2 and b 2 j are vectors and a j is a scalar. a j is recovered by
running a weighted regression of b 2 j on p 2 . The regression is run without a constant and the weights are the
absolute value of the t-statistics of the parameters contained in b 2 j .
28 Individuals with less than 100 peers in their metropolitan area were dropped in all regressions.
18

Figure 14 and 15 show the estimated alphas for each age of immigration. As there is some noise in alpha,
the red line displays the locally weighted regression of alpha on age of immigration. Panel A and B provide
the linear approximation of this regression for men and women, respectively. 29
For men two things are apparent across all racial/ethnic groups. First, the smoothed regression line of
a lies in between 0 and 1, which is in line with the original interpretation of an assimilation parameter or
weight on native characteristics. Second, the smoothed regression is generally downward sloping with a
decreasing with age of immigration. Neither of the two hold consistently for women, suggesting that the
model of social interactions in terms of labor force participation fits men’s behavior better than women’s.
This seems plausible since woman’s labor force participation is more likely to be complicated by decisions
involving child-bearing and raising.
The linear approximation of the relation between age of immigration and a yields that black and Hispanic
men have the greatest negative gradient between a and age of immigration (Figure 16a and Table 11).
If we interpret a as a parameter ranging from 0 to 1, increasing age of immigration by one year, decreases
the estimated a by .7 percentage points for blacks and .78 for Hispanics. This gradient is .57, and .29 percentage
points for Asian, and white men, respectively. Considering the small unit (one year), these estimates
are very large. As Figure 14 shows, this parameter reaches from close to .4 for blacks who immigrate as
young children to close to 0 for blacks who immigrate at 50.
Case 2: Group 2 Members Are Affected by Behavior of All Metropolitan Area Residents
Similar to Case 1, a j is overidentified as p 2 , b 2 j , and b 3 j are vectors and a j is a scalar. Instead of 6, we now
have 12 identifying equations. a j is recovered by running a weighted regression of b 2 j on p 2 . The regression
is run without a constant and the weights again are the absolute value of the t-statistics of the parameters
contained in b 2 j .
Figure 17 and 18 show the alphas by age of immigration. As before, the red line displays the locally
weighted regression of alpha on age of immigration. For men, the smoothed regression line again lies in
between 0 and 1, and is generally downward sloping. The linear approximation of the negative gradient,
however, is significantly smaller for all races than in the case 1 (Figure 19a and Table 12). For black men
it drops by .31 percentage points to .39. For Hispanic, Asian, and white men it drops by .65, .43, and .12
percentage points, respectively. Note that black men are the only group whose alpha age of immigration
gradient is highly significant in both specifications.
29 This estimation is using as backed out of previous regressions. The variables in those regressions were themselves coefficients in
previous regressions. Hence the standard errors need to be adjusted in a non-trivial way. Bootstrapped errors using 500 iterations are
displayed in Table 11 and 12.
19

8 Conclusion
This paper adds to the existing literature on three dimensions. First, it lays out the fact that black immigrants
have become a large part of the black population in the last decades. Second, it demonstrates that the
characteristics of immigrant blacks are very different from those of native blacks. Third, it determines how
much immigrants blacks converge to native blacks across and within a generation. To put these patterns into
perspective, they are compared to those of other races/ethnicities.
The share of black immigrants among the black population in the US has increased from 1% in the
1970s to 11% in 2011. Black immigrant males’ earnings and wages are higher than those of native blacks
but the premium is small once we condition on being employed. Employment, education, incarceration,
and marriage outcomes confirm that we are dealing with two completely different subsets of the population.
Black immigrants are much more likely than native blacks to be employed, married, highly educated and not
incarcerated.
In summary, first generation black immigrants undoubtedly have better labor market outcomes than black
natives. They directly pass on some characteristics such as high human capital investments to the next generation.
The transmission of labor force attachment seems to be weaker, suggesting that the second generation
is converging to natives. Black immigrants do not only converge to black natives across generations but
also within a generation. For Asians and Hispanics, earnings premia decrease monotonically with age of
immigration. For blacks, the earnings-age of immigration profile is upward sloping for those immigrating
before the age of 15. Convergence across generations is mostly driven by low-educated second generation
blacks that drop out of the labor force in greater numbers than low-educated first generation immigrants do.
Similarly, convergence within a generation is driven by the fact that black first generation immigrants who
arrive at an early age have a weaker labor force attachment than immigrants who arrive in the US when they
are older. A social interactions model with an assimilation parameter that varies by age of immigration helps
explain this phenomenon for men. When making their labor force participation decision, male immigrants
generally place more weight on the characteristics of natives the earlier they immigrate.
References
[1] Abramitzky, Ran, Leah Boustan, and Katherine Eriksson. 2012. “A Nation of Immigrants: Assimilation
and Economic Outcomes in the Age of Mass Migration”.
[2] Becker, Gary S., and Nigel Tomes. 1979. “An equilibrium theory of the distribution of income and
intergenerational mobility”. Journal of Political Economy 87: 1153–1189.
[3] Bhattacharya, Debopam, and Bhashkar Mazumder 2007. “Nonparametric analysis of intergenerational
income mobility with application to the United States”. Federal Reserve Bank of Chicago WP 2007-12.
20

[4] Black, Sandra, and Paul Devereux. 2011. "Recent Developments in Intergenerational Mobility". In
Orley Ashenfelter and David Card (Eds.), Handbook of Labor Economics 4: 1487-1541.
[5] Blume, Lawrence, William Brock, Steven Durlauf, and Yannis Ioannides. 2010. “Identification of Social
Interactions”. In J. Benhabib, A. Bisin, and M. Jackson (Eds.), Handbook of Social Interactions.
[6] B¨öhlmark, Anders, Lindquist, Matthew J.. 2006. “Life-cycle variations in the association between
current and lifetime income: replication and extension for Sweden”. Journal of Labor Economics 24:
879–896.
[7] Bollinger Christopher R., and Barry T. Hirsch. 2006. "Match Bias from Earnings Imputation in the
Current Population Survey: The Case of Imperfect Matching." Journal of Labor Economics 24(3):
483-520.
[8] Borjas, George. 1992. “Ethnic Capital and Intergenerational Mobility”. Quarterly Journal of Economics
107(1): 123-150.
[9] Borjas, George. 1993. “The Intergenerational Mobility of Immigrants”. Journal of Labor Economics.
11(1): 113-135.
[10] Bowles, Samuel, and Herbert Gintis. 2002. “The inheritance of inequality”. Journal of Economic Perspectives
16: 3–30.
[11] Brock, William, and Steven Durlauf. 2001a. “Discrete Choice with Social Interactions.” Review of
Economic Studies 68: 35-60.
[12] Brock, William and Steven Durlauf. 2001b. “Interaction-Based Models.” In Handbook of Econometrics,
vol. 5, J. Heckman and E. Leamer, eds. Amsterdam: North-Holland.
[13] Butcher, Kristin. 1994. “Black Immigrants in the United States: A Comparison with Native Blacks and
Other Immigrants.” Industrial and Labor Relations 47(2).
[14] Butcher, Kirsten, and Anne Morrison Piehl. 1998. “Recent Immigrants: Unexpected Implications for
Crime and Incarceration,” Industrial and Labor Relations Review 51(4): 654- 679.
[15] Camarota, Steven. 2012. Immigrants in the United States: A Profile of America’s Foreign-Born Population
(Washington: Center for Immigration Studies).
[16] Camarota, Steven. 2011. Welfare Use by Immigrant Households with Children: A Look at Cash, Medicaid,
Housing, and Food Programs (Washington: Center for Immigration Studies).
[17] Card David, John DiNardo, and Eugena Estes. 2000. “The More Things Change: Immigrants and the
Children of Immigrants in the 1940s, the 1970s, and the 1990s”. In George Borjas (Ed.), Issues in the
Economics of Immigration, Chicago: University of Chicago Press for NBER.
21

[18] Carliner, Geoffrey. 1980. “Wages, Earnings, and Hours of First, Second, and Third Generation American
Males”. Economic Inquiry 18: 87-102.
[19] Chadwick, Laura, and Gary Solon. 2002. “Intergenerational income mobility among daughters”. American
Economic Review 92: 335–344.
[20] Chandra, Amitabh. 2003. “Is the convergence in the racial wage gap illusory?” NBER Working Paper
9476.
[21] Chevalier, Arnaud, Kevin Denny, and Dorren McMahon. 2009. “A Multi-country study of intergenerational
educational mobility”. In Dolton, Peter, Asplundh, Rita, Barth, Erling (Eds.), Education
and inequality Across Europe. Edward Elgar, London.
[22] Chiswick, Barry R. 1977. “Sons of Immigrants: Are they at a an Earnings Disadvantage?” American
Economic Review, Papers and Proceedings 67: 376-80.
[23] Chiswick, B. 1978, “The Effect of Americanization on the Earnings of Foreign-born Men”, Journal of
Political Economy 86(5): 897-921.
[24] Cortes, Kalena. 2004. “Are Refugees Different from Economic Immigrants? Some Empirical Evidence
on the Heterogeneity of Immigrant Groups in the United States”. The Review of Economics and
Statistics 86(2): 465-480.
[25] Fix, Michael, and Ron Haskins. 2002. “Welfare Benefits for Non-citizens”. Series CCF Briefs.
[26] Friedberg, Rachel. 2000. “You Can’t Take It with You? Immigrant Assimilation and the Portability of
Human Capital”, Journal of Labor Economics 18(2): 221-251.
[27] Gans, Herbert. 1992. “Second-Generation Decline: Scenarios for the Economic and Ethnic Futures of
the Post-1965 American Immigrants.” Ethnic and Racial Studies 15: 173–92.
[28] Grawe, Nathan D., and Casey Mulligan. 2002. Economic interpretations of intergenerational correlations.
Journal of Economic Perspectives 6: 45–58.
[29] Haider, Steven J., and Gary Solon, 2006. “Life-cycle variation in the association between current and
lifetime earnings”. American Economic Review 96: 1308–1320.
[30] Heckman, James, Lance Lochner, and Petra Todd. 2006. “Earnings Functions, Rates of Return, and
Treatment: The Mincer Equation and Beyond.” In Hanushek, Eric, Welsh, Finis (Eds.), Handbook of
Economics of Education, vol. 1: 307-458.
[31] Hellerstein, Judith, and Melinda Morrill. 2008. “Dads and daughters: the changing impact of fathers
on women’s occupational choices”. Mimeo.
22

[32] Hertz, Tom, Tamara Jayasundera, Patrizio Piraino,Sibel Selcuk, Nicole Smith, and Alina Verashchagina.
2007. “The inheritance of educational inequality: international comparisons and fifty-year trends”.
The B.E. Journal of Economic Analysis & Policy 7, Article 10.
[33] Hirvonen, Lalaina. 2008. “Intergenerational earnings mobility among daughters and sons: evidence
from Sweden and a comparison with the United States”. The American Journal of Economics and
Sociology 67(5): 777–826.
[34] Jäntti, Markus, Bernt Bratsberg, Knut Røed, Oddbjørn Raaum, Robin Naylor, Eva Osterbacka, Anders
Björklund, and Tor Eriksson. 2006. “American exceptionalism in a new Light: a comparison of intergenerational
earnings mobility in the Nordic countries, the United Kingdom and the United States”.
Discussion paper no. 1938, Institute for the Study of Labor (IZA), Bonn.
[35] Johnson, Rucker, and Steven Raphael. 2006. “The Effects of Male Incarceration Dynamics on AIDS
Infection Rates among African American Women and Men,” Working Paper 22, National Poverty
Center, University of Michigan.
[36] Kasinitz, Philip, Juan Battle, and Inés Myares. 2001. “Fade to Black? The West Indian Immigrants
in Southern Florida”. In Rubén Rumbaut, and Alejandro Portes (Eds.), Ethnicities: Children of Immigrants
in America, University of California Press. Russel Sage Foundation, New York.
[37] Kaushal, Neeraj, Robert Kaestner, and Cordelia Reimers. 2007. “Labor Market Effects of September
11th on Arab and Muslim Residents of the United States”. Journal of Human Resources 42(2): 275-
308.
[38] Krashen, Stephen D., Michael A. Long, and Robin Scarcella. 2011. “Age, Rate and Eventual Attainment
in Secondary Language Aquisition.” TESOL Quarterly 13(4): 573-582.
[39] Krauth Brian. 2006. Simulation-based estimation of peer effects. Journal of Econometrics 133:
243–271.
[40] Ku, Leighton, and Brian Bruen. 2013. “The Use of Public Assistance Benefits by Citizens
and Non-citizen Immigrants in the United States,” Cato Working Paper, February
19, 2013, http://www.cato.org/publications/working-paper/usepublic-assistance-benefits-citizens-noncitizen-immigrants-united.
[41] Li, Ji, and Lung-Fei Lee. 2009. “Binary Choice Under Social Ineractions: An Empirical Study With
and Without Subjective Data on Expectations”. Journal of Applied Econometrics 24: 257-281.
[42] Mazumder, Bhashkar. 2005. “Fortunate sons: new estimates of intergenerational mobility in the US
using social security earnings data”. Review of Economics and Statistics 87: 235–255.
[43] Murray, Charles. Losing Ground: American Social Policy, 1950–1980. New York: Basic, 1984.
23

[44] Neal, Derek, and William Johnson. 1996. “The Role of Pre-Market Factors in Black-White Wage
Differences”. Journal of Political Economy 104(5): 869-895.
[45] O’Neill, Donal, Sweetman, Olive, van de Gaer, Dirk, 2007. “The effects of measurement error and
omitted variables when using transition matrices to measure intergenerational mobility”. Journal of
Economic Inequality 5: 159–178.
[46] Orrenius, Pia, and Madeline Zavodny. 2006. “Did 9/11 Worsen the Job Prospects of Hispanic Immigrants?”
[47] Park, Julie, and Dowell Myers. 2010. “Intergenerational Mobility in the Post-1965 immigration era:
Estimates by an Immigrant Generation Cohort Method. Demography 47(2): 396-392.
[48] Raaum, Oddbjørn, Bernt Bratsberg, Knut Røed,Eva Österbacka, Tor Eriksson, Markus Jäntti, and
Robin Naylor. 2007. “Marital sorting, household labor supply and intergenerational earnings mobility
across countries”. The B.E. Journal of Economic Analysis and Policy Article 7.
[49] Rauh, Alison. 2013. “Succesful Black Immigrants Narrow Black-White Achievement Gaps”. Working
Paper.
[50] Reville, Robert T.. 1995. “Intertemporal and life cycle variation in measured intergenerational earnings
mobility”. RAND working paper, Santa Monica, CA: RAND Corporation.
[51] Soetevent Adriaan R., and Peter Kooreman. 2007. A discrete-choice model with social interactions:
with an application to high school teen behavior. Journal of Applied Econometrics 22: 599–624.
[52] Solon, Gary, 1992. “Intergenerational income mobility in the United States”. American Economic
Review 82: 393–408.
[53] Solon, Gary, 1999. “Intergenerational mobility in the labor market”. In Ashenfelter, Orley, Card, David
(Eds.), Handbook of Labor Economics, vol. 3. North-Holland, Amsterdam: 1761–1800.
[54] Solon, Gary, 2004. “A model of intergenerational mobility variation over time and place”. In Corak,
Miles (Ed.), Generational income mobility in North America and Europe, Cambridge University Press,
Cambridge: 38–47.
[55] Suárez-Orozco, Carola, and Marcelo Suárez-Orozco. 2001. Children of Immigration. Cambridg, MA:
Harvard University Press.
[56] Tamer Elie. 2003. Incomplete simultaneous discrete response model with multiple equilibria. Review
of Economic Studies 70: 147–165.
[57] Waters, Mary. 1999. Black Identities: West Indian Immigrant Dreams and American Realities. New
York: Russell Sage Foundation.
24

[58] Wilson, William Julius. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public
Policy. Chicago: University of Chicago Press.
Figures and Tables
Figure 1: Black Immigrants as Share of Black US Population (by Origin)
0 .02 .04 .06 .08 .1
1940 1950 1960 1970 1980 1990 2000 2010
Census year
Africa
South America
All Origins
Source: IPUMS− Census and ACS with probability weights
West Indies
Other
25

Figure 2: Race Premia/Penalties for Men (including 2001-2003)
First Generation
Second Generation
Natives
−.4 −.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
−.4 −.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
−.4 −.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
Black
Asian
Hispanic
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to white native men. All regressions control for education,
potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time spent in
the US.
Figure 3: Race Premia/Penalties for Women (including 2001-2003)
First Generation
Second Generation
Natives
−.2 −.1 0 .1 .2
−.2 −.1 0 .1 .2
−.2 −.1 0 .1 .2
1990 1995 2000 2005 2010
Census year
1990 1995 2000 2005 2010
Census year
1990 1995 2000 2005 2010
Census year
Black
Asian
Hispanic
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to white native women. All regressions control for education,
potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time spent in
the US.
26

Figure 4: Race Premia/Penalties for Men
First Generation
Second Generation
Natives
−.4 −.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
−.4 −.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
−.4 −.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
Black
Asian
Hispanic
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. Individuals
are restricted to 21-65 year olds. Earnings premia by race are compared to white native men. All regressions control for education,
potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time spent in
the US.
Figure 5: Race Premia/Penalties for Women (Log Earnings)
First Generation
Second Generation
Natives
−.2 −.1 0 .1 .2
1990 1995 2000 2005 2010
Census year
−.2 −.1 0 .1 .2
1990 1995 2000 2005 2010
Census year
−.2 −.1 0 .1 .2
1990 1995 2000 2005 2010
Census year
Black
Asian
Hispanic
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to white native women. All regressions control for education,
potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time spent in
the US.
27

Figure 6: Earnings Premium, by Age of Immigration (Log Earnings)
(a) Men
−.3 −.2 −.1 0 .1 .2
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
(b) Women
−.4 −.3 −.2 −.1 0 .1
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to have been in
the US more than 10 years. Earnings premia are compared to natives of the same race. All regressions control for education, potential
experience, time spent in the US, state and year fixed effects.
28

Figure 7: Earnings Premium by Age of Immigration for Low-Educated Immigrants (Log Earnings)
(a) Men
−.4 −.2 0 .2 .4
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
(b) Women
−.4 −.3 −.2 −.1 0 .1
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to have been in
the US more than 10 years. Earnings premia are compared to natives of the same race. All regressions control for education, potential
experience, time spent in the US, state and year fixed effects. The sample is restricted to individuals with a high school degree or less.
29

Figure 8: Earnings Premia by Age of Immigration & Employment Status (Log Earnings)
(a) Employed Men
(b) Employed Women
−.4 −.2 0 .2
−.5 −.4 −.3 −.2 −.1 0
5 10 15 20 25 30 35 40
Age of Immigration
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Black
Asian
Hispanic
White
(c) Unemployed Men
(d) Unemployed Women
−.2 0 .2 .4 .6
−.4 −.2 0 .2 .4
5 10 15 20 25 30 35 40
Age of Immigration
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Black
Asian
Hispanic
White
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to have been in
the US more than 10 years. Earnings premia are compared to natives of the same race. All regressions control for education, potential
experience, time spent in the US, state and year fixed effects. The sample split into employed individuals in (a) and (b) and unemployed
individuals in (c) and (d).
30

Figure 9: Earnings Premium by Age of Immigration, Origin, and Refugee Status, English Language Status
(Log Earnings)
(a) Men by Origin
(b) Women by Origin
−.5 0 .5 1
−.5 0 .5 1
5 10 15 20 25 30 35 40
Age of Immigration
5 10 15 20 25 30 35 40
Age of Immigration
Black African
Hispanic
White
Black Caribbean
Asian
Black African
Hispanic
White
Black Caribbean
Asian
(c) Men by Refugee Status
(d) Women by Refugee Status
−.4 −.2 0 .2 .4 .6
5 10 15 20 25 30 35 40
Age of Immigration
−1.5 −1 −.5 0 .5
5 10 15 20 25 30 35 40
Age of Immigration
Black Refugee
Hispanic
White
Black Non−Refugee
Asian
Black Refugee
Hispanic
White
Black Non−Refugee
Asian
(e) Men by English Language Status
(f) Women by English Language Status
−.3 −.2 −.1 0 .1 .2
−.8 −.6 −.4 −.2 0
5 10 15 20 25 30 35 40
Age of Immigration
5 10 15 20 25 30 35 40
Age of Immigration
Black English
Hispanic
White
Black Non−English
Asian
Black English
Hispanic
White
Black Non−English
Asian
Note: The sample consists of 21-65 year olds in the 2000 census31
and the 2001-2011 ACS. Immigrants are restricted to have been in
the US more than 10 years. Earnings premia are compared to natives of the same race. All regressions control for education, potential
experience, time spent in the US, state and year fixed effects. In panels a) and b) blacks are split into those from African and Caribbean
countries. Blacks with other origins are not not plotted. In panels c) and d) blacks are split into those with refugee status and those
without. In panels e) and f) blacks are split into those that come from countries that list English as an official language and those who
do not.

Figure 10: Labor Force Participation Probability Premium by Age of Immigration
(a) Men
−.1 0 .1 .2 .3
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
(b) Women
−.1 0 .1 .2
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who
have been in the US more than 10 years. Labor Force Participation premia are compared to natives of the same race. All regressions
control for education, potential experience, time spent in the US, state and year fixed effects.
32

Figure 11: Labor Force Participation Probability Premium for Low-Educated Immigrants by Age of Immigration
(a) Men
−.1 0 .1 .2 .3
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
(b) Women
−.2 −.1 0 .1 .2 .3
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who
have been in the US more than 10 years. Labor Force Participation premia are compared to natives of the same race. All regressions
control for education, potential experience, time spent in the US, state and year fixed effects. The sample is restricted to individuals
with a high school degree or less. 33

Figure 12: Labor Force Participation Probability Premium by Refugee Status and Age of Immigration
(a) Men
−.1 −.05 0 .05 .1 .15
5 10 15 20 25 30 35 40
Age of Immigration
Black Refugee
Hispanic
White
Black Non−Refugee
Asian
(b) Women
−.1 −.05 0 .05 .1 .15
5 10 15 20 25 30 35 40
Age of Immigration
Black Refugee
Hispanic
White
Black Non−Refugee
Asian
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who
have been in the US more than 10 years. Labor Force Participation premia are compared to natives of the same race. All regressions
control for education, potential experience, time spent in the US, state and year fixed effects. Blacks are split into those with refugee
status and those without.
34

Figure 13: Parental Outcomes by Age Premium by Age of Immigration of the Child
(a) Head of Household Log Earnings
−.6 −.4 −.2 0 .2
3 6 9 12 15
Age of Immigration
Black
Asian
Hispanic
White
(b) Family Log Income
−.6 −.4 −.2 0 .2 .4
3 6 9 12 15
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 0-15 year olds in the 1990 census. Outcomes are compared to natives of the same race. All regressions
control for time spent in the US.
35

Figure 14: Case 1: Male Assimilation Parameter a by Age of Immigration
Blacks
Hispanics
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
Outliers are excluded for graphing purposes. They are included in the regression sample of Table 11 as well as in Figure A.7.
36

Figure 15: Case 1: Female Assimilation Parameter a by Age of Immigration
Blacks
Hispanics
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
Outliers are excluded for graphing purposes. They are included in the regression sample of Table 11 as well as in Figure A.8.
37

Figure 16: Case 1: Assimilation Parameter a by Age of Immigration (Linear Approximation)
(a) Men
alpha
0 .1 .2 .3 .4 .5
5 10 15 20 25 30 35 40 45 50
Age of Immigration
Blacks Hispanics Asians
Whites
95% CI
(b) Women
alpha
−.5 0 .5 1
5 10 15 20 25 30 35 40 45 50
Age of Immigration
Blacks Hispanics Asians
Whites
95% CI
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of
immigration. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects.
The underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who
have been in the US more than 10 years. Bootstrapped standard errors using 500 iterations are used to construct the 95% confidence
intervals. 38

Figure 17: Case 2: Male Assimilation Parameter a
Blacks
Hispanics
−.5 0 .5
alpha
−.5 0 .5
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−.5 0 .5
alpha
−.5 0 .5
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
Outliers are excluded for graphing purposes. They are included in the regression sample of Table 11 as well as in Figure A.9.
39

Figure 18: Case 2: Female Assimilation Parameter a
Blacks
Hispanics
−.5 0 .5
alpha
−.5 0 .5
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−.5 0 .5
alpha
−.5 0 .5
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
Outliers are excluded for graphing purposes. They are included in the regression sample of Table 11 as well as in Figure A.10.
40

Figure 19: Case 2: Assimilation Parameter a by Age of Immigration (Linear Approximation)
(a) Men
alpha
0 .05 .1 .15 .2
5 10 15 20 25 30 35 40 45 50
Age of Immigration
Blacks Hispanics Asians
Whites
95% CI
(b) Women
alpha
−.5 0 .5 1
5 10 15 20 25 30 35 40 45 50
Age of Immigration
Blacks Hispanics Asians
Whites
95% CI
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of
immigration. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects.
The underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to have been
in the US more than 10 years. Bootstrapped standard errors using 500 iterations are used to construct the 95% confidence intervals.
41

Table 1: Summary Statistics
42
1980 2011
All Blacks All Blacks
Whites Blacks Native Blacks Immigrant Blacks Whites Blacks Native Blacks Immigrant Blacks
Panel A: Men
Annual Earnings 16,400 9,460 9,460 9,370 48,500 24,060 23,500 32,900
Employed 0.84 0.71 0.71 0.75 0.76 0.59 0.57 0.76
Cond. Annual Earnings* 18,500 12,100 12,170 11,600 61,700 39,700 39,300 42,100
Age 39.95 37.98 38.05 36.49 43.63 41.13 41.07 41.52
Married (percent) 73.06 54.20 54.01 58.60 57.32 36.27 33.95 53.36
Incarcerated (percent) 0.69 3.29 3.38 1.26 1.30 7.56 8.35 1.74
HS (percent) 35.72 33.46 33.55 31.52 36.59 44.64 46.00 36.64
Some College (percent) 19.54 17.15 16.88 23.19 25.46 26.88 26.82 27.34
College (percent) 11.47 4.60 4.47 7.42 20.96 11.00 9.97 18.63
Grad School (percent) 10.96 3.93 3.67 9.81 11.05 4.87 3.95 11.63
Sample Size 2,490,000 296,000 283,500 12,500 605,300 93,500 83,900 9,570
Panel B: Women
Annual Earnings 5,250 5,230 5,2003 6,160 27,600 22,100 21,500 26,500
Employed 0.56 0.58 0.58 0.67 0.68 0.64 0.63 0.71
Cond. Annual Earnings* 8,580 8,260 8,250 8,480 39,500 33,300 32,900 36,000
Age 40.56 38.49 38.52 37.71 43.99 41.70 41.70 41.69
Married (percent) 71.86 44.87 44.66 50.34 58.99 30.03 28.08 45.15
Incarcerated (percent) 0.29 0.44 0.45 0.18 0.30 0.65 0.72 0.14
HS (percent) 43.96 35.26 35.20 36.83 33.31 37.45 37.89 34.02
Some College (percent) 18.98 17.16 17.00 21.26 27.72 32.22 32.59 29.32
College (percent) 9.19 5.08 5.07 5.43 22.42 13.82 13.32 17.67
Grad School (percent) 5.76 3.71 3.68 4.56 12.06 7.16 6.86 9.44
Sample Size 2,600,000 357,000 346,000 13,400 621,000 102,000 91,500 10,600
The sample consists of 21-65 year olds in the 1980 census and the 1-in-100 Public Use Sample of the American Community Survey for 2011.
*Conditional outcomes only refer to employed individuals. High school dropouts are the omitted group in the schooling categories.

Table 2: 2008-2012 Summary Statistics
Whites Native Blacks Imm. Blacks 1st Imm. Blacks 2nd Weighted Imm. Blacks 2nd
Panel A: Men
Annual Earnings 50,700 28,100 34,100 28,400 36,700
Labo Force 0.84 0.75 0.86 0.79 0.81
Employed 0.78 0.63 0.76 0.66 0.70
Cond. Annual Earnings* 62,000 40,900 42,200 39,700 49,300
Age 43.35 41.14 40.78 32.44 40.78
Married (percent) 62.17 44.72 59.93 25.65 43.23
HS (percent) 30.80 38.35 29.07 22.28 22.36
Some College (percent) 28.75 30.22 29.23 41.32 36.96
College (percent) 22.53 12.90 19.85 22.13 23.48
Grad School (percent) 11.37 5.60 11.11 7.72 11.83
Sample Size 182,300 28,900 3616 797 797
43
Panel B: Women
Annual Earnings 27,900 22,900 24,800 25,500 31,500
Labor Force 0.73 0.72 0.76 0.76 0.77
Employed 0.69 0.64 0.68 0.65 0.68
Cond. Annual Earnings* 38,700 33,500 34,400 36,000 43,600
Age 43.64 41.52 41.36 33.48 41.36
Married (percent) 64.57 39.28 56.96 28.99 37.90
HS (percent) 27.47 31.76 28.10 16.70 17.49
Some College (percent) 31.84 34.57 30.90 40.07 36.82
College (percent) 23.65 15.14 19.74 25.46 24.29
Grad School (percent) 11.85 7.15 8.19 11.15 14.25
Sample Size 191,500 37,200 4102 837 837
The sample consists of 21-65 years olds in the March CPS samples from 2008-2012. Second generation black immigrants in column (5) are
reweighted to have the same age distribution as the first generation in column (3). High School dropouts are the omitted group in the schooling
categories. *Conditional outcomes only refer to employed individuals.

Table 3: Male ’Fading to black’ Across Generations (Log Annual Earnings)
(1) (2) (3) (4) (5) (6)
First Generation 0.000491 0.0487 ⇤⇤⇤ 0.0356 ⇤⇤ 0.0326 ⇤ 0.0266 ⇤ 0.0119
(0.0137) (0.0165) (0.0181) (0.0181) (0.0160) (0.0153)
Second Generation 0.0292 0.0283 0.0184 0.0179 0.0272 0.0485
(0.0334) (0.0334) (0.0340) (0.0341) (0.0328) (0.0314)
Immigrants

Table 4: Female ’Fading to black’ Across Generations (Log Annual Earnings)
(1) (2) (3) (4) (5) (6)
First Generation 0.0549 ⇤⇤⇤ 0.118 ⇤⇤⇤ 0.0519 ⇤⇤⇤ 0.0476 ⇤⇤⇤ 0.0933 ⇤⇤⇤ 0.0750 ⇤⇤⇤
(0.0130) (0.0148) (0.0164) (0.0164) (0.0145) (0.0143)
Second Generation 0.0692 ⇤⇤ 0.0674 ⇤⇤ 0.0120 0.0115 0.0702 ⇤⇤ 0.0652 ⇤⇤
(0.0287) (0.0287) (0.0290) (0.0290) (0.0276) (0.0272)
Immigrants

Table 6: Female Fading Across Generations (Log Annual Earnings)
(1) (2) (3) (4)
Blacks Hispanics Asians Whites
First Generation 0.118 ⇤⇤⇤ -0.0604 ⇤⇤⇤ -0.0239 0.0247 ⇤⇤
(0.0148) (0.00989) (0.0220) (0.0109)
Second Generation 0.0674 ⇤⇤ 0.0600 ⇤⇤⇤ 0.0350 0.0586 ⇤⇤⇤
(0.0287) (0.0112) (0.0282) (0.00864)
Observations 77881 78878 26265 489589
Adjusted R 2 0.177 0.138 0.145 0.107
Standard errors in parentheses ⇤ p < 0.10, ⇤⇤ p < 0.05, ⇤⇤⇤ p < 0.01
All regressions control for schooling, a quadratic in potential experience, and year fixed effects.
The sample consists of 21-65 years olds in the March CPS samples from 1994-2012.
Immigrants are restricted to have been in the US more than 10 year.
Table 7: Male ’Fading to black’ Across Generations (Labor Force Probability)
(1) (2) (3) (4) (5) (6)
First Generation 0.0806 ⇤⇤⇤ 0.0945 ⇤⇤⇤ 0.105 ⇤⇤⇤ 0.105 ⇤⇤⇤ 0.161 ⇤⇤⇤ 0.0939 ⇤⇤⇤
(0.00474) (0.00566) (0.00613) (0.00614) (0.00987) (0.00660)
Second Generation -0.0196 ⇤ -0.0226 ⇤ -0.00908 -0.00940 -0.0218 -0.0241 ⇤⇤
(0.0119) (0.0119) (0.0120) (0.0120) (0.0228) (0.0119)
Immigrants

Table 8: Female ’Fading to black’ Across Generations (Labor Force Probability)
(1) (2) (3) (4) (5) (6)
First Generation 0.0303 ⇤⇤⇤ 0.0588 ⇤⇤⇤ 0.0726 ⇤⇤⇤ 0.0724 ⇤⇤⇤ 0.112 ⇤⇤⇤ 0.0687 ⇤⇤⇤
(0.00493) (0.00565) (0.00610) (0.00611) (0.00949) (0.00654)
Second Generation -0.0242 ⇤⇤ -0.0268 ⇤⇤⇤ -0.0143 -0.0146 -0.00787 -0.0277 ⇤⇤⇤
(0.0103) (0.0103) (0.0103) (0.0104) (0.0216) (0.0103)
Immigrants

Table 10: Female Convergence Across Generations (Labor Force Participation for Citizens)
(1) (2) (3) (4)
Blacks Hispanics Asians Whites
First Generation 0.0687 ⇤⇤⇤ 0.0145 ⇤⇤⇤ -0.000431 -0.0244 ⇤⇤⇤
(0.00654) (0.00398) (0.00768) (0.00401)
Second Generation -0.0277 ⇤⇤⇤ 0.0149 ⇤⇤⇤ -0.0107 -0.00597 ⇤⇤
(0.0103) (0.00392) (0.00976) (0.00283)
Observations 111217 94369 31459 662135
Adjusted R 2 0.115 0.095 0.070 0.082
Standard errors in parentheses ⇤ p < 0.10, ⇤⇤ p < 0.05, ⇤⇤⇤ p < 0.01
All regressions for schooling, a quadratic in potential experience, and year fixed effects.
The sample consists of 21-65 years olds in the March CPS samples from 1994-2012.
Immigrants are restricted to citizens who have been in the US more than 10 years.
Table 11: Case 1: Assimilation Parameter Alpha
(1) (2) (3) (4)
Blacks Hispanics Asians Whites
Panel A: Men
Age of Immigration -0.00702 ⇤⇤⇤ -0.00786 ⇤⇤⇤ -0.00577 ⇤⇤ -0.00292 ⇤
(0.00272) (0.00237) (0.00292) (0.00157)
Constant 0.540 ⇤⇤⇤ 0.361 ⇤⇤⇤ 0.234 ⇤⇤ 0.227 ⇤⇤⇤
(0.0972) (0.00841) (0.117) (0.0534)
Panel B: Women
Age of Immigration -0.00171 ⇤ -0.00393 ⇤⇤ -0.0155 ⇤⇤⇤ 0.0119 ⇤⇤
(0.000915) (0.00170) (0.00785) (0.00519)
Constant 0.113 ⇤⇤⇤ 0.163 ⇤⇤⇤ 0.859 ⇤⇤⇤ -0.350 ⇤
(0.0337) (0.0623) (0.289) (0.180)
Observations 50 50 50 50
as are backed out of equation (11) and then regressed on age of immigration.
Bootstrapped standard errors using 500 iterations are in parentheses
⇤ p < 0.10, ⇤⇤ p < 0.05, ⇤⇤⇤ p < 0.01
48

Table 12: Case 2: Assimilation Parameter Alpha
(1) (2) (3) (4)
Blacks Hispanics Asians Whites
Panel A: Men
Age of Immigration -0.00388 ⇤⇤⇤ -0.00138 -0.00148 -0.00168 ⇤⇤⇤
(0.00112) (0.000897) (0.00145) (0.000454)
Constant 0.182 ⇤⇤⇤ 0.0740 ⇤⇤ 0.109 ⇤⇤ 0.0740 ⇤⇤⇤
(0.0461) (0.0351) (0.0817) (0.0153)
Panel B: Women
Age of Immigration -0.000142 0.00147 ⇤⇤ -0.00119 0.00296 ⇤⇤⇤
(0.000385) (0.000618) (0.00264) (0.000944)
Constant 0.0145 -0.0523 ⇤⇤ 0.163 ⇤ -0.0616 ⇤
(0.0128) (0.0218) (0.0956) (0.0319)
Observations 50 50 50 50
as are backed out of equation (16) and (17) and then regressed on age of immigration.
Bootstrapped standard errors using 500 iterations are in parentheses
⇤ p < 0.10, ⇤⇤ p < 0.05, ⇤⇤⇤ p < 0.01
A Supplemental Figures and Tables
Figure A.1: Race Premia/Penalties for Employed Men
First Generation
Second Generation
Natives
−.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
−.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
−.3 −.2 −.1 0 .1
1990 1995 2000 2005 2010
Census year
Black
Asian
Hispanic
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to employed white native men. All regressions control for
education, potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time
spent in the US.
49

Figure A.2: Race Premia/Penalties for Employed Women
First Generation
Second Generation
Natives
−.2 −.1 0 .1 .2
−.2 −.1 0 .1 .2
−.2 −.1 0 .1 .2
1970 1980 1990 2000 2010
Census year
1994 1999 2004 2009
Census year
1970 1980 1990 2000 2010
Census year
Black
Hispanic
Asian
Note: The premia/penalties are compared to employed white natives.
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to employed white native women. All regressions control
for education, potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for
time spent in the US.
Figure A.3: Race Premia/Penalties for Men (from 1970 onwards)
First Generation
Second Generation
Natives
−.4 −.2 0 .2
−.4 −.2 0 .2
−.4 −.2 0 .2
1970 1980 1990 2000 2010
Census year
1994 1999 2004 2009
Census year
1970 1980 1990 2000 2010
Census year
Black
Hispanic
Asian
Note: The premia/penalties are compared to employed white natives.
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to white native men. All regressions control for education,
potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time spent in
the US.
50

Figure A.4: Race Premia/Penalties for Women (from 1970 onwards)
First Generation
Second Generation
Natives
−.2 −.1 0 .1 .2
1970 1980 1990 2000 2010
Census year
−.2 −.1 0 .1 .2
1994 1999 2004 2009
Census year
−.2 −.1 0 .1 .2
1970 1980 1990 2000 2010
Census year
Black
Hispanic
Asian
Note: The premia/penalties are compared to employed white natives.
Note: Panel 1 and 3 use the 1990 and 2000 census and ACS data from 2001-2011. Panel 2 uses CPS data from 1994 to 2012. The
sample consists of 21-65 year olds. Earnings premia by race are compared to white native women. All regressions control for education,
potential experience, and state fixed effects. In addition, the regression for first generation immigrants also controls for time spent in
the US.
51

Figure A.5: Earnings Premia by Age of Immigration for High-Educated Immigrants (Log Earnings)
(a) Men
−.6 −.4 −.2 0 .2
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
(b) Women
−.6 −.4 −.2 0 .2
5 10 15 20 25 30 35 40
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to have been in
the US more than 10 years. Earnings premia are compared to natives of the same race. All regressions control for education, potential
experience, time spent in the US, state and year fixed effects. The sample is restricted to individuals with some college or more.
52

Figure A.6: Parental Outcomes by Age Premium by Age of Immigration of the Child (No Years in the US
Control)
(a) Head of Household Log Earnings
−.6 −.4 −.2 0 .2
3 6 9 12 15
Age of Immigration
Black
Asian
Hispanic
White
(b) Family Log Income
−.6 −.4 −.2 0 .2 .4
3 6 9 12 15
Age of Immigration
Black
Asian
Hispanic
White
Note: The sample consists of 0-15 year olds in the 1990 census. Outcomes are compared to natives of the same race. with no controls.
53

Figure A.7: Case 1: Male Assimilation Parameter a with Outliers
Blacks
Hispanics
−.5 0 .5 1 1.5 2
alpha
−.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
54

Figure A.8: Case 1: Female Assimilation Parameter a with Outliers
Blacks
Hispanics
−.5 0 .5 1
alpha
−1 −.5 0 .5
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−2 0 2 4 6 8
alpha
−1.5−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
55

Figure A.9: Case 2: Male Assimilation Parameter a with Outliers
Blacks
Hispanics
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−1 −.5 0 .5 1
alpha
−1 −.5 0 .5 1
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
56

Figure A.10: Case 2: Female Assimilation Parameter a with Outliers
Blacks
Hispanics
−1 0 1 2
alpha
−1 0 1 2
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
Asians
Whites
−1 0 1 2
alpha
−1 0 1 2
alpha
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
5 10 15 20 25 30 35 40 45 50
Age of Immigration
bandwidth = .8
alpha lowess: alpha
Note:a parameters are backed out by comparing labor force participation parameters for natives and immigrants by different ages of immigration. The locally weighted
regression line is based on a 0.8 bandwidth. The underlying regressions control for education, potential experience, time spent in the US, state and year fixed effects. The
underlying sample consists of 21-65 year olds in the 2000 census and the 2001-2011 ACS. Immigrants are restricted to citizens who have been in the US more than 10 years.
57

Table A.1: Summary Statistics by Origin
1980 2011
West Indians Africans South Americans Other Origin West Indians Africans South Americans Other Origin
Panel A: Men
Annual Earnings 10,400 7,010 10,700 8,470 31,300 33,800 38,500 34,000
Employed 0.81 0.57 0.82 0.72 0.74 0.80 0.78 0.68
Cond. Annual Earnings* 11,900 10,900 12,400 10,900 41,100 41,280 48,400 48,800
Age 38.02 30.53 36.52 36.71 43.47 39.90 43.44 38.38
Married (percent) 65.75 50.73 67.93 47.61 54.52 54.89 58.01 39.20
Jail (percent) 0.40 0.27 0.57 3.54 1.79 1.05 0.96 4.80
HS (percent) 35.52 11.14 33.40 34.38 42.92 24.40 45.58 34.65
Some College (percent) 19.66 43.10 26.19 18.90 25.86 28.61 21.00 31.72
College (percent) 6.41 15.47 7.97 4.99 14.33 23.84 17.83 17.02
Grad School (percent) 7.02 26.93 13.09 5.56 6.52 18.68 5.83 8.06
Sample Size 6,700 1,850 530 350 4,580 3,630 410 950
58
Panel B: Women
Annual Earnings 6,720 3,970 7,180 5,230 27,400 24,300 27,900 28,000
Employed 0.72 0.47 0.72 0.61 0.72 0.69 0.70 0.72
Cond. Annual Earnings* 8,750 7,170 9,280 7,890 36,500 34,000 39,300 37,700
Age 38.41 29.87 38.31 37.62 44.03 38.48 44.11 38.59
Married (percent) 51.78 63.81 51.67 44.92 42.29 52.70 47.18 34.53
Jail (percent) 0.06 0.00 0.00 0.47 0.18 0.11 0.00 0.11
HS (percent) 38.14 22.93 40.91 36.08 37.20 30.77 38.68 25.37
Some College (percent) 20.94 35.36 20.00 19.59 29.03 29.10 26.79 32.79
College (percent) 5.12 15.06 5.61 4.31 14.35 21.47 18.52 22.64
Grad School (percent) 4.37 13.26 6.52 3.06 8.65 9.84 6.77 13.70
Sample Size 7,930 720 660 48,000 5,900 3,200 528 1,030
The sample consists of 21-65 year olds in the 1980 census and the 1-in-100 Public Use Sample of the American Community Survey for 2011.
Weights are used to reflect sampling. *Conditional outcomes only refer to employed individuals.

Table A.2: Male ’Fading to black’ Across Generations (Employment Probability Conditional on Labor Force
Participation)
(1) (2) (3) (4) (5) (6)
First Generation 0.0284 ⇤⇤⇤ 0.0228 ⇤⇤⇤ 0.0291 ⇤⇤⇤ 0.0288 ⇤⇤⇤ 0.0404 ⇤⇤⇤ 0.0294 ⇤⇤⇤
(0.00397) (0.00470) (0.00521) (0.00522) (0.00793) (0.00529)
Second Generation -0.00903 -0.00845 -0.000731 -0.000992 0.00724 -0.00818
(0.0114) (0.0114) (0.0115) (0.0115) (0.0224) (0.0114)
Immigrants

Table A.3: Female ’Fading to black’ Across Generations (Employment Probability Conditional on Labor
Force Participation)
(1) (2) (3) (4) (5) (6)
First Generation 0.00700 ⇤ 0.0132 ⇤⇤⇤ 0.0138 ⇤⇤⇤ 0.0139 ⇤⇤⇤ 0.0273 ⇤⇤⇤ 0.0134 ⇤⇤⇤
(0.00365) (0.00398) (0.00439) (0.00439) (0.00700) (0.00454)
Second Generation -0.00200 -0.00215 -0.000902 -0.00121 -0.0103 -0.00209
(0.00846) (0.00846) (0.00857) (0.00859) (0.0221) (0.00846)
Immigrants

Table A.5: Female Fading Across Generations (Employment Probability Conditional on Labor Force Participation
for Citizens)
(1) (2) (3) (4)
Blacks Hispanics Asians Whites
First Generation 0.0134 ⇤⇤⇤ 0.00173 -0.00392 -0.00692 ⇤⇤⇤
(0.00454) (0.00252) (0.00436) (0.00225)
Second Generation -0.00209 -0.00681 ⇤⇤ -0.00985 ⇤ -0.00400 ⇤⇤⇤
(0.00846) (0.00278) (0.00594) (0.00151)
Observations 79978 65818 23829 493471
Adjusted R 2 0.040 0.023 0.010 0.013
Standard errors in parentheses ⇤ p < 0.10, ⇤⇤ p < 0.05, ⇤⇤⇤ p < 0.01
All regressions for schooling, a quadratic in potential experience, and year fixed effects.
The sample consists of 21-65 years olds in the March CPS samples from 1994-2012.
Immigrants are restricted to have been in the US more than 10 year.
B Data Description
Data Limitations
There are some very unfortunate data limitations with the CPS, ACS, and census data. The CPS asks respondents
about the birthplace of their parents starting in 1994. In 2012 the sample includes over 200,000
individuals. Roughly 22,700 of the sample identified themselves as non-Hispanic black, 2100 as first generation
immigrant non-Hispanic black, and only 1200 as second generation non-Hispanic blacks. If we restrict
our sample to the working age population the number of first and second generation members in the sample
drops to 1600 and 340 respectively. Substantial black immigration is a very recent phenomenon and it accelerated
massively over the past 30 years. Hence many immigrants have not had children that are old enough
to be in the sample. Consequently the sample size is very small for calculating average annual wage, employment
status, and educational attainment rates. Furthermore, the CPS coverage does not include military
and institutional populations.
The sample size for the ACS is much larger averaging about 3 million addresses per year. It includes
both military and institutional populations. Neither the census not the ACS, however, ask respondents about
their parents’ place of birth after 1970 so I can infer very little about second- or later generations from the
ACS and census and am limited to first generation immigrants. My method is to mostly use the ACS and
census due to its sample size and more inclusive coverage, but infer second generation characteristics from
the CPS.
Besides data limitations there are technical problems with both the CPS and the ACS/census. First,
earnings are capped at different values in different years. This biases wage gaps since disproportionately
many whites are high earners. Second, both the CPS and ACS oversample certain populations and sample
61

weights are necessary. Third, both the CPS and ACS/census have started a hot-decking procedure that
imputes earnings and other characteristics for individuals who do not report the relevant statistics. Dropping
these observations would delete a disproportionate number of young low-skill blacks and cause illusory
convergence (Chandra 2003), leading to the inclusion of imputed data by most researchers. Bollinger and
Hirsch (2006) show that this introduces substantial bias due to mismatch in the imputation process.
In this paper, I estimated my results including imputed data. The ACS/census is a much larger data set
than the CPS and the bias is smaller.
Census and ACS Data
This paper uses data from the decennial censuses from 1970 to 2000. The 1970 sample consist of two
IPUMS self-weighting subsamples: the 1% State sample (5% form) and the 1% State sample (15%) form,
which together represent 2% of the population. The 1980, 1990, and 2000 samples are 5% Public use
Samples. While the 1980 sample is self-weighting, the 1990 and 2000 sample are not and weights are used
to return to random proportions. The 2001-2011 ACS sample represent about .4% of the population in years
2001-2004 and 1% thereafter.
Sample Restrictions
• Age: The sample includes individuals aged 21-65 in order to focus on the working-age population.
• Race: Individuals are coded as black, Asian, and white if they are non-Hispanic. Hence Hispanics are
comprised of different races in this sample.
• Earnings: In the results presented I exclude persons with negative earnings. Only a very small fraction
of individuals have negative earnings.
Constructed Variables
• Annual Earnings: For the 1990-2000 censuses and the ACS annual earnings is available. For the 1970
and 1980 census comparable measures are derived by adding business and farm income.
• Schooling: High school dropouts are those with less than 12 years of schooling for the 1970 and
1980 sample and those with less than a high school degree in the 1990 to 2011 samples. High school
(College) graduates are those with 12 (16) or more completed years of schooling for the 1970 and
1980 sample and those with a high school (bachelors) degree or higher in the 1990 to 2011 samples.
Graduate school attendants/graduates are those with more than 16 years of schooling for the 1970 and
1980 sample and those with schooling beyond college in the 1990 to 2011 samples. “Some college”
individuals have in between 13 and 15 years of schooling for the 1970 and 1980 census and a high
school but no college degree in the 1990 to 2011 samples.
62

• Labor force participation and employment are two important variables in the analysis. Both variables
are coded referring to the individual ’s status in the “previous week”. This does not have to be the same
week for all respondents, since the census or survey was taken over a period of time. In all years, labor
force participants consist of employed and unemployed persons. Individuals are considered employed
under three different circumstances:
1. They worked at least one hour for pay or profit during the reference period.
2. They worked at least 15 hours as "unpaid family workers".
3. They had a job from which they were temporarily absent (e.g., because of illness or vacation time).
• Incarceration: To estimate the proportion incarcerated I use the group-quarters identifier included
in the PUMS data. The decennial Census enumerates both the institutionalized as well as the noninstitutionalized
population. See Butcher and Piehl (1998) and Johnson and Raphael (2005) for studies
that also use the group quarter variable to identify the incarcerated.
CPS Data
This paper pools the 2008-2012 March CPS samples. Each sample year includes about 200,000 individuals
from the non-institutionalized population. Weights are included to reflect sampling. Observations with
imputed values are included but dropping them has minimal effects on the results.
Sample Restrictions
• Age: The sample includes individuals aged 21-65 in order to focus on the working-age population.
• Race: Individuals are coded as black, Asian, and white if they are non-Hispanic. Hence Hispanics are
comprised of different races in this sample.
• Earnings: In the results presented I excluded persons with negative earnings. Only a very small
fraction of individuals have negative earnings.
Constructed variables
• Annual earnings: Annual earnings include earnings, farm and business income.
• Schooling: Data on high school, college, and graduate school degrees are available. Individuals with
more than an high school degree and less than a college degree fall into the “some college” category.
Individuals with more than a college degree (whether they complete graduate school or not) count as
graduate school attendants/ completers.
• Labor force participation and employment are comparable to those in the ACS/census.
63

Refugee status
The world fact book provides estimates of refugees and internally displaced persons for each country. An
individual was coded as a refugee if he or she immigrated from a country in unrest in a year in which many
refugees fled the country. For more recent years (2000-2013) the Office of Refugee Resettlement has detailed
data on refugee arrival by source country from 2000-2012, which was used to infer refugee status.
C Mathematical Proofs of Section 7
Assumptions on the Errors
Following Blume et al. (2010), we make two standard assumptions on the error terms:
1. Let’s assume that the expected value of e i is 0, conditional on the information set (x i, ¯x g ,y g ,i 2 g)
8i, gE(e i |x i , ¯x g ,y g , i 2 g)=0
2. Let’s assume that for each i, j,,g, h such that i 6= j or g 6= h
cov(e i e j |x i , ¯x g ,y g ,i 2 g,x j , ¯x h ,y h , j 2 h)=0
Condition (2) rules out conditional variation between errors. Including group memberships eliminates any
link between the identity of the group and model errors, which allows groups to be treated as exchangeable.
Model Derivation
The choice of an immigrant in racial group g 1 with immigrant background g 2 is denoted as w i,g1,2 = aw i,g1 +
(1 a)w i,g2 . Assume that g 1 evolves as follows
w i,g1 = k + cx i + d ¯x g1 + Jm e i,g 1
+ e i
Restricting the evolution of choices to be consistent with the model, ignoring the intercept, and taking
the limit as the number of group members approaches infinity, we get the following equation
Solving for ¯w g1 , gives us
¯w g1 = c ¯x g1 + d ¯x g1 + ¯ Jw g1 + ē g1 (18)
64

Plugging (19) back into (3) gives us
(c + d)
¯w g1 =
1 J ¯x g 1
+ ēg 1
1 J
(19)
or
w i,g1 = k + cx i + d + Jc
1 J ¯x g 1
+ J
1 J ēg 1
+ e i (20)
w i,g1 = p 0 + p 1 x i + p 2 ¯x g1 + n i,g1 (21)
Case 1: Group 2 Members Are only Affected by own Characteristics
Recall that the decision rule of group 2 members in case 1 evolves as follows
Returning to the joint choice probability, we get
w i,g2 = k + cx i + v i,g2 = p 0 + p 1 x i + v i,g2 (22)
Simplifying, this gives us
w i,g1,2 = a i (p 0 + p 1 x i + p 2 ¯x g1 + n i,g1 )+(1 a i )(p 0 + p 1 x i + n i,g2 ) (23)
where a i measures the degree of assimilation.
w i,g1,2 = p 0 + p 1 x i + a i p 2 ¯x g1 + v i,g1,2 (24)
Case 2: Group 2 Members Are Affected by Average Characteristics of Entire Metropolitan
Area
Now, both group 1 and group 2 members’ decisions evolve as follows
or
w i,g j
= k + cx i + d ¯x g j
+ Jm e i, j + e i , j = 1, 2
w i,g j
= p 0 + p 1 x i + p 2 ¯x g j
+ v i,g j
, j = 1, 2
Returning to the joint choice probability for immigrants, we get
w i,g1,2 = a i (p 0 + p 1 x i + p 2 ¯x g1 + n i,g1 )+(1 a i )(p 0 + p 1 x i + p 2 ¯x g2 + n i,g2 ) (25)
65

Simplifying, this gives us
where a i measures the degree of assimilation.
w i,g1,2 = p 0 + p 1 x i + a i p 2 ¯x g1 +(1 a i )p 2 ¯x g2 + v i,g1,2 (26)
D Welfare Eligibility Requirements for Non-Citizens
The programs discussed here are the Supplementary Security Income (SSI), Supplemental Nutrition Assistance
Program (SNAP), the Temporary Assistance for Needy Families (TANF), Medicaid, and State Children’s
Health Insurance Program (SCHIP). When determining welfare eligibility for non-citizens, two broad
criteria are generally taken into account. 30 The most general screen is the “qualified” versus “not qualified”
alien. Qualified immigrants include legal permanent residents (LPRs), refugees, asylees, Cuban/Haitian
entrants, and a few smaller categories. Non-qualified immigrants mostly include illegal and temporary immigrants.
Non-qualified immigrants are largely ineligible for all except emergency benefits. The second
factor determining eligibly is the date of entry into the United States. Up until 2002, qualified immigrants
who entered after August 22, 1996 were barred from SSI and SNAP until they become citizens and from
TANF, Medicaid, and SCHIP for five years after entry. Immigrants who entered before 1996 were eligible
for TANF, Medicaid, SCHIP, and SSI but food stamp eligibility remained more restricted. Today, lawful
permanent residents must wait at least five years to be eligible for benefits, but states have the option of providing
them earlier. Refugees and asylees are generally eligible for public benefits if they meet the eligibility
criteria.
Camarota (2011, 2012) finds that immigrant-headed households use more Medicaid than native-headed
households with children and had higher food assistance, but lower cash assistance. This study, however, did
not adjust for income. In other words, the percent of immigrants receiving benefits is higher in part because
a greater percent of immigrants are low-income, and more likely to be eligible for public benefits. Ku and
Bruen (2013) find that low-income non-citizen children and adults (below the 200% poverty line) generally
use Medicaid, SNAP, cash assistance, and SSI at a lower rate than comparable low-income native-born
children and adults.
30 This discussion draws from Fix and Haskins (2002) and Ku and Bruen (2013).
66