econ questions
wjccrsg720
NBER WORKING PAPER SERIES
THE CAUSES AND CONSEQUENCES OF DISTINCTIVELY BLACK NAMES
Roland G. Fryer, Jr. Steven D. Levitt
Working Paper 9938 http://www.nber.org/papers/w9938
NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue
Cambridge, MA 02138 August 2003
We are grateful to George Akerlof, Susan Athey, Gary Becker, David Card, John Donohue, Esther Duflo, Shoshana Grossbard-Shechtman, Daniel Hamermesh, Lawrence Katz, Glenn Loury, Douglas McAdam, Andrei Shleifer, and Steve Tadelis for helpful comments and suggestions. We also thank seminar participants too numerous to mention for their comments and suggestions. Michael Roh provided excellent research assistance. Financial support provided by the National Science Foundation. This paper was written while Levitt was a fellow at the Center for Advanced Study in the Behavioral Sciences in Stanford and Fryer was an NSF Post-Doctoral Fellow at the University of Chicago. The views expressed herein are those of the authors and not necessarily those of the National Bureau of Economic Research.
©2003 by Roland G. Fryer, Jr. and Steven D. Levitt. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.
The Causes and Consequences of Distinctively Black Names Roland G. Fryer, Jr. and Steven D. Levitt NBER Working Paper No. 9938 August 2003 JEL No. J0, J7, Z1
ABSTRACT
In the 1960's, Blacks and Whites chose relatively similar first names for their children. Over a short
period of time in the early 1970's, that pattern changed dramatically with most Blacks (particularly
those living in racially isolated neighborhoods) adopting increasingly distinctive names, but a subset
of Blacks actually moving toward more assimilating names. The patterns in the data appear most
consistent with a model in which the rise of the Black Power movement influenced how Blacks
perceived their identities. Among Blacks born in the last two decades, names provide a strong signal
of socio-economic status, which was not previously the case. We find, however, no negative causal
impact of having a distinctively Black name on life outcomes. Although that result is seemingly in
conflict with previous audit studies involving resumes, we argue that the two sets of findings can
be reconciled.
Roland G. Fryer, Jr. NBER 1050 Massachusetts Avenue Cambridge, MA 02138 and Harvard Society of Fellows [email protected]
Steven D. Levitt Department of Economics University of Chicago 1126 East 59th Street Chicago, IL 60637 and NBER [email protected]
1
On May 17, 1954, the landmark Supreme Court decision in Brown v. Board of Education
of Topeka, Kansas ruled, unanimously, that segregation in public schools was unconstitutional.
This ruling paved the way for the fall of Jim Crow and large-scale desegregation. In the 1960s, a
series of further government actions were taken with the goal of achieving racial equality and
integration, most notably the Civil Rights Act of 1964, Executive Order 11246 in 1965, and the
Fair Housing Act of 1968. The civil rights movement arguably represents one of the most
profound social transformations in American history (Woodward 1974 and Young 1996).
Nonetheless, an enormous racial divide persists. There are large disparities between
Blacks and Whites in the United States on many indicators of social and economic welfare
including income (Bound and Freeman 1992, Chandra 2003, Heckman, Lyons and Todd 2000,
and Smith and Welch 1989), educational achievement (Jencks and Phillips 1998), out-of-
wedlock childbearing (Ventura and Bachrach 2000), health (see Chandra and Skinner,
forthcoming), and criminal involvement (Reno et.al. 1997). The degree of residential
segregation by race, though lower today than in 1970, remains high (Cutler, Glaeser, and Vigdor
1999, Massey 2001).
Racial differences also persist, and in some cases have become even more pronounced,
on a wide range of cultural dimensions including musical tastes (Waldfogel, forthcoming),
linguistic patterns (Wolfram and Thomas 2002), and consumption choices. For instance, the
cigarette brand Newport has a 75 percent market share among Black teens, but just 12 percent
among White teen smokers; 65 percent of white teens smoke Marlboro compared to only 8
percent of blacks. Seinfeld, one of the most popular sitcoms in television history among whites,
never ranked in the top 50 among blacks. Indeed, of the top ten shows with the highest
2
viewership among Whites during the 1999-2000 television season, only one show was also
among the top ten for blacks: NFL Monday Night Football.
Understanding whether cultural differences are a cause of continued economic disparity
between races is a question of great social importance. Cultural differences may be a cause of
Black economic struggle if Black culture interferes with the acquisition of human capital or
otherwise lowers the labor market productivity of Blacks (as argued in the culture of poverty
paradigm in sociology, see Hannerz 1969, Lewis 1966, Riessman 1962, and implicitly, Anderson
1990). For instance, high-achieving Black children may be ostracized by their peers for “acting
white,” potentially leading to lower investment in human capital (Fordham and Ogbu 1986,
Austen-Smith and Fryer 2003). Speaking “Ebonics” may interfere with the ability to interact
with White co-workers and customers, or disrupt human capital acquisition more directly.1 On
the other hand, the presence of a Black culture may simply be the consequence of past and
current segregation and economic inequality, but play no role in perpetuating economic disparity.
If differences in tastes do not influence human capital acquisition or labor market productivity,
then there is little reason to believe that such tastes will have a causal negative economic impact
on Blacks. For example, “soul food” (Counihan and Van Esterik 1997) and traditional African
American spirituals (Jackson 1944) can be traced to the social conditions endured during slavery,
but are unlikely to be causes of current poverty. Eliminating cultural differences in this scenario
would have no overall impact on Black welfare relative to Whites.
1 Orr (1989) writes “I did not even know there was something called Black English when I began to realize
that many of the difficulties that my students were having were rooted in language. It was the incongruence of the obvious intelligence and determination of these students with the unusual kinds of misunderstandings that persisted in their work that drove me to find answers....For the students whose first language was Black English vernacular, language can be a barrier to success in mathematics and science.”
3
A primary obstacle to the study of culture has been the lack of quantitative measures. In
this paper, we focus on one particular aspect of Black culture --the distinctive choice of first
names—as a way of measuring cultural investments. Our research builds upon a growing
literature by economists devoted to understanding a diverse set of social and cultural phenomena
(Akerlof and Kranton 2000, Berman 2000, Fryer 2003, Glaeser, Laibson, and Sacerdote 2002,
Ianaconne 1992, Lazear 1999). In contrast to these earlier papers, however, our contribution is
primarily empirical.
Using data that cover every child born in California over a period of four decades, our
analysis of first names uncovers a rich set of facts. We first document the stark differences
between Black and White name choices in recent years. 2 For example, more than forty percent
of the Black girls born in California in recent years received a name that not one of the roughly
100,000 White girls born in California in that year was given. Even among popular names,
racial patterns are pronounced. Names such as DeShawn, Tyrone, Reginald, Shanice, Precious,
Kiara, and Deja are quite popular among Blacks, but virtually unheard of for Whites.3 Connor,
Cody, Jake, Molly, Emily, Abigail, and Caitlin are distinctively White names. Each of those
names appears in at least 2,000 cases, with less than two percent of the recipients Black.4
Overall, Black choices of first names differ substantially more from Whites than do the names
chosen by native born Hispanics and Asians.
More surprising, perhaps, is the time series pattern of Black first names. In the 1960s, the 2 There are multiple dimensions along which a name can be considered “black” or “white.” For example, Lieberson and Mikelson (1995) study distinctive patterns of phonemes that are characteristic of Black names. In this paper, we study only one dimension of the issue: the relatively frequency with which Blacks and Whites choose a given name for their children. 3 For example, there are 405 children named DeShawn, 403 of whom are Black. The name Tyrone is given to 445 Black boys and only 12 Whites. 274 out of 277 Shanice’s are Black, as are 377 out of 388 girls named Precious, and 521 out of 539 girls named Deja. 4 The most extreme case is for the name Molly, in which only 6 of 2,328 children given the name are Black.
4
differences in name choices between Blacks and Whites were relatively small, and factors that
predict distinctively Black names in later years (single mothers, racially isolated neighborhoods,
etc.) have much lower explanatory power in the 1960’s. At that time, Blacks who lived in highly
racially segregated neighborhoods adopted names that were almost indistinguishable from
Blacks in more integrated neighborhoods and similar to Whites. Within a seven-year period in
the early 1970’s, however, a profound shift in naming conventions took place, especially among
Blacks in racially isolated neighborhoods. The median Black female in a segregated area went
from receiving a name that was twice as likely to be given to Blacks as Whites to a name that
was more than twenty times as likely to be given to Blacks. Black male names moved in the
same direction, but the shift was less pronounced. Among a subset of Blacks, encompassing
about one-fourth of Blacks overall and one-half of those in predominantly White neighborhoods,
name choices actually became more similar to those of Whites during this period.
We argue that these empirical patterns are most consistent with a model in which the rise
of the Black Power movement influenced Black identity. Other models we consider, such as
ignorance on the part of Black parents who unwittingly stigmatize their children with such
names, simple price theory models, and signaling models, all contradict the data in important
ways.
The paper concludes by analyzing the causal impact of distinctively Black names on life
outcomes. Previous studies have found that distinctively black names are viewed negatively by
others (e.g. Busse and Seraydarian 1977). Most persuasive are audit studies in which matched
resumes, one with a distinctively Black name and another with a traditionally White name, are
mailed to potential employers (Jowell and Prescott-Clarke 1970, Hubbick and Carter 1980,
5
Brown and Gay 1985, Bertrand and Mullainathan 2002). Such studies repeatedly have found
that resumes with traditional names are substantially more likely to lead to job interviews than
are identical resumes with distinctively Black names. These results suggest that giving one’s
child a Black name may impose important economic costs on the child. In our data, however, we
find no compelling evidence of a causal impact of Black names on a wide range of life outcomes
after controlling for background characteristics.
Although seemingly in conflict with the audit study findings, we believe the two sets of
results can be reconciled. To the extent that Black names are used simply as signals of race by
discriminatory employers, it is unlikely that names would be correlated with job outcomes
beyond the interview stage since the employer directly observes the applicant’s race once an
interview takes place. In the face of discriminatory employers, it is actually in the interest of
both employee and employers for Blacks to signal race, either via a name or other resume
information, rather than undertaking a costly interview with little hope of receiving a job offer.
More generally, we show that Black names are correlated with family background characteristics
that may predict labor market productivity, even after controlling for the type of information
available to employers on resumes. If that is the case, then it may be efficient (albeit illegal) for
employers to use names in screening applicants. Once an individual is personally known to us,
names wane dramatically in importance, as a moment of reflection about one’s own oddly named
acquaintances and colleagues will likely bear out. Because of this, and the fact that the cost of
changing one’s name is low, it is hard to imagine how names could plausibly have a large impact
on life outcomes, even if resume callbacks are somewhat reduced. Therefore, we conclude that
the stark differences in naming patterns among Blacks and Whites is best explained as a
6
consequence of continued racial segregation and inequality, rather than a cause that is
perpetuating these factors.
The remainder of the paper is structured as follows. Section II describes the data used in
the analysis. Section III summarizes the basic patterns observed in the data. Section IV attempts
to reconcile the stylized facts with a range of potential theories. Section V analyzes the causal
impact that names have on life outcomes and reconciles our results with previous audit studies.
Section VI concludes. A data appendix describes the details of our sample construction.
II. The Data
The data used in this paper are drawn from the Birth Statistical Master File maintained by
the Office of Vital Records in the California Department of Health Services. These files provide
information drawn from birth certificates for all children born in California over the period 1961-
2000 – over sixteen million births. With the approval of the California Committee for the
Protection of Human Subjects, we have been able to supplement the information contained on
the public use versions of these data sets with personal identifiers including mother’s first name,
mother’s maiden name, and child’s full name.
The information included varies by year and has generally been increasing over time.
For our entire sample, we have information on the baby’s first name, race, gender, date of birth,
hospital of birth, and birth weight, as well as the mother’s maiden name, parental ages, and
inferred marital status. Beginning in 1989, information on parental education, residential zip
code, and source of payment were added to the data. Also starting in 1989 we know the
mother’s first name and exact date of birth, two critical elements for linking information from a
7
woman’s own birth certificate to that of her children in those cases in which a woman is both
born in California and later gives birth there. This allows us to look at the relationship between
circumstances at a woman’s birth (e.g. her own name, her mother’s level of education, her
mother’s marital status, racial segregation, etc.) and the situation in which that same woman lives
in when she gives birth decades later.5 It also enables us to link information from all births to the
same mother that took place in California, which permits the comparison of naming patterns
controlling for mother-fixed effects.
We restrict our sample to non-Hispanic Blacks (referred to simply as Blacks).6 In
determining how Black a name is, our comparison group is non-Hispanic Whites. Summary
statistics for the Black babies are provided in Table 1. We divide the sample into four sets of
years: 1961-1967, 1968-1977, 1978-1988, and 1989-2000. The cutoffs for these groups have
been chosen with three goals in mind: (1) roughly equalizing the time periods in each group, (2)
matching the cutoffs to breaks in trend in the aggregate data, and (3) grouping years in which
similar sets of covariates are available. Excluded from the data set are a small percentage of
observations missing information on names. When other variables are missing, we opted to
leave the observation in the analysis, including an indicator variable for the missing value. A
few notable patterns are apparent in the summary statistics. The fraction of Black children born
out-of-wedlock rises over the sample, from 31 percent in the early period to 58 percent in the
latest years. The fraction of interracial births also rises sharply, almost doubling over the period. 5 Unfortunately, the father’s first name is not included in the data, so a parallel analysis cannot be performed for males. 6 There is some ambiguity in racial and ethnic categorizations when children are born to parents of different races. We use the classification of a child’s race and ethnicity on the birth certificate to assign these categories, except in the years 1999 and 2000 when the information on child’s race is missing. In those years, a child is considered Black if either parent is Black. Before 1989, we do not have explicit identifiers for Hispanics. Using information from later years of data, however, we are able to effectively identify Hispanic surnames and drop from the sample any child born to a parent with a surname or maiden name that is more than ten percent Hispanic.
8
White women are five times more likely to give birth to children with Black fathers than are
Black women with White fathers. The fraction of teenage births initially rises, but then falls to
the lowest levels (18 percent) in the period 1989-2000. Shifting attention to the aggregate
variables in the bottom portion of the table, the Black share of non-Hispanic births steadily rises
over the sample, going from 12.85 percent initially to 27.72 percent.
The bottom row of Table 1 presents our summary measure of how distinctively Black
first names are. The measure we choose, which we term our “Black name index” (BNI) is:
( )
( ) ( ) 100,Pr,Pr ,Pr
, ∗+ =
tWhitenametBlackname tBlackname
BNI tname (1)
where name reflects a particular first name and t reflects the year of birth. The index ranges
from 0 to 100. If all children who receive a particular name are White, then BNI takes on a value
of 0. If only Black children receive a name, BNI is equal to 100. If Whites and Blacks are
equally likely to choose a name, BNI equals 50. If Blacks are four time as likely to select a
name, then BNI takes on a value of 80 (e.g., 4/ (4+1)*100=80).7 This measure is invariant to the
fraction of the population that a particular minority group comprises, and to the overall
popularity of a name. The bottom row of Table 1 shows that the mean BNI for Blacks rises from
60.9 in the early period (implying that the mean Black name is given to Blacks about 50 percent
more often than it is given to Whites) to 71.0 in the last period (implying these names are given
to Blacks about two and one-half times more frequently than Whites).
Figure 1 more clearly demonstrates the dramatic differences between Black and White
name choices. The figure presents a smoothed plot of the probability distribution function of
7 In computing BNI for a particular child, we exclude that child from the calculation. When a name is unique, meaning that only one child receives that name in a particular year, we assign the name a value of 0 if the person getting the name is White and 1 if the baby is Black. We explore unique names in greater detail later in the paper.
9
Black and White names. The horizontal axis reflects how Black an individual’s name is. The
vertical axis measures the density of names chosen by race. More than forty percent of Whites
are given names that are at least four times as likely to be given to Whites (between 0-20). The
fraction of Whites steadily shrinks as one moves from left to right in the figure. More than half
of all Blacks have names that are at least four times as likely to be given to Blacks (between 81-
100). For both races there is very little weight in the middle of the distribution (41-60), implying
that there are relatively few individuals carrying names that are similarly likely for Blacks and
Whites.
One might suspect that the sharp differences across races in Figure 1 may in part be an
artifact of how we construct our measure of Black names using the observed empirical
distribution. In other words, we might mis-categorize a name as being distinctively Black or
White simply because for many names we observe only a few individuals with that name.8
Figure 2, which is identical to Figure 1 except that it compares the naming patterns of Whites to
that of American-born Asians, demonstrates that this is not the case. With the exception of a
small fraction (approximately ten percent) of the Asian population adopting names that are rare
among Whites, name choices of American-born Asians strongly parallel White name choices. A
comparison of native-born Hispanics and Whites in Figure 3 shows differences in naming
patterns among these two groups, although there is still substantially more overlap than for
Blacks and Whites.9
Another difference between Black and White naming patterns is the greater usage of
8 Although, arguing against this conjecture, the results in Figure 1 only change slightly when we exclude names received by twenty or fewer children in the sample. 9 We have also compared the names chosen by Whites with different levels of education. There are systematic differences in name choices (larger, in fact, than between Asians and Whites overall), but these differences are much smaller than either the Black-White or Hispanic-White gap.
10
unique or nearly unique names in the Black community. Figures 4a and 4b report, by race and
gender, the number of children born in California in that same year (regardless of race) with that
child’s name. Remarkably, nearly 30 percent of Black girls receive a name that is unique among
the hundreds of thousands of children born annually in California. Among Whites, that fraction
is only five percent. Similarly, the fraction of unique names among Black boys is six times
higher than for White boys, although only about half the rate of Black girls. The median Black
child shares his or her name with 23 other children; the number is almost fifteen times greater for
Whites (351).10
Perhaps the most interesting findings in the data are the changes in the distribution of
name choices over time. For children born in each year between 1961 and 2000, we compute
our Black name index and then rank order the Blacks in our sample according to how Black their
name is. Figure 5 presents the mean BNI by year for each of the four quartiles of the
distribution. The top quartile is very close to 100 throughout the entire time period (i.e. almost
one-quarter of Blacks had names virtually never given to Whites throughout the sample) and thus
exhibits little time-series variation. For the other three quartiles, Black naming patterns were
largely stable throughout most of the 1960s. Beginning in the late 1960s, the second quartile
from the top experiences a sharp rise in how Black the name choices are. Between 1968 and
1977, the mean BNI within this quartile goes from roughly 75 (meaning the name was three
times as likely to be given to a Black baby as a White baby) to almost 95 (15 times more likely
to be given to a Black baby). The third quartile also rises over that time period, but not as
sharply, and also steadily increases over the period 1985-2000. The bottom quartile, in contrast,
10 The differences between Blacks and Whites in figures 4 are not primarily due to the fact that there are many more Whites than Blacks born in California each year. If we randomly select a subset of Whites equal in size to the number of Blacks born each year, a similar pattern of results persists.
11
remains almost unchanged throughout the sample period.
Figure 6 is identical to Figure 5, except that it disaggregates the data by the degree of
racial segregation in the hospital in which a Black baby is born. We show only the top and
bottom deciles (i.e. the 10 percent of Blacks giving birth in hospitals with the greatest percent
Black and the lowest percent Black respectively) to highlight the extremes of the distribution.
The most important observation emerging from Figure 6 is the widening gap between the names
given to Black babies born in predominantly Black and predominantly White hospitals over time.
At the beginning of the sample, name choices differ little by the degree of racial segregation.
For example, in the bottom quartile, Blacks born in racially segregated hospitals have a mean
BNI of 26 versus 23 for Blacks born in predominantly White hospitals. By 1978, these gaps
have widened substantially, particularly in the third quartile where Blacks in Black hospitals
have shifted more than twenty points relative to the beginning of the sample, but names of
Blacks in White hospitals are essentially unchanged. Interestingly, among the Black babies
given the least distinctively Black names (the bottom quartile), those born in White hospitals
actually see a discernible decrease in how Black their names are, in contrast to the rest of the
distribution.11
Figure 7 presents the time path of the median Black name for the eight California
counties with the greatest number of Blacks in 1990. With the exception of Orange County, the
other counties follow remarkably similar time-series patterns. Orange County, which has a much
smaller fraction Black than any of the other counties shown and has the highest per capita
incomes, follows a path more similar to the 10th percentile Black names in other counties. The
11 The choice of names by Whites over time does not exhibit this pattern of bifurcation that is present in Black naming in Figures 5 and 6. The BNI for Whites is almost uncorrelated with the racial segregation of the hospitals in which they give birth throughout the entire time period.
12
similarities across counties suggest that the forces that are operating on Black names are being
driven by macro forces, rather than locally isolated phenomena.
The sharp rise in the prevalence of distinctively Black names is driven in part by the
increasing use of unique names among Blacks. Figures 8a and 8b present time-series data on the
percentage of babies with unique first names by gender and racial mix of the birth hospital,
where a unique name is defined as a name that no other baby born in California that year shares.
The fraction of Black baby girls receiving unique names rose for the first fifteen years of the
sample. Initially rates of unique naming were similar among Blacks in predominantly Black and
predominantly White neighborhoods (around 10 percent); that gap grows over time.
Nonetheless, even for Blacks in White neighborhoods, almost one-quarter of baby girls received
unique names. Among Whites, the rates are around five percent, although this (as well as the
Black numbers) is likely to be an upper bound due to typographical errors being counted as
unique names. Figure 8b documents a parallel, though less pronounced, phenomenon for Black
boys. In 1961 the percentage of unique names ranged from 5-6%, irrespective of hospital. By
2000, one-fifth of Black boys born in predominantly Black hospitals in 2000 receive unique
names.
Unique names, while an important part of the explanation for the divergence in Black and
White naming patterns, are not the entire story. Figure 9 presents time-series evidence on the
fraction of Blacks that have non-unique names that are above 80 on the BNI index, as well as the
fraction of Blacks with unique names. The share of babies given distinctively Black, but non-
unique names rises from 18 percent to 30 percent from the beginning of the sample to the peak in
1978. Unique names rise from 9 percent to 23 percent over that same period. Thus, the absolute
13
increase in Black babies receiving unique names and non-unique distinctively Black names is
roughly the same between 1961 and 1978. Note, however, that after a dip in the 1980s, unique
names gain in prominence in the 1990s, whereas non-unique distinctively Black names do not
rise in popularity in the later years.
Table 2 examines the relationship between parental characteristics, birth circumstances,
and the names given to children at four different points in time. We estimate equations of the
form:
ichii XBNI εθγαξβ +++= ,
where iX represents an array of individual level background variables, hξ represents hospital
level variables, and cγ denotes county level variables. The first four columns of the table
correspond to different sample periods. Column 1 reflects Black children born between 1961
and 1967, before the sharp changes in Black naming behavior. Column 2 has births occurring
between the years 1968 and 1977, the years in which the transition occurred. Columns 3 and 4
capture the periods 1978-1988 and 1989-2000 respectively. We separate these two time periods
because of the availability of a more complete set of covariates after 1989. Also, after 1989 we
have the necessary information to link multiple births to the same women, allowing for the
inclusion of mother fixed effects in Column 6. All of the controls included in the regression are
taken from a child’s birth certificate and thus are determined at the time of birth. For the entire
period, covariates available include parents’ marital status, age, and race, mother’s place of birth,
her total number of children, months of prenatal care, the baby’s birth weight, the percent Black
at the birth hospital, the percent black in the county, and whether the child is born in a county
hospital. In the later years of the sample, the set of included controls is much richer; they
14
include income and percent Black at the ZIP code level, highest grade completed by mother and
father, and expected source of payment.
Columns 1-4 hold the set of covariates constant and compare the relative importance of
these variables in predicting BNI over the four periods. In almost all cases, variables associated
with low socio-economic status are also associated with Blacker names. Moreover, the link
between low socio-economic status and Black names becomes much stronger over time. The
coefficients on mother’s age, birth weight, and single mother are less than one-half as large in the
first period (column 1) as they are in the last period (column 4). Father’s age has no impact in
the early period, but does have a negative coefficient in the later periods. Note also that the R-
squared in the regressions increases steadily over time, meaning that these characteristics explain
a growing fraction of the variation in names. A final implication of the table is that names are
becoming more distinctively Black over the course of the sample. For instance, a woman with
the mean sample characteristics in the 1960s would be predicted to choose a name with a BNI of
60.9 in the 1961-1967 period. A woman with those same characteristics giving birth in the
1990s will on average choose a name with a BNI of 71.9. Likewise, for a woman with the
average traits of mothers giving birth in the 1990s, the predicted BNI of the name is 71.0 for a
baby born in the 1990s compared to 61.0 if that baby had been born in the 1960s.
The specification estimated in column (5) and (6) are similar to those in the previous
columns of the table, except that a much richer set of covariates are available in this latter period.
The additional variables added to the specification carry coefficients that are quite interesting:
Blacker name choices are associated with residing in lower-income zip codes, lower levels of
parental education, not having private insurance, and having a mother who herself has a Blacker
15
name.
The last column of table 2 adds mother-fixed effects to the specification. Thus, the
identification comes from changes in a mother’s situation across different births. To the extent
that mothers anticipate the changes that occur in their circumstances between births and factor
these expectations into the name given to their first child, the fixed-effects coefficients will be
biased towards zero. The estimates in column (6) are generally substantially smaller in
magnitude than in the earlier columns, with the exception of variables reflecting how Black the
neighborhood is and the coefficient on female. Interestingly, the coefficient on mother’s age
flips sign, implying that women who have babies early in life tend to choose Black names
relative to other women, but that a given women picks slightly Blacker names for her children as
she grows older. In addition, the inclusion of mother-fixed effects raises the R-squared from
.092 to .773 -- there is a great degree of continuity in the names mothers choose for their
children, implying that person-specific tastes are quite important determinants of name choice.
The results with mother-fixed effects suggest either that temporary changes in circumstances
have a relatively small impact on name choices, or that our measures of current circumstances
are noisy.
III. Understanding the patterns in the data through the lens of economic theory
A number of stylized facts emerge from the analysis of the preceding sections. Blacks,
much more than other minorities, choose distinctive names for their children. The
distinctiveness of Black names has risen greatly over time, most notably in the late 1960s and
early 1970s. These shifts in naming patterns have not, however, been uniform. In particular,
16
among the quarter of the Black population choosing the names most common among Whites, the
opposite pattern is evident. Further, Blacks living in highly segregated Black communities today
are much more likely to have distinctively Black names than those in integrated communities,
whereas this was not the case in the early 1960’s. Finally, until the late 1970s, the choice of
Black names was only weakly associated with socio-economic status; in the 1980s and 1990s
distinctively Black names have come to be increasingly associated with mothers who are young,
poor, unmarried, and have low education.
In this section, we consider the extent to which existing theories can successfully account
for this disparate set of facts. The four models we compare differ substantially. In the first
model, Black parents are misinformed about how others perceive names, leading distinctively
Black names to be costly mistakes. The second model is a simple price theory model in which
parents are optimally choosing names for their children, trading off benefits of having a Black
name in a Black community and benefits of having a White name in the White community. In
the third class of models, names serve as signals of otherwise unobservable personal
characteristics. The final model considered is the identity model of Akerlof and Kranton
(2000).12
12 All of these models stand in contrast to the taste-based frameworks of previous research such as Lieberson (2000), who argues that names are driven by popular fads and fashions. Without further restrictions, there are few testable empirical predictions of the taste-based model. One would imagine, however, that a conceivable prediction of a taste-based approach would be that naming patterns should demonstrate local variation, as there are certain to be variations in local tastes and styles or at least variations in the timing of adopting fads in naming trends. Figure 7 demonstrates that this is not an important feature of our data. Everywhere except Orange County, Black naming conventions follow remarkably similar patterns.
17
The Ignorance Model
One explanation for the prevalence of distinctively Black names is ignorance on the part
of Black parents, who fail to appreciate the costs they are imposing on their children through
such choices. Audit study results, for instance, suggest that Black names may be punished in the
labor market.
This theory fails to adequately explain the sharp increase in distinctively Black names in
the late 1960s and early 1970s. There is little reason to believe that Black parents became
systematically less informed about the consequences of distinctively Black names at that time.
This theory also is at odds with the fact that those adopting Black names in the early 1970s were,
for the most part, representative of the Black community, not a small subset of parents likely to
be particularly misinformed.13
Price Theory Model
Consider the following skeletal outline of a price theory model of names, which we
derive formally in an earlier version of this paper (Fryer and Levitt 2002). Parent’s give names
to their children at birth to maximize the child’s expected utility. Individuals are born into
neighborhoods that differ in racial composition. People live and work in the same
13 On the other hand, one cannot a priori rule out that this explanation has potential relevance for
explaining at least some of the patterns of the last two decades. If it became apparent in the 1980s that there were costs to having a Black name, one might expect that parents who were likely to be best informed about these costs (e.g., older working parents living in less racially isolated neighborhoods) would choose such names less frequently, whereas learning on this dimension might be more gradual for teenage mothers in the inner city. In absolute terms, this explanation fails, because the adoption of Black names rose throughout the 1990s. At least in relative terms, however, there was a shift in the 1980s and 1990s towards an increasing concentration of distinctively Black names among parents least likely to be well informed about the potential stigma of such names in broader society. In light of results presented later in the paper that call into question the costs of adopting distinctively Black names, however, we are quite skeptical of this theory as an explanation for the observed phenomena.
18
neighborhoods. Moving between neighborhoods is costly. The returns to ability are assumed to
be higher in predominantly White neighborhoods. White names provide benefits in interactions
with Whites; Black names are beneficial when interacting with Blacks. We also assume that the
value of a White name is increasing in ability, motivated by the audit study literature which
suggests that the primary cost of a distinctively Black name is via the labor market. A child’s
ability is not known with certainty at birth, although the distribution from which the child’s
ability will be drawn is known.
In this model, parents will opt for Whiter names when (i) their children are more likely to
be high ability, (ii) the cost of moving to predominantly White neighborhoods falls, (iii) returns
to ability rise in the labor market, (iv) the relative cost of having a White name when interacting
with Blacks falls, and (v) the benefit of having a White name when interacting with Whites rises.
The predictions of this model face mixed success in terms of the patterns observed in the
data. Consistent with the theory is the fact that those in racially isolated communities are
especially likely to adopt distinctively Black names in recent decades, and that such names are
most common among groups likely to face the greatest barriers to participating in traditional
labor markets. This theory, however, does quite poorly in explaining the sharp rise in
distinctively Black names in the late 1960s and early 1970s – a period immediately following
the passage of national Fair Housing laws in 1968, falling social barriers to integration, and
increased economic opportunity for Blacks. Empirically, urban racial segregation which had
been rising began to fall around this time and has steadily declined for three decades (Cutler,
Glaeser, and Vigdor 1999).14 In light of these apparent reductions in the costs of switching
14 Although the level of residential racial segregation has declined at a relatively slow rate, to the extent that these persistent declines were anticipated and parents are concerned with the impact a name will have on their children as
19
between neighborhoods, one would have expected a shift away from distinctively Black names at
precisely the point where such names were becoming most prevalent.15
Additionally, it is not clear that the price theory model provides an adequate explanation
as to why Black names are much more distinctive than Asian or Hispanic names, when
presumably many of the same tradeoffs might also exist among those minority groups.
A Signaling Model
In a third attempt to explain our disparate set of facts, we develop a simple signaling
model in which distinctively Black names serve as a signaling device, but are otherwise not
productive, along the lines of prior research by Iannacone (1992), Berman (2000), and Fryer
2003).16 Imagine a predominantly Black neighborhood populated by Black individuals of one or
two types: those who have a strong affinity for the Black community (the “black” type) and those
who do not (the “white” type). An individual’s type is fixed at birth and cannot be changed.
Each individual knows her own type, but type is not observable to others. An individual’s utility
is determined by his type and the neighborhood’s perception about his type. Regardless of one’s
actual type, social interactions in the Black community yield higher utility if others believe that
adults in the labor market, one would nonetheless expect to see an immediate, abrupt shift in naming patterns. 15 A proponent of this theoretical framework might argue that a more nuanced application of the theory could account for the time series patterns of the early 1970s. Wilson (1987) argues that those who escaped inner cities improved their position relative to whites, whereas those left behind became even more marginalized. Thus, the bifurcation of Black name choices reflects this bifurcation, with those trapped in the inner cities adopting Blacker names because their life prospects have worsened and those leaving the inner city doing the opposite. That conjecture, however, is inconsistent with the patterns of who actually chooses Black names. In the 1970’s, variables that are associated with inner city residency (e.g. being a single mother) are not strong predictors of black names. The 1970’s shift to distinctive black names was a very broad (in terms of the set of blacks participating) change in naming patterns by a diverse set of parents. The “ghettoization” of black names did not begin until the 1980’s, when distinctive black names became synonymous with low income, single motherhood, and low birth-weight babies. 16 For a formal treatment of this model, see the earlier version of this paper Fryer and Levitt (2002).
20
one has an affinity for the Black community.17 Thus, all else being equal, both types prefer to be
perceived as being a “black” type.
One way in which an individual signals his or her type is by the names that they choose
for their children. In the model, each parent has one child and bestows a name on that child.
The (suitably anthropomorphized) community observes the name each parent gives his or her
child, and based on that signal, draws unbiased inferences about the parent’s type. As in the price
theory model, we assume that White names provide labor market benefits. The total cost and
marginal cost of giving a child a Blacker name is lower for a parent with the “black” type.
Either separating or pooling equilibria may arise in such models. In any separating
equilibrium, parents whose type is “black” are willing to give otherwise costly names to their
children simply because it allows them to distinguish themselves from parents who identify as
“white,” and derive more utility from social interactions. As a result, peers come to regard Black
names as a signal of community loyalty. In any pooling equilibrium, the payoff to social
interactions is determined by the underlying population distribution of types. Pooling is likely to
occur, in our model, when the marginal rates of substitution for both types of parents are similar.
The predictions of the signaling model fit some of the basic patterns observed in the data.
Black name choices in the 1960s, for instance, look roughly consistent with a pooling
equilibrium in which one’s name is not a strong signal. Any mechanism that amplified the
differences in the cost of signaling between the two types could cause a bifurcation in the
distribution of Black names such as occurred in the early 1970s, with “black” types moving to
distinctively Black names and “white” types shifting towards Whiter names (if they previously
17 For example, those with an affinity to the Black community may be more likely to contribute to local public goods than those with “white” identities, inducing better treatment from their neighbors. For instance, Anderson (1999) talks about the importance of having others “watch your back.”
21
had been in a pooling equilibrium) or continuing to choose traditional white names (if they
formerly were in a separating equilibrium). Such a change could have come as a result of the
increased opportunities for integration due to the civil rights movement (Fryer 2003). Those
Blacks with the most to gain from integration would opt for Whiter names. Somewhat counter-
intuitively, for those Blacks who continue to signal a “black” type, to sustain the new
equilibrium requires an even costlier (i.e. Blacker) name choice since the outside option is now
more attractive.
The signaling model appears to fall short in explaining three dimensions of the data.
First, in the 1980s and 1990s, individuals who are most likely to adopt distinctively Black names
(young, unmarried women in predominantly Black neighborhoods) are those with the least need
to signal affinity with the Black community or their commitment to remaining in the
neighborhood. Second, the signaling model has a hard time explaining why Black names
became more prevalent among many different types of Blacks, including those who live in
mostly White communities. Third, in order for the Black names signal to be credible, it must
impose costs on those who carry the names. In Section IV of this paper, however, we are unable
to detect evidence that Black names are costly. Finally, the signaling model provides little
rationale as to why Blacks might engage much more extensively in signaling than do either
Asians or Hispanics.
An Identity Model
The primary shortcoming of the signaling model in explaining the data is that, for a large
segment of the Black community, distinctively Black names appear to be viewed as a benefit
22
rather than a cost. In the signaling model, an investment must be costly to provide a credible
signal. Using a similar framework, but allowing Black names to be a benefit for those with the
“black” type and a cost for those with the “white” type converts the signaling model into a
simple identity model.
In particular, we have in mind the framework of Akerlof and Kranton (2000). In their
language, identities are accompanied by certain “prescriptions” that define appropriate behaviors
for a person of that type. When an individual takes actions in line with these prescriptions (e.g. a
“black” type choosing a distinctively Black name, or a “white” type choosing a White name),
there is a utility benefit.
To justify the patterns in the time series circa 1970 through this model, there must be a
shock to the identity prescriptions (i.e. what it means to be Black) around this time period. The
rise of Black Power appears to be precisely that sort of shock. The underlying philosophy of the
Black Power movement was to encourage Blacks to accentuate and affirm black culture and fight
the claims of black inferiority (Van Deburg 1992). Within the Black community, there were
widespread changes in hair styles and the rising popularity of afro-centric clothing between
1968-1975 (Van Deburg 1992, Woodward 1974). The adoption of distinctively Black names
would be completely consistent with these other cultural phenomena. The identity model may
also help to explain why naming patterns among Blacks are quite distinctive from Whites, but
Asians name their children in much the same manner as Whites. For instance, if Asian
“prescriptions” stress financial success and assimilation, Asian names would be expected to
mirror those of Whites.
Another fact consistent with a Black Power explanation is that the concentration of
23
Blacks by county is a much stronger predictor of Black names in the 1968-1977 period than the
rest of the sample. In column 2 of Table 2, the coefficient on that variable is .152 (standard error
of .017) for 1968-1977, but never larger than .061 in any other period. The county-level trends
in Figure 7 also exhibit a basic consistency with a Black Power story. Alameda (in which
Oakland is located), Los Angeles, and San Francisco counties were the centers of the Black
Power movement. Black names increased earlier and to a greater extent in these three counties
than the rest of the state.
In summary, although the evidence in favor of the identity model is far from
overwhelming, it is the only theory examined which does not yield predictions that are directly at
odds with the observed patterns in the data.
IV. Do names have a causal impact on life outcomes?
In light of audit studies documenting the use of names as a screening device by
employers, one might expect that having a distinctively Black name should be causally linked to
worse economic outcomes, holding all else equal. In this section, we link information from a
woman’s own birth certificate to her adult circumstances as reflected in the information on the
birth certificates of her children at the time she gives birth.
In order to make this linkage, a woman must be born in California and later give birth in
California. We focus our analysis on Black women born in 1973 and 1974 -- the earliest years
for which we have the necessary information to make reliable links. Of the estimated 46.4
percent of women still residing in California who have given birth, we appear to match at a high
24
rate (over 90 percent). 18 The subset of women that we successfully link is not representative of
all women born in 1973-1974. In particular, the women we eventually link are themselves born
to slightly younger, unmarried mothers and are less likely to be biracial. The differences
between the whole sample and the linked subset are consistent with higher fertility rates and
lower rates of cross-state mobility among these groups. Importantly, however, after controlling
for socio-economic characteristics, there is no systematic difference between the first names of
the women who are or are not successfully linked. If we regress an indicator variable for a
successful link (equal to one if a link occurs and zero otherwise) on the woman’s BNI and
background characteristics at the time of the woman’s birth, the coefficient on the woman’s BNI
is both substantively and statistically insignificant.
Our approach to testing for a causal impact of names involves predicting adult life
outcomes as a function of everything known about a woman and her parents at the time of her
own birth, including her name:
i birth ih
birth ii
adult i XBNIy ελβ ++Γ+= (3)
where i indexes women, h represents hospitals, and the superscripts adult and birth correspond to
the time at which the variable is measured. We analyze a wide selection of outcome variables as
the dependent variable. All of the covariates included in the earlier analysis of the cohorts born
in the 1970s are also in this specification. We limit the sample to the last birth that we observe
for a particular woman in order to most closely approximate the long-run outcomes of the
women, although the results are not sensitive to this restriction. The clear weakness of this 18 According to the 1990 Census, roughly 20 percent of Black women born in California live outside the state during their childbearing years. Also based on 1990 Census data, 42 percent of Black women born in California have no children by age 27. If leaving California is independent of the decision to have a child by age 27, then 46.4 percent of Black women born in California should give birth in California in our sample. (1-.2)*.58. For details of how this linkage is performed, see the data appendix.
25
empirical approach is that if unobserved characteristics of the woman are similarly correlated
both with life outcomes and her name, our approach is likely to exaggerate the true causal link
between a woman’s name and her life outcomes.19
The results of estimating equation (3) for a number of different outcomes are presented in
the top two rows of Table 3. The top row does not include any controls; the second row includes
the full set of controls. In both cases, only the coefficient on the woman’s own Black Name
Index is presented in the table. In the absence of controls for background characteristics, Blacker
names are uniformly associated with worse adult outcomes. Given the correlation between
Blacker names, growing up in segregated neighborhoods, and more difficult home environments,
this relationship is expected. What is surprising, especially in light of the biases discussed
above, is how limited the impact of a woman’s name is on her life outcomes once we control for
other factors that are present at the time of her birth. When we include covariates in the basic
specification (row 2), we find statistically significant coefficients for only four outcomes: percent
Black in the hospital the mother gives birth, whether the woman is unmarried at the time of the
birth, per capita income among Blacks living in her ZIP code, and how Black a name the woman
chooses for her own child. Even in these cases of statistical significance, the magnitude of the
coefficients associated with the BNI is substantively small. Changing the BNI from 50 to100
raises the percent Black in the hospital where the mother gives birth by less than one percentage
point, the probability of an unmarried birth two-tenths of one percentage point, and per capita
income for Blacks in the ZIP code by $100. The largest impact that a woman’s name appears to
19 One logical solution to this problem would be to use the name of the woman’s mother as an instrument. The mother’s name is a strong predictor of her daughter’s name, but one might plausibly argue that controlling for a wide range of covariates at the time of the daughter’s birth, the mother’s name will have no impact on her daughter’s adult outcomes. Unfortunately, the mother’s first name is not included in our data set until 1982, so this instrumental variables strategy will not be feasible until more years of time have passed.
26
have is on how she names her children. An increase of 50 in the BNI raises the BNI of her
children by about 3 – a little more than half the impact that being a single mother has on naming,
and the same impact as having the child in a hospital in which an extra 30 percent of the births
are to Blacks. None of the other variables considered yield statistically significant coefficients:
years of education of the woman or the father of her child, mother’s age at first birth, private
insurance coverage, her baby’s birth weight, and number of total children born to date.20
In the remaining rows of Table 3, we replace our continuous measure of BNI with a
series of mutually exclusive indicator variables corresponding to having a unique name, a BNI
greater than 80 but not unique, a BNI between 50 and 80, and the omitted category which is a
BNI less than 50. The rationale for analyzing the data in this form is that the imposition of a
linear relationship over the whole range of the Black name index may be inappropriate. For
instance, a move from 25 to 50 may be inconsequential to a woman’s life outcomes, but a move
from 75 to 100 may have important implications. There is little in the results to challenge the
findings discussed above. In almost all cases, in the raw data the presence of a Blacker name is
associated with worse outcomes. Once we control for other variables, however, the impact of
names tends to diminish or evaporate.21 Particularly in light of the biases likely to be pushing the
20 The absence of a causal impact of names on our outcome measures might lead one to suspect that our outcome measures are noisy or not capturing relevant phenomena. As a test of this hypothesis, we tried including a woman’s years of education as an explanatory variable in the regressions reported in Table 3. A woman’s education has a large impact with the expected sign on the relevant outcomes such as ZIP code income, suggesting that it is not simply bad outcome measures that account for our null findings with respect to names. 21 We have also explored whether other aspects of naming have a causal impact on life outcomes. For example, one might expect that having the “wrong” kind of name (i.e. a White name in a Black neighborhood and vice-versa) will, at least in terms of utility, adversely affect life outcomes. We attempted to identify people with the “wrong” names in two different ways. In the first approach, we calculate the deviation between the BNI that a person actually received and the BNI that we predict they should have received based on their birth circumstances. The absolute magnitude of the deviation represents how “wrong” their name is from an ex ante perspective. An alternative way of characterizing someone as having the “wrong” name is to look at cases in which a woman’s name is mismatched ex post with the actual neighborhood in which she lives as an adult. In neither case do we find a systematic link between having the “wrong” name and the types of outcomes we are measuring.
27
results towards finding a spurious negative relationship between names and outcomes, we
conclude that there is little evidence that how Black one’s name is causally impacts life
outcomes.
An important question is how our results can be reconciled with the audit-studies that
report lower interview rates for resumes with distinctively Black names (Jowell and Prescott-
Clarke 1970, Hubbick and Carter 1980, Brown and Gay 1985, Bertrand and Mullainathan 2002).
The first point to note is that it is unlikely that a Black name could have a large impact on one’s
labor market success at any other step in the process. Once an employer has met a candidate in
person, race is directly observable. A person’s manner of speaking, dress, interview responses,
and on-the-job performance no doubt provide far better signals of productivity than a name. A
second important point is that it is not that costly to change one’s name, either legally or with
respect to what name is put on a resume. If job applicants understand the costs of having
distinctively Black names, one would expect to frequently observe name changes of this kind,
and yet they appear to be rare.22 Third, even if some employers select interview candidates
based on Black names, there are other employers who do not. In an efficient market with
sufficiently many non-discriminatory employers, the presence of discrimination need not result
in worse outcomes for Blacks (Becker 1957). Fourth, in the face of a discriminatory employer, it
is actually beneficial for the Black worker to signal race with a name. Job interviews are time
consuming and potentially involve missing work for those searching while employed. If a
22 When comparing Black women born in California in 1973-74 who also later give birth in California, we are able to identify women with last names and dates of birth that uniquely match, but first names that do not match. Excluding obvious typographical errors, such cases represent a trivial percent of all women. Among this group, the apparent name changes are almost equally distributed between women reporting names that are more distinctively Black when they give birth as when they were themselves born, and vice-versa. The name a woman reports on her child’s birth certificate, however, may differ from what she lists on a resume.
28
discriminatory employer will not hire the Black applicant anyway, it is more efficient to make
one’s race apparent prior to the interview stage. Blacks with distinctive names may actually do
better than Blacks with traditional names by avoiding these needless interview costs. Finally,
empirically a relatively small fraction of the jobs actually obtained are through this formal
resume-based process. Granovetter (1974) reports that approximately 10% of respondents in a
survey report obtaining their job through job advertisements.
A different, but related question is why employers less frequently give interviews to
applicants with distinctively Black names. One possibility is animus towards Blacks involving
racial discrimination unrelated to productivity differences. An alternative interpretation is that
Black names, because of self-selection among Black parents, provide useful signals of human
capital to employers, even controlling for race itself and other information available on resumes.
Audit study data are equally consistent with either of these hypotheses.23
We cannot directly test between these two hypotheses for two reasons. First, our data set
lacks clean measures of productivity. We do, however, observe worker characteristics that might
be correlated with labor productivity (see, e.g. Heckman and Carneiro 2003), but are typically
not included on a resume (and are for the most part illegal for employers to inquire about):
whether a woman is a single mother, was born to a teenage mother, was raised in a single-parent
household, or the degree of racial segregation into which she was born. The second difficulty in
23 From a legal perspective, the use of names as a basis for hiring decisions is likely to be a violation of current law, regardless of whether there are underlying productivity differences. In the 1971 decision Griggs v. Duke Power, the Supreme Court ruled unanimously that policies that are neutral on their face, but have a disparate impact on Blacks are not allowed. This doctrine was codified in the Civil Rights Act of 1991 after the 1989 case Antonio v. Wards Cove Packing modified the Griggs ruling by stating that policies with disparate impacts were allowable as long as they served a legitimate business interest. Indeed, even the use of names to choose a subset of Blacks from an entirely Black hiring pool would probably be viewed as illegal under Connecticut v. Teal, in which it was found that even policies that have no disparate impact on a protected group, but harm some individuals within that group, violate the law.
29
testing this hypothesis is that we do not have in our data all of the information on a resume (e.g.,
particular schools attended, complete work history, misspellings).
Our empirical strategy, in light of these shortcomings, is to estimate a number of
specifications in the general spirit of equation (2), varying the set of covariates included in the
regression. A critical difference between the earlier estimates and those presented below are that
we now include information about the adult circumstances of a woman as covariates, since such
information is on the resume. The results are presented in Table 4. Columns of the table
correspond to different factors that might be correlated with labor productivity due to their
influence on human capital: segregation in one’s birth hospital, mother’s marital status, and
whether one’s mother was a teenager. The entries in the table are the coefficient on the woman’s
Black name index. Each table entry is from a different regression. The two columns of the table
correspond to specifications with and without controls. The second column includes the
woman’s age, years of education, whether or not she has private health insurance (the best
measure we have with respect to the quality of her current employment), zip code of residence
fixed effects, and year dummies. These variables capture some of the information on a resume.
In all cases, our sample is the set of women who we observe both at their birth in 1973-74 and as
adults in 1989-2000.
The results in Table 4 suggest that a woman’s first name is indeed a useful predictor of
the circumstances in which she grow up, which may in turn be correlated with labor productivity.
Comparing columns (1) and (2), the types of information available to employers on a resume
does little to reduce the value of this signal. In column (2), a woman with a BNI equal to one
(implying a name that no Whites have) is 10 percentage points more likely to have been born to a
30
teenage mother and 9 percentage points more likely to have been born out-of-wedlock than a
Black woman living in the same zip code with the same age and education, but carrying a name
that is equally common among Whites and Blacks. The woman with a Black name is also more
likely to have been born in a Black neighborhood and to herself be unmarried. Thus, while we
cannot rule out animus on the part of employers, we find evidence supporting a potential
productivity-related statistical discrimination motive for employers to base interview decisions
on first names.
V. Conclusion.
We document stark differences in naming patterns between Blacks and Whites, and
demonstrate that these patterns change sharply over time. While most Blacks have shifted
towards more distinctively Black names (particularly in the late 1960s and early 1970s), the
fraction of Blacks choosing starkly White names has also increased. How distinctively Black
one’s name was appears to have signaled little in the early 1960s. Naming conventions differed
modestly across parental characteristics or neighborhood types. The last two decades, however,
have led to a “ghettoization” of distinctively Black names, namely, a distinctively Black name is
now a much stronger predictor of socio-economic status. Among the theories we consider,
models in which the rise of the Black Power movement triggers important changes in Black
identity appear to be most consistent with our data. In contrast with prior audit studies of Black
names on resumes, we find little evidence that names have a causal impact on adult life
outcomes.
More generally, this paper takes first steps towards an attempt to understand what role
31
Black culture might play in explaining continued poverty and racial isolation. With respect to
this particular aspect of distinctive Black culture, we conclude that carrying a black name is
primarily a consequence rather than a cause of poverty and segregation
32
Figu re 1: 1989 - 2000 BNI distribu tion for Black vs Wh ite
0
5
10
15
20
25
30
35
40
0 20 40 60 80 100
BNI (Black Nam e Inde x )
%
W hite Black
33
Figu re 2: 1989 - 2000 ANI distribu tion for Asian vs Wh ite
0
5
10
15
20
25
30
35
40
0 20 40 60 80 100
ANI (Asian Name Index)
%
W hite Asian
34
Figu re 3: 1989 - 2000 HNI distribu tion for Hispan ic vs Wh ite
0
5
10
15
20
25
30
35
40
0 20 40 60 80 100
HNI (Hispanic Nam e Inde x )
%
W hite Hispanic
35
Figure 4a: Distribution of Male Babies by How Many Share a Name (among children of all races born in California in a year)
0
5
10
15
20
25
30
35
40
45
50
55
1 2-5 6-50 51-500 500+
Number of male babies born in California in the same year with the same name as this baby
P er
ce nt
o f
ba bi
es , b
y ra
ce White Male Black Male
Figure 4b: Distribution of Female Babies by How Many Share a Name (among children of all races born in California in a year)
0
5
10
15
20
25
30
35
40
45
50
55
1 2-5 6-50 51-500 500+
Number of female babies born in California in the same year with the same name as this baby
P er
ce nt
o f
ba bi
es , b
y ra
ce
White Female Black Female
36
Figu re 5: Black Naming Pattern s, 1961 - 2000 by Quartile of Black Name In dex
0
10
20
30
40
50
60
70
80
90
100
1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000
Year
B la
ck N
am e
In de
x (
0= al
l W hi
te , 1
00 =a
ll B
la ck
)
Bottom Quartile Third Quartile Second Quartile Top Quartile
37
Blacks in Whitest
Blacks in Blackest
Bottom Quartile
Third Quartile
Second Quartile Top Quartile
Figure 6: Changes in Black Naming Patterns by Racial Composition of Neighborhood and by Quartile of Black Name Index
0 5
10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
100
1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 Y e ar
B la
ck N
am e
In de
x (0
=a ll
W hi
te , 1
00 =a
ll B
la ck
)
38
Figure 7: Trend in Black Names by County, 1961 - 2000 50th Percentile of Black Name Index
40
50
60
70
80
90
100
1961 1966 1971 1976 1981 1986 1991 1996 Year
B la
ck N
am e
In de
x (0
=a ll
W hi
te , 1
00 =a
ll B
la ck
)
Alemeda Fresno Los Angeles Orange Sacramento San Diego San Francisco Solano
39
Figure 8a: Percentage of Unique Girl Names by Racial Mix of the Hospital
0
5
10
15
20
25
30
35
1961 1966 1971 1976 1981 1986 1991 1996 Year
%
Black in Black Hospital
Black in White Hospital
White
Figure 8b: Percentage of Unique Boy Names by Racial Mix of the Hospital
0
5
10
15
20
25
30
35
1961 1966 1971 1976 1981 1986 1991 1996 Year
%
Black in Black Hospital
Black in White Hospital
White
40
Figure 9: Percent of Black Babies Given Unique Names and Non-Unique Distinctively Black Names, 1961-2000
0
5
10
15
20
25
30
35
40
45
50
55
60
1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 Year
% o
f a ll
B la
ck b
ab ie
s Unique
BNI>80
41
Table 1: Summary Statistics Years 1961-1967 1968-1977 1978-1988 1989-2000
Variables
Female Percentage 49.4% 49.2% 49.1% 49.2%
24.43 23.26 24.45 25.71 Age of mother at time of birth (6.12) (5.55) (5.47) (6.16)
28.47 26.86 27.75 28.73 Age of father at time of birth (7.49) (7.20) (7.28) (7.61)
Mother born in California 0.14 0.39 0.51 0.62 (0.35) (0.49) (0.50) (0.48)
0.31 0.46 0.56 0.58 Mother unmarried at time of birth (0.46) (0.50) (0.50) (0.49)
3098.76 3127.67 3181.99 3223.71 Birth weight (in grams) (581.17) (598.15) (587.47) (591.03)
3.20 2.29 2.11 2.29 Total number of children (2.36) (1.68) (1.34) (1.50)
5.87 6.77 7.08 7.28 Months of prenatal care (1.95) (1.68) (1.59) (1.59)
0.23 0.29 0.20 0.18 Teen mother at time of birth (0.42) (0.45) (0.40) (0.38)
.42 0.16 0.11 0..09 County Hospital (0.49) (0.37) (0.32) (0.28)
- 0.10 0.13 0.19 White mother - (0.30) (0.33) (0.40)
- 0.02 0.02 0.04 White father - (0.14) (0.15) (0.18)
- - - 12.59 Mother’s years of education - - - (2.00)
- - - 12.61 Father’s years of education - - - (2.55)
- - - 0.43 Privately insured - - - (0.49)
Per capita income in ZIP code - - 12073.68 13171.02 (All residents in 1989, 1989 dollars) - - (5296.38) (5276.93)
Per capita income in ZIP code - - 11032.02 11582.49 (Blacks in 1989, 1989 dollars) - - (3850.33) (4189.81)
- - 33.36 37.85 Percent of Black population in ZIP code - - (37.78) (34.53)
12.85 17.78 23.86 27.72 Percent of black babies in birth county (4.19) (6.82) (8.98) (10.10)
37.97 32.29 40.37 41.87 Percent of black babies in birth hospital (28.03) (24.15) (27.50) (27.58)
60.93 66.90 68.45 71.00 Black Name Index (27.79) (31.40) (31.42) (31.62)
Number of observations 165,226 259,231 409,069 500,932
Notes: All data are drawn from California birth certificate records, except for the information on ZIP codes which combines birth certificate information with data from the 1990 Census. The sample in columns 1 - 4 represent all Blacks born in California between 1961-1967, 1967-1977, 1978-1988, and 1989-2000, respectively. See the data appendix for further details of the construction of the samples and variables.
42
Table 2: Determinants of Name Choices Among Blacks Dependent Variable: Black Name Index Given to Child at Birth
Variable\Years 1961-1967 1968-1977 1978-1988 1989-2000 1989-2000
Full Covs. 1989-2000
Mother FE’s Female 1.31 3.22 4.73 3.14 3.28 4.16
(0.14) (0.14) (0.11) (0.10) (0.10) (0.07)
Percent of Black babies in birth hospital 0.045 0.059 0.079 0.090 0.027 0.017
(0.003) (0.005) (0.004) (0.004) (0.004) (0.002)
Percent of Black babies in County 0.064 0.151 0.042 0.015 0.012 0.003
(0.020) (0.017) (0.010) (0.010) (0.010) (0.008)
-2.17 0.45 -0.12 0.33 0.01 - Mother Born in California
(0.35) (0.16) (0.11) (0.11) (0.10) -
-0.35 -0.52 -0.61 -0.73 -0.50 0.18 Mother’s age at time of birth
(0.02) (0.02) (0.01) (0.01) (0.01) (0.03)
0.00 -0.17 -0.13 -0.18 -0.15 -0.06 Father’s age at time of birth
(0.02) (0.01) (0.01) (0.01) (0.01) (0.01)
-0.07 0.28 -0.03 -0.17 -0.04 -0.01 Months of Prenatal Care
(0.04) (0.04) (0.03) (0.03) (0.03) (0.02)
County Hospital 1.13 -0.11 0.65 -0.03 -0.97 -0.59
(0.20) (0.29) (0.19) (0.30) (0.29) (0.17)
-0.43 -1.08 -1.88 -1.71 -0.54 -0.01 Birth weight in kilograms
(0.12) (0.11) (0.09) (0.08) (0.08) (0.08)
Total Number of Children 0.21 0.70 1.28 1.80 0.82 -0.05
(0.04) (0.05) (0.05) (0.04) (0.04) (0.05)
2.05 4.15 5.15 4.62 2.88 0.21 Single Mother
(0.28) (0.15) (0.13) (0.12) (0.11) (0.10)
Percentage of Black Babies in Zip Code - - - - -0.0036 -0.0094
- - - - (0.0024) (0.0021)
Per Capita Income (All residents, 1989) - - - - -0.000058 -0.000009
- - - -
(0.000016) (0.000016)
Per Capita Income (Black residents, 1989) - - - - -0.000228 -0.000070
- - - - (0.000019) (0.000019)
Mother’s years of education - - - - -0.822 0.042
- - - - (0.030) (0.034)
Father’s years of education - - - - -0.274 -0.039
- - - - (0.022)
(0.020)
Privately Insured - - - - -2.94 -0.54
- - - - (0.15) (0.11)
White Mother - - - - -12.34 -
- - - - (0.13) -
White Father - - - - -18.06 -6.57
- - - - (0.26) (0.38)
Constant 66.59 75.86 81.03 88.04 103.96 3.74
(0.78) (0.60) (0.51) (0.51) (0.63) -
R-squared 0.0116 0.0346 0.0437 0.0564 0.0919 0.7733
43
Number of Observations 165226 257622 409069 500932 500932 500932
Notes: The dependent variable in all columns is the Black Name Index of the name given to the child at birth. The Black Name Index ranges from zero for names that are only given to Whites to 100 for names that are only given to Blacks. All variables included on the right-hand side of the regression are known at the time of a baby’s birth. Included in the regression, but not reported in the table are indicator variables for missing values. Standard errors are clustered to take into account correlation across people born in the same hospital in the same year.
44
Table 3: The Relationship Between Names and Adult Outcomes
Notes: The dependent variable is listed at the top of each column. The values reported in the table are the coefficient on the measure of how Black an individual’s name is. The bold horizontal lines denote coefficients from different specifications. In the top two rows, the key regressor is the Black Name Index. In the bottom four rows, three indicator variables are included denoting whether a name is unique, greater than 80 on the Black Name Index, or between 50 and 80. The omitted category is less than 50 in those specifications. Results are presented with no controls and including the full set of covariates used in Table 2. The sample is the set of Black women for whom we are able to successfully match their own birth certificate information to that of their children when the women themselves give birth. Standard errors, in parentheses, are clustered to take into account multiple births to the same woman. All of the covariates in the regressions are known as of 1973-74 when the women are born. All of the outcome variables correspond to the circumstances when they give birth as adults in the period 1989-2000.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Percent Black in
Residential ZIP code as
an adult
Percent Black in hospital
where mother gives birth as
an adult
Percent minority in
hospital where mother gives
birth as an adult
Years of education
(the woman herself)
Years of education (father of woman’s
child)
Woman’s age at
first birth
Baby’s birth
weight in grams
Unmarried when baby
born
Privately insured at
time of birth
Total children born so
far
Per capita ZIP code income
(all residents)
Per capita ZIP code income (Black
residents)
Black Name
Index of name
chosen for her child
43.29 47.43 57.01 12.37 12.22 19.58 3219.42 0.71 0.37 2.01 12587.01 11170.08 77.84 Mean Standard
Dev. (35.04) (27.22) (23.33) (1.59) (2.25) (2.97) (560.36) (0.45) (0.48) (1.16) (4697.46) (3551.65) (29.23)
Coefficient on: Other controls?
0.0602 0.0543 0.0458 -0.00057 -0.00055 -0.00203 -0.27 0.00072 -0.00039 0.00087 -7.14 -4.96 0.0810 Black Name Index (linear)
No
(0.0099) (0.0080) (0.0101) (0.00049) (0.00072) (0.00081) (0.16) (0.00014) (0.00014) (0.00032) (1.39) (1.08) (0.0087)
0.0155 0.0172 0.0201 0.00079 -0.00004 0.00087 -0.04 0.00035 -0.00008 0.00045 -2.21 -1.94 0.0620 Black Name Index (linear)
Yes
(0.0096) (0.0072) (0.0092) (0.00046) (0.00072) (0.00073) (0.17) (0.00013) (0.00014) (0.00030) (1.06) (1.39) (0.0092)
Unique Name No 3.65 3.09 3.10 -0.027 -0.043 -0.166 -20.52 0.054 -0.036 0.080 -457.09 -391.76 6.65
(0.90) (0.68) (0.85) (0.042) (0.060) (0.072) (14.20) (0.012) (0.012) (0.028) (124.95) (91.87) (0.77)
5.39 4.64 3.25 0.016 -0.043 -0.140 -19.80 0.040 -0.031 0.057 -545.11 -307.71 5.14 Black Name Index>80
No
(0.85) (0.70) (0.84) (0.044) (0.060) (0.077) (14.48) (0.012) (0.012) (0.029) (116.25) (91.11) (0.77)
2.22 2.04 1.39 0.012 -0.054 -0.030 0.36 0.012 -0.040 0.064 -164.80 -175.72 1.70 50>Black Name Index>80
No
(1.02) (0.81) (0.98) (0.047) (0.073) (0.096) (16.75) (0.014) (0.015) (0.034) (135.96) (101.88) (0.87)
Unique Name Yes 0.45 0.34 0.98 0.074 -0.008 0.033 -6.13 0.028 -0.013 0.043 -77.90 -191.39 5.30
(0.86) (0.63) (0.78) (0.039) (0.061) (0.064) (14.56) (0.012) (0.012) (0.026) (124.56) (90.84) (0.80)
2.25 1.99 1.73 0.120 0.000 0.078 0.10 0.011 -0.008 0.022 -169.94 -96.50 3.64 Black Name Index>80
Yes
(0.83) (0.63) (0.78) (0.040) (0.060) (0.070) (14.88) (0.012) (0.013) (0.028) (115.55) (89.59) (0.79)
0.66 0.67 0.61 0.038 -0.046 0.015 10.68 0.000 -0.031 0.038 -18.91 -105.17 1.07 50>Black Name Index>80
Yes
(0.95) (0.73) (0.89) (0.042) (0.070) (0.084) (16.66) (0.013) (0.014) (0.031) (133.43) (99.04) (0.86)
45
Table 4: The Signaling Value of Names to Employers (1) (2) Coefficient on Black Name Index when the following set of controls are included in the regression: Dependent variable No controls included Controls similar to information on a resume
0.0710 0.0544 Percent of black babies in birth hospital (0.0059) (0.0082) 0.00247 0.00243 Woman born to a teenage mother
(0.00010) (0.00015) 0.00201 0.00182 Woman born to an unmarried mother
(0.00010) (0.00016) 0.00043 0.00025 Woman herself unmarried at time she gives birth
(0.00010) (0.00009) Notes: The relevant dependent variable is listed on the left-hand side of the table. The values reported in the table are the coefficients on the Black Name Index of the woman. Each table entry is from a separate regression. In column 1, no controls are included. In column 2, controls for a woman's age, years of education, whether she is privately insured, ZIP code of residence, and year dummies are included. These variables proxy for some of the types of information available to employers on resumes. The sample used in all regressions is the set of women for whom we successfully link information from their own birth certificate to their child’s birth certificate. Unlike earlier regressions, these specifications control for choices that a woman makes over the course of her life, such as years of education. The number of observations is 11,257 in all cases.
46
Data Appendix In this data appendix, we describe the data sources that we draw upon in our analysis, give a description of some data limitations and other data issues we confronted, provide useful definitions and describe the process used to link mothers. A. DATA SOURCES Birth Data: Our empirical analyses are mainly based on data drawn from the Birth Statistical Master File maintained by the Office of Vital Records in the California Department of Health Services. These files provide information drawn from birth certificates for virtually all children born in California over the period 1961-2000. With the approval of the California Committee for the Protection of Human Subjects, we have been able to supplement the information contained on the public use versions of these data sets with personal identifiers including mother’s first name, mother’s maiden name, and child’s full name. Zip Code Statistics: U.S. Bureau of the Census, American Fact Finder, 1990 Census Demographic and Socioeconomic Data: For an accuracy check on mother-children links, data were obtained from the American Community Survey 5 Percent Public Use of Microdata Samples (PUMS) for California and United States(All States), U.S. Bureau of the Census Arabic First Names: To identify and discard Arabic names from our data, we used Bruce Lansky (1999); Baby Names Around the World. B. DATA LIMITATIONS AND OTHER DATA ISSUES
Race/Ethnicity: From 1998 to 2000, information of a child’s race is missing. For these three years, a child is considered Black if either parent is Black. In other years, over 98 percent of children with at least one Black parent are coded as Black. Prior to 1982, most Hispanics are included with whites and it is not possible to directly calculate statistics for Non-Hispanics. Beginning in 1982, an identifier for Hispanic ethnicity was added to the birth certificate in California. Using information from this latter period, we drop anyone in the earlier time period with a surname or mother’s maiden name that is more than 10 percent Hispanic. Link of birth certificate data for women born in 1973-1974 to when they give birth themselves in the period 1989-2000: In order to uniquely link information from a woman’s own birth certificate to those of her children when she gives birth, a woman must both be born in California and later give birth there. Only after 1989 do we have the mother’s date of birth and full first name, two critical pieces of information for making the linkage. We thus limit the sample to babies born 1989-
47
2000 to Black women themselves born in California. In order to identify women whose birth histories are likely to be most fully captured by the period 1989-2000, we restrict the matching to women who were themselves born in 1973-1974. We restrict the sample to Black women born in California. We then carry out five different matching procedures in which we use a mother’s exact date of birth and permutations of her first name and maiden name (match only on maiden name, only on first name, on first name and maiden name, on first initial and maiden name, and on first name and maiden name initial). We keep only cases where the match between mother and child is unique, i.e. there is no other woman born in California in 1973-74 who has information that could possibly match the characteristics of the mother on the birth certificate. Because of typographical errors, using only a subset of the information in a woman’s name somewhat increases the number of successful matches.
48
References
Akerlof, George., and Kranton, Rachel. 2000. Economics and Identity. Quarterly Journal of
Economics, 115 (3), 715 - 753 Anderson, Elijah. 1990. Streetwise: Race, Class, and Change in an Urban Community.
Chicago: University of Chicago Press. Anderson, Elijah. 1999. Code of the Street: Decency, Violence, and the Moral Life of the Inner
City. New York: W. W. Norton and Company, Inc. Austen-Smith, David., and Fryer, Roland. 2003. “The Economics of ‘Acting White’.”
Unpublished Manuscript. Harvard University. Becker, Gary. 1957. The Economics of Discrimination. Chicago: University of Chicago Press.
Berman, Eli. 2000. “Sect, Subsidy, and Sacrifice: An Economist's View of Ultra-Orthodox
Jews.” Quarterly Journal of Economics, 115(3), 905-953
Bertrand, Marianne, and Mullainathan, Sendhil. 2002. “Are Emily and Brendan More Employable Than Latoya and Tyrone? Field Experiment Evidence on Racial Discrimination in the Labor Market.” Working paper. MIT
Bound, John., and Freeman, Richard. 1992. “What Went Wrong? The Erosion of Relative
Earnings and Employment Among Black Men in the 1980's.” Quarterly Journal of Economics Vol. 107 (1), 201-232.
Brown, C. and Gay, P. 1985. Racial Discrimination 17 Years After the Act, London: Policy
Studies Institute. Busse, Thomas V., and Seraydarian, Louisa. 1977. “Desirability of First Names, Ethnicity and
Parental Education.” Psychological Reports Vol 40(3, Pt 1), 739-742. Chandra, Amitabh. 2003. “Is the Convergence of the Racial Wage Gap Illusory?” National
Bureau of Economic Research Working Paper #9476 Chandra, Amitabh., and Skinner, John. 2003. “Geography and Racial Health Disparities.” in
Ethnic Disparities in Health Care. National Academy of Sciences. Counihan, Carol., and Van Esterik, Penny. 1997. Food and Culture, A Reader, Routledge: New
York. Cutler, David., Glaeser, Edward, and Vigdor, Jake. 1999. “The Rise and Decline of the Amercan
49
Ghetto,” Journal of Political Economy, Vol. 107, no. 3: 455-506. Fordham, S. and Ogbu, J. 1986. “Black Students' School Successes: Coping with the Burden of
‘Acting White’,” The Urban Review 18 (3): 176-206. Fryer, Roland. G. 2003. “An Economic Approach to Cultural Capital.” Unpublished Manuscript.
University of Chicago. Fryer, Roland., and Levitt, Steven. 2002. The Causes and Consequences of Distinctively Black
Names. Unpublished manuscript. University of Chicago. Glaeser, Edward., Laibson, David., and Sacerdote, Bruce. 2002. “An Economic Approach to
Social Capital.” Economic Journal, 112, p. 437-458. Granovetter, Mark. 1974. Getting a Job: A Study of Contacts and Careers. Chicago: University
of Chicago Press. Hannerz, Ulf. 1969. Soulside: Inquiries into Ghetto Culture and Community. New York:
Columbia University Press. Heckman, James., Lyons, Tom., and Todd, Petra. 2000. “Understanding Black-White Wage
Differentials, 1960-1990.” American Economic Review. Vol. 90 (2), 344-49. Heckman, James., and Carneiro, Pedro. 2003. “Human Capital Policy.” National Bureau of
Economic Research Working Paper #9495. Hubbick, J. and Carter, S. 1980. Half a Chance? A Report on Job Discrimination Against Young
Black Males in Nottingham. London: Commission for Racial Equality. Iannaccone, Laurence. 1992. “Sacrifice and Stigma: Reducing Free-riding in Cults, Communes,
and Other Collectives.” Journal of Political Economy, Vol 100 (2), 271-291. Jackson, George. 1944. White and Negro spirituals, their life span and kinship, tracing 200 years
of untrammeled song making and singing among our country folk, with 116 songs as sung by both races. New York: J. J. Augustin
Jencks, Christopher., and Phillips, Meredith. 1998. The Black-White Test Score Gap. Washington
DC: The Brookings Institute. Jowell, R. and Prescott-Clarke, P. 1970. “Racial Discrimination and White-Collar Workers in
Britain.” Race, Vol 11, pp.397-417. Lazear, Edward. 1999. “Culture and Language.” Journal of Political Economy, 107 (6), S95-
S126.
50
Lewis, Oscar. 1966. La Vida: A Puerto Rican Family in the Culture of Poverty--San Juan and
New York. New York: Random House. Lieberson, Stanley. 2000. A Matter of Taste: How Names, Fashions, and Culture Change. New
Haven: CT Yale University Press. Lieberson, Stanley., and Mikelson, Kelly. 1995. “Distinctive African American Names: An
Experimental, Historical, and Linguistic Analysis of Innovation.” American Sociological Review, Vol 60 (6), 928-946.
Massey, Douglas. 2001. “Residential Segregation and Neighborhood Conditions in U.S.
Metropolitan Areas.” in America Becoming: Racial Trends and Their Consequences, Vol. 1, 391-434. Washington DC: The National Academies Press.
Orr, Eleanor. 1989. Twice as Less: Black English and the Performance of Black Students in
Mathematics and Science. New York, W. W. Norton. Reno, Janet. et. al. 1997. Correctional Populations of the United States. Washington: Bureau of
Justice Statistics. Riessman, Frank. 1962. The Culturally Deprived Child. New York: Harper & Bros. Smith, James., and Welch, Finis. 1989. “Black Economic Progress After Myrdal.” Journal of
Economic Literature, Vol. 27 (2), 519-564. Van Deburg, W.L. 1992. New Day in Babylon: The Black Power Movement and American
Culture, 1965-1975. Chicago: University of Chicago Press. Ventura, Stephanie., and Bachrach, Christine. 2000. Non-marital Childbearing in the United
States, 1940-1999. National Vital Statistics Reports, Vol. 48(16). Hyattsville, MD: National Center for Health Statistics.
Waldfogel, Joel. “Preference Externalities: An Empirical Study of Who Benefits Whom in
Differentiated Product Markets.” The Rand Journal of Economics, forthcoming. Wilson, William. 1987. The Truly Disadvantaged: The Inner City, the Underclass and Public
Policy. Chicago: University of Chicago Press. Wolfram, Walt., and Thomas, Erik. 2002. Development of African American English. Blackwell
Publishers. Woodward, Vann, C. 1974. The Strange Career of Jim Crow. New York: Oxford University Press.
51
Young, Andrew. 1996. An Easy Burden: The Civil Rights Movement and the Transformation of
America. HarperCollins Publishers.