Poverty among minorities in the U'S': Explaining the racial poverty gap for Blacks and Latinos

1
Poverty among minorities in the U.S. Explaining
the racial poverty gap for Blacks and Latinos

• Carlos Gradín
• Universidade de Vigo
• Visiting Scholar at Cornell University

2
Motivation

• large socioeconomic gap between Whites and other
  racial/ethnic groups in the US (Blacks and
  Latinos)
• income distribution
• higher risk of unemployment, low-paid
  occupations, lacking health care coverage,
• minorities continue to increase
• non Hispanic Blacks 13
• Hispanics 10 (1994), 15 (2007), 25 (2050)
• why are Blacks and Hispanics more likely to be
  poor (at least twice)?
• poor characteristics (characteristics effect)
• geography
• demographic factors
• number of children
• family type
• labor force performance (participation, hours,
  occupation)
• education
• different impact on poverty risk (coefficients
  effect)

3
Data and definitions

• Data Current Population Survey, 1994, 2002, and
  2007 Annual Social and Economic (ASEC) March
  Supplement (US Census Bureau)
• Races non-Hispanic Whites, non-Hispanic Blacks
  and Hispanics (of any race)
• Poverty US official poverty measure

4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
Figures 1a. Racial and ethnic distribution by
family income deciles Total family income divided
by the poverty threshold
8
Figures 1b. Racial and ethnic distribution by
family income deciles Total family income divided
by the poverty threshold
9
Poverty among minorities in the US

• Raw racial poverty gaps are large in 2006
  (despite decline during the 1990s decline)
• 15.9 percentage points (Blacks)
• 12.5 percentage points (Hispanics)

10
Figure 2. Poverty rates and racial poverty gaps
in the US
11
Blacks and Hispanics (compared with Whites)

• Are overrepresented in states with higher poverty
  rates
• south central east (BH)- west (H)
• but also in large MSA
• Their families are different
• more lone-mothers
• more other female-headed families (B)
• more children
• family heads younger, less-educated
• more foreign family heads (H)

12
(No Transcript)
13
(No Transcript)
14
Blacks and Hispanics(compared with Whites)

• Different labor market performance
• Higher employment rates (H males)
• similar employment rates (B females)
• lower employment rates (unskilled young B males,
  H females)
• Jobs
• segregation overrepresented in non managerial
  and professional jobs in the private sector,
• the number of hours worked by family heads is
  similar, but lower number of hours worked by
  other family members
• lower earnings

15
(No Transcript)
16
(No Transcript)
17
Methodology Estimation
Poverty estimation
likelihood of being poor (person i in group g)
household per capita income
poverty line
vector of household characteristics
vector of coefficients estimates
The head-count ratio of poverty in group g
population in group g
18
Methodology Aggregate decomposition
Gap in poverty rates
Aggregate decomposition (reference group 0)
Extending Oaxaca-Blinder to deal with non-linear
regressions
Blacks or Hispanics
Whites
Characteristics effect (explained)
Coefficients effect (unexplained)
19
Methodology Aggregate decomposition

• Poverty levels
• White-Black differences in Brazil Gradín (IZA
  DP, 2007)
• inter-group differences in India and Kosovo
  Gang, Sen and Yun (IZA DP, 2006) or Bhaumik, Gang
  and Yun (IZA DP,2006)
• inter-country differences Biewen and Jenkins
  (Emp. Ec, 2004) and Quintano and DAgostino (RIW,
  2006)
• Parametric model of income
• Other issues (non-linear)
• changes in the employment of married women
  Gomulka and Stern (Economica, 1989)
• inter-country differences in duration of
  unemployment Ham, Svejnar and Terrell (AER,
  1998)
• racial gap in the transition rate into
  self-employment and in computer ownership Farlie
  (JLE, 1999, JESM, 2005)
• gender gap in formal sector employment and child
  labor incidence Nielsen (Ec Let. 1998, Ap. Ec.
  Let. , 2000)
• employment success of immigrants Bevelander and
  Nielsen (JPE, 2000)
• attitudes towards foreigners in the EU Gang,
  Rivera-Batiz and Yun (IZA DP, 2002).

20
Methodology Detailed decomposition

• Linearized models Linear approximation
• Even and Macpherson (JHR, 1993), generalized by
  Yun (Ec. Let. 2004)
• transparent, simple to compute coefficients and
  characteristics sample means,
• no path-dependency and no assumption needed to
  match individuals (sequential models)
• the original Oaxaca-Blinder approach is a
  particular case

21
Methodology Yun (Economics Letters, 2004)
detailed decomposition
Linearization of
around
22
Methodology normalized regression

• Identification problem in detailed decompositions
  of the coefficients effect (Oaxaca and Ransom,
  1999)
• Normalized regressions invariant to the
  lef-out reference category in computing the
  contribution of dummies in the detailed
  coefficients effect (Gardeazábal and Ugidos, 2005
  and Yun, 2005)
• they do not alter neither the detailed
  characteristics effect nor the contribution of
  continuous variables to the coefficients effect.

Original regression
L continuous variables X M sets of categorical
variables D, the mth set has Km categories and
Km-1 dummy variables in the equation (ref. group
1st dummies in each set)
Normalized regression
where
23
LOGIT REGRESSION Explanatory variables Area of
residence geographical region, MSA size Type of
family (couple, single-male, single-female
with/without children) Family head sex, age,
education, labor status and occupation,
immigration Other family members dependents
(age), employed (age, education, occupation av.
number of hours worked, female), others receiving
income (education / female)
Benchmark person lived in large city (5
million) in the Middle Atlantic region, in a
married-couple family, where the head was a 15-24
years old male, born in the US with American
parents, with only primary school education,
working full-time in the private sector in a non
managerial or professional occupation, and did
not change residence in the previous year
24
(No Transcript)
25
Regression results, 2006

• Coefficients are similar in sign for all groups,
  although they differ in magnitude and statistical
  significance.
• Poverty risk increases with
• Living in the Pacific region and small MSA
• the lowest risk of being poor in New England
  (Blacks), East North Central (Latinos), and in
  Mountain (Whites)
• Families other than married couples were more
  likely to be poor in all three groups, (higher
  effect for Whites, lower for Blacks).
• Families with many dependents (especially adults)
  faced a higher risk of being poor.
• The older and more educated the family head, the
  lower the probability of being poor in all cases.
  
• While the effect associated with age was higher
  among Whites, the impact of College education was
  larger among both minorities.
• The effect associated with high school was
  similar in all three groups.
• The risk of falling into poverty increased for
  Blacks and Latinos when the head was non citizen,
  while this characteristic appears to be non
  significant in the case of Whites, for whom the
  poverty risk was however lower for second
  generation immigrant family heads.
• Those families who changed their residence during
  the previous year were also more likely to be
  poor in all cases.

26
Regression results, 2006

• Lower poverty the head of the family was
  unemployed or worked in a part-time job,
• while it decreased when he or she worked in
  managerial or professional occupations in the
  private sector.
• Working more weeks (especially Whites and Blacks)
  and weekly hours (Latinos) by the family head
• The presence of more employed adults in the
  family was generally associated with lower
  poverty.
• This latter effect increased with their attained
  education and was higher in more skilled
  occupations and lower in the case of females and
  self-employed.
• The presence of young workers without College
  studies had a significant effect on reducing
  poverty in the case of Latinos but not in the
  case of Blacks.
• More weeks worked by other family members also
  reduced poverty risk, while their average of
  weekly hours worked was only significant and
  negative in the case of Blacks.
• The effect of other nonlabor income receivers was
  large and significant in all groups.

27
(No Transcript)
28
Characteristics effect Blacks

• Raw racial poverty gap (15.9) largely explained
  by the characteristics effect 12.2 (76.6)
• the conditional racial poverty gap was 3.7 (23).
• differences in education attainment and in labor
  market related variables of all family members
  37
• the level of education and labor activity of the
  family head 13.5
• lower number of hours worked (6 percent) and
  education (5.5 )
• the education and labor activity of other family
  members 23.3.
• especially important the lower number of hours
  worked (15.3) and the occupation of those
  employed (8.4 ).
• non-labor incomes 4.3
• demographic factors 36
• dependent children (13.5), family type and sex
  of the family head (9.4 2.6), and the younger
  age of family heads (7).
• Geography as a whole appears to be irrelevant

29
Characteristics effect Hispanics

• Raw racial poverty gap (12.5) largely explained
  by the characteristics effect 9.5 (76.2)
• the conditional racial poverty gap was 3.0
  (23.8).
• the underlying reasons are substantially
  different
• sociodemographic characteristics more than ½
• larger number of dependent children (25)
• immigration profile (15)
• young age of the family head was similarly
  important compared with Blacks (7)
• but not the type of family and the sex of the
  head (less relevant, 3.6 and 0.1).
• On the other side, education and labor activity
  of family members explained about 20
• gap in the level of education of family heads
  (18 vs. 5.5 Blacks)
• the actual jobs in which they were employed (5),
  mainly in low-paid occupations.
• But, unlike Blacks, the larger number of hours
  worked (-3)
• other family members, not too much due to high
  employment rates of males,
• non-labor income (pensions) 7
• Geography also irrelevant.

30
Temporal trends

• Decline in the racial poverty gap during the
  economic boom of the 1990s
• due to larger decreases in the poverty rate among
  US minorities than among Whites.
• Decline was driven in both cases by the
  characteristics effect, with the conditional
  racial poverty gaps oscillating around 3-4
  (Blacks) and 2-3 (Latinos)
• 1993-2001
• Blacks (9 points) due to number of hours worked
  by family head (4), family type (1.1) and number
  of dependent children (0.9).
• increasing employment rate of single-mothers and
  the ongoing decline in the number of children
• Latinos (8 points) due to education attainment
  gap, employment and number of hours worked by
  head (jointly 3.6), mobility status (1.5), number
  of hours worked by other family members (1) and
  number of dependent children (1.5) and mobility
  (1.6).
• 2001-2006 the evolution of the racial poverty
  gap has been different
• Blacks the trend in the gap was slightly
  reversed, (1.4) due to the higher contribution
  of employment of non-head family members (0.9).
• Latinos continued to decrease due to the lower
  contribution of the number of dependent children
  (1.5).

31
(No Transcript)
32
Coefficients Effect

• Despite the fact that observed characteristics
  explained a large proportion of the racial
  poverty gap for both Blacks and Latinos, the
  impact of certain attributes on poverty risk
  among BH was different than among Whites.
• For example the hours worked by Hispanic family
  heads and other family members were less
  effective than those worked by Whites in
  protecting them from being poor, as both
  coefficient effects jointly explained about 15
  percent of the gap.
• Similarly, the family type and the age of the
  family head of both Latinos and Blacks are less
  effective in preventing poverty in these groups
  than in the case of Whites.
• In the case of Blacks, it appears also that the
  work of other family members was however much
  more effective than in the case of Whites.

33
Conclusions

• Racial poverty gaps are largely explained by
  differences in family characteristics
• but the main reasons diverge in both cases
• Blacks
• half the explained gap was attributed to their
  demographic characteristics, especially the large
  number of dependent children, the family type and
  the age of the family head
• the other half education and performance in the
  labor market, especially the low participation of
  family members other than the family head.
• Hispanics
• 2/3 of the explained gap (gt½ raw gap) attributed
  to demographic characteristics, with even more
  relevance of the number of children, and with a
  especial role played by their predominant
  immigration status
• the labor market related characteristics played a
  less fundamental role and was almost fully
  accounted by their larger educational gap
• region of residence played no role in explaining
  their higher poverty rates.

34
Conclusions

• Conditional poverty gap (Coefficients effect)
• 3.7 (Blacks) and 3.0 (Latinos)
• some attribute are less effective in preventing
  minorities from poverty
• hours worked by family heads and other family
  members (H)
• family type and the age of the family head (HB)
• The decline in the raw racial poverty gap during
  the 1990s
• due to the Characteristics effect (labor market)
• while the Coefficients effect remained constant