Explorations in Economic History 38, 93–122 (2001)
doi:10.1006/exeh.2000.0759, available online at http://www.idealibrary.com on
The Impact of the New Deal on Black and White Infant
Mortality in the South 1
Price V. Fishback
University of Arizona and NBER
Michael R. Haines
Colgate University and NBER
and
Shawn Kantor
University of Arizona and NBER
The New Deal led to unprecedented involvement by the federal government in the
provision of relief. Although New Deal officials argued that they sought to be nondiscriminatory in providing access to programs, various scholars have suggested that blacks
received fewer benefits than whites. One method for testing the effectiveness of the New
Deal at reaching households with low incomes is to examine its impact on infant mortality.
In this paper we use county-level data from three subperiods during the 1930s to examine
the relationship between several New Deal spending programs and black and white infant
mortality in the South. Some New Deal programs contributed to a reduction in infant
mortality, but other factors may have been strong contributors to a long-term secular
decline in infant mortality. Meanwhile, some of the New Deal programs appear to have
been nondiscriminatory or even more beneficial for blacks than for whites, while others
appear to have had differential effects that favored whites. © 2001 Academic Press
Since the New Deal the United States federal government, in cooperation with
state and local governments, has implemented a myriad of programs designed to
1
This paper was prepared for the conference “One Kind of Freedom Reconsidered: African
American Economic Life in the Segregation Era” at Lehigh University, September 17–19, 1999. We
thank Howard Bodenhorn, participants at the conference, and two anonymous referees for their
helpful comments. Special thanks to Kari Beardsley, Amanda Ebel, Michael Hunter, Angela Phillips,
and Jeffrey Taylor for their help in computerizing the data. Fishback and Kantor’s work on this paper
was supported by National Science Foundation Grant Number SBR-9708098. Any opinions stated
here are the authors’ and not the position of the NSF.
93
0014-4983/01 $35.00
Copyright © 2001 by Academic Press
All rights of reproduction in any form reserved.
94
FISHBACK, HAINES, AND KANTOR
improve the living standards of children in America. Yet some critics believe that
the United States has far to go in alleviating poverty among children and
improving their access to health care. For example, even though infant mortality
rates today are one-tenth their 1920s levels, American rates are significantly
higher than those in many other developed countries. Further, significant concerns have been raised about racial differences in the access to public programs,
in part because black infant mortality rates today remain substantially higher than
those of whites (14.7 deaths per thousand live births versus 6.1).
The federal government’s efforts to improve the welfare of black children in
the South dates back to the Reconstruction Era. Yet as Roger Ransom and
Richard Sutch argue in One Kind of Freedom, the Freedman’s Bureau had
modest success in providing relief and education for newly freed black children.
The federal government’s first broad-based effort to provide social welfare
occurred during the New Deal, first through emergency assistance and then
through the establishment of long-term programs such as Aid to Dependent
Children and Social Security. In this paper we examine the effectiveness of
various New Deal programs in improving the socioeconomic status of black and
white children during the 1930s. We focus specifically on infant mortality.
Franklin Roosevelt and the Congress responded to the Great Depression by
establishing a number of programs that potentially could have contributed to
lowering infant mortality rates. As part of the New Deal, the federal government
provided funds for direct relief and work relief through the Federal Emergency
Relief Administration (FERA), and work relief under the Civil Works Administration (CWA) and eventually the Works Progress Administration (WPA). The
Public Works Administration (PWA) was supposed to alleviate unemployment
problems by providing additional employment, while enhancing the economy by
building social capital, including sanitation projects in many cities. The Agricultural Adjustment Administration (AAA) was designed to help raise the
incomes of struggling farmers. Under the Social Security Act of 1935, the
long-term programs for old-age assistance, aid to dependent children, and aid to
the blind were established. Public housing was created to try to upgrade the
quality of housing for families in slums. The administrators of all of these
programs saw them as means to put resources into the hands of families in need,
precisely the group where infant mortality was most likely to be a problem.
There has been some empirical controversy whether the stated goals of the
New Deal—relief, recovery, and reform—were actually met. In studies of the
distribution of funds across the states, Wright (1974), Wallis (1998), and Couch
and Shughart (1998), for example, show that the distribution of New Deal funds
does not seem to fit a reform motive. Couch and Shughart (1998) question
whether the funds were distributed in ways that promoted relief or recovery,
claiming that the political agendas of the president and congressmen were more
important determinants (see also Anderson and Tollison (1991) and Wright
(1974)). John Wallis (1998), on the other hand, finds some evidence that New
Deal money was distributed to areas with relatively higher unemployment. All of
95
NEW DEAL AND INFANT MORTALITY
TABLE 1
Infant Mortality Rate by Race for the Birth Registration Area (1915–1932), the United States
(1933–1940), and Selected States (1915–1940)
Current birth registration area
(1915–1932) or United States
(1933–1940)
Year
Total
White
Nonwhite
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
99.9
101.0
93.8
100.9
86.6
85.8
75.6
76.2
77.1
70.8
71.7
73.3
64.6
68.7
67.6
64.6
61.6
57.6
58.1
60.1
55.7
57.1
54.4
51.0
48.0
47.0
98.6
99.0
90.5
97.4
83.0
82.1
72.5
73.2
73.5
66.8
68.3
70.0
60.6
64.0
63.2
60.1
57.4
52.8
52.0
53.6
51.9
52.9
50.3
47.1
44.3
43.2
181.2
184.9
150.7
161.2
130.5
131.7
108.5
110.0
117.4
112.9
110.8
111.8
100.1
106.2
102.2
99.9
93.1
92.1
101.4
105.8
83.2
87.6
83.2
79.1
74.2
73.8
Black
Virginia
South Carolina
Total
White
Nonwhite
97.8
102.9
91.0
83.6
78.7
76.8
84.0
77.6
80.8
83.7
75.5
75.9
78.8
77.3
76.3
67.2
68.5
72.6
69.6
73.9
69.7
66.2
59.7
58.5
80.4
85.8
77.9
71.7
68.1
65.4
70.5
65.9
67.5
71.9
62.4
64.1
66.9
65.1
63.9
57.7
59.2
62.2
58.8
62.6
59.2
57.0
50.3
50.5
136.5
141.5
119.6
109.8
102.5
102.1
114.7
104.4
110.9
110.8
106.4
104.5
107.3
107.2
107.8
90.1
90.0
98.4
96.1
101.8
95.2
89.6
83.9
79.5
Total
White
Nonwhite
113.1
115.8
96.1
92.9
96.3
101.6
75.9
83.4
69.5
66.7
69.5
77.1
149.3
147.9
122.6
119.0
125.1
127.0
96.5
91.0
88.7
81.0
77.2
78.2
83.0
79.3
80.8
75.6
80.3
66.5
68.2
77.7
72.1
69.0
58.9
61.5
60.9
67.3
61.6
62.1
62.7
64.4
54.1
50.8
115.2
109.6
108.1
102.0
91.9
94.6
98.3
95.8
98.9
87.9
95.9
78.8
86.1
Mississippi
Total
White
Nonwhite
68.4
68.0
68.1
71.3
68.5
70.0
66.8
73.8
72.1
67.7
55.9
53.6
63.6
64.8
53.9
58.2
58.9
56.7
56.6
54.4
52.8
56.0
53.1
54.9
52.9
58.7
55.3
61.1
58.4
51.4
43.9
44.4
50.9
54.8
47.8
50.3
50.0
50.4
50.4
46.6
84.6
79.2
82.3
88.0
83.3
80.8
77.9
86.1
84.9
83.0
67.1
61.8
74.7
73.6
59.0
64.7
66.0
61.7
61.8
60.6
184.3
162.5
134.3
135.6
110.7
111.5
119.9
114.1
112.0
112.1
99.9
105.9
101.5
99.5
92.7
84.1
85.4
91.0
81.9
86.1
82.0
77.9
73.2
72.9
Sources. U.S. data are from National Center for Health Statistics (1996). The individual state data
are from Linder and Grove (1947).
these discussions focus on the politics and economics of distributing federal
largesse. Surprisingly, very little research has attempted to address the important
question of how successful these programs were in reducing socioeconomic
problems.
One means of measuring the success of federal relief programs during the New
Deal is to examine their effect on infant mortality rates. Since children under the
age of one are probably the societal group with the fewest defenses against the
problems of disease, malnourishment, and maltreatment, they are probably
affected most by negative socioeconomic shocks to the household. Infant mortality rates have been shown by numerous studies to be strongly associated with
socioeconomic status broadly defined (see Antonovski and Bernstein (1977),
Clifford and Brannon (1978), Fordyce (1976/1977), and Waldman (1992)).
Because social insurance programs and relief programs are designed to help
people in the lower end of the income distribution, their impact on the infant
mortality rate is one important measure of their success. In fact, the social
96
FISHBACK, HAINES, AND KANTOR
programs of the New Deal might well have been extremely important. During the
worst depression in American history, the infant mortality rate continued to
decline through 1932, rose slightly in 1933 and 1934, and then declined sharply
through the rest of the 1930s (see Table 1). Such a drop is surprising since we
might expect that lower incomes would contribute to greater infant mortality,
particularly if lower income influenced both prenatal care and nutrition and the
degree of crowding in households.
Both black and white infant mortality rates followed similar patterns during
the 1930s, although the black rate remained substantially higher and had larger
fluctuations. Part of the difference probably stems from differences in household
incomes, but there might also have been differential access to the benefits from
the government programs. Federal New Deal administrators claimed that they
sought to treat the races equally when they distributed funds. Many New Deal
programs, however, were administered in partnerships with state and local
officials, and there was plenty of room for local administrative discretion in the
distribution of funds. Since state and local governments in the South had
implemented policies that had significantly retarded black progress for decades
(see, for example, Kousser (1974), Higgs (1977), Ransom and Sutch (1977), and
Margo (1990)), there may well have been racial differences in the access to and
subsequent effects of the New Deal. 2
In this paper we describe the long-term infant mortality decline for both blacks
and whites during the early 1900s, its continuation through the Great Depression,
and the nature of New Deal programs. Using a sample of over 700 counties from
the South with separate information for blacks and whites, we then examine the
effect of New Deal spending on black and white infant mortality.
MORTALITY TRENDS
One of the most significant trends in America since 1850 has been the great
mortality transition (Easterlin, 1996, Chap. 6). Even more dramatic has been the
decline in infant mortality for both white and black populations. In 1850, the
expectation of life at birth (e(0)) in the United States was 38.3 years (39.5 for the
white population), and the infant mortality rate (IMR: deaths before age 1 per
1,000 live births) was 229 (217 for whites) (Haines, 1998a, Table 1). The IMR
for the black population (most of whom were slaves in 1850) may have been as
high as 350 and the e(0) as low as 23 years (Steckel, 1986). The mortality
transition essentially began in the 1870s and was well underway by the turn of
the century. In 1900 the expected life span at birth for the overall population was
50.1 (51.8 for whites and 41.8 for blacks). The corresponding IMR had fallen to
2
Richard Sterner (1943, pp. 213–323) portrays a very complex picture of the racial differences in
access to many of the New Deal relief programs. Warren Whatley (1983) has suggested that spending
by the Agricultural Adjustment Administration might have led to greater income inequality as
landlords demoted sharecroppers and tenants to wage workers.
NEW DEAL AND INFANT MORTALITY
97
120 for the overall population (111 for whites and 170 for blacks) (Haines,
1998a, Table 1).
A complex set of factors contributed to this achievement, including improvements in diet, nutrition, clothing, and shelter; expanded public health measures to
provide clean water supplies, proper sewerage and refuse disposal, quarantine,
pure milk, unadulterated food and drink, health education, and vaccinations and
inoculations; improved health behavior (e.g., washing and personal hygiene, food
preparation); medical advances; and even possibly exogenous changes in the
virulence and etiology of specific diseases (e.g., scarlet fever). The improvements
in infant mortality were assisted, in part, by the decline in fertility, which had
proceeded throughout the 19th century (Haines, 2000; Meckel, 1990; Troesken,
1999).
During the Progressive Era there was a heightened interest in controlling and
reducing morbidity and mortality among infants, children, and mothers. For
instance, the U.S. Children’s Bureau was founded in 1912 “to investigate and
report upon all matters pertaining to the welfare of children and child life among
all classes of our people” (Bremner, 1971, p. 1525; Lindenmeyer, 1997, Chap. 1).
Julia Lathrop, first head of the agency, “selected a noncontroversial issue,
reducing infant mortality, as the agency’s first project” (Lindenmeyer, 1997, p.
30). The Bureau did much to promote measurement of levels, trends, differentials, and causes of infant mortality. For example, the Children’s Bureau sponsored early studies (1911–1915) of matched birth-infant death records for several
U.S. cities (see Woodbury (1926)).
Perhaps more importantly, the Children’s Bureau successfully lobbied for the
extension of the Death Registration Area (DRA) and the Birth Registration Area
(BRA). The original DRA included only the six New England states plus New
York, New Jersey, Michigan, Indiana, and the District of Columbia. This
preliminary set of states comprised only 26% of the American population, was
more urban, had a higher proportion of the foreign born, and had relatively fewer
blacks (4.4% of the U.S. total) compared to the nation as a whole. In addition,
most of the blacks in the DRA were urban (82%), in contrast to the nation overall,
where 20% of blacks were urban (Preston and Haines, 1991, Table 2.1). It took
even longer to form the BRA. It was put together in 1915 and consisted of the
six New England states plus New York, Pennsylvania, Michigan, Minnesota, and
the District of Columbia. Both the DRA and the BRA did not include the entire
continental United States until 1933, when Texas was admitted. Thus, for the
time period considered in this paper, the vital statistics coverage was not
complete at the beginning of the period, although births, deaths, and infant deaths
were reported for about 95% of the population by 1929 (U.S. Bureau of the
Census, 1975, p. 44).
The Progressive Era saw great strides in the reduction of infant mortality.
Between 1900 and 1920, the IMR declined by about 35%, from 111 to 82, for the
white population and by about 29%, from 170 to 132, for the black population
(Haines, 2000, Table 3). The efforts of the progressives eventually led to the
98
FISHBACK, HAINES, AND KANTOR
passage of the Sheppard–Towner Act in November 1921, which remained in
effect until June 1929. The law appropriated about seven million dollars in
federal money for grants-in-aid to states for the promotion of maternal and infant
health and welfare (Bremner, 1971, pp. 1003–1025; Meckel, 1990, Chap. 8;
Lindenmeyer, 1997, Chap. 4). By 1930 the Children’s Bureau had reported that
the legislation led to, among other things, expansion of the BRA and the DRA,
establishment of state child-hygiene bureaus and divisions, establishment of
permanent state health centers for mothers and children, and, perhaps most
important, an accompanying increase in state appropriations for infant and
maternal health (U.S. Children’s Bureau, 1930, pp. 1–3). Thus, even though the
Sheppard–Towner legislation had expired, much public health infrastructure was
already in place by the late 1920s, both of a general sort—purification of water,
improvements in the disposal of sewerage and refuse, quarantine, health education—and of a type that specifically assisted infants and mothers—milk pasteurization, vaccinations and inoculations, visiting nurses, maternal care, and education.
The IMR continued to decline through the 1930s. 3 As shown in Table 1,
between 1930 and 1940, the IMR for whites declined by 28% while that for
blacks fell by 27%, although both rates rose somewhat between 1932 and 1934.
The declines were virtually identical to the percentage declines of the 1920s
(27% for each group). In the southern states of Virginia, South Carolina, and
Mississippi (see Table 1), black and white infant mortality rates followed the
national trends, although the levels were different from state to state.
Given that infant mortality tends to be correlated with socioeconomic status,
the continuation of the downward trend for both blacks and whites during the
1930s is surprising because America experienced its most severe macroeconomic
disturbance during that decade. Between 1929 and 1933, nominal GNP fell by
46% and real GNP by 30%. The unemployment rate rose from 3.2 to 24.9% of
the civilian labor force (U.S. Bureau of the Census, 1975, Series F1,5, D 86). The
1929 unemployment rate was not reached again until World War II, and the level
of real GNP per capita in 1929 was not achieved until 1940. Thus, despite a
horrendous downturn and a decade of lost economic growth, infant and maternal
health improved, except for a short-lived jump upward in 1933 and 1934 during
the trough of the Depression. Part of the improvement over the course of the
decade may have stemmed from the continuation of practices that had been
developed in the 1920s. Another possible reason might be that the federal
programs under the New Deal helped sustain the earlier progress and helped
prevent those struck hardest by the Depression from suffering too sharp a decline
in their material well-being.
3
In fact, dramatic declines continued throughout the 20th century. By 1997 the IMR had declined
to 7.2 infant deaths per 1000 live births overall. The white rate had fallen to 6.0 and the rate for blacks
to 14.2. See National Center for Health Statistics (1999, Table 27).
NEW DEAL AND INFANT MORTALITY
99
TABLE 2
Per Capita Spending on Various New Deal Programs in the South and Non-South
Agricultural Adjustment Administration, Rental
Benefits and Conservation Payments
Civil Works Administration
Federal Emergency Relief Administration
Public Works Administration
Aid to Dependent Children, Social Security
Administration
PWA/U.S. Housing Administration Housing Projects
Works Progress Administration
Non-South
South
$13.85
6.58
25.10
18.91
$21.99
5.25
13.39
14.76
0.42
1.12
58.05
0.32
0.83
24.44
Source. New Deal spending is from the U.S. Office of Government Reports, Statistical Section,
Report No. 10, Volume 1, “County Reports of Estimated Federal Expenditures March 4, 1933–June
30, 1939.” The per capita figures were created by dividing by the population in 1930 from Historical,
Demographic, Economic, and Social Data: The United States, 1790 –1970 (ICPSR tape number
0003, as corrected by Michael Haines).
Note. The South includes the states of Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana,
Maryland, Oklahoma, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia.
NEW DEAL PROGRAMS
Prior to the 1930s providing financial relief for people struck by misfortune—
poverty, unemployment, injury, etc.—was not considered to be a federal responsibility. State and local governments provided varying degrees of aid through
mothers’ aid programs, poor relief, workers’ compensation, and institutions for
the blind and disabled. When the Depression hit, however, attitudes toward the
federal role were altered, as people saw the Depression as a problem that was
national in scope (Wallis, 1984). The federal government began loaning funds for
relief through the Reconstruction Finance Corporation in 1932, during the
Hoover Administration. Activity at the federal level exploded when Franklin
Roosevelt and the Democratic Congress established the New Deal in March
1933. A profusion of programs were established with goals ranging from direct
relief, to employment, to the building of public works. The federal government
spent approximately 5% of GDP on New Deal programs during the 1930s. This
level of federal government spending during the 1930s was unprecedented and
set the stage for a dramatic shift in the financial responsibilities of the national,
state, and local governments. New Deal programs raised federal spending as a
proportion of all government spending from 30% in 1932 to 46% in 1940 (Wallis
1984, pp. 141–142).
Table 2 shows per capita New Deal spending in the South and the rest of the
nation for several key programs that might have influenced infant mortality
between March 1933 and June 1939. The information comes from the U.S. Office
of Government Reports (1940), which reported federal spending, loans, and
insurance for all of the counties in the United States. The programs listed in Table
100
FISHBACK, HAINES, AND KANTOR
2 potentially influenced infant mortality by reducing economic distress in both
direct and indirect ways. The Federal Emergency Relief Administration (FERA)
provided a combination of work relief and direct relief for the unemployed and
needy families between 1933 and 1935. During the winter of 1933–1934, the
Civil Works Administration (CWA) provided extensive amounts of work relief.
The Works Progress Administration (WPA), later the Works Projects Administration, superceded the FERA in the latter part of 1935 and provided work relief
for “employables,” while “unemployables” once again became the responsibility
of local and state governments. The federal government still provided some
direct relief for dependent children, the blind, and the elderly through programs
established by the Social Security Act of 1935. The Public Works Administration
built public works, while aiding the unemployed by providing construction
employment and potentially some economic stimulus in the area. The Public
Works Administration, in particular, might have had a more direct effect on
public health because it contributed to the building of sewage control and
waterworks facilities in hundreds of communities (Public Works Administration,
1939, pp. 139 –180). The Public Works Administration and later the U.S.
Housing Authority began building public housing projects that were designed to
improve housing and reduce overcrowding in slum areas. Payments to farmers
under the Rental and Benefit program and later the Conservation programs of the
Agricultural Adjustment Administration were designed to increase farm incomes.
The North/South comparisons in Table 2 show that per capita federal New
Deal spending tended to be smaller in the South than in the rest of the country
for most programs. Thus, just by living in the South, blacks had access to a
smaller share of New Deal spending. Part of the difference in spending across the
country may have been driven by the fact that the South was less urban than the
rest of the nation. Many of the relief programs were more focused on resolving
issues related to urban unemployment. The South fared much better with respect
to the AAA programs, although, as we will see below, the distribution of AAA
funds might have contributed to greater income inequality and more problems for
the disadvantaged.
THE DETERMINANTS OF INFANT MORTALITY
The costs and benefits of public programs are never distributed equally across
society. During the 20th century the federal government seems to have been less
discriminatory than southern states and local governments; therefore, we might
expect that New Deal federal government programs were likely to provide more
equal access to blacks and whites than the earlier state and local relief groups.
The Federal Works Agency (1940, p. 23) argued that its programs—including the
Public Works Administration and the Works Projects Administration—actively
sought to ensure no racial discrimination in employment and in the distribution
of benefits. Although the federal government may have attempted to create equal
access to these programs based on race or socioeconomic status, its oversight was
NEW DEAL AND INFANT MORTALITY
101
limited because nearly all of the programs were administered in conjunction with
state and local authorities in some way. Thus, while the federal government may
have been a positive force in aiding blacks in a number of situations, state and
local governments in the South had developed policies that had significantly
retarded black progress for decades (see, for example, Kousser (1974), Higgs
(1977), Ransom and Sutch (1977), and Margo (1990)). Racial differences in
program participation may also have led to an unequal distribution of program
funds. Blacks with limited education may have faced more obstacles in determining their eligibility for relief programs. In addition, past experiences with
local public programs may have discouraged them from applying.
Richard Sterner (1943, pp. 213–323) used various surveys to develop a
complex picture of the extent to which black families participated in New Deal
relief programs. There appear to have been racial differences in the participation
in New Deal programs that varied across programs, and varied from state to state
and probably from county to county within most programs. Sterner found from
surveys in 1933 and 1935 that the share of the black population receiving relief
was higher than the white share of the population in southern cities, but it was
lower in southern rural areas. Black families seem to have fared the worst from
the Aid to Dependent Children program, which was largely administered by state
and local agencies. Lieberman (1998, p. 128) and Sterner (1943, pp. 282–286)
found that the percentage of black children accepted for ADC in the late 1930s
in nearly every southern state was smaller than the black percentage of children
under age 16, even though black families were more likely to have low incomes.
Meanwhile, ADC benefits per child recipient were lower for blacks than for
whites.
In some cases the introduction of New Deal programs could have had perverse
effects on low-income and minority populations. Warren Whatley (1983) has
argued, for example, that sharecroppers and share tenants tended to be demoted
to wage laborers as a result of the Agricultural Adjustment Administration
(AAA). The AAA payments gave landlords incentive to displace tenants, even
though the intent of the legislation was for tenants to share in the benefits of the
farm programs. Since blacks were found more often at the lower end of the tenure
ladder, they may have actually suffered from the introduction of the AAA.
Only limited race-specific information on the division of New Deal spending
between black and white families is available (see Sterner (1943)). However, we
can draw some inferences by looking at how New Deal spending affected black
and white infant mortality rates in the various counties of the United States. The
various New Deal programs potentially might have reduced infant mortality by
providing employment to the unemployed, in some cases by providing direct
relief, by building sanitation projects, and by stimulating the economy either
directly or indirectly with an influx of spending. To measure the impact of the
New Deal, we developed a data set for over 700 counties with race-specific
information on births and deaths throughout the 1930s. The regression model we
estimate is relatively straightforward:
102
FISHBACK, HAINES, AND KANTOR
IMR i ⫽ f共GFR i , ND, X, e i 兲,
where IMR i is the infant mortality rate for race i (black or white), GFR is the
general fertility rate for race i, ND is a group of variables describing New Deal
spending per capita for various programs, X is a group of variables describing the
economic and geographic characteristics of the counties, and e i is a set of
race-specific stochastic error terms. We have race-specific information on the
infant mortality rate and birth rates, but we do not have race-specific information
on the other variables in the equation.
In addition to the New Deal programs, we are able to include several variables
that capture key factors that might have influenced infant mortality. Geographic
differences in infant mortality based on the ethnicity and/or racial mix of the
population are captured in the regression with variables measuring the percentages of black and foreign-born individuals in each county. Racial and ethnic
differences in infant mortality may have arisen from a disparity in the socioeconomic status of various population groups or cultural differences in the use of or
access to medical care in different areas. Carson (1994) cites several studies
showing that the foreign-born and southern blacks were much more likely to use
midwives than were black and white women in northern cities. She also notes
that there were substantial differences in blacks’ and whites’ access to hospitals.
Moreover, since there was substantial residential segregation at the time, there
were likely to be racial and ethnic differences in the extent of residential
crowding and access to sanitation services, even within the same county.
The regression includes the percentage of the population living in urban areas
(defined as residing in towns with more than 2,500 people) and rural, nonfarm
areas. These variables capture the impact of greater population density and
urbanization on the potential spread of epidemics among infants and on access to
health care. To control for differences in the overall economic condition across
the counties, we include retail sales per capita. This variable is designed to proxy
personal income at the local level, which is not available.
We also included a measure of literacy because more highly educated families
were likely to have better information about infant care and health, more
generally. The census reported illiteracy for people aged 10 and above in 1930
and then reported the number of years of school completed for people over 24
years in 1940. We used the illiteracy rate from 1930 and then calculated illiteracy
rates for 1940 for people over 24 years old. The 1940 estimate is based on 1947
information from the U.S. Bureau of the Census (1948, p. 7) on the illiteracy of
males and females over 24 who had no schooling and 1 to 4 years of schooling. 4
4
The illiteracy rate for those with no schooling was 78.2% for males and 80.7% for females. The
illiteracy rate among those with 1 to 4 years of schooling was 22.5% for males and 16.7% for females.
We assumed those with more than 5 years of schooling were all literate. See U.S. Bureau of the
Census (1948).
NEW DEAL AND INFANT MORTALITY
103
CONSTRUCTION OF THE DATA SET
Infant deaths and births were collected from the U.S. Bureau of the Census
(1934a, 1934b, 1934c, 1936a, 1936b, 1938, 1939, 1940, 1941, 1942). From 1930
to 1934 the information on births, deaths, infant deaths, and stillbirths was
reported separately for towns and cities above 10,000 in population and for the
rural sections of each county. We determined the counties in which each of the
town and cities were located and then aggregated the rural and city information
to the county level. In some cases cities were located in two counties. In
situations where over 90% of the population of the city was in one county, we
combined the city information with that county’s information. In several other
situations we combined counties. 5 We also had to combine a number of counties
because of the way the New Deal information was reported. 6
Separate information for “colored persons” and whites was provided for
between one-fourth of the counties in the early 1930s and one-third of the
counties in the late 1930s. The counties with separate information were primarily
located in the South. Colored persons includes some Asians and others who were
not of African descent, although the vast majority of the nonwhites in the South
were African-Americans. Our sample is limited to a subset of all southern
counties for several reasons. In a number of counties, typically with populations
that were less than 10% black, the Census did not report separate information for
blacks and whites. In quite a few settings, the Census reported black and white
information for the urban sector but not the rural sector of the county. Since the
information on the determinants of infant mortality rates was reported for the
entire county level, we felt that we should focus on the counties with full infant
mortality information for the entire county. After ensuring that we had full
information on infant mortality for blacks and whites for the southern counties,
our sample includes 723 counties in the years 1930 to 1932, 845 counties in the
years 1933 to 1935, and 1192 counties for the years 1936 to 1939. In the
difference regressions, which compare the later years with the period 1930 to
1932, the sample is limited to the number of counties for which we have data in
5
Therefore, we combined Russell and Lee counties in Alabama (Phenix City); DeKalb, Fulton,
Milton, and Campbell in Georgia (Atlanta); Edgcomb and Nash in North Carolina (Rocky Mount);
and James City and York in Virginia (Williamsburg).
6
In Virginia we combined the following districts that were reported separately in the Census:
Albemarle county and Charlottesville city; Allegheny county and Clifton Forge city; Augusta county
and Staunton city; Campbell county and Lynchburg city; Dinwiddie county and Petersburg city;
Elizabeth City county and Hampton city; Frederick county and Winchester city; Henrico county and
Richmond city; Henry county and Martinsville city; James City county and Williamsburg city;
Montgomery county and Radford city; Nansemond county and Suffolk city; Norfolk county with
Norfolk city, South Norfolk city, and Portsmouth city; Pittsylvania county and Danville city; Prince
George county and Hopewell city; Roanoke county and Roanoke city; Rockbridge county and Buena
Vista city; Rockingham county and Harrisonburg city; Spotsylvania county and Fredericksburg city;
Warwick county and Newport News city; Washington county and Bristol city; Arlington county and
Alexandria city.
104
FISHBACK, HAINES, AND KANTOR
the earliest period. One reason for the smaller number of observations in 1930 to
1932 is that Texas (253 counties) was not included in the BRA or DRA at this time.
Census information on population and many of the socioeconomic variables
were derived from ICPSR data set 0003, as corrected by Michael Haines (ICPSR,
0003, no date). The information on the different New Deal programs was
reported by the United States Office of Government Reports (the “Office”). In
1940 the Office compiled a detailed statistical description of the federal government’s expenditures across the United States during the New Deal era. Data were
collected on 31 different New Deal-era programs for 3,069 counties in the United
States for the period March 1933 through June 1939. For each county, aggregate
expenditures over the entire time period were reported for each of the programs.
We focus our attention here on the spending programs that would have been most
directly beneficial in preventing infant mortality, including the major work relief
programs, Aid to Dependent Children, the AAA, and the PWA public works and
public housing projects.
Although the Office reported information aggregated for the entire period from
1933 to 1939, we were able to use the reports of the various agencies to roughly
identify the timing of the programs. This allowed us to break the New Deal
spending into two time periods, 1933 through 1935 and 1936 through 1939. In
brief, the funds spent by the Federal Emergency Relief Administration (FERA),
the Civil Works Administration (CWA), and the Agricultural Adjustment Administration’s (AAA) Rental and Benefits program were largely exhausted prior
to 1936. The Social Security Aid to Dependent Children program, the Public
Works Administration’s (PWA) public housing program (later transferred to the
U.S. Housing Authority), the AAA Conservation payments, and 95% of the
WPA programs had their effects after 1935. We used a variety of sources to split
the PWA public works programs into the two periods. 7 To translate the New Deal
spending into per capita figures, we divided by the county’s population in 1930.
7
The Federal Emergency Relief Administration programs allocated well over 90% of its funds
between May of 1933 and June of 1935. There was a small amount of additional spending by FERA
as the program wound down through March 1937. The Civil Works Administration operated between
November 1933 and April 1934. Because the WPA did not start until July 1935 and less than 6% of
the employment occurred in 1935 (see Federal Works Agency, 1940, p. 413), we placed all of the
WPA spending in the 1936 –1939 period. The spending on Aid to Dependent Children was distributed
after 1935. The U.S. Housing Authority was established on November 11, 1937, but took over
housing projects that had been built or started by the Public Works Administration. The timing of the
projects suggests that people did not begin to live in the projects until the end of 1935, so we placed
the impact of the spending in the 1936 –1939 period (Public Works Administration, 1939, pp.
207–217, 283–284). We distributed the PWA funds for federal and nonfederal projects using
employment information from the Federal Works Agency (1940, p. 307). Between 1933 and 1935,
77.9% of the employed worked on federal projects, and 22.1% did from 1936 on. Thirty-two percent
worked on nonfederal projects from 1933–1935, 68% did from 1936 to 1939. The Office of
Government Reports (1940) provided information on “rental and benefit” payments under the AAA
from 1933 and 1935 and for “conservation” payments from 1936 and 1937, which allowed us to
divide the information between time periods accordingly.
NEW DEAL AND INFANT MORTALITY
105
Given the timing of our information on the New Deal programs, we have
created three time groupings, 1930 –1932, 1933–1935, and 1936 –1939. For each
period we summed the infant deaths and births and then calculated an infant
mortality rate. When we calculated the general fertility rate for each race we used
the number of women of each race aged 15 to 44 in the denominator. There was
no New Deal spending in the 1930 –1932 period for the programs that we
considered. We used the census information from 1930 and 1940 and straightline interpolations for the midpoint of each period (1931 for 1930 –1932, 1934 for
1933–1935, and 1937/38 for 1936 –1939) for the following variables: percent
black, percent foreign-born, percent illiterate, percent urban, and percent ruralnonfarm. For retail sales per capita, we used information on retail sales in 1929
for the 1930 –1932 period, we averaged retail sales for 1933 and 1935 for
1933–1935, and we came up with a midpoint estimate for retail sales in 1936 –
1939 using the retail sales information from 1935 and 1939. 8 For each period we
divided the retail sales by a population estimate based on a linear interpolation
between 1930 and 1940. The retail sales information for 1933 and 1935 comes
from the Consumer Market Data Handbooks for 1936 and 1939 published by the
U.S. Department of Commerce, Bureau of Foreign and Domestic Commerce
(1936, 1939).
RESULTS OF THE ANALYSIS
The analysis involves several steps. Given that many prior infant mortality
studies have used cross-section regressions, the cross-sectional regressions for
the three time periods are reported. We lack measures of some of the variables
that might have influenced infant mortality, such as public health expenditures,
hospitals, or household sanitation practices. We know that public health expenditures and access to medical care were important factors, and these varied
substantially across states. In the cross-sectional regressions state dummy variables are included to control for differences in such variables as public health
spending, state relief programs, cost of living, and other factors that were
common to all counties in each state.
The state dummies, however, cannot control for unmeasured differences
across counties. Numerous econometric studies have shown that mistaken inferences can be drawn by not controlling for unmeasured heterogeneity across the
observations. As a step toward controlling for the unmeasured heterogeneity
across counties, we use two separate difference regressions, one for 1933–1935
minus 1930 –1932 and another for 1936 –1939 minus 1930 –1932. The difference
regression helps control for factors that remained constant within the same
counties over time but varied across counties. We are using 1930 –1932 as a base
year without New Deal programs for the comparisons. In essence, the difference
8
Using the average of retail sales in 1929 and 1933 to represent the retail sales in 1930 –1932 does
not change the qualitative results reported here.
106
FISHBACK, HAINES, AND KANTOR
regressions are another method of controlling for fixed effects. We do not run
differences for 1936 –1939 and 1933–1935 because for some of the New Deal
programs like the PWA, which spent money in both time periods, we would have
to make assumptions about the timing of the distribution of the funds. We
avoided estimating the difference between 1936 –1939 and 1933–1935 because
we did not want our results to be driven by our assumptions about the timing of
the New Deal spending.
Finally, we worried about potential simultaneity bias in the estimates of the
New Deal variables. That is, it is possible that increased infant mortality (or a
slowing in the decline of infant mortality) might have been a signal used by New
Deal administrators to deliver more funds for relief and employment. Therefore,
in all of the regressions we use instruments for the New Deal programs to control
for possible endogeneity. The instruments are predictions from regressions with
a variety of economic and political variables similar to the ones used in the
literature summarized by Wallis (1998) and Couch and Shughart (1998). We
estimated separate regressions for each program, with expenditures regressed on
the following socioeconomic and political variables (for 1930 unless otherwise
specified): the inverse of population, the percent black, the percent urban, the
decline in retail sales from 1929 to 1933, the ratio of tax returns filed in 1935 to
the population in 1930, the per capita value of farms, the percentage of land in
farms, average farm size, the ratio of unemployed in 1930 to gainfully employed
in 1930, the ratio of unemployed in 1937 to gainfully employed in 1930, mean
and standard deviation of the percent voting Democrat from 1896 to 1932, the
difference between the percent voting for Roosevelt in 1932 and the mean from
1896 to 1932, electoral votes per capita, percent voting in 1932, a series of
variables describing representation on House committees during the 1930s, and
a series of state dummy variables. We then used the predictions from these
equations as instruments for the New Deal programs in the infant mortality
regressions. The correlation between the instruments for the New Deal programs
and the actual values are typically above 0.5, with the lowest correlation at about
0.37. 9 Since the instruments were all estimated using the same sets of correlates,
we also examined the correlations between the instruments. The overwhelming
majority of correlations were below 0.5 in absolute value. The highest correlation
was between FERA and CWA (0.73). There was a 0.73 correlation between the
FERA and WPA, but they do not appear in the same equations.
The inclusion of the birth rate in the infant mortality equation also creates
potential endogeneity problems. Demographers have found a positive relationship between infant deaths and birth rates, which could be explained as either
replacement or insurance (Haines, 1998b). As couples seek a target family size,
a baby lost in infancy would be replaced with another child. Alternatively, some
9
The correlations between the instruments and the actual per capita funds in the programs are
FERA (0.64), WPA (0.74), PWA (0.37), CWA (0.68), SSA aid to dependent children (0.73), AAA
programs (0.89), PWA housing (later USHA housing) (0.78).
107
NEW DEAL AND INFANT MORTALITY
TABLE 3
Summary Statistics of Variables Used in the Difference Equations
Variables
Infant mortality rates:
Whites, 1930–1932
Whites, 1933–1935
Whites, 1936–1939
Blacks, 1930–1932
Blacks, 1933–1935
Blacks, 1936–1939
General fertility rates:
Whites, 1930–1932
Whites, 1933–1935
Whites, 1936–1939
Blacks, 1930–1932
Blacks, 1933–1935
Blacks, 1936–1939
Percent black, 1930
Percent black, 1940
Percent illiterate, 1930
Percent illiterate, 1940
Percent urban, 1930
Percent urban, 1940
Percent foreign born, 1930
Percent foreign born, 1940
Percent rural-nonfarm, 1930
Percent rural-nonfarm, 1940
Retail sales per capita, 1929
Retail sales per capita, 1933
Retail sales per capita, 1935
Retail sales per capita, 1939
Per capita spending on New Deal programs:
Federal Emergency Relief Administration
Civil Works Administration
Public Works Administration, 1933–1935
Agricultural Adjustment Administration,
1933–1935
Works Progress Administration
Agricultural Adjustment Administration,
1936–1937
Aid to Dependent Children under Social
Security Administration
PWA/USHA Housing Projects
Public Works Administration, 1936–1939
Mean
Standard deviation
55.2
56.9
52.0
89.9
85.4
79.1
16.8
17.1
14.9
37.7
34.1
26.7
93.1
90.4
91.0
98.4
100.2
104.3
38.2
36.7
11.6
10.9
15.9
17.9
0.6
0.5
26.7
27.5
179.6
88.1
115.7
149.0
34.5
33.2
32.5
34.5
42.8
66.3
18.0
18.2
5.8
4.8
20.7
20.9
1.2
1.0
16.3
16.6
93.9
48.6
63.8
81.3
12.9
4.9
7.0
17.4
2.8
35.5
14.4
18.2
12.0
15.3
8.6
6.1
0.3
0.1
6.2
0.3
1.5
17.8
Sources. For New Deal spending and demographic sources, see Table 2. Retail sales in 1933 and
1935 are from U.S. Department of Commerce, Bureau of Foreign and Domestic Commerce (1936,
1939). The infant mortality data are from U.S. Bureau of the Census (1934a, 1934b, 1934c, 1936a,
1936b, 1937, 1938, 1939, 1940, 1941, 1942).
Note. Sample is restricted to southern counties.
108
FISHBACK, HAINES, AND KANTOR
families may have had more children as insurance against the future loss of a
child. 10 To determine the relative effects of the birth rate variable, we have
estimated a reduced-form version of the mortality equations that did not include
the birth rate, and the results for the remaining variables, which are the focus of
the study, are similar to those reported here.
Table 3 reports summary statistics of New Deal spending for the various
programs and of the socioeconomic measures in 1930 and 1940 for the southern
sample of 723 counties used in the difference regressions. Comparisons of the
southern sample with national averages for all counties show that the percent
black and the percent illiterate are substantially higher in the southern sample.
Meanwhile, nominal retail sales per capita, the percent foreign-born, the percent
urban, and the percent rural-nonfarm were all substantially lower.
We ran OLS regressions and used White’s (1980) method to correct the
standard errors for potential heteroskedasticity. We chose OLS over a logit
transformation of the infant mortality rate for ease of interpretation of the
coefficients. Using a logit transformation does not affect the statistical inferences.
Tables 4 and 5 show the results of cross-section regressions for 1930 –1932,
1933–1935, and 1936 –1939 for blacks and whites, respectively. The independent
variables explain roughly 36% of the variation in 1930 –1932, 31% in 1933–
1935, and 14% in 1936 –1939 in the black cross-section regressions. In contrast,
the percentage of variation explained for the white observations rises to 39% by
the last period. 11 The percentages explained by the variables in the regression are
actually impressive given that we could not obtain county-level information on
many factors that influence infant mortality, such as specific behaviors of
families with small children, breast-feeding practices, use of hospitals, and access
to other forms of medical care.
Most of the results are consistent with expectations about how various factors
influence infant mortality. Black and white infant mortality rates were higher in
areas where the illiteracy rate was higher, consistent with views that the absence
of education contributed to an environment where small children were subject to
greater illness. Black and white infant mortality rates were also higher in urban
and rural nonfarm areas. It appears that in more densely settled regions the
problems of contagion and/or public health issues with people living closer
together were not overcome by extant sanitation practices or access to more
doctors in more populated areas.
We included a measure of retail sales per capita as a proxy for consumption in
10
Haines (1998b) estimated a replacement effect in the United States of between one-fourth to
one-third of a child per child death. For more discussion of the issue, see the edited volume by
Montgomery and Cohen (1998).
11
The R 2 values in regressions with only the state dummies were 0.295 for blacks and 0.128 for
whites in 1930 –1932, 0.144 for blacks and 0.0497 for whites in 1933–1935, and 0.078 for blacks and
0.047 for whites in 1936 –1939. The R 2 values in regressions with all of the variables except the state
dummies are 0.139 for blacks and 0.069 for whites in 1930 –1932, 0.232 for blacks and 0.146 for
whites in 1933–1935, and 0.104 for blacks and 0.324 for whites in 1936 –1939.
109
NEW DEAL AND INFANT MORTALITY
TABLE 4
Cross-Sectional Regression Results of Black Infant Mortality
1930–1932
Variables
1933–1935
1936–1939
Coeff.
t-stat.
Coeff.
t-stat.
Coeff.
t-stat.
Intercept
85.78
General fertility rate of blacks
⫺0.10
Percent black
⫺0.23
Percent illiterate
1.13
Percent urban
0.42
Percent foreign-born
⫺1.17
Percent rural nonfarm
0.54
Retail sales per capita
0.0004
Instruments for New Deal spending
per capita:
Federal Emergency Relief
Administration (FERA)
Works Progress Administration
(WPA)
Civil Works Administration (CWA)
Public Works Administration
(PWA)
Agricultural Adjustment
Administration (AAA)
Aid to Dependent Children (ADC)
under Social Security
PWA/USHA Housing Projects
States dummy variables:
Alabama
⫺21.45
Arkansas
⫺42.57
Florida
⫺21.43
Georgia
⫺14.24
Louisiana
⫺24.20
Mississippi
⫺21.61
North Carolina
⫺4.02
South Carolina
⫺8.30
Texas
Kentucky
54.95
Maryland
33.54
Oklahoma
⫺9.88
Tennessee
⫺9.70
N
723
R-squared
0.36
R-bar-squared
0.35
7.96
⫺1.81
⫺2.66
3.90
4.48
⫺0.93
5.48
0.02
70.88
0.001
⫺0.13
1.54
0.78
2.35
0.70
⫺0.02
6.13
2.04
⫺1.48
4.60
6.06
2.01
5.39
⫺0.34
68.52
⫺0.004
⫺0.29
0.86
0.76
0.63
0.45
0.03
5.78
⫺0.38
⫺2.25
1.61
4.53
0.37
3.50
0.64
⫺0.23
⫺1.11
⫺5.26
⫺8.13
⫺4.12
⫺3.38
⫺4.52
⫺5.51
⫺0.93
⫺1.67
3.77
4.34
⫺0.90
⫺1.50
0.83
1.15
⫺6.90
⫺2.22
0.04
0.31
⫺0.42
⫺1.05
0.68
4.28
0.67
1.87
⫺45.61
⫺22.46
⫺1.46
⫺1.89
⫺3.34
⫺20.08
⫺19.20
⫺7.83
14.58
⫺23.45
⫺5.05
13.67
⫺11.37
7.03
58.38
52.22
19.01
1192
0.14
0.13
⫺0.30
⫺2.09
⫺3.00
⫺0.95
0.56
⫺3.55
⫺0.59
1.40
⫺1.24
0.67
1.66
1.99
0.96
⫺13.62
⫺37.20
⫺8.39
⫺18.58
⫺37.89
⫺30.36
⫺16.41
⫺12.55
⫺12.83
⫺7.56
5.14
28.15
⫺9.47
845
0.31
0.29
⫺2.34
⫺4.99
⫺0.72
⫺3.05
⫺4.10
⫺3.39
⫺2.14
⫺1.37
⫺1.87
⫺0.62
0.42
1.68
⫺1.19
Source. See Table 3 and the discussion of the New Deal instruments in the text.
Note. The t-statistics are based on heteroskedasticity-corrected standard errors using White’s
(1980) method.
110
FISHBACK, HAINES, AND KANTOR
TABLE 5
Cross-Sectional Regression Results of White Infant Mortality
1930–1932
1933–1935
1936–1939
Variables
Coeff.
t-stat.
Coeff.
t-stat.
Coeff.
t-stat.
Intercept
General fertility rate of whites
Percent black
Percent illiterate
Percent urban
Percent foreign-born
Percent rural nonfarm
Retail sales per capita
Instruments for New Deal spending
per capita:
Federal Emergency Relief
Administration (FERA)
Works Progress Administration (WPA)
Civil Works Administration (CWA)
Public Works Administration (PWA)
Agricultural Adjustment Administration
(AAA)
Aid to Dependent Children (ADC)
under Social Security
PWA/USHA Housing Projects
State dummy variables:
Alabama
Arkansas
Florida
Georgia
Louisiana
Mississippi
North Carolina
South Carolina
Texas
Kentucky
Maryland
Oklahoma
Tennessee
N
R-squared
R-bar-squared
48.74
0.01
⫺0.07
0.68
0.18
⫺2.80
0.02
0.01
11.84
0.51
⫺1.46
4.31
3.39
⫺4.17
0.38
1.16
56.66
⫺0.02
⫺0.17
1.50
0.24
0.28
0.07
0.00
9.49
⫺0.93
⫺3.42
9.06
3.49
0.46
1.17
0.18
25.35
0.14
⫺0.09
1.26
0.36
3.03
0.17
0.01
7.34
10.29
⫺2.18
8.13
6.99
8.80
4.37
0.75
0.50
1.50
⫺3.13
⫺0.09
⫺2.22
⫺1.37
⫺0.11
⫺2.08
⫺0.52
⫺3.97
0.09
1.76
0.12
1.75
⫺3.31
⫺8.61
⫺0.35
⫺2.20
⫺2.44 ⫺6.83
⫺1.61 ⫺4.64
⫺1.83 ⫺15.64
⫺0.57 ⫺3.66
⫺4.68 ⫺17.71
⫺2.35 ⫺6.82
⫺1.34 ⫺1.56
⫺0.98
2.39
⫺1.74 ⫺6.23
⫺1.61
0.91
⫺1.09 ⫺1.31
⫺0.38
3.12
⫺1.17
3.22
1192
0.39
0.38
⫺1.88
⫺1.51
⫺6.41
⫺1.42
⫺2.21
⫺2.85
⫺0.58
0.68
⫺2.49
0.34
⫺0.12
0.40
0.54
⫺8.34
⫺12.70
⫺3.35
⫺1.70
⫺9.80
⫺12.47
4.24
⫺1.93
4.11
13.48
⫺10.09
0.42
723
0.19
0.17
⫺2.95 ⫺8.82
⫺3.99 ⫺7.06
⫺1.17 ⫺10.79
⫺0.73 ⫺1.92
⫺3.53 ⫺22.37
⫺4.78 ⫺11.13
1.63 ⫺5.56
⫺0.64 ⫺4.79
⫺6.15
1.18 ⫺8.83
3.27 ⫺6.24
⫺2.40 ⫺1.84
0.12 ⫺5.21
845
0.20
0.17
Source. See Table 3 and the discussion of the New Deal instruments in the text.
Note. The t-statistics are based on heteroskedasticity-corrected standard errors using White’s
(1980) method.
NEW DEAL AND INFANT MORTALITY
111
these areas. Although many studies find an inverse relationship between socioeconomic status and infant mortality, we cannot reject the hypothesis that retail
sales had no effect in either the black or the white regression. We also included
the variable percent black as one means of examining the impact of the black
group economy on infant mortality. A higher black percentage of the population
was associated with lower black infant mortality rates. In the highly segregated
southern society, access to hospitals and medical care was subject to discrimination. As the percentage black in a county rose, there was the potential for a
hospital or public health services devoted to the black community that otherwise
would be unavailable in an area where there were relatively few blacks. It is
interesting to note that white infant mortality was also lower in areas with a larger
percent black population. 12
We included a measure of the birth rate under the hypothesis that more births
might raise the probability of infant deaths by increasing the number of infants
at risk or by increasing pressure on the existing health-care network, holding
other factors constant. The positive relationship is present for blacks in 1933–
1935 and for whites in 1936 –1939, but in the remaining periods we could not
reject the hypothesis of no effect.
We also included a series of state dummy variables to capture the effects of
factors that were common to the counties in those states during that time period.
Such factors might include state public health policies, cost of living differences,
epidemics, and other factors. The results for all of the years show that there was
substantial variation in infant mortality across the states, after holding the other
factors constant. The differences across states were similar across time periods
and races, as the correlations of the coefficients of the state dummies from the
regressions for different time periods ranged from a low of 0.46 between the state
coefficients in the black regressions for 1930 –1932 and 1933–1935 to a high of
0.72 between the black regressions for 1933–1935 and 1936 –1939. The correlation of state coefficients across time in the white regressions ranged from a low
of 0.3 between the white state coefficients in 1930 –1932 and in 1933–1935 to a
high of 0.81 between the white state coefficients in 1933–1935 and 1936 –1939.
The correlation between the state coefficients in the black regressions and in the
white regressions in the same year were relatively high at 0.78 in 1930 –1932,
0.57 in 1933–1935, and 0.44 in 1936 –1939.
The cross-sectional regressions offer an initial look at the impact of key New
Deal spending programs while controlling for differences in state policies. We
12
There might be a couple of explanations for this result. First, it might be that the reduction in
infant mortality among blacks benefited whites by reducing the contagion of infant diseases,
particularly in cases where black domestic servants were in daily contact with relatively high-income
whites. A second potential explanation for why whites benefited when the black population was
higher may be that income was distributed more unevenly in those counties. In other words, a
relatively large black population in the South typically translated into relatively more farm tenants,
sharecroppers, and wage laborers. Perhaps in such counties the white population was wealthier, which
may have been manifest in lower infant mortality rates.
112
FISHBACK, HAINES, AND KANTOR
are hesitant to put much emphasis on the New Deal results from the crosssectional regressions because we believe that the inclusion of the state dummies
has not fully controlled for unmeasured heterogeneity across counties in the
sample. One reason to report the cross-sectional results is to show how some of
the statistical inferences change when we switch to a difference method that
controls for unmeasured factors that are constant for each county over the years
but vary across counties. The results in Tables 4 and 5 show that among the relief
programs, the CWA was associated with statistically significant reductions in
infant mortality for blacks and whites, the FERA was associated with statistically
insignificant increases in infant mortality for both races, and the WPA lowered
infant mortality for both blacks and whites, although statistically significant only
for whites. The Social Security Administration’s Aid to Dependent Children
program was associated with large reductions in infant mortality for blacks and
whites, but we cannot reject the hypothesis of no effect. The PWA’s public
housing program, later under the U.S. Housing Administration, also lowered
infant mortality. The PWA sewer projects and other public works projects had
statistically insignificant and small effects in 1933–1935 and then was associated
with lower infant mortality, particularly for whites, in 1936 –1939. Finally, the
AAA funds distributed to farmers in Rental and Benefit payments and later as
Conservation payments were associated with higher levels of infant mortality for
both blacks and whites in both 1933–1935 and 1936 –1939.
Although we have controlled for differences in state policy in the crosssectional regressions, there still may have been unmeasured heterogeneity across
counties that we have not controlled for at this point. Even within the same state
there may have been social differences in the birthing and raising of children
across counties. There certainly were unmeasured differences in the extent and
reach of county health programs, and we can anticipate that epidemics or drought
conditions hit some counties in the same state harder than others. Therefore, we
analyzed the data again in difference form to try to control for the unmeasured
factors that remained the same over the two time periods in each county but
differed across counties. There are two sets of difference regressions, one set
focuses on the difference between infant mortality and the independent variables
between the periods 1930 –1932 and 1933–1935 and the other focuses on the
differences between the periods 1930 –1932 and 1936 –1939. The two sets of
difference regressions contain different lists of New Deal programs because
programs like the FERA and the CWA essentially terminated by the end of 1935,
while other programs were started after 1935. Some programs like the AAA
covered both periods.
In the difference regressions the differences in the percent black, illiterate,
rural nonfarm, urban, and foreign-born between 1930 –1932 and 1933–1935 and
between 1930 –1932 and 1936 –1939 are essentially linear transformations of the
trends between 1930 and 1940. All monetary variables in the equation are in
current dollars because the New Deal spending programs were spread over
different time periods. The consumer price index (1967 ⫽ 100) for the entire
113
NEW DEAL AND INFANT MORTALITY
TABLE 6
Difference-Regression Results of Black and White Infant Mortality, 1936 –1939 minus 1930 –1932
Black
Variables
Coeff.
Intercept
⫺10.53
General fertility rate of blacks
0.03
General fertility rate of whites
Percent black
1.24
Percent illiterate
⫺0.07
Percent urban
0.39
Percent foreign-born
⫺3.53
Percent rural nonfarm
0.17
Retail sales per capita
⫺0.04
Instruments for New Deal spending
per capita:
Works Progress Administration (WPA)
⫺0.05
Public Works Administration (PWA)
⫺0.81
Agricultural Adjustment Administration
(AAA)
⫺0.24
Aid to Dependent Children (ADC)
under Social Security
9.00
PWA/USHA Housing Projects
2.81
N
723
R-squared
0.02
R-bar-squared
0.00
Black difference
minus white
difference
White
t-stat.
Coeff.
t-stat.
Coeff.
t-stat.
⫺2.20 ⫺18.01
1.28
0.14
1.59
0.56
⫺0.13
0.10
0.82 ⫺0.09
⫺0.59 ⫺3.41
0.46 ⫺0.10
⫺0.89
0.00
⫺2.82
1.94
1.23
0.27
⫺0.34
⫺1.02
⫺0.45
⫺0.16
11.78
0.01
⫺0.17
0.78
⫺0.26
0.50
0.57
0.26
⫺0.03
1.75
0.75
⫺3.08
0.91
⫺0.46
0.95
0.10
0.62
⫺0.76
⫺0.39
⫺2.37
0.13
⫺0.16
2.24
⫺1.02
⫺0.16
⫺0.65
⫺1.25
⫺1.82
⫺0.74
0.13
0.99
⫺0.35
⫺1.07
1.63
0.54
⫺1.58
0.62
⫺0.72
0.22
10.44
0.87
723
0.04
0.02
1.91
0.17
0.08
0.07
Source. See Table 3 and the discussion of the New Deal instruments in the text.
Note. The t-statistics are based on heteroskedasticity-corrected standard errors using White’s
(1980) method.
United States fell 6.7% per year, from 51.3 in 1929 to 38.8 in 1933, rose 2.6%
per year, to 43 in 1937, and then fell 0.8% per year, to 42 in 1940 (U.S. Bureau
of Census, 1975, series E-135, pp. 210 –211).
As is typically the case, the R 2 values and adjusted R 2 values are substantially
lower in the difference regressions reported in Tables 6 and 7 than in the
cross-sectional regressions in Tables 4 and 5. In the cross-sectional regressions,
illiteracy, percent black, percent urban, and percent rural nonfarm had strong and
statistically significant effects. In the difference regressions, we cannot reject the
hypothesis of no effect. There are two potential reasons for the difference in these
effects. One possibility is that the illiteracy, urbanization, and percent black
coefficients in the cross-section regressions were correlated with unmeasured
features, like sanitation and public health programs, that influenced infant mortality. Thus, the coefficients in the cross-section regressions might have reflected
the bias of omitting the unmeasured variables. A second possibility is that there
was very little variation in the percent illiterate, percent urban, and percent black
114
FISHBACK, HAINES, AND KANTOR
TABLE 7
Difference-Regression Results of Black and White Infant Mortality, 1933–1935 minus 1930 –1932
Black
Variables
Intercept
General fertility rate of blacks
General fertility rate of whites
Percent black
Percent illiterate
Percent urban
Percent foreign-born
Percent rural nonfarm
Retail sales per capita
Instruments for New Deal spending
per capita:
Federal Emergency Relief Administration
(FERA)
Civil Works Administration (CWA)
Public Works Administration (PWA)
Agricultural Adjustment Administration
(AAA)
N
R-squared
R-bar-squared
Coeff.
Black minus
white
White
t-stat.
⫺27.77 ⫺3.46
0.001
0.01
Coeff.
t-stat.
Coeff.
t-stat.
⫺3.10
⫺0.23
0.27
0.11
⫺0.26
⫺0.31
⫺0.62
⫺0.13
⫺0.78
1.38
⫺0.10
⫺0.39
⫺7.82
⫺0.79
⫺0.02
0.68
⫺0.10
⫺0.32
⫺0.62
⫺0.82
⫺0.55
⫺1.25 ⫺0.54 ⫺26.28
⫺0.02
0.02
0.31
0.05
1.21
1.08
0.28
0.25
0.27 ⫺0.37
⫺0.05 ⫺0.09 ⫺0.40
1.06
0.17 ⫺8.68
⫺0.70 ⫺1.48 ⫺0.13
0.01
0.79 ⫺0.03
⫺1.13
6.74
⫺0.32
⫺2.48
3.25
⫺2.82
⫺0.25 ⫺1.27
1.28
1.66
⫺0.04 ⫺0.70
⫺0.91 ⫺1.75
5.53
2.39
⫺0.28 ⫺2.18
0.39
722
0.05
0.04
2.62
0.13
722
0.02
0.01
0.27
722
0.03
0.01
1.96
1.60
Source. See Table 3 and the discussion of the New Deal instruments in the text.
Note. The t-statistics are based on heteroskedasticity-corrected standard errors using White’s
(1980) method.
between 1930 and 1940. The correlations between 1930 and 1940 for the sample
of 722 counties was 0.99 for percent black, 0.84 for percent illiterate, 0.97 for
percent urban, 0.97 for percent foreign-born, and 0.92 for percent rural nonfarm.
Thus, much of the variation in the cross section for these variables is eliminated
by the differencing method.
When looking at the impact of the New Deal, however, we can eliminate the
second possibility that there was little variation in New Deal spending because
there was no spending until 1933 in the programs we examine and the amounts
spent varied tremendously across counties. Therefore, the primary reason for
differences between the econometric results for the cross sections and the
difference regressions for the New Deal program will be driven by our controls
for unmeasured heterogeneity across counties in the difference regressions.
In the early New Deal the FERA and CWA were the major relief programs.
Spending on the FERA, which primarily ran from 1933 through 1935, lowered
black infant mortality by a statistically significant 1.13 infant deaths per 1000
births for every per capita dollar spent. A one-standard-deviation increase in per
NEW DEAL AND INFANT MORTALITY
115
capita FERA spending of $17.35 would have been associated with a reduction in
black infant mortality of nearly 20 deaths per 1000, which would have cut the
black infant mortality rate to nearly the level of white infant mortality. Spending
on the FERA also had a negative effect on white infant mortality, but the effect
was much smaller and not statistically significant. 13
The reductions in infant mortality associated with FERA spending were
somewhat offset by increases in infant mortality associated with Civil Works
Administration spending on work relief between November 1933 and April 1934.
In the difference regressions in Table 7, the CWA effect is positive and statistically significant at the 10% level for both blacks and whites. Further, CWA
spending had a larger effect on black infant mortality than on white infant
mortality.
The WPA, the major work relief program of the later New Deal, had a
statistically insignificant and small negative effect on black infant mortality and
a positive and statistically significant effect on white infant mortality. The
difference in the impact between the FERA, the CWA, and the WPA may have
been driven by the explicit goals of the programs. The FERA had both direct
relief and work relief components to it, while the CWA was specifically focused
on work relief. Williams (1968, pp. 120, 130) notes that the FERA used
deficiencies in the families’ budgets to determine the eligibility of recipients for
either work or direct relief as well as the number of hours that recipients could
work. Further, many of the FERA rules for projects and work opportunities
limited relief opportunities for more highly skilled and employable workers, and
thus targeted more low-income people. In contrast, even though the CWA was
largely administered by FERA administrators, the CWA basically followed the
rules established for work on PWA projects, and half of the CWA workers,
though unemployed, were without relief status. When the WPA was established
in the latter half of 1935, the WPA explicitly focused on work relief and left
responsibility for the “unemployables” to state and local governments. Families
with an employable member, even if unemployed at the time, were likely to have
more resources than those eligible for direct relief. Thus, it appears that the
FERA offered more benefits to the types of families with the least resources, and
thus, the funds were directed to families where infant health was likely most
threatened.
The Public Works Administration was less oriented toward work relief than
the building of public works projects, which included a significant number of
sewage treatment plants in smaller cities. The PWA has a negative and statistically significant effect on black infant mortality in both time periods, and the
effect strengthens over time. An additional dollar of PWA spending in 1933–
13
We experimented with including interactions between the percent urban and the New Deal
variables because Sterner (1943) had suggested that the relative share of relief spending for blacks
was higher in urban areas than in rural areas. However, none of the interaction terms were statistically
significant and most of the other results were unaffected.
116
FISHBACK, HAINES, AND KANTOR
1935 was associated with a statistically significant reduction of about one-third
of a death per 1000 births in black infant mortality. The effect is even stronger
in 1936 –1939 at a reduction of 0.8 of a death per 1000 births. In both periods the
PWA had a small negative and statistically insignificant association with white
infant mortality. The t-tests in the black minus white regressions imply that the
PWA had a more beneficial impact on infant mortality for blacks than for whites.
The strength of the PWA effect is probably more from the building of sewage
facilities and other public works than from the work relief provided. The PWA
tended to be more focused on skilled workers, whose families were more likely
to be at income levels where the risk of infant mortality was diminished. In fact,
the stronger reductions in infant mortality in 1936 –1939 than in 1933–1935 are
consistent with a strong lagged effect of the PWA. The effects of better sanitation
and public works were not likely to take effect until the later period, after the
projects had been finished and put into operation. The results here are also
consistent with earlier results by Werner Troesken, who found that black mortality due to typhoid and other water-borne diseases was sharply reduced by the
development of public water treatment plants at the turn of the century (Troesken, 1999).
At the end of 1935 the Public Works Administration began building a series
of public housing projects for the poor that were later administrated by the U.S.
Housing Administration. Although these public housing projects have marred
reputations today, projects like Techwood in Atlanta and projects in Birmingham
appeared more promising when new. While we would anticipate that improved
housing for the poor would reduce infant mortality, we cannot reject the hypothesis of no effect in the difference regressions. It may have been that the housing
was not sufficiently better to have made a difference. It is possible that New Deal
administrators chose families to live in the new projects with an eye toward
ensuring the success of the project and thus did not populate them with the types
of families for whom infant mortality was a problem. Or dislocations from the
housing torn down to build the new housing might have offset any salutary
effects.
The Social Security Administration’s aid to dependent children (ADC) program was the primary program in the later New Deal targeted specifically for
children. The impact of ADC spending was negative, although not statistically
significant, for whites, and positive and large, but statistically insignificant, for
blacks. We have no definitive explanation for why ADC spending did not lower
infant mortality for either blacks or whites. The amount spent per person was
small relative to the relief programs, and most of the spending came late in the
New Deal, after the long-term downward trend for infant mortality had been
reestablished.
The major agricultural spending program we examine is the Agricultural
Adjustment Administration’s payments to farmers. Although the AAA payments
were supposed to go to farm owners, croppers, and tenants, Warren Whatley
(1983) has suggested that the owners relegated a number of tenants and croppers
NEW DEAL AND INFANT MORTALITY
117
to wage labor status, as they sought to obtain a larger share of the AAA payments
for themselves. Thus, even though the AAA brought more income into the
counties, there is the possibility that the AAA may have engendered an income
redistribution that left croppers and tenants in a worse income position than
before. The results here are consistent with this hypothesis, particularly in the
1933–1935 period. In this period, AAA spending was associated with higher
infant mortality for both blacks and whites. After the AAA was reorganized in
response to the Supreme Court decision declaring the program unconstitutional,
the effects on infant mortality are no longer statistically significant in either the
black or white regressions. The absence of the effect in the 1936 –1939 period
may have been the result of a change in the attentiveness of the AAA to
protecting tenant rights. Whatley (1983, p. 927) claims that by 1936 and 1938 the
share of benefits going to tenants had increased, thus lower-income groups might
have shared in the benefits of the AAA, which in turn would have meant the
AAA would have helped contribute to lowering infant mortality. Another possible interpretation of the results is that the damage to croppers and tenants had
already been done in the earlier period, so the AAA neither degraded the situation
further nor made it better.
In comparisons of the cross-sectional regressions and the difference regressions, we see some substantial swings in the estimated coefficients on the New
Deal programs. The most striking are the reversal of the effect of the CWA from
lowering infant mortality to raising it and the switch from a strong lowering
effect of the PWA/USHA housing projects to a statistically insignificant effect.
There was also a switch in sign for the ADC coefficient in the black regressions,
although none of the coefficients in the tables were statistically significant. We
believe that the changes in the results as we move from the cross-sectional to the
difference analysis are the result of inadequate accounting for unmeasured
heterogeneity across counties in the cross-sectional regressions.
So what was the source of this heterogeneity? There appears to have been a
negative omitted-variables bias. We are looking for unmeasured factors that
would have been negatively associated with lower infant mortality but at the
same time positively correlated with per capita spending on the CWA, the ADC,
and PWA/USHA. We can only speculate, but one possible factor is the variation
in the strength of local relief networks. Williams (1968) and Howard (1943)
argue that there was substantial variation in the strength and extent of local relief
networks. The networks were an agglomeration of people, churches, private
groups, and local government programs. The stronger and more established local
relief networks were in a better position to seek out families and help them obtain
access to benefits. Such help would be extremely important in notifying families
of the possibilities of new programs like the ADC program. These networks also
played an important role in creating and lobbying for projects to attract more
New Deal funds. These more established and successful groups might have
already had more success in combating problems that led to infant mortality and
at the same time may have had more success in helping families obtain funds and
118
FISHBACK, HAINES, AND KANTOR
in attracting New Deal projects. The combination potentially creates the negative
correlation with infant mortality and the positive association with New Deal
funds that led to the negative bias in the cross-sectional regressions.
The FERA situation is one where there was a swing from essentially no effect
on infant mortality to a negative and statistically significant one. This swing
suggests a positive bias in the FERA coefficient that might have been driven by
unmeasured factors that are positively associated with the FERA as well as with
infant mortality. Given that the FERA spending included direct relief to families
with “unemployable” household heads, it is possible that FERA spending was
targeted at long-term pockets of poverty. Such pockets of poverty might have
gone unmeasured by an average income proxy like retail sales per capita, which
captures the average in the county but not a highly unequal distribution of income
that might be associated with a significant group with high incomes and another
group with low incomes. Thus, the combination of a long-term pocket of poverty
attracting FERA funds and the association of higher infant mortality with such
poverty in the county might have led to a positive bias in the FERA coefficient
in the cross-sectional regressions.
Although the coefficients for blacks and whites have been discussed separately, one of the issues we would like to address is the difference in access
to New Deal funds for blacks and whites. Our metric here is the comparisons
of how black and white infant mortality responded to New Deal programs.
The far right columns of Tables 6 and 7 report the results of estimating a
difference-in-difference analysis for blacks and whites. The dependent variable is the difference in black infant mortality rates minus the difference in white
infant mortality rates. The advantage of the difference-in-difference analysis is
that the t-tests on the per capita New Deal spending coefficients are direct
statistical tests of the hypothesis that the impact on black and white infant
mortality was the same. The results suggest that all but two of the programs were
at worst nondiscriminatory and sometimes even benefited black families more
than white families, at least from the perspective of infant mortality. The FERA
and the PWA appear to have reduced infant mortality more for blacks than for
whites. We cannot reject the hypothesis of no difference in the effects of the
AAA and the WPA. The CWA was associated with even higher infant mortality
for blacks than for whites, although it was unexpected to see a positive association for both.
The difference in the effect of ADC was statistically significant and more
positive for blacks than for whites. On the other hand, the ADC coefficient in the
black–white difference equation in Table 6 is positive and statistically significant.
This finding is consistent with Robert Lieberman’s (1998, pp. 118 –120) and
Richard Sterner’s (1943, pp. 281–285) findings that blacks had less access to
ADC benefits than did poor whites and that blacks received lower amounts per
child on average. Yet this is an odd situation because whites appear to have
received a greater share of the ADC benefits, yet the infant mortality rates of
NEW DEAL AND INFANT MORTALITY
119
whites were not reduced much if at all by additional spending on aid to dependent
children.
CONCLUSIONS
During the worst depression in American economic history, black and white
infant mortality rates continued to decline except for a slight blip upward in
1933–1934. Normally, we might anticipate that such a large economic dislocation that lasted a decade would have manifested itself in a more substantial rise
in infant mortality rates. The results of our analysis suggest that the New Deal
had a mixed effect on the secular decline in infant mortality rates. The combination of direct relief and work relief administered by the FERA was associated
with lower infant mortality. The PWA programs appear to have lowered infant
mortality through the building of large numbers of sewage facilities throughout
the United States and by providing extra employment in communities. Not all of
the relief programs were associated with lower infant mortality, however. The
focus of the CWA and the WPA on “employables” may have limited the impact
of work relief on infant mortality. Neither Aid to Dependent Children nor the
PWA Housing projects had any statistically measurable effect.
On the other hand, the AAA, which put substantial amounts of income into the
hands of farmers, was associated with higher infant mortality. The incentives
established by the AAA, which led landowners to transfer the status of tenants
and croppers to wage workers, may have led to a decline in income for some
families that increased the likelihood of infant mortality.
It is important to note, however, that over the course of the decade infant
mortality continued to follow a long-term trend that had been established in
earlier decades. Despite the small jumps upward in 1933 and 1934, the decline
in infant mortality proceeded at a fairly rapid pace through the end of the decade.
Thus, the specific New Deal programs that had had some success in contributing
to the reductions in infant mortality were only a subset of the factors that explain
the continued decline.
Several New Deal agencies claimed that they sought to avoid racial discrimination in the provision of employment and benefits. On the other hand, scholars
such as Joseph Lieberman (1998) have argued that some programs, largely run
by administrators from the local community, appear to have been actively
discriminating against blacks. Our metric for testing the extent of discrimination
was to determine the relative impact of the programs on infant mortality for
blacks and whites. Many of the programs appear to have been nondiscriminatory.
Some, like the FERA’s combination of direct and work relief and the PWA’s
sanitation works and other public works, even lowered black infant mortality
more than they lowered white infant mortality. On the other hand, the CWA work
relief programs and the relief to children under the Social Security Administration’s ADC program seem to have been associated with relatively higher infant
mortality for blacks and whites. In the final analysis we cannot say that the New
Deal as a whole was discriminatory or nondiscriminatory. Any effects describing
120
FISHBACK, HAINES, AND KANTOR
the aggregate New Deal would cover the differences that we see program by
program.
REFERENCES
Anderson, G. M., and Tollison, R. D. (1991), “Congressional Influence and Patterns of New Deal
Spending, 1933–1939.” Journal of Law and Economics 34, 161–175.
Antonovsky, A., and Bernstein, J. (1977), “Social Class and Infant Mortality,” Social Science and
Medicine 11, 453– 470.
Bremner, R. H. (Ed.) (1971), Child and Youth in America: A Documentary History. Volume II:
1866 –1932. Cambridge, MA: Harvard Univ. Press.
Carson, C. L. (1994), “And the Results Showed Promise . . . Physicians, Childbirth, and Southern
Black Migrant Women, 1916 –1930; Pittsburgh as a Case Study,” Journal of American Ethnic
History 14, 32– 64.
Clifford, W. B., and Brannon, Y. S. (1978), “Socioeconomic Differentials in Infant Mortality: An
Analysis Over Time,” Public Use Data 6, 29 –37.
Couch, J. F., and Shughart, W. II (1998), The Political Economy of the New Deal. New York: Edward
Elgar.
Easterlin, R. A. (1996), Growth Triumphant: The Twenty-First Century in Historical Perspective.
Ann Arbor, MI: Univ. of Michigan Press.
Federal Works Agency (1940), First Annual Report, 1940. Washington, DC: U.S. Govt. Printing
Office.
Fleck, R. K. (1994), “Essays on the Political Economy of the New Deal.” Unpublished Ph.D. thesis,
Stanford University.
Fordyce, E. J. (1976/1977), “Early Mortality Measures as Indicators of Socio-Economic Well-Being
for Whites and Non-Whites: A Re-Appraisal.” Sociology and Social Research 61, 125–137.
Haines, M. R. (1998a), “Estimated Life Tables for the United States, 1850 –1910.” Historical
Methods 31, 149 –169.
Haines, M. R. (1998b), “The Relationship between Infant and Child Mortality and Fertility: Some
Historical and Contemporary Evidence for the United States.” In Mark R. Montgomery and
Barney Cohen (Eds.), From Death to Birth: Mortality Decline and Reproductive Change.
Washington, DC: National Academy Press, 227–253.
Haines, M. R. (2000), “The Population of the United States, 1790 –1920.” In Stanley Engerman and
Robert Gallman, (Eds.), The Cambridge Economic History of the United States, Volume II. New
York: Cambridge Univ. Press, Chap. 4.
Higgs, R. (1977), Competition and Coercion. New York: Cambridge Univ. Press.
Howard, D. S. (1943), The WPA and Federal Relief Policy. New York: Russell Sage Foundation.
ICPSR (Inter-University Consortium for Political and Social Research), “Historical, Demographic,
Economic, and Social Data: The United States, 1790 –1970,” tape 0003, as corrected by Michael
Haines, Department of Economics, Colgate University, Hamilton, NY.
ICPSR, “United States Historical Election Returns, 1824 –1968,” tape 0001.
Kousser, J. M. (1974), The Shaping of Southern Politics: Suffrage Restriction and the Establishment
of the One-Party South, 1880 –1910. New Haven, CT: Yale Univ. Press.
Lieberman, R. C. (1998), Shifting the Color Line: Race and the American Welfare State. Cambridge,
MA: Harvard Univ. Press.
Lindenmeyer, K. (1997), A Right to Childhood: The U.S. Children’s Bureau and Child Welfare,
1912–1946. Urbana, IL: Univ. of Illinois Press.
Linder, F. E., and Grove, R. D. (1947), Vital Statistics Rates in the United States, 1900 –1940.
Washington, DC: U.S. Govt. Printing Office. Tables 26 and 28.
Margo, R. A. (1990), Race and Schooling in the South, 1890 –1940. Chicago: Univ. of Chicago Press.
Meckel, R. A. (1990), “Save the Babies”: American Public Health Reform and the Prevention of
Infant Mortality, 1850 –1929. Baltimore: Johns Hopkins Univ. Press.
NEW DEAL AND INFANT MORTALITY
121
Montgomery, M. R., and Cohen, B. (Eds.) (1998), From Death to Birth: Mortality Decline and
Reproductive Change. Washington, DC: National Academy Press.
National Center for Health Statistics (1996), Vital Statistics of the United States, 1992, Vol. II,
“Mortality,” Part A. Washington, DC: Public Health Service. Table 2-2.
National Center for Health Statistics (1999), “Deaths: Final Data for 1997.” National Vital Statistics
Reports, 47,(19) (June 30).
Preston, S. H., and Haines, M. R. (1991), Fatal Years: Child Mortality in Late Nineteenth Century
America. Princeton, NJ: Princeton Univ. Press.
Public Works Administration (1939), America Builds: The Record of the PWA. Washington, DC:
U.S. Govt. Printing Office.
Ransom, R., and Sutch, R. (1977), One Kind of Freedom: The Economic Consequences of Emancipation. New York: Cambridge Univ. Press.
Steckel, R. H. (1986), “A Peculiar Population: The Nutrition, Health, and Mortality of American
Slaves from Childhood to Maturity.” Journal of Economic History 46, 721–741.
Sterner, R. (1943), The Negro’s Share: A Study of Income, Consumption, Housing, and Public
Assistance. New York: Harper and Brothers.
Troesken, W. (1999), “Race, Disease, and the Provision of Water in American Cities, 1889 –1921.”
Unpublished working paper, Department of History, Univ. of Pittsburgh.
U.S. Bureau of the Census (1934a), Birth, Stillbirth, and Infant Mortality Statistics for the Birth
Registration Area of the United States, 1930, 16th Annual Report. Washington, DC: U.S. Govt.
Printing Office.
U.S. Bureau of the Census (1934b), Birth, Stillbirth, and Infant Mortality Statistics for the Birth
Registration Area of the United States, 1931, 17th Annual Report. Washington, DC: U.S. Govt.
Printing Office.
U.S. Bureau of the Census (1934c), Birth, Stillbirth, and Infant Mortality Statistics for the Birth
Registration Area of the United States, 1932, 18th Annual Report. Washington, DC: U.S. Govt.
Printing Office.
U.S. Bureau of the Census (1936a), Birth, Stillbirth, and Infant Mortality Statistics for the Continental United States, The Territory of Hawaii, and the Virgin Islands, 1933. 19th First Annual
Report. Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1936b), Birth, Stillbirth, and Infant Mortality Statistics for the Continental United States, The Territory of Hawaii, and the Virgin Islands, 1934. 20th Annual Report.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1937), Birth, Stillbirth, and Infant Mortality Statistics for the Continental
United States, The Territory of Hawaii, and the Virgin Islands, 1935, 21st Annual Report.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1938), Birth, Stillbirth, and Infant Mortality Statistics for the Continental
United States, The Territory of Hawaii, and the Virgin Islands, 1936, 22nd Annual Report.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1939), Birth, Stillbirth, and Infant Mortality Statistics for the Continental
United States, The Territory of Hawaii, and the Virgin Islands, 1937, 23rd Annual Report.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1940), Birth, Stillbirth, and Infant Mortality Statistics for the Continental
United States, The Territory of Hawaii, and the Virgin Islands, 1938, 24th Annual Report.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1941), Birth, Stillbirth, and Infant Mortality Statistics for the Continental
United States, The Territory of Hawaii, and the Virgin Islands, 1939, 25th Annual Report.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1942), Birth, Stillbirth, and Infant Mortality Statistics for the Continental
United States, The Territory of Hawaii, and the Virgin Islands, 1940, 26th Annual Report.
Washington, DC: U.S. Govt. Printing Office.
122
FISHBACK, HAINES, AND KANTOR
U.S. Bureau of the Census (1975), Historical Statistics of the United States: Colonial Times to 1970.
Washington, DC: U.S. Govt. Printing Office.
U.S. Bureau of the Census (1948), “Illiteracy in the United States, October 1947,” Current Population
Reports: Population Characteristics, Series P-20, no. 20. Washington, DC: U.S. Bureau of the
Census.
U.S. Children’s Bureau (1930), “Eighteenth Annual Report of the Children’s Bureau.” Washington,
DC: U.S. Bureau of the Census.
U.S. Department of Commerce, Bureau of Foreign and Domestic Commerce (1936), Consumer
Market Data Handbook, 1936. Washington, DC: U.S. Govt. Printing Office.
U.S. Department of Commerce, Bureau of Foreign and Domestic Commerce (1939), Consumer
Market Data Handbook, 1939. Washington, DC: U.S. Govt. Printing Office.
U.S. Office of Government Reports, Statistical Section (1940), Report No. 10, Volume 1, “County
Reports of Estimated Federal Expenditures March 4, 1933–June 30, 1939.” [Mimeo].
Waldmann, R. J. (1992), “Income Distribution and Infant Mortality.” Quarterly Journal of Economics 107, 1284 –1302.
Wallis, J. J. (1984), “The Birth of the Old Federalism: Financing the New Deal, 1932–1940.” Journal
of Economic History 44, 139 –159.
Wallis, J. J. (1998), “The Political Economy of New Deal Spending Revisited, Again: With and
Without Nevada.” Explorations in Economic History 35, 140 –170.
Whatley, W. (1983), “Labor for the Picking: The New Deal in the South.” Journal of Economic
History 43, 905–929.
White, H. (1980), “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test
for Heteroskedasticity.” Econometrica 48, 817– 838.
Williams, E. A. (1968), Federal Aid for Relief. New York: AMS Press. [Reprint]
Woodbury, R. M. (1926), Infant Mortality and Its Causes. Baltimore: Williams and Wilkins Co.
Wright, G. (1974), “The Political Economy of New Deal Spending: An Econometric Analysis.”
Review of Economics and Statistics 56, 30 –38.