The Kindergarten Movement and the US Demographic Transition
Philipp Ager
Francesco Cinnirella
Peter Sandholt Jensen
Economic History Seminar at UC Davis
September 20th, 2016
1 / 36
.9
.8
.7
School Attendance Rate
1
Fertility and Schooling in U.S. Cities
.6
Children under 5 years old per 1,000 Women 15 to 49
300
400
500
600
700
200
The Demographic Transition and Schooling in US Cities
1840
1860
1880
1900
Census year
Fertility
1920
1940
Schooling
2 / 36
Fertility and Early Childhood Education
I
Negative correlation between education and fertility
I
What is the role of early childhood education for the
demographic transition?
I
I
Quantity-Quality trade-off or alternative theories
I
Long-term consequences (lower fertility and higher income
levels)
The kindergarten was transplanted to America by German
migrants during the second half of the 19th century
3 / 36
Questions
I
I
I
How does the spread of the kindergarten affect fertility?
I
Cost of education
I
Cost of raising children
I
Labor market opportunities
Who was affected by the establishment of kindergartens?
I
Fecund mothers
I
Children age 4 to 6
Role of German communities for kindergarten diffusion
4 / 36
Main findings
I
Kindergarten diffusion lead to an overall fertility decline
I
Mothers of 4-6-year old that attended kindergarten had less
children
I
German communities play important role for kindergarten
diffusion and contribute to the fertility decline in US cities
I
Higher school attendance rates of 4-6-year old children
exposed to kindergartens
I
Children exposed to kindergartens have higher earnings and
school attainment later in life
5 / 36
Related Theories
I
Kindergartens as investment in quality of children
Kindergartens lower cost of education ⇒ quantity-quality
trade-off mechanism (Becker and Lewis, 1973; Galor and
Weil, 1999; Galor and Moav, 2006)
I
Standard theory (Becker and Lewis, 1973)
Public child care ⇒ mothers return to work sooner after
childbirth ⇒ lower opportunity costs of having children ⇒
increase in fertility
I
Kindergartens affect female labor supply (Gelbach, 2002)
Enter the workforce ⇒ relative wage of women increases ⇒
fertility drops (Galor and Weil, 1996)
6 / 36
Related Literature
I
US Demographic Transition: Easterlin (1971), Haines
(1979), Greenwood and Seshadri (2002), Jones and Tertilt
(2008)
I
Quantity-Quality Trade-off in the US: Bleakley and Lange
(2009), Aaronson et al. (2014)
I
The Role of Kindergartens in the US: Cascio (2009),
Heckman et al. (2010), Haimovich (2014)
7 / 36
History of Kindergarten in the United States (1)
German-American Interaction (Allen, 1997)
I
Kindergarten a German “cultural institution” (Froebel)
I
M. Schurz: first US Kindergarten in 1855 (Watertown, WI)
I
E. Peabody: first English speaking kindergarten in 1860
(Boston, MA)
I
S. Blow: First public kindergarten in 1873 (St. Louis, MO)
I
Peabody facilitated the spread of Kindergartens to a broader
English-speaking constituency
I
Peabody trained with female German migrants the first
generation of American (mainly female) kindergarten teachers
8 / 36
History of Kindergarten in the United States (2)
I
Mission: “the public school kindergarten stands as a
mediating element, in which it is sought to provide for the
children of the people the best kind of nurturing and scientific
care, to give them the best kind of physical, mental, social,
and spiritual training”
Take-away:
⇒ Kindergarten very popular among educated, middle-class
American women
⇒ German influences important in each stage of the
Kindergarten Movement: “transplanted institution”
9 / 36
History of Kindergarten in the United States (3)
Figure: Susan Blow (Source: The State Historical Society of Missouri)
10 / 36
Kindergartens in the Q-Q Trade-off Framework
Returns to Education/Cost of Schooling:
I
Ellen M. Quigley, Troy, NY: [. . . ] “little children who have had
the great privilege of being trained in a kindergarten by a
skillful, enthusiastic kindergartner have many advantages over
those who come from even the best homes directly to first
grade (Bureau of Education, Bulletin 1914, 6, p.107)”
I
Nellie Walton Ford, St. Paul, MN: “Kindergarten children
possess greater self-control, are more mature, less timid, pay
better attention, take commands more intelligently, do better
handwork, and have a better vocabulary (Bureau of
Education, Bulletin 1914, 6, p.112)”
I
Survey of school teachers (Palmer, 1915): Overall
kindergarten children have better school habits, are more
fluent in language and are better to work with others
11 / 36
Division of Time in Kindergartens
Figure: General Practice in Kindergarten Education in the United States,
Davis (1925, p.28)
12 / 36
National Kindergarten Movement 1890-1910
Figure: National Entrollment in Kindergartens (Haimovich, 2014, mimeo)
13 / 36
Enrollment Rates by State in 1912
Figure: Entrollment Rates by State 1912
Source: Bureau of Education, Bulletin, 1914, No. 6, p.12
14 / 36
Relevance of Kindergarten Education
I
Almost no kindergarten enrollment of children age 4-6 in 1880,
around 5% in 1902, increased to about 9 percent in 1912
I
In 1912, 87% were public kindergartens, 85% of the enrolled
children attended public kindergartens
I
Large variation across states in 1912: in New Jersey 28% of
children between 4-6 years were enrolled in kindergarten; in
West Virginia only 0.1%
Questions:
1. How does the establishment of kindergartens affect fertility?
2. Is there a fertility reduction due to higher returns to education
as sketched in the historical narrative (Q-Q trade-off)?
15 / 36
Kindergarten: Data Collection (1)
I
We have digitized data on Kindergarten from 1880 to 1910
mainly from statistical reports commissioned by the Bureau of
Education
I
Information digitized: name, location (municipality), county,
state, year of establishment, number of pupils, type
(public/private)
I
Match data on kindergarten with city code from IPUMS
I
I
I
1880: 10%
1900: 5%
1910: 1%
⇒ In total we have about 104,000 women, age 20-49 listed as
spouse or household head at the time of the Census.
16 / 36
Kindergarten: Data Collection (2)
Figure: Enrollment Rates by State 1912
Source: Bureau of Education, Bulletin, 1914, No. 6, p.24
17 / 36
Map of Kindergartens by US Cities in 1910
18 / 36
Outcome Variables
Fertility Measures:
I
Children less than 5 years old
Education Data:
I
School attendance 4 to 17 years old
I
Educational attainment from Iowa 1915 (Goldin and Katz,
2010)
I
Returns to kindergarten education (Goldin and Katz, 2010)
19 / 36
Estimation Strategy
Fertilityict = ac + bt + δKGct + Controlsict + ict
1. Fertilityict : individual fertility measures
2. KGct : share of kindergartens per capita
3. Controlsict : individual controls (race, quadratic in age, white,
birthplace, marital status, literacy)
4. City fixed effects (ac )
5. Time fixed effects (bt )
20 / 36
The Diffusion of Kindergartens and Fertility
Fertility and Age in US Cities
Number of children below age 5
.4
0
.2
.6
.8
1880--1910
20
30
40
50
age
Kindergarten Dummy =0
Kindergarten Dummy =1
21 / 36
The Diffusion of Kindergartens and Fertility (1880-1910)
Dependent Variable: Number of Children below Age 5
Whole Sample
Share Kindergarten
Kindergartens Established 10 Years Earlier Late Kindergarten Adopters 10 Years Earlier
-0.996**
(0.477)
1.445
(0.902)
1.132
(1.212)
Share Kindergartens x Age 20-24
-0.651
(0.650)
-0.107
(1.036)
0.659
(1.330)
Share Kindergartens x Age 25-29
-1.760***
(0.634)
-0.562
(0.869)
0.627
(1.113)
Share Kindergartens x Age 30-34
-0.730
(0.679)
0.815
(0.721)
1.551*
(0.919)
Share Kindergartens x Age 35-39
-1.007*
(0.565)
-0.290
(0.666)
-1.261
(0.837)
Share Kindergartens x Age 40-44
-0.384
(0.490)
-0.438
(0.583)
-0.455
(0.903)
City FE
Time FE
City x Time FE
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Observations
104,188
104,188
51,514
51,514
15,670
15,670
R-squared
0.192
0.198
0.190
0.195
0.198
0.208
The sample spans females (household head or spouse) age 20 to 49. Individual controls are dummies for being literate, white, married, birthplace, and a quadratic in age, and
dummies for age cohorts. Robust standard errors clustered at the city level in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
22 / 36
Wrap-up Results
I
On average, moving from zero kindergartens to the mean (to
the 95th percentile) would decrease the number of children
below 5 by around 2% (6%)
I
Interacting the share of kindergartens with the different age
cohorts reveals that the most significant decline in fertility is
for the cohort age 25 to 29
I
Placebo tests do not reveal any significant fertility pre-trends
for kindergarten cities
23 / 36
The Effect on the Treated
I
We can differentiate mothers with 4-6-year old children that
indicated school attendance in the Census
I
We can compare them to mothers with 4-6-year old children
living in cities with no kindergarten (control group 1)
I
We can compare them to mothers with 4-6-year old children
living in the same city (control group 2)
24 / 36
The Effect of Kindergartens on the Treated (1880-1910)
Dependent Variable:
==1 if child < 4yrs
== 1 if 4-6 years old in Kindergarten
Sample
-0.258***
(0.00957)
Number of Children below Age 4
-0.556***
(0.0131)
Control Group 1 & 2 Control Group 1 & 2
-0.190***
(0.0403)
-0.556***
(0.0130)
-0.586***
(0.0300)
Control Group 1
Control Group 2
Grandparents in HH
Control Group 1 & 2
Observations
30,286
30,261
13,074
25,730
8,691
R-squared
0.158
0.216
0.189
0.209
0.280
City FE
Yes
Yes
Yes
Yes
Yes
Time FE
Yes
Yes
Yes
Yes
Yes
City x Time FE
No
Yes
No
Yes
Yes
The sample spans females (household head or spouse) age 20 to 49 with at least one 4 to 6 years old. Individual controls are dummies for being literate,
white, married, birthplace, and a quadratic in age, and dummies for age cohorts. Robust standard errors clustered at the city level in parentheses: ***
p<0.01, ** p<0.05, * p<0.1.
25 / 36
Wrap-up Results
I
Mothers that send their 4-6 years old to the kindergarten have
fewer additional children
I
Fertility decline also present in households with lower
opportunity costs of child rearing “(grandparents as nannies)”
26 / 36
The Role of German Communities
.2
.4
Kindergarten Dummy
.8
.6
1
1.2
Diffusion of Kindergartens and German Communities
0
.1
.2
.3
.4
German Community (Share) in 1880
.5
27 / 36
Diffusion of Kindergartens (1880--1910)
Dependent Variable: ==1 if Kindergarten
German Community 1880 x (== 1 if year == 1880)
0.590**
(0.241)
0.588**
(0.243)
0.557**
(0.245)
German Community 1880 x (== 1 if year == 1900)
0.294**
(0.116)
0.274**
(0.114)
0.263**
(0.121)
Irish Community 1880 x (== 1 if year == 1880)
-0.161
(0.241)
-0.135
(0.242)
Irish Community 1880 x (== 1 if year == 1900)
0.111
(0.134)
0.0265
(0.141)
Yes
Yes
Yes
Yes
Yes
Yes
City FE
Year FE
Controls
Yes
Yes
No
Yes
Yes
Yes
Observations
1,040
1,040
1,040
1,040
R-squared
0.767
0.770
0.767
0.770
Dependent variable equals one if city has kindergarten in year t . Controls (columns 2-4) include ln(city size), share of literate,
average family size, and the share of white collar workers in the labor force. Robust standard errors clustered at the city level in
parentheses: *** p<0.01, ** p<0.05, * p<0.1.
28 / 36
The Spread of Kindergartens, German Communities, and Fertility (1880--1910)
Number of Children below Age 5
== 1 if Kindergarten
-0.0111
(0.0137)
-0.00920
(0.0144)
== 1 if Kindergarten x German Community in 1880
-0.360***
(0.0912)
-0.346***
(0.0917)
== 1 if Kindergarten x Irish Community in 1880
0.0990
(0.0696)
Share Kindergartens
-1.358***
(0.455)
-1.382***
(0.471)
Share Kindergartens x GC in 1880
-6.774***
(2.312)
-7.211***
(2.245)
Share Kindergartens x Irish Community in 1880
City FE
Time FE
2.829*
(1.650)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Observations
104,188
104,188
104,188
104,188
R-squared
0.192
0.194
0.192
0.194
The sample spans females (household head or spouse) age 20 to 49. Individual controls are dummies for being literate, white, married, birthplace, German
ancestry, Irish ancestry (Columns 2 and 4), a quadratic in age, and dummies for age cohorts. Robust standard errors clustered at the city level in parentheses:
*** p<0.01, ** p<0.05, * p<0.1.
29 / 36
Wrap-up Results
I
German communities mattered for the diffusion of
kindergartens
I
Kindergartens mattered via German communities for the
fertility transition
30 / 36
School Attendance of the Relevant Group
Kindergartens and School Attendance Age 4 to 6
.2
School Attendance Age 4 to 6
.25
.3
.35
.4
1880--1910
-.02
0
.02
Share of Kindergartens
.04
.06
31 / 36
The Spread of Kindergartens and School Attendance (1880--1910)
Dependent Variable: School Attendance
Age 4-17
Age 4-6
Share Kindergartens x Age 4 to 6
Share Kindergartens
City FE
Time FE
City x Time FE
Age 4-17
1.572***
(0.464)
0.725***
(0.275)
1.977***
(0.476)
Yes
Yes
No
Yes
Yes
No
Yes
Yes
Yes
Observations
158,294
37,881
158,294
R-squared
0.415
0.354
0.424
The sample spans children age 4 to 17. Individual controls are dummies for being white,
birthplace, a quadratic in age, and dummies for age cohort 4 to 6 (Column 3). Robust standard
errors clustered at the city level in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
32 / 36
The Effect of Kindergarten Exposure on Educational Attainment and Earnings (Iowa 1915)
Dependent Variable:
Total Years of School Attended
ln (Earnings)
== 1 if Exposed
0.350***
(0.108)
0.354***
(0.117)
0.0593**
(0.0275)
0.0936***
(0.0304)
0.0609**
(0.0287)
==1 if US Born
2.095***
(0.134)
2.092***
(0.134)
0.261***
(0.0300)
0.259***
(0.0298)
0.0957***
(0.0276)
== 1 if White
2.851***
(0.251)
2.851***
(0.248)
0.475***
(0.0441)
0.474***
(0.0432)
0.259***
(0.0466)
==1 if Female
0.136**
(0.0644)
0.132**
(0.0644)
-0.529***
(0.0220)
-0.526***
(0.0220)
-0.567***
(0.0212)
Years of Schooling
City FE
Qudratic in Age
Age FE
0.0752***
(0.00343)
Yes
Yes
No
Yes
No
Yes
Yes
Yes
No
Yes
No
Yes
Yes
No
Yes
Observations
6,959
6,959
4,049
4,049
4,049
R-squared
0.075
0.078
0.310
0.319
0.408
The sample spans individuals that turned 6 in the 5 years before the first kindergarten establishment (preexposed) and individuals that turned 4 in the five years
after the first kindergarten establishment (exposed). Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
33 / 36
Evidence on the Q-Q: Wrap-up
I
We already showed a negative effect of the diffusion of
kindergartens on fertility
I
We find that the school attendance rate of 4 to 6 years old is
significantly higher in places with a higher kindergarten share
I
The Iowa 1915 state census data reveal that individuals
exposed to kindergartens in their early life had significant
gains in earnings and years of schooling compared to
individuals who were just too old to attend
34 / 36
VARIABLES
Share Kindergartens
Alternative Mechanisms
Dependent Variable:
==1 if works
-0.143
(0.152)
Share Kindergartens x GC in 1880
-0.0370
(0.187)
-0.743
(0.792)
== 1 if 4-6 years old in Kindergarten
City FE
Year FE
City x Time FE
Child Mortality
0.836
(0.807)
0.789
(1.042)
0.437
(5.363)
Yes
Yes
No
Yes
Yes
No
0.00344
(0.00592)
Yes
Yes
No
Yes
Yes
No
Yes
Yes
Yes
Observations
104,188
104,188
12,392
66,301
66,301
R-squared
0.287
0.287
0.289
0.127
0.127
The sample spans females (household head or spouse) age 20 to 49 for the years 1880 to 1910. Individual controls
are dummies for being literate, white, married, birthplace, German ancestry, and a quadratic in age, and dummies for
age cohorts. Column (3) considers only mothers with at least one 4-6 years old, but no younger children. For
Columns (4)-(5) the sample spans married females (household head or spouse) age 20 to 49 for the years 1900 and
1910. Robust standard errors clustered at the city level in parentheses: *** p<0.01, ** p<0.05, * p<0.1.
35 / 36
Conclusions
I
We showed that the diffusion of kindergartens in the United
States increased educational attainment and reduced fertility
I
Our evidence supports the quantity-quality trade-off as
mothers saw kindergartens as an opportunity to invest in the
quality of their children
I
The establishment of kindergartens seems not to have had a
significant impact on female labor supply and child mortality
during 1880 to 1910
I
Our results indicate that German cultural values related to
child education have diffused into American communities via
kindergartens
36 / 36