Labour force participation of
married women:
The United States, 1860-2010
Richard Zijdeman (IISH)
Valencia, Spain
Aula 5, NIVEL 0
March 31, 2016
H-7 The causes and consequences of women’s empowerment
Introduction
Post WW II research shows major increase in
FLFP
• So, when did it start?
• How did this change occur?
– many hypotheses on change in FLFP
• Level of economic development (GDP)
• Reputation (social status)
FLFP = Female Labour Force participation
U-shape female labour force function
• U-shaped relation between country’s level of
development and FLFP:
– Higher at lower and higher levels of development
– Lower at mediocre levels of development
Left side of U-shape
• Rise in income, due to expanded markets or
introduction of new technology
– barriers preventing women (social custom,
employer preference)
• Reduction in the relative price of home
produced goods
• Decrease in the demand for women in
agriculture
Center of U-shape
• No explicit arguments (for U vs. V-shape)
• U-shape maybe explained by:
– regional dispersion of e.g. technology
– slow change in social behaviour
Right side of U-shape
• Improvement of women’s education,
particularly higher education
• Improvement of women’s wages
More in-depth on reputation
Formal barriers:
- e.g. marriage bars
Informal barriers:
– Employer preference
– Social norms or stigmas
Within-family-competition
Within-family-competition
– Disruptive rivalry between partners (Parsons ’49,
’54, also see Oppenheimer ’77)
– The higher the husband’s status, the bigger the
range of non-rivalrous jobs (lower and mediocre)
Ergo: the higher a husband’s occupational status,
the higher the probability of FLFP
Between-family-competition
• Competition between families, NOT within
families
– Reduce risk of economic hardship (two earners)
– Enhance socio-economic position
• But 19th century: few higher occupational
positions for women, so women more likely to
work when married to lower status husband
Ergo: the lower a husband’s occupational status, the higher
the probability of FLFP
What this papers adds
• Increased time period at both ends
• Test of theories at individual level…
• Taking regional (state) variation into account
• Census data: comparability of different age
groups and characteristics
Data
• IPUMS USA census data 1860-2000
– 1, 5 or 10 per cent samples
– 1970 excluded (for now)
• 2010 + 2013: American Community Survey
• married women whose husband is in the
household at time of the census
• N = 11,773,133
• NHGIS: for total population at state level
• GDP in GK dollars from CLIO-INFRA
Key variables
Micro (individual):
• Status husband (Duncan SEI)
• Family size
• # children under age 5
Macro (state by census year):
• Proportion of couples living at a farm
• Population per million
• Proportion in education (5-16)
• Proportion in education (16-20)
Methods
• Hierarchical generalized linear model (binomial)
– Nested observations
– Clustering of observations within states and time
• LME4 package in R
Descriptive results
• Regional variation in FLFP
• U-shaped curve between GDP and FLFP?
Summary of regional descriptives
• From ‘random’ (1860 – 1880)
• To horse shoe (1900 – 1930)
• To coasts (1940 – 1960)
• To Great Lakes (1980-2000)
• To ‘random’ (2010)?
U-shaped?
Explanatory results
Model with just time and cubic time effect:
• Non-linear effect indeed
• Bottom of U at 1820, not 1920 (Goldin 1994)
Explanatory results
Random effects:
Variance: 0.2815 Std.Dev.
Std.Dev: 0.5305
Number of obs: 11773133, groups: stime, 655
Estimate Std. Error z value
(Intercept) -2.293e+00 -2.293e+00 -75.3
age(center) -5.057e-02 -5.057e-02 -778.6
SEI husband (center) 1.964e-03 1.964e-03 66.9
family size -5.301e-02 -5.301e-02 -103.7
# children <5 -8.112e-01 -8.112e-01 -579.8
decades since 1800 2.374e-01 2.374e-01 28.8
(dec since 1800)^2 3.857e-03 3.857e-03 9.1
population (millions) -1.236e-02 -1.236e-02 -2.3
prop. living at farm -1.131e+00 -1.131e+00 -31.0
prop. fem.at.school (6-15) -4.032e+00 -4.032e+00 -140.1
prop.fem.at.school (16-20) 1.849e+00 1.849e+00 68.4
AIC BIC logLik deviance df.resid
11742052 11742224 -5871014 11742028 11773121
Conclusions
• On national level no evidence for U-shape
• Mechanisms underlying the U-shape appear
to be correct though:
– Inverse relation between FLFP and agriculture
– Increased FLFP with higher secondary education
• but: ‘white collar work’ or ‘cultural indicator’
– Inconclusive results for within or between family
status hypotheses
Caveats
• Different definitions of and instructions on
‘being in the labor force’ over time
– starting age
– e.g. 1910 census data
• So far rather imprecise measures:
– e.g. no sectorial information used
• No information on income -> SSHA 2016

Labour force participation of married women, US 1860-2010

  • 1.
    Labour force participationof married women: The United States, 1860-2010 Richard Zijdeman (IISH) Valencia, Spain Aula 5, NIVEL 0 March 31, 2016 H-7 The causes and consequences of women’s empowerment
  • 2.
    Introduction Post WW IIresearch shows major increase in FLFP • So, when did it start? • How did this change occur? – many hypotheses on change in FLFP • Level of economic development (GDP) • Reputation (social status) FLFP = Female Labour Force participation
  • 3.
    U-shape female labourforce function • U-shaped relation between country’s level of development and FLFP: – Higher at lower and higher levels of development – Lower at mediocre levels of development
  • 4.
    Left side ofU-shape • Rise in income, due to expanded markets or introduction of new technology – barriers preventing women (social custom, employer preference) • Reduction in the relative price of home produced goods • Decrease in the demand for women in agriculture
  • 5.
    Center of U-shape •No explicit arguments (for U vs. V-shape) • U-shape maybe explained by: – regional dispersion of e.g. technology – slow change in social behaviour
  • 6.
    Right side ofU-shape • Improvement of women’s education, particularly higher education • Improvement of women’s wages
  • 7.
    More in-depth onreputation Formal barriers: - e.g. marriage bars Informal barriers: – Employer preference – Social norms or stigmas
  • 8.
    Within-family-competition Within-family-competition – Disruptive rivalrybetween partners (Parsons ’49, ’54, also see Oppenheimer ’77) – The higher the husband’s status, the bigger the range of non-rivalrous jobs (lower and mediocre) Ergo: the higher a husband’s occupational status, the higher the probability of FLFP
  • 9.
    Between-family-competition • Competition betweenfamilies, NOT within families – Reduce risk of economic hardship (two earners) – Enhance socio-economic position • But 19th century: few higher occupational positions for women, so women more likely to work when married to lower status husband Ergo: the lower a husband’s occupational status, the higher the probability of FLFP
  • 10.
    What this papersadds • Increased time period at both ends • Test of theories at individual level… • Taking regional (state) variation into account • Census data: comparability of different age groups and characteristics
  • 11.
    Data • IPUMS USAcensus data 1860-2000 – 1, 5 or 10 per cent samples – 1970 excluded (for now) • 2010 + 2013: American Community Survey • married women whose husband is in the household at time of the census • N = 11,773,133 • NHGIS: for total population at state level • GDP in GK dollars from CLIO-INFRA
  • 12.
    Key variables Micro (individual): •Status husband (Duncan SEI) • Family size • # children under age 5 Macro (state by census year): • Proportion of couples living at a farm • Population per million • Proportion in education (5-16) • Proportion in education (16-20)
  • 13.
    Methods • Hierarchical generalizedlinear model (binomial) – Nested observations – Clustering of observations within states and time • LME4 package in R
  • 14.
    Descriptive results • Regionalvariation in FLFP • U-shaped curve between GDP and FLFP?
  • 28.
    Summary of regionaldescriptives • From ‘random’ (1860 – 1880) • To horse shoe (1900 – 1930) • To coasts (1940 – 1960) • To Great Lakes (1980-2000) • To ‘random’ (2010)?
  • 29.
  • 30.
    Explanatory results Model withjust time and cubic time effect: • Non-linear effect indeed • Bottom of U at 1820, not 1920 (Goldin 1994)
  • 31.
    Explanatory results Random effects: Variance:0.2815 Std.Dev. Std.Dev: 0.5305 Number of obs: 11773133, groups: stime, 655 Estimate Std. Error z value (Intercept) -2.293e+00 -2.293e+00 -75.3 age(center) -5.057e-02 -5.057e-02 -778.6 SEI husband (center) 1.964e-03 1.964e-03 66.9 family size -5.301e-02 -5.301e-02 -103.7 # children <5 -8.112e-01 -8.112e-01 -579.8 decades since 1800 2.374e-01 2.374e-01 28.8 (dec since 1800)^2 3.857e-03 3.857e-03 9.1 population (millions) -1.236e-02 -1.236e-02 -2.3 prop. living at farm -1.131e+00 -1.131e+00 -31.0 prop. fem.at.school (6-15) -4.032e+00 -4.032e+00 -140.1 prop.fem.at.school (16-20) 1.849e+00 1.849e+00 68.4 AIC BIC logLik deviance df.resid 11742052 11742224 -5871014 11742028 11773121
  • 32.
    Conclusions • On nationallevel no evidence for U-shape • Mechanisms underlying the U-shape appear to be correct though: – Inverse relation between FLFP and agriculture – Increased FLFP with higher secondary education • but: ‘white collar work’ or ‘cultural indicator’ – Inconclusive results for within or between family status hypotheses
  • 33.
    Caveats • Different definitionsof and instructions on ‘being in the labor force’ over time – starting age – e.g. 1910 census data • So far rather imprecise measures: – e.g. no sectorial information used • No information on income -> SSHA 2016

Editor's Notes

  • #28 Range from yellow = .64 to upper red = .76