What drives female labor force participation? Comparable micro-level - - PowerPoint PPT Presentation

▶
what drives female labor force participation comparable
SMART_READER_LITE
LIVE PREVIEW

What drives female labor force participation? Comparable micro-level - - PowerPoint PPT Presentation

Oct 29, 2022 213 likes •608 views

What drives female labor force participation? Comparable micro-level evidence from eight developing and emerging economies Stephan Klasen 1 , 3 , Tu Thi Ngoc Le 1 , Janneke Pieters 2 , 3 , Manuel Santos Silva 1 1 University of G ottingen 2

slide-1
SLIDE 1

What drives female labor force participation? Comparable micro-level evidence from eight developing and emerging economies

Stephan Klasen1,3, Tu Thi Ngoc Le1, Janneke Pieters2,3, Manuel Santos Silva1

1University of G¨

  • ttingen

2Wageningen University 3IZA

WIDER Development Conference, Sept 11, 2019

slide-2
SLIDE 2

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Motivation

  • In the last two decades, in the developing world:

◮ rising female education, ◮ declining fertility, ◮ economic growth,

  • favorable background for rising FLFP rates everywhere.

Manuel Santos Silva FLFP: micro evidence 2 / 34

slide-3
SLIDE 3

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Female labor force participation rates, age 15+

Figure 1: Source: ILO, modeled estimates

Manuel Santos Silva FLFP: micro evidence 3 / 34

slide-4
SLIDE 4

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Puzzle

  • Klasen and Pieters (2015) on India: “Against this background,

it is puzzling to see that the reported female labor force participation rate in urban India has stagnated at around 18 percent since the 1980s.”

  • Schaner and Das (2016) on Indonesia: “Why, in the face of so

much change, has Indonesian women’s labor force participation remained so stagnant?”

  • Majbouri (2018) on MENA region: “Fertility and the Puzzle of

Female Employment in the Middle East”

  • Gasparini and Marchionni (2015): in LA, slowdown in the

growth of female labor supply since the 2000s

  • etc...

Manuel Santos Silva FLFP: micro evidence 4 / 34

slide-5
SLIDE 5

Motivation Methods & Data FLFP correlates Decompositions Conclusion

What we do

We use comparable microdata from 8 low and middle-income countries, covering the period 2000–2014, to ask:

1 How are women’s (and their households’) characteristics

associated with FLFP, and what are the key commonalities and differences across countries?

2 What drives FLFP changes over time within countries? 3 What explains differences in FLFP rates between countries and

how has this changed over time?

Manuel Santos Silva FLFP: micro evidence 5 / 34

slide-6
SLIDE 6

Motivation Methods & Data FLFP correlates Decompositions Conclusion

How we do it

1 We estimate FLFP models for each country and year, 2 We decompose changes in FLFP over time for each country, 3 We decompose gaps in FLFP between countries.

Manuel Santos Silva FLFP: micro evidence 6 / 34

slide-7
SLIDE 7

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Our contribution

  • richer data than in cross-country analyses → heterogeneity

across space and time,

  • unified empirical framework → direct comparison between

countries and over time,

  • robust FLFP correlates over large samples and several periods.

Manuel Santos Silva FLFP: micro evidence 7 / 34

slide-8
SLIDE 8

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Empirical model

  • We follow the specification of Klasen and Pieters (2015):
  • Population: married women of ages 25-54 living in urban areas.
  • Probit model:

P(LFPict = 1) = Φ

  • αct +
  • E

βE

ctDE ict + Xictγct + δrct

  • , (1)

Manuel Santos Silva FLFP: micro evidence 8 / 34

slide-9
SLIDE 9

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Explanatory variables

  • DE

ict: woman’s education attainment dummies.

  • Xict - individual and household level:

◮ age, age2, ◮ ethnic or religious group, ◮ per capita household income excluding the woman’s earnings (log), ◮ education attainment of household head, ◮ at least one male household member has wage employment (dummy), ◮ number of children 0–2, 3–5, boys 6–14, girls 6–14.

  • δrct - region fixed effects.

Manuel Santos Silva FLFP: micro evidence 9 / 34

slide-10
SLIDE 10

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Interpretation

  • reduced-form correlates,
  • not causal, not structural (no own-wage effects),
  • supply-side focus,
  • (local) demand conditions captured by regional fixed effects.

Manuel Santos Silva FLFP: micro evidence 10 / 34

slide-11
SLIDE 11

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Data

  • Large scale repeated cross-sectional surveys for:
  • Bolivia, Brazil, India, Indonesia, Jordan, South Africa,

Tanzania, Vietnam,

  • 32 surveys, ∼ 800,000 urban married women (prime-age),
  • Period: roughly 2000-2014.

Manuel Santos Silva FLFP: micro evidence 11 / 34

slide-12
SLIDE 12

Motivation Methods & Data FLFP correlates Decompositions Conclusion

FLFP (prime-age) vs. income, 2014

Bolivia Indonesia India Jordan Tanzania Vietnam South Africa Brazil 20 40 60 80 100 FLFP (%) 6 8 10 12 ln(GDP per capita)

Manuel Santos Silva FLFP: micro evidence 12 / 34

slide-13
SLIDE 13

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Result 1

  • No universal relationship between a woman’s education and her

LFP status:

  • strong, positive, and linear in Brazil and SA,
  • U- or J-shape in India, Indonesia, and Jordan,
  • Mixed in Bolivia, Tanzania, and Vietnam.

Manuel Santos Silva FLFP: micro evidence 13 / 34

slide-14
SLIDE 14

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effects of own education: Brazil

−.1 .1 .2 .3 .4 .5 Average marginal effect < Prim Elem (1−4) Elem (5−8) High school Tertiary 2002 95% CI (2002) 2013 95% CI (2013)

Manuel Santos Silva FLFP: micro evidence 14 / 34

slide-15
SLIDE 15

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effects of own education: India

−.1 .1 .2 .3 .4 .5 Average marginal effect Illiterate Literate Primary Middle school Secondary Tertiary 1999 95% CI (1999) 2011 95% CI (2011)

Manuel Santos Silva FLFP: micro evidence 15 / 34

slide-16
SLIDE 16

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effects of own education: Vietnam

−.1 .1 .2 .3 .4 .5 Average marginal effect < Prim Primary Secondary High school Tertiary 2002 95% CI (2002) 2014 95% CI (2014)

Manuel Santos Silva FLFP: micro evidence 16 / 34

slide-17
SLIDE 17

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Result 2

  • The negative effect of fertility is stronger in richer countries.

Manuel Santos Silva FLFP: micro evidence 17 / 34

slide-18
SLIDE 18

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effect of young children: Brazil

−.1 −.05 .03 Average marginal effect 2002 2005 2009 2013 Year Children 0−2 95% CI Children 3−5 95% CI Boys 6−14 95% CI Girls 6−14 95% CI Manuel Santos Silva FLFP: micro evidence 18 / 34

slide-19
SLIDE 19

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effect of young children: Tanzania

−.1 −.05 .03 Average marginal effect 2000 2006 2014 Year Children 0−2 95% CI Children 3−5 95% CI Boys 6−14 95% CI Girls 6−14 95% CI Manuel Santos Silva FLFP: micro evidence 19 / 34

slide-20
SLIDE 20

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Result 3

  • Household circumstances lose their grip on FLFP in richest

countries: Brazil and SA.

  • Negative household income effects very strong in India,

Indonesia, and Bolivia,

  • Same for household head education.

Manuel Santos Silva FLFP: micro evidence 20 / 34

slide-21
SLIDE 21

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effect of log income: Indonesia

−.08 −.06 −.04 −.02 .02 .04 Average marginal effect 2000 2004 2007 2014 Log income 95% CI

Figure 2: Notes: income is the household per capita earnings from main job excluding woman’s own earnings

Manuel Santos Silva FLFP: micro evidence 21 / 34

slide-22
SLIDE 22

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Average marginal effect of log income: South Africa

−.08 −.06 −.04 −.02 .02 .04 Average marginal effect 1995 2001 2003 2010 2014 Log income 95% CI

Figure 3: Notes: income is the household per capita earnings from main job excluding woman’s own earnings

Manuel Santos Silva FLFP: micro evidence 22 / 34

slide-23
SLIDE 23

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Robustness

Correlates are robust to:

  • PSU fixed effects (Brazil, Bolivia, SA, Tanzania),
  • trends in marriage rates and urbanization,
  • controlling for rural-urban migration directly (Tanzania) and

indirectly (Brazil, Bolivia).

  • Details

Manuel Santos Silva FLFP: micro evidence 23 / 34

slide-24
SLIDE 24

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Within-country decompositions: results

Explained (composition effect) vs. unexplained (coefficients and unobservables) changes in FLFP:

1 composition effect explains FLFP changes relatively well in

India, Brazil, and Jordan,

2 coefficients and unobservables account for most of the change

in Bolivia, Indonesia, and Vietnam,

3 composition and unexplained term cancel each other out in

South Africa,

4 results depend on the choice of coefficients in Tanzania.

Manuel Santos Silva FLFP: micro evidence 24 / 34

slide-25
SLIDE 25

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Composition effect vs. unexplained term

−.1 −.05 .05 .1 .15 .2

Tanzania Bolivia India South Africa Vietnam Jordan Brazil Indonesia

− f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r − f i r s t y e a r l a s t y e a r

Labor force: difference Composition effect Unexplained term Manuel Santos Silva FLFP: micro evidence 25 / 34

slide-26
SLIDE 26

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Within-country decompositions: results

Contribution of variable groups:

1 rising female education and falling fertility contribute positively

everywhere,

2 but the magnitude of these contributions varies across

countries,

3 in all but richest 3 countries (Jordan, SA, Brazil) positive

education and fertility contributions are offset by rising household income,

4 other factors contribute only marginally.

Manuel Santos Silva FLFP: micro evidence 26 / 34

slide-27
SLIDE 27

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Contribution of variable groups

−.05 −.025 .025 .05 .075

Tanzania Bolivia India South Africa Vietnam Jordan Brazil Indonesia

f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r f i r s t y e a r l a s t y e a r

Own education Children Log income Hh head educ Male salaried emp. Age Pop group Region dummies Survey waves Manuel Santos Silva FLFP: micro evidence 27 / 34

slide-28
SLIDE 28

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Between-country decompositions

  • Brazil’s coefficients as reference,
  • decompose FLFP gap of each country viz a viz Brazil,
  • decompose gaps around 2000, and around 2014.

Manuel Santos Silva FLFP: micro evidence 28 / 34

slide-29
SLIDE 29

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Between-country decompositions: results

1 covariates cannot explain FLFP gaps between countries, 2 for some countries, the composition effect even has the

“wrong” sign,

3 coefficients and unobservables account for the bulk of FLFP

variation between countries.

Manuel Santos Silva FLFP: micro evidence 29 / 34

slide-30
SLIDE 30

Motivation Methods & Data FLFP correlates Decompositions Conclusion

A thought experiment

  • Imagine there is a single, fictional, labor market, where:

1 all women face Brazil’s coefficients and unobservables,

irrespective of their country origin,

2 but, otherwise, each woman has her own observable

characteristics as given in the data.

  • What would be the FLFP rates in this “Brazilian”-like market?

Manuel Santos Silva FLFP: micro evidence 30 / 34

slide-31
SLIDE 31

Motivation Methods & Data FLFP correlates Decompositions Conclusion

“Brazilian”-like labor market (c. 2014)

.2 .4 .6 .8

Brazil Jordan India Indonesia Bolivia South Africa Tanzania Vietnam

Real FLFP Simulated FLFP at Brazil’s coefficients

Manuel Santos Silva FLFP: micro evidence 31 / 34

slide-32
SLIDE 32

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Conclusion

  • Participation-returns to women’s own characteristics and family

circumstances differ substantially across countries,

  • In fact, heterogeneity in returns to these characteristics explains

most of the between-country differences in participation rates.

Manuel Santos Silva FLFP: micro evidence 32 / 34

slide-33
SLIDE 33

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Policy message

  • Economic growth alone or further improvements of women’s

labor market characteristics FLFP rates ⇈,

  • unless: removal of barriers and constraints to female

employment both at the household and at the labor market level in each country.

Manuel Santos Silva FLFP: micro evidence 33 / 34

slide-34
SLIDE 34

Motivation Methods & Data FLFP correlates Decompositions Conclusion

Thank you for your attention

Manuel Santos Silva FLFP: micro evidence 34 / 34

slide-35
SLIDE 35

Sample selection bias

  • Sample composition effect due to trends in:

1 urbanization rates 2 marriage rates

  • problem if selection into urban areas and marriage is correlated

with labor force attachment. Solution (following Blau and Kahn 2007):

  • estimate parsimonious probit models to predict urban and

marriage probabilities (age, age2, education, region, children)

  • create “artificial” samples with constant urbanization and

marriage rates by excluding the women with the lowest urban and marriage propensity.

Manuel Santos Silva FLFP: micro evidence 1 / 5

slide-36
SLIDE 36

Selection bias in the education

  • Massive expansion in education attainment in some countries,
  • Rising education levels in our 25-54 group: more educated

younger cohorts replacing less educated older cohorts.

  • If the older cohorts were positively selected on education then,

the decreasing AMEs for secondary and tertiary education could be due to a decline in this positive selection over time. Our solution (following Klasen and Pieters 2015):

  • Estimate an upper bound on this selection effect
  • By weighting the AMEs of the first period by the changes in

the distribution of education attainment

Manuel Santos Silva FLFP: micro evidence 2 / 5

slide-37
SLIDE 37

Distribution of educational attainment: example South Africa

.1 .2 .3 .4 .5 .6 .7 .8 .9 1 2014 1995 < Prim Primary < Sec Secondary Tertiary

Manuel Santos Silva FLFP: micro evidence 3 / 5

slide-38
SLIDE 38

Selection into education

0.207 0.142 0.141 0.361 0.300 0.213 0.431 0.379 0.373

.1 .2 .3 .4 India Indonesia South Africa Estimated: India 1999, Indonesia 2000, SA 1995 Estimated: India 2011, Indonesia 2014, SA 2014 Reweighted: India 2011, Indonesia 2014, SA 2014

return Manuel Santos Silva FLFP: micro evidence 4 / 5

slide-39
SLIDE 39

Decomposition analysis for nonlinear models

  • Fairlie’s (2006) extension of the Oaxaca-Blinder decomposition

analysis,

  • Groups: A, B (two years, or two countries)
  • Decomposition at group A’s coefficients:

LFPB−LFPA ≈

  • NB

Φ(XB ˆ βA) NB −

  • NA

Φ(XA ˆ βA) NA

  • +
  • NB

Φ(XB ˆ βB) NB −

  • NB

Φ(XB ˆ βA) NB

  • ,

(2)

  • Equally valid: decomposing at group B’s coefficients.
  • We show results using both coefficient vectors (A and B).

Manuel Santos Silva FLFP: micro evidence 5 / 5