IFPRI- Caste, Female Labor Supply and Gender Wage-Kanika Mahajan

CASTE, FEMALE LABOR SUPPLY AND THE GENDER WAGE GAP IN INDIA: BOSERUP REVISITED

Kanika Mahajan, Bharat Ramaswami

Conference on Gender Just and Food Nutrition in India, IFPRI, Delhi

29th August, 2016

2

Motivation Ratio of female to male agriculture

wages lower in the southern states compared to northern states. Ester Boserup (Women’s Role in Economic

Development, 1970) noticed the same pattern in Indian data from the mid-1950s.

3

Motivation: Geographical variation in female/male wage ratio, 2004

Hypothesis Why?

Is there greater discrimination against women in the southern states?

Are women less productive (relative to men) in the southern states?

Variation in gender segregation by task where `female’ tasks are paid less than `male’ tasks.

Boserup’s hypothesis : There are more women workers (relative to men) in the southern states.

5

Boserup’s Hypothesis: Correlations (State level)

PunjabHaryana

RajUP

Bihar

AssamWB

Orissa

MP

Guj

MahaAP

Kar

Kerala

TN

.5.6

.7.8

.9W

age

ratio

0 1 2 3 4Female employment in agriculture

Source: NSS, Schedule 10, 2004-05

Our question The paper examines the role of female and

male labor supply together with comprehensive controls for infrastructure, agro-climatic endowments and cropping patterns. Besides the Boserup hypothesis, the effect of

male labor supply on the wage gap is also of interest.

Since men have greater access to non-farm work opportunities, do women working as agricultural labor gain from growth in the non-farm sector?

7

What we do Estimate district level inverse demand

functions that relate female and male agricultural wages to exogenous variation in female and male labor supply to agriculture.

Data is cross-section from 2004/5 (NSS employment surveys).

We propose instruments for both female and male labor supply.

8

Empirical strategy For observed levels of female and male

employment in agriculture, we estimate the inverse total demand for labor functions:

where F, M indexes female and males respectively, i indexes district, W is log of real wage, L is log of labor employed in agriculture, X are other control variables.

9

Identification: Female labor supply to Agriculture Instrument: proportion of district population that is

SC, ST and OBC. `High caste’ women refrain from work participation

because of `status’ considerations (Aggarwal, 1994; Beteille, 1971; Boserup, 1970; Chen, 1995). Could this just be an income effect?

Eswaran, Ramaswami and Wadhwa (2013) show that `higher’ caste households have lower female labor supply even when there are controls for male labor supply, female and male education, family wealth, family composition, and village level fixed effects that control for local labor market conditions and local infrastructure.

Low caste households and female employment in agriculture

-2-1

01

2Lo

g fe

mal

e em

ploy

men

t in

agric

ultu

re

.2 .4 .6 .8 1Proportion of low caste households

11

Identification: Male labor supply to agriculture

Instrument: district proportion of men (in the age group 15-59) employed in non-farm manufacturing and mining units with a workforce of at least 20.

The competition from non-farm jobs reduces the labor supply to agriculture and increases wages (Lanjouw and Murgai (2009))

Male employment in agriculture and large scale industrial employment

01

23

Log

mal

e em

ploy

men

t in

agric

ultu

re

0 .05 .1 .15 .2Proportion of men in mining and manufacturing enterprises with at least 20 workers

13

Validity of Instruments Pitfalls in the use of both instruments addressed

by inclusion of comprehensive controls.

But are our controls good enough? Hard to be totally sure in a cross-sectional study. Additional Test in the paper.

14

Data Employment and Unemployment survey of 2004/05

conducted by National Sample Survey (NSS) The survey contains labour force participation and

earnings details for the reference period of a week Census 2001; Land use statistics, 2004; Fertiliser

Association of India 2004-05; Area, Production and yield statistics 2004-05; India Water Portal 2004-05; Livestock Census 2003; Agro Ecological Zones- Compiled by Richard Palmer-Jones and Kunal Sen

The analyses includes 15 major states in the sample 279 districts

15

Estimation District-level regressions weighted by district

population and the standard errors are robust and corrected for clustering at state-region level. To avoid measurement error, the districts for which

number of wage observations for either males or females was less than 5 were dropped from the analyses.

Two stage least squares First stage regression Second stage IV estimates Robustness checks

16

First stage estimatesMale LS Female LS Male LS Female LS Male LS Female LS

(1) (2) (3)Low caste -0.11

(0.19) 0.70** (0.27) -0.15

(0.20) 0.66** (0.26) -0.22

(0.19) 0.79*** (0.27)

Industry -3.86***(0.53) -0.58 (0.77) -3.68***

(0.55) -0.29 (0.89) -3.33***

(0.59) -0.26 (0.97)

R-Square 0.69 0.53 0.70 0.54 0.71 0.54

Observations 279 279 279 279 279 279

17

Second stage IV estimates : Aggregate demand for total labor in agriculture, System2SLS District Controls: Agriculture

District Controls: Agriculture District Controls: Infrastructure

District Controls: Agriculture District Controls: InfrastructureDistrict Controls: Education & Urbanization

Male wage Female wage Male wage Female wage Male wage Female wage(1) (2) (3)

Female LS -0.08(0.17) -0.49*

(0.27) -0.11

(0.17) -0.54*

(0.31) -0.13

(0.15) -0.52**

(0.25)

Male LS-

0.29***(0.09)

-0.35***

(0.12)

-0.23***

(0.09)

-0.36***

(0.14) -0.28***

(0.09) -0.37**

(0.15)

Irrigation 0.21*(0.12) 0.30*

(0.17) 0.28**

(0.12) 0.41**

(0.19) 0.31**

(0.12) 0.41**

(0.20)

Gini -0.52(0.37) -1.28**

(0.54) -0.64*

(0.34) -1.33**

(0.56) -0.65*

(0.33) -1.30**

(0.51)

Rainfall -0.00(0.01) 0.01

(0.01) 0.00

(0.00) 0.01

(0.01) 0.00

(0.01) 0.01

(0.01)

Paved roads 0.43***(0.10) 0.05

(0.25) 0.47***

(0.11) 0.08

(0.23)

Electrified-

0.55***(0.17) -0.41*

(0.25) -0.61***

(0.18) -0.44*

(0.24)

Commercial bank 0.04

(0.20) -0.01

(0.21) 0.04

(0.17) -0.00

(0.21)

Primary-Mid female -0.01

(0.27) -0.15

(0.54)

Secondary female 0.39

(0.35) 0.39

(0.66)

Primary-Middle male -0.28

(0.26) -0.20

(0.40)

Secondary male -0.16

(0.24) 0.04

(0.45)

Urban percent -0.15**(0.08) -0.08

(0.16)

Constant 4.50***(0.37) 4.64***

(0.49) 4.85***

(0.41) 5.08***

(0.69) 5.10***

(0.49) 5.16***

(0.76)

AEZ Yes Yes YesCrop composition Yes Yes YesObservations 279 279 279 279 279 279

Results…1 The null of equality of coefficient of female

labor supply on male and female wages is rejected.

10% increase in female labor supply decreases female wages by 5.2% and male wages by 1.3%.

Boserup hypothesis is validated: A 10% increase in female labor supply decreases relative female wage by 4%.

Results…2 Effect of male labor supply is significant

for both male and female wages. The null that the effects are the same for

males and females is not rejected. Thus, there is an asymmetry: male labor

supply affects female wages but female labor supply does not affect male wages. Why?

Explaining the North-south differential gap in wages

Aggregate demand equations for Southern states can be written as

Subtracting 1 from 2 we get:

Similarly for northern states we get:

Subtracting 4 from 3 we get:

Explained difference in wage gap between northern and southern states

Variable Proportion wage gap explained

Female LS 55%Paved roads 36%Rice 29%Horticulture 10%Gini 10%Rainfall 7%Irrigation 5%Primary-Middle female 2%Commercial bank 1%Secondary female 0%Primary-Middle male 0%Cotton -2%Urban percent -2%Oilseeds and Pulses -2%Secondary male -2%Electrified -13%Male LS -14%Coarse Cereals -22%

22

Conclusion We confirm the Boserup hypothesis: Increase in female labor

supply reduces relative female wage in rural India Attributing the gender wage gap to only individual characteristics or

discrimination is incomplete Shows that female and male labor are imperfect substitutes.

Male labor supply has sizeable effects on male as well as female wages. Females gain despite limited direct access to non-farm employment

Creating jobs for women in non-farm sector enabling them to earn a greater wage it can reduce the gender wage gap in the agriculture sector

Thank You

Appendix

25

Theoretical Framework Consider a competitive agricultural labor market

with exogenously determined labor supply and three factors of production – Land (A), Male labor (Lm) and Female labor (Lf).

The production function is homogenous,

continuous and differentiable. There exist diminishing returns to each factor and in the short run the amount of land is fixed.

The profit function is given by:

Chapter 1: Caste, Female Labor Supply and the Gender Wage Gap in India: Boserup Revisited

26

Comparative Statics…1 In a competitive equilibrium all factors are paid

their marginal products

The own and cross price inverse demand elasticities


27

Comparative Statics…2 Effect of female labor supply on the gender wage

gap

Not possible to sign the above if males and females are substitutes in production

The relative magnitude of the cross wage elasticities can however be obtained. This is clearly greater than one.


28

Asymmetry The elasticity of female wages with respect to

male labor supply relative to the similar cross elasticity for male wages is the product of two ratios:

The sample estimate of the above is 2.63 The estimate obtained by the econometric

estimation is 2.84


Checks and Robustness Check for weak instruments. Include more controls – fertilizers,

machinery, health. Missing districts because of few wage

observations. Allowing hired and family labor to have

unequal efficiency. Individual level regressions.


Individual wages regressed on gender dummy, task dummies and other controls

Wage Wage (1) (2)

Female -0.35***(0.03

) -0.33***(0.03

)

Age 0.02***(0.00

) 0.02***(0.00

)

Age square -0.00***(0.00

) -0.00***(0.00

)

Below primary 0.06***(0.02

) 0.06**(0.02

)

Primary 0.05*(0.02

) 0.05*(0.02

)

Middle 0.03(0.03

) 0.02(0.03

)

Secondary 0.04(0.03

) 0.04(0.03

)

Senior secondary and above -0.03(0.03

) -0.03(0.03

)

Married -0.02(0.02

) -0.01(0.02

)

Widowed -0.06**(0.03

) -0.05(0.03

)

Divorced -0.13***(0.04

) -0.11**(0.05

)

Sowing -0.17**(0.06

)

Transplanting -0.04(0.05

)

Weeding -0.20***(0.04

)

Harvesting -0.12***(0.04

)

Other cultivation -0.11***(0.03

)

Constant 3.37***(0.05

) 3.50***(0.06

) Observations 14,190 14,190R-square 0.21 0.22

Check for weak instruments

Male wage Female wage Male wage Female

wage Male wage Female wage

(1) (2) (3)Low caste -0.02

(0.11) -0.31**

(0.13) -0.04

(0.10) -0.30**

(0.13) -0.04

(0.10) -0.34**

(0.13)

Industry1.15**

*(0.35)

1.63***

(0.42)

0.89***

(0.33)

1.47***

(0.44) 0.98***

(0.34) 1.37***

(0.48)

R-Square 0.62 0.61 0.68 0.62 0.68 0.63

Observations 279 279 279 279 279 279

Include more controls – fertilizers, machinery

Male wage Female wage Male wage Female wageFemale LS -0.10 (0.14) -0.46** (0.23) -0.12 (0.15) -0.52** (0.26)Male LS -0.31*** (0.10) -0.44*** (0.15) -0.29*** (0.09) -0.37** (0.15)Irrigation 0.25** (0.11) 0.27 (0.17) 0.31** (0.13) 0.40** (0.20)Gini -0.66** (0.33) -1.31*** (0.48) -0.64* (0.34) -1.28** (0.51)Rainfall 0.00 (0.00) 0.01 (0.01) 0.00 (0.01) 0.01 (0.01)Paved roads 0.52*** (0.11) 0.18 (0.20) 0.49*** (0.12) 0.09 (0.23)Electrified -0.60*** (0.18) -0.43* (0.24) -0.62*** (0.19) -0.45* (0.24)Commercial bank -0.02 (0.19) -0.15 (0.19) 0.04 (0.18) 0.00 (0.21)Primary-Middle female -0.04 (0.26) -0.23 (0.52) -0.02 (0.27) -0.16 (0.54)Secondary female 0.07 (0.40) -0.35 (0.65) 0.36 (0.33) 0.37 (0.65)Primary-Middle male -0.24 (0.25) -0.13 (0.37) -0.28 (0.26) -0.20 (0.40)Secondary male -0.05 (0.25) 0.30 (0.47) -0.14 (0.24) 0.06 (0.45)Urban percent -0.23*** (0.09) -0.27 (0.17) -0.15** (0.07) -0.08 (0.15)Fertilizer 0.04** (0.02) 0.10*** (0.03)Implements 0.08 (0.10) 0.06 (0.12)Constant 5.13*** (0.50) 5.23*** (0.75) 5.06*** (0.50) 5.13*** (0.76)AEZ Yes YesLand allocation to crops Yes YesObservations 279 279 279 279Under-id (p-val) 0.01 0.01 0.01 0.01F(excluded instruments) LS

F 4.86 4.86 4.60 4.60F(excluded instruments) LS

M 15.81 15.81 17.06 17.06

Include more controls – health Male wage Female wage

Female LS -0.16 (0.16) -0.53* (0.28)Male LS -0.28*** (0.10) -0.37** (0.16)Irrigation 0.33*** (0.12) 0.39** (0.19)Gini -0.75*** (0.29) -1.20** (0.47)Rainfall 0.00 (0.01) 0.01 (0.01)Paved roads 0.35*** (0.13) 0.13 (0.26)Electrified -0.59*** (0.21) -0.50* (0.30)Commercial bank -0.04 (0.16) -0.01 (0.23)Primary-Middle female 0.04 (0.27) -0.15 (0.55)Secondary female 0.38 (0.35) 0.34 (0.66)Primary-Middle male -0.29 (0.27) -0.21 (0.42)Secondary male -0.16 (0.25) 0.11 (0.48)Urban percent -0.11 (0.08) -0.09 (0.17)BMI (Female) -0.00 (0.01) -0.01 (0.02)BMI (Male) -0.01 (0.01) 0.01 (0.02)Constant 5.73*** (0.60) 4.91*** (0.87)AEZ YesLand allocation to crops YesObservations 279 279Under-id (p-val) 0.01 0.01F(excluded instruments) LS

F 3.957 3.957F(excluded instruments) LS

M 17.25 17.25

Missing districts because of few wage observations

Male wage Female wageFemale LS -0.05 (0.06) -0.53** (0.24)Male LS -0.36*** (0.13) -0.34** (0.16)Irrigation 0.22** (0.10) 0.42** (0.19)Gini -0.46** (0.20) -1.32** (0.53)Rainfall -0.01 (0.01) 0.01 (0.01)Paved roads 0.40*** (0.12) 0.09 (0.22)Electrified -0.60*** (0.20) -0.47* (0.24)Commercial bank 0.06 (0.22) -0.03 (0.22)Primary-Middle female 0.08 (0.22) -0.24 (0.51)Secondary female 0.20 (0.30) 0.29 (0.64)Primary-Middle male -0.21 (0.20) -0.16 (0.37)Secondary male 0.11 (0.26) 0.14 (0.42)Urban percent -0.16* (0.09) -0.01 (0.15)Constant 5.09*** (0.50) 5.22*** (0.77)AEZ YesLand allocation to crops YesObservations 359 288Under-id (p-val) 0.02 0.02F (excluded instruments) LS

F 8.76 5.54F (excluded instruments) LS

M 6.69 17.03

35

Allowing hired and family labour to have unequal efficiency

θ= 0.5 0.7 0.9 1Male Wage log female LS -0.12 (0.15) -0.13 (0.15) -0.13 (0.15) -0.13 (0.15)log male LS -0.37*** (0.13) -0.32*** (0.11) -0.29*** (0.10) -0.28*** (0.09)

Female Wage log female LS -0.47* (0.26) -0.50** (0.25) -0.52** (0.25) -0.52** (0.25)log male LS -0.58*** (0.22) -0.47*** (0.18) -0.40** (0.16) -0.37** (0.15)

Consider the possibility of hired and family labor having unequal efficiency (Family labor may be more efficient)

In terms of efficiency units of family labor, the total labor supply is , where and are the aggregate labor supply to the home farm and to outside farms.

Individual regressionsMale wage Female wage

log female LS -0.06 (0.23) -0.55** (0.28)log male LS -0.39*** (0.13) -0.40* (0.20)Irrigation 0.31** (0.15) 0.71** (0.28)Gini -0.66 (0.47) -1.36*** (0.49)Rainfall 0.01 (0.01) 0.02* (0.01)Coarse Cereals 0.01 (0.29) 0.85** (0.43)Cotton -0.01 (0.41) 0.98* (0.59)Oilseeds and Pulses -0.04 (0.25) 0.48 (0.35)Rice 0.09 (0.38) 1.12** (0.52)Horticulture -0.05 (0.36) 0.92 (0.61)Paved roads 0.41*** (0.13) 0.12 (0.22)Electrified -0.34 (0.26) -0.46 (0.30)Commercial bank 0.49 (0.33) 0.19 (0.25)Urban percent -0.13 (0.11) -0.13 (0.19)Primary-Middle female -0.12 (0.31) -0.42 (0.63)Secondary female 0.17 (0.51) -0.17 (0.67)Primary-Middle male -0.21 (0.40) -0.18 (0.45)Secondary male -0.03 (0.25) 0.52 (0.49)AEZ 18 0.21 (0.18) 0.24 (0.26)Constant 4.74*** (0.60) 4.55*** (0.65)Observations 7,812 6,378Under-id (p-val) 0.00 0.00F(excluded instruments) LS

F 3.71 5.34F(excluded instruments) LS

M 12.96 13.14

Summary statistics of variables across northern and southern states

Variable Mean Standard deviation Mean Standard

deviationNorthern states Southern states

Female LS 0.54 0.73 0.98 0.60Male LS 1.70 0.61 1.19 0.53Irrigation 0.52 0.27 0.34 0.22Gini 0.66 0.10 0.71 0.09Rainfall 9.21 4.73 7.12 6.11Paved roads 0.53 0.23 0.83 0.13Electrified 0.75 0.27 0.99 0.02Commercial bank 0.06 0.03 0.14 0.17Primary-Middle female 0.23 0.10 0.27 0.11Secondary female 0.09 0.05 0.15 0.07Primary-Middle male 0.36 0.09 0.36 0.10Secondary male 0.21 0.09 0.25 0.08Urban percent 0.23 0.18 0.32 0.18Coarse Cereals 0.09 0.13 0.24 0.22Cotton 0.08 0.12 0.09 0.11Oilseeds and Pulses 0.22 0.20 0.30 0.19Rice 0.39 0.28 0.25 0.25Horticulture 0.03 0.03 0.10 0.17Male wage 3.77 0.25 3.88 0.30Female wage 3.63 0.29 3.43 0.29

38

Female to male agricultural wage ratio across Indian states across years

State 1983 1993 1999 2004Assam 86% 81% 78% 90%Gujarat 88% 98% 89% 90%West Bengal 93% 88% 89% 88%Bihar 84% 87% 88% 87%Haryana 97% 85% 90% 84%Madhya Pradesh 85% 83% 85% 83%Punjab 81% 108% 94% 83%Uttar Pradesh 79% 75% 78% 83%Rajasthan 65% 75% 80% 81%Orissa 75% 73% 79% 72%Karnataka 71% 73% 68% 69%Andhra Pradesh 66% 72% 67% 65%Maharashtra 59% 63% 65% 63%Kerala 65% 70% 63% 59%Tamil Nadu 55% 57% 58% 54%All India 69% 72% 72% 70%Source: NSS Schedule 10, 1983, 1993, 1999, 2004

39

Sectoral distribution of off-farm employment

Industry

Percentage in units with 20 or more workers

Percentage in units with 9 or less workers

(1) (2)Agriculture and allied activities 1% 7%Fishing 0% 1%Mining 7% 1%Manufacturing 44% 20%Construction 11% 17%Trade and hotels 3% 28%Transport 9% 12%Finance and real estate 3% 2%Public administration 22% 11%Domestic services 0% 1%

Notes: The above figures are calculated from the usual status activity status of respondents in NSS 2004 Schedule 10 for men aged 15-59

40

Alternate wage ratio measure The wage ratio can be computed as the

weighted mean across tasks given by:

Here, is the proportion of females working in task ‘j’ in state ‘s’

Purging wage ratio of the effect of the across-state variation in the gender division of labor by taking a benchmark state

41

Female labor supply and the re-weighted female-male wage ratio

Punjab

Haryana

Raj

UPBihar

Assam

WB

Orissa

MP

Guj

MahaAP

Kar

KeralaTN

.5.6

.7.8

.9W

age

ratio

rew

eigh

ted

0 1 2 3 4Female labour supply

42

Agro Ecological Regions

Literature Blau and Kahn (2003): Look at gender wage gaps for

22 countries (mostly OECD) and find that they are smaller whenever women are in shorter supply. Estimates do not fully correct for endogeneity of labor supply.

Acemoglu, Autor and Lyle (2004): The spurt in female labor force participation during WW II increased gender wage gaps in US. Male mobilization rates used as instrument for female labor supply. Male labor supply is not instrumented.

Rosenzweig (1978): Estimates district-level labor

demand functions for 1960-61. Increase in female labor supply decreases both female and male wages. Boserup hypothesis is not supported.


44

Points of departure from Rosenzweig(1978)

Rosenzweig(1978) This paperData Agricultural wages in

India (1960)NSS(2004)

Instruments for labor supply

proportion of population living in urban areas in the district, non-farm economy, percentage of Muslims in the district

Percentage low caste, Non-farm manufacturing economy

Definition of labor Demand for hired agricultural labor

Log of Demand for total agricultural labor per unit land

Control variables - Crop composition, soil and climate, and infrastructure


45

Validity of InstrumentsAdditional Test

Since we estimate two equations, we have an additional test of the validity of the instruments.

Suppose conditional on our controls, the instrument is still correlated with omitted variables that affect the demand for agricultural labor. Then the caste composition also ought to have an effect on the demand for male labor. First stage regression for male employment.

Similarly, in the first stage regression for female employment, we can check for the significance of non-farm employment in large enterprises.

IFPRI- Caste, Female Labor Supply and Gender Wage-Kanika Mahajan

Government & Nonprofit