Springer Nature is making SARS-CoV-2 and COVID-19 research free. View research | View latest news | Sign up for updates

Why are fewer married women joining the work force in rural India? A decomposition analysis over two decades

Abstract

In contrast with global trends, India has witnessed a secular decline in women’s employment rates over the past few decades. We investigate this decline in rural areas, where the majority of Indian women reside. Using parametric and semi-parametric decomposition techniques, we show that changes in individual and household attributes fully account for the fall in women’s labor force participation in 1987–1999 and account for more than half of the decline in 1999–2011. Our findings underscore increasing education levels among rural married women and the men in their households as the most prominent attributes contributing to this decline. We provide suggestive evidence that changes in more educated women’s relative returns to home production compared with market production may have adversely affected female labor force participation in rural India.

This is a preview of subscription content, log in to check access.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Notes

  1. 1.

    India’s fertility rate declined from 4.12 in 1987 to 2.60 in 2011. (http://databank.worldbank.org/data/reports.aspx?source=2&country=IND). GDP grew at an average rate of 5.94% during 1987–1999 and 7.19% in 1999–2011. (http://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG?locations=IN).

  2. 2.

    70% of India’s population continues to reside in rural areas (Census 2011).

  3. 3.

    Das et al. (2015) analyze the relationship between labor market rigidities and female labor force participation, distinguishing between formal and informal sector employment, using NSS data for 1993–1994 to 2011–2012.

  4. 4.

    The NSS follows a two-stage sampling design: In rural areas, the first stratum is a district. Villages are the primary sampling units (PSUs), picked randomly in a district with equal number of households surveyed in each quarter (over an entire agricultural year of July to June) to ensure equal spacing of observations across the year. The households are randomly chosen in the selected PSUs.

  5. 5.

    We define the labor force participation rate (LFPR) as the proportion of people currently working or seeking work. Besides UPSS, the NSS provides another measure of labor force participation status—“daily status”—the number of days worked in the preceding week before the survey date. We do not use this measure in our analysis because the daily status employment rates in 1987 are not comparable to 1999 and 2011 due to a change in survey methodology. While these changes are unlikely to affect the employment rate using UPSS, they can artificially increase the employment figures by daily status. However, none of our conclusions change if we use daily status as our measure of employment.

  6. 6.

    The decline in LFPR is 6 and 3 percentage points for rural and urban males, respectively, during this period. Urban women’s LFPR declined by 3 percentage points between 1987 and 2011. The proportion of rural women looking for work has not changed during this period (NSS, various years). This suggests that it is not unemployment which is the cause of decline in women’s LFPR.

  7. 7.

    LFPR never married females have increased between 1987 and 2011. The share of married women in 1987, 1999, and 2011 was 82.5, 85.4, and 87.5%, respectively (NSS survey rounds). The small but significant increase in this proportion is attributable to a lower proportion of widowed women due to falling mortality rates in India.

  8. 8.

    Domestic work in the NSS includes domestic chores and not-for-wages collection of goods (vegetables, roots, firewood, cattle feed, etc.), sewing, tailoring, weaving, etc. for household use. The difference between women’s LFPR and the share of women in domestic work is the share of women unemployed in the previous year.

  9. 9.

    The consumption variable as a proxy to capture the income effect includes women’s income. Consequently, the contribution of income in explaining the decline in female LFPR in our analyses is likely to be a lower bound on the true negative income effect.

  10. 10.

    To illustrate, the first decile in 1987 contains households having a monthly per capita expenditure of less than Rs. 76. In nominal terms. Rs 76 in 1987 is equivalent to Rs 213 and Rs 429 in 1999 and 2011, respectively. Rs 213 is then defined to be the cutoff for the first decile in 1999. Similarly, Rs 429 is defined to be the cutoff for the first decile in 2011. Our results are unchanged when we include household consumption expenditure as a continuous, non-linear variable.

  11. 11.

    The NSS provides data on the highest level of completed education and not years of schooling of household members. Therefore, to avoid measurement error in calculating the average years of schooling of men in the household, we use the maximum level of male education. However, our results do not change if we use average education years. Other household characteristics which could possibly explain changes in women’s employment, such as household size, share of children under age 5, share of male members, caste, and religion, have not been included in the main regressions since they do not alter our main conclusions. Also, some of these characteristics (e.g., fertility) can be endogenous to the labor force participation decision. The decomposition results including these variables are shown in robustness checks in Appendix B.

  12. 12.

    The correlation between education of 18–35-year-old daughters-in-law in the household with the highest education of married males who are sons of the household head has increased from 0.54 in 1987 to 0.63 in 2011. We reach the same conclusion of a rise in positive assortative mating on education if we use the average level of education of males in the household.

  13. 13.

    DiNardo (2002) shows that the DFL method is identical to Blinder-Oaxaca decomposition when the variable of interest is the mean of the outcome variable and there is a single categorical explanatory variable. While this technique has been used to decompose wage and earning differentials (Leibbrandt et al. 2010; Biewen 2001; Butcher and DiNardo 2002; Hyslop and Mare 2005; Daly and Valletta 2006), only a handful of papers have used it to decompose differences in other outcomes, such as employment (Black et al. 2011) and health (Geruso 2012).

  14. 14.

    For example, in the above case, we re-weight observations in year 1987 so that the distribution of observed characteristics in 1987 is identical to that in 1999. If real income is higher in 1999, individuals belonging to households with higher incomes in 1987 are weighted up so that the percentage of individuals in each income decile after re-weighting is identical across years.

  15. 15.

    In urban India, the proportion of 25–65-year-old married women with higher secondary and graduate education has risen dramatically—from 13 to 25% and 6 to 18%, respectively—between 1987 and 2011. In contrast, Table 2 shows the percent of rural women with higher secondary schooling or above rose from only 2 to 11% over the same period.

  16. 16.

    In additional analyses (available on request), we investigated relationships between fertility choices, female education, and female work. We controlled for the proportion of household members who are in the 0–5 and 6–14 age group as well as their interactions with woman’s education. As expected, the higher the share of young kids in the household, the lower the female LFPR is, in all years (insignificant in 2011). This was particularly so for women in the 25–45 age group. Moreover, this negative correlation between young children and female employment is larger for women with higher levels of education. The share of children in the older age groups has a positive effect on female LFP. This could be because of older children providing a substitute for mother’s time. It is, however, difficult to interpret these results causally since fertility decisions are jointly determined with woman’s LFP. There are two opposing effects here: if women, who derive greater utility from raising children, choose to have more children, then this would bias the coefficient downwards. On the contrary, if women, who derive greater utility from higher quality of children, choose to have fewer children, then this would result in a positive bias on the coefficient. Our results indicate that the latter channel of deriving greater satisfaction from quality of children may be dominant. Unfortunately, given that there exists no exogenous variation in fertility in our study, we cannot estimate the true effect of fertility on women’s labor force participation. We hope to address this issue in future work.

  17. 17.

    Socio-economic factors which show minor or no change in distribution (for example, social group, religion, number of male members in household) or that exhibit a change in a direction that cannot explain the fall in women’s LFPR (e.g., number of children, household size) have not been included in our specifications in Tables 3 and 4. We show that our main results are robust to including these additional variables in Appendix B, Table 8.

  18. 18.

    For instance, we interact age group indicators with education, land owned, income, male education, and own education separately.

  19. 19.

    In specification (6), the re-weighting function is unable to match the age-group composition at statistically significant levels, for the decomposition in 1987–1999, but the absolute differences are not large. For example, in panel A, specification 6, the re-weighted observations in 1987 have an age group composition of 22, 20, 17, 13, 11, 8, and 10% for age groups 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, and 55–65, respectively. The corresponding numbers for actual population in 1999 provided in Table 1 are very close to these proportions.

  20. 20.

    Rounding off errors in Table 5.

  21. 21.

    The proportion of explained variation falls when additional variables are included in Appendix B, Table 8. The female characteristics continue to explain the entire fall in female LFPR between 1987 and 1999, but the explained proportion between 1999 and 2011 falls to 48%. This is because household size and number of children under age 5 have fallen over time. This change in quantity of children should increase the female LFPR. Social group membership and male members do not contribute much to the explained proportion. The only additional characteristic which contributes to the decline in female LFPR between 1999 and 2011 is the change in religious composition. This is because the proportion of population that is Muslim has increased, and Muslim women tend to have lower participation rates in the labor market. Our main conclusion remains: individual and household characteristics (in particular education) play the most important role in explaining declining LFPR.

  22. 22.

    Eswaran et al. (2013) suggest that the decline in women’s work force participation and increase in their engagement in “status”-related activities are well predicted by rising household incomes in rural India since status is a normal good.

  23. 23.

    The results at 1999 (for change during 1987–1999) and 2011 (for change during 1999–2011) coefficients and using the DFL decomposition of domestic work give us qualitatively similar results, hence have been omitted for brevity.

  24. 24.

    The rise in women’s and men’s education explains 17 and 75% of the increase in domestic work using 1987 regression coefficients, respectively (specification 2 of Table 6) during 1987–1999. Similarly, women’s and men’s education explains 5 and 18% of the increase in domestic work using 1999 regression coefficients (specification 4 of Table 6) during 1999–2011.

  25. 25.

    Time use data were collected from 18,591 households across six states of India by the same nodal agency that conducts the NSS to assess the economic contribution of women. The selection of states was purposive. One state from each region of India was chosen (north—Haryana, center—Madhya Pradesh, west—Gujarat, east—Orissa, south—Tamil Nadu, and northeast—Meghalaya), to capture the diversity in gender norms and culture (http://mdgs.un.org/unsd/Demographic/sconcerns/tuse/Country/India/sourceind99b.pdf). While the NSS collects data on aggregate domestic work, the time use survey allows us to break down domestic work into various components, other than leisure.

  26. 26.

    The relative starting position of rural and urban women on the ‘U’ curve may therefore account for the different relationships between education and changes in female LFPR between urban (Klasen and Pieters 2015) and rural areas (this paper).

  27. 27.

    The LFPR of 25–45-year-old, married rural women declined 3.4 and 11.2 percentage points during 1987–1999 and 1999–2011, respectively. The corresponding numbers for 46–65-year-old women were 1.8 and 7.5.

  28. 28.

    Authors’ calculations show that in 2011, the daily wage in agriculture, manufacturing, construction, and services sectors were Rs 100, Rs 119, Rs 116, and Rs 209, respectively.

  29. 29.

    The observation that the distribution of caste and religious groups—important predictors of social norms regarding women’s work force participation in India (Eswaran et al. 2013)—has not changed significantly during 1987–2011 suggests that it is unlikely that social norms either changed significantly or were a significant unexplained determinant of the decline in women’s LFPR during the period of our study.

References

  1. Aromolaran AB (2004) Female schooling, non-market productivity and labor market participation in Nigeria. In: Center Discussion Paper No 879. Yale University, USA

  2. Azam M (2012) Changes in wage structure in urban India, 1983–2004: a quantile regression decomposition. World Dev 40(6):1135–1150

  3. Behrman JR, Foster AD, Rosenweig MR, Vashishtha P (1999) Women's schooling, home teaching, and economic growth. J Polit Econ 107(4):682–714

  4. Biewen M (2001) Measuring the effects of socio-economic variables on the income distribution: an application to the east German transition process. Rev Econ Stat 83(1):185–190

  5. Biewen M (2012) Additive decompositions with interaction effects. IZA DP No 6730, IZA, Bonn, Germany

  6. Black D, Tseng Y-P, Wilkins R (2011) Do changes in demographic characteristics explain declining male employment rates? Examination of the Australian case using a propensity score re-weighting decomposition approach. Appl Econ 43(28):4215–4226

  7. Blinder AS (1973) Wage discrimination: reduced form and structural estimates. J Hum Resour 8:436–455

  8. Butcher KF, DiNardo J (2002) The immigrant and native-born wage distributions: evidence from United States censuses. Ind Labor Relat Rev 56(1):97–121

  9. Cameron LA, Dowling JM, Worswick C (2001) Education and labor market participation of women in Asia: evidence from five countries. Econ Dev Cult Chang 49(3):459–477

  10. Chandrasekhar CP and Ghosh J (2011) “Latest employment trends from the NSSO”, the Hindu business line, 12 July http://www.thehindubusinessline.com/opinion/columns/c-p-chandrasekhar/latest-employment-trends-from-the-nsso/article2219107.ece. Accessed 18 Nov 2013

  11. Chatterjee U, Murgai R, Rama M (2014) Job opportunities along the rural-urban gradation and female labor force participation in India. Mimeo, New York

  12. Chowdhury S (2011) Employment in India: what does the latest data show? Econ Polit Wkly 34(32):23–26

  13. Dalmia S (2011) Testing Becker’s efficient marriage market hypothesis and its implications for spouse selection and marital transfers in India. J Bus Econ Res 2(4):1–8

  14. Daly MC, Valletta RG (2006) Inequality and poverty in United States: the effects of rising dispersion of men’s earnings and changing family behavior. Economica 73(289):75–98

  15. Das MB, and Desai S (2003) Are educated women less likely to be employed in India? Social Protection Discussion Paper No. 313, World Bank, Washington DC

  16. Das S, Jain-Chandra S, Kochchar K, and Kumar N (2015) Women workers in India: why so few among so many? IMF Working Paper WP/15/55

  17. DiNardo J (2002) Propensity score reweighting and changes in wage distributions. Mimeo, University of Michigan, Ann Arbor

  18. DiNardo J, Fortin NM, Lemieux T (1996) Labor market institutions and the distribution of wages, 1973-1992: a semiparametric approach. Econometrica 64(5):1001–1044

  19. Eswaran M, Ramaswami B, Wadhwa W (2013) Status, caste, and the time allocation of women in rural India. Econ Dev Cult Chang 61(2):311–333

  20. Fairlie RW (2005) An extension of the Blinder-Oaxaca decomposition technique to logit and probit models. J Econ Soc Meas 30(4):305–316

  21. Gaddis I, Klasen S (2014) Economic development, structural change and women’s labor force participation. J Popul Econ 27(3):639–681

  22. Geruso M (2012) Black-white disparities in life expectancy: how much can the standard SES variables explain? Demography 49(2):553–574

  23. Goldin C (1995) The U-shaped female labor force function in economic development and economic history. In: Paul Schultz T (ed) Investment in Women's Human Capital. University of Chicago Press, Chicago, pp 61–90

  24. Himanshu (2011) Employment trends in India: a re-examination. Econ Polit Wkly 46(37):43–59

  25. Hyslop DR, Mare DC (2005) Understanding New Zealand’s changing income distribution, 1983–1998: a semi-parametric analysis. Economica 72(287):469–495

  26. Kannan KP, Raveendran G (2012) Counting and profiling the missing labor force. Econ Polit Wkly 47(6):77–80

  27. Kapsos S, Silberman A, and Bourmpoula E (2014) Why is female labour force participation declining so sharply in India? ILO Research Paper No. 10

  28. Klasen S, Pieters J (2015) What explains the stagnation of female labor force participation in urban India? World Bank Econ Rev 29(3):449–478

  29. Lahoti R, and Swaminathan H (2013) Economic growth and female labor force participation in India. Working paper no. 414, Indian Institute of Management, Bangalore

  30. Lam D, Duryea S (1999) Effects of schooling on fertility, labor supply, and investments in children, with evidence from Brazil. J Hum Resour 34(1):160–192

  31. Leibbrandt, Murray, Levinsohn, James A., and Justin McCrary. (2010). Incomes in South Africa after the fall of apartheid. J Globalization Dev 1(1)

  32. Lincove JA (2008) Growth, girls' education, and female labor: a longitudinal analysis. J Developing Areas 41(2):45–68

  33. Mammen K, Paxson C (2000) Women’s work and economic development. J Econ Perspect 14(4):141–164

  34. Mendiratta P, and Gupt Y (2013) Private returns to education in India by gender and location: a pseudo-panel approach http://www.econcaluniv.ac.in/arthanitiweb/book/2013/PM.pdf. Accessed on 11 Nov 2015

  35. National Sample Survey Office, (2015) Key Indicators of Social Consumption in India Education http://mospi.nic.in/Mospi_New/upload/nss_71st_ki_education_30june15.pdf. Accessed 10 Nov 2015

  36. Neff D, Sen K, and Kling V (2012) The puzzling decline in rural women’s labor force participation in India: a re-examination. GIGA working paper no. 196, GIGA Research Unit, Institute of Asian Studies

  37. Oaxaca R (1973) Male-female wage differentials in urban labor markets. Int Econ Rev 14(3):693–709

  38. Sorsa P et al( 2015) Determinants of the low female labour force participation in India, OECD Economics Department Working Papers, No. 1207, OECD Publishing, Paris

  39. Tam H (2011) U-shaped female labor participation with economic development: some panel data evidence. Econ Lett 110(2):140–142

  40. Topalova P (2010) Factor immobility and regional impacts of trade liberalization: evidence on poverty from India. Am Econ J Appl Econ 2(4):1–41

  41. United Nations (2013) Millennium Development Goals Report http://www.un.org/millenniumgoals/pdf/report-2013/mdg-report-2013-english.pdf. Accessed 12 Oct 2015

  42. World Bank (2012) World Development Report 2012: Gender Equality and Development http://siteresources.worldbank.org/INTWDR2012/Resources/7778105-1299699968583/7786210-1315936222006/Complete-Report.pdf. Accessed 18 Nov 2013)

Download references

Acknowledgements

The authors thank the International Growth Center (IGC)—India Central and the Planning and Policy Research Unit (PPRU) at the ISI (Delhi), for financial support. Abhiroop Mukhopadhyay, Bharat Ramaswami, and two anonymous referees provided valuable inputs. The authors are responsible for any remaining errors.

Funding

This study was funded by IGC—India, Project Code 1-VRA-VINC-VXXXX-89217 and PPRU project “Women and Work in Rural India.”

Author information

Correspondence to Farzana Afridi.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Additional information

Responsible editor: Junsen Zhang

Appendices

Appendix 1: Definition of labor force participation rate

The National Sample Survey uses three reference periods for the employment survey: (i) 1 year, (ii) 1 week, and (iii) each day of the previous week. This paper employs the Usual Principal and Subsidiary Status (UPSS) definition. The activity status on which a person spent relatively longer time (major time criterion) during the 365 days preceding the date of survey is considered the Usual Principal Activity Status of the person. Persons are first categorized as those in the labor force and those out of the labor force depending on the major time spent during the 365 days preceding the date of survey. For persons belonging to the labor force, the broad activity status of either “working” (employed) or “not working but seeking and/or available for work” (unemployed) is then determined based on the time criterion. After determining the principal status, the economic activity on which a person spent 30 days or more during the reference period of 365 days preceding the date of survey is recorded as the Subsidiary Economic Activity Status of a person. In case of multiple subsidiary economic activities, the major activity and status based on the relatively longer time spent criterion is considered. If a person is defined to be in the labor force in either the principal activity status or the subsidiary activity status then she is defined to be in the labor force according to the UPSS. A woman who reports her primary activity is domestic production is classified as out of the labor force.

Appendix 2: Explaining the role of interactions in the decomposition

The three-way linear Blinder-Oaxaca decomposition for change in mean outcome of employment (Y) between 1999 and 2011 can be written as

\( {\displaystyle \begin{array}{l}{\overline{Y}}^{1999}-{\overline{Y}}^{2011}\\ {}={X}^{1999}{\beta}^{1999}-{X}^{2011}{\beta}^{2011}\\ {}=\left({X}^{1999}-{X}^{2011}\right){\beta}^{2011}+{X}^{2011}\left({\beta}^{1999}-{\beta}^{2011}\right)+\left({X}^{1999}-{X}^{2011}\right)\left({\beta}^{1999}-{\beta}^{2011}\right)\end{array}} \)

Here, the first term is the ceteris paribus effect of a change in characteristics, the second term is the ceteris paribus effect of a change in coefficients, and the third term is the interaction effect between the changing characteristics and the changing coefficients (Biewen 2012). The corresponding non-linear decomposition for the change in mean employment (Y) is

$$ {\displaystyle \begin{array}{l}{\overline{Y}}^{1999}-{\overline{Y}}^{2011}\\ {}=\sum \limits_{i=1}^{N^{1999}}\frac{F\left({\boldsymbol{X}}_i^{1999}{\widehat{\beta}}^{1999}\right)}{N^{1999}}-\sum \limits_{i=1}^{N^{2011}}\frac{F\left({\boldsymbol{X}}_i^{2011}{\widehat{\beta}}^{2011}\right)}{N^{2011}}\\ {}=\left(\sum \limits_{i=1}^{N^{1999}}\frac{F\left({\boldsymbol{X}}_i^{1999}{\widehat{\beta}}^{2011}\right)}{N^{1999}}-\sum \limits_{i=1}^{N^{2011}}\frac{F\left({\boldsymbol{X}}_i^{2011}{\widehat{\beta}}^{2011}\right)}{N^{2011}}\right)\\ {}+\left(\sum \limits_{i=1}^{N^{2011}}\frac{F\left({\boldsymbol{X}}_i^{2011}{\widehat{\beta}}^{1999}\right)}{N^{2011}}-\sum \limits_{i=1}^{N^{2011}}\frac{F\left({\boldsymbol{X}}_i^{2011}{\widehat{\beta}}^{2011}\right)}{N^{2011}}\right)\\ {}+\left[\left(\sum \limits_{i=1}^{N^{1999}}\frac{F\left({\boldsymbol{X}}_i^{1999}{\widehat{\beta}}^{1999}\right)}{N^{1999}}-\sum \limits_{i=1}^{N^{2011}}\frac{F\left({\boldsymbol{X}}_i^{2011}{\widehat{\beta}}^{1999}\right)}{N^{2011}}\right)-\left(\sum \limits_{i=1}^{N^{1999}}\frac{F\left({\boldsymbol{X}}_i^{1999}{\widehat{\beta}}^{2011}\right)}{N^{1999}}-\sum \limits_{i=1}^{N^{2011}}\frac{F\left({\boldsymbol{X}}_i^{2011}{\widehat{\beta}}^{2011}\right)}{N^{2011}}\right)\right]\end{array}} $$

The last term in the square brackets is the interaction effect, which is equal to the explained component at 2011 coefficients subtracted from the explained component at the 1999 coefficients. We show an example of including the interaction term in the decomposition in Table 7 below.

Table 7 Blinder-Oaxaca decomposition of change in women’s LFPR (three way)
Table 8 Blinder-Oaxaca decomposition of change in women’s LFPR with additional controls

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Afridi, F., Dinkelman, T. & Mahajan, K. Why are fewer married women joining the work force in rural India? A decomposition analysis over two decades. J Popul Econ 31, 783–818 (2018). https://doi.org/10.1007/s00148-017-0671-y

Download citation

Keywords

  • Female labor force participation
  • Decomposition analysis
  • Education
  • India

JEL classification

  • J21
  • J22