Spousal Violence and Women's Employment in India

ABSTRACT

This study analyzes women's experiences of physical or sexual spousal violence as a correlate of their employment. Based on the 2005–6 National Family Health Survey III, a nationally representative dataset from India, the analysis illustrates that married women who experienced spousal violence are more likely to be employed and are also more likely to work for cash remuneration and be employed year-round. These results may appear to suggest that spousal violence is associated with higher likelihood of married women seeking financial self-reliance. However, investigation of who decides how to spend women's earnings reveals that Indian women who experienced spousal violence are less likely to have a say in that vital decision, which suggests that women who experience spousal violence may also be more susceptible to financial exploitation. The evidence further indicates a need for caution among analyses that uniformly embrace employment as a financial empowerment tool for women.

KEYWORDS:

• Spousal violence
• women's employment
• India

JEL Codes:

• J12
• J16
• O53

NOTES ON CONTRIBUTOR

Haimanti Bhattacharya is Assistant Professor of Economics at the University of Utah. She is an applied micro-economist. Her research interests are in the fields of environment and resource economics, development economics, and experimental and behavioral economics. The geographic focus of her research is South Asia. Her research has addressed issues pertaining to environmental sustainability, the role of status in varied socioeconomic decision making, and women's well-being.

Notes

1 Data source for India's economy's rank (GDP statistics) for the year 2010: http://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG/countries?display=defaulthttp://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG/countries?display=default.

2 Source: http://www.trust.org/trustlaw/womens-rights/dangerpoll/. The ranking is based upon opinion of gender experts about six risk categories: sexual violence, nonsexual violence, cultural or religious factors, discrimination and lack of access to resources, and trafficking.

3 China and India ranked first and second with populations of about 1.3 and 1.2 billion respectively, in 2012; data source: https://www.cia.gov/library/publications/the-world-factbook/rankorder/2119rank.htmlhttps://www.cia.gov/library/publications/the-world-factbook/rankorder/2119rank.html. Data source for labor force participation in 2010: http://data.worldbank.org/indicator/SL.TLF.CACT.FE.ZS. It is well recognized that the definition of women's labor force participation used to compute statistics for international comparison has a major limitation as it excludes several types of productive work done by women all over the world. Hence, women's labor force participation statistics are downward biased for all countries, including India.

4 The NFHS put in substantial effort to address the concerns about reliability of survey data on sensitive issues like domestic violence. For example, the NFHS reported the data on domestic violence only for those women who had completed the interview without any outside observers or interruptions (IIPS and Macro International 2007). NFHS is the only data source in India that provides spousal violence data with national coverage, using internationally tested questionnaires that minimize ambiguity in the interpretation of the questions. Hence, it is the best available data for the purpose of this analysis. Yet, we recognize that there may still exist some reporting bias on part of responders due to cultural stigma associated with reporting sensitive issues.

5 Historical origin of the caste system involved the classification of Hindu society based upon patriarchal/hereditary-rankable occupations. Even though the link between caste and occupation is weakening in modern India, caste still plays a prominent role in socio-cultural practices.

6 The surveys or census in India collect data only about the following caste classifications: SC, ST, OBC, and all other classes. While SC is at the bottom of the ladder in the Hindu caste hierarchy, ST is outside the realm of traditional Hindu society. Data is collected about SC, ST, and OBC because they are eligible for affirmative action policies, which were put in place with the objective of bridging the education and economic gaps between these historically marginalized groups and the rest of the population in India. No data are available about the distribution of the main castes.

7 The “svy” option in STATA is used to estimate survey weight adjusted logit and multinomial logit regressions. Clustering of errors by regions (like states) is not feasible when using the svy option for these regressions.

8 Following standard practice (John Bound, David A. Jaeger, and Regina M. Baker 1995), we assess the instrument's strength from its performance in a first stage regression of woman's experience of spousal violence on all exogenous explanatory variables and the instrument.

9 The conventional 2SPS estimation approach uses the instrumental variable(s) and the other exogenous explanatory variables of the model to predict the endogenous variable in the first stage; and then substitutes the original endogenous variable with the predicted endogenous variable from the first stage to estimate the regression model in the second stage. The 2SRI also uses the instrumental variable(s) and the other exogenous explanatory variables of the model to predict the endogenous variable in the first stage. However, in the second stage, instead of substituting the original endogenous variable with the predicted endogenous variable from the first stage, it uses the residual from the first stage as an additional variable to estimate the model.

10 In this analysis, 2SPS provided qualitatively similar results as those obtained from 2SRI.

11 A Supplementary Appendix, available online on the publisher's website, presents the detailed regression results for interested readers.

12 Recent studies by Bhattacharya, Bedi, and Chhachhi (2011) and Chin (2012), which accounted for the endogeneity in the relationship between spousal violence and employment, show that women's employment reduces the risk of spousal violence in India. Hence, we can consider the expected sign of the marginal effect of employment on likelihood of spousal violence to be negative. And, if the sign of the marginal effect of spousal violence on the likelihood of employment is positive, then the expected direction of the simultaneity bias will be negative. Thus the downward bias in the exogenous estimates in Table 2 appears to align with the existing evidence from the recent literature on India.

13 Panda and Agarwal (2005), based on data from the Thiruvananthapuram district of Kerala, show that 65.7 percent of women did not own any property. Kerala is a state where several communities have a tradition of matrilineal property inheritance, hence property inheritance by women in Kerala is expected to be higher than many other states of India. In a sample from rural Uttar Pradesh, Bhattacharya, Bedi, and Chhachhi (2011) found that only 9 percent of women owned a house.