Presentation on theme: "S TATUS, C ASTE AND T IME A LLOCATION OF W OMEN IN I NDIA Bharat Ramaswami Indian Statistical Institute South Asia Development Workshop, University of."— Presentation transcript:
S TATUS, C ASTE AND T IME A LLOCATION OF W OMEN IN I NDIA Bharat Ramaswami Indian Statistical Institute South Asia Development Workshop, University of New South Wales, September 15-16, 2011
C O - AUTHORS Mukesh Eswaran, University of British Columbia Wilima Wadhwa, ASER Centre, Delhi
S OME WELL KNOWN TRENDS Increasing spread of education among women Declining fertility rates
T OTAL FERTILITY RATE (W ORLD BANK INDICATORS ) Source: n_tfrt_in&hl=en&dl=en#ctype=l&strail=false&nselm=h&met_y=sp_dyn_tfrt_in&scale_y=lin&ind_y=false&rdi m=country&idim=country:IND&ifdim=country&hl=en&dl=en
A P UZZLE (M AITREYI D AS, 2006) Yet, work participation rates of women low in India (South Asia generally) – around 35%. Over time ( ), this rate has not moved – in fact, a marginal decline of about 2% percentage points over so. Incomes have increased; so is it all because of an income effect?
C ROSS - COUNTRY E VIDENCE ON L ABOR F ORCE P ARTICIPATION AND GDP (G OLDIN 1995)
C ULTURE AND S TIGMA – G OLDIN (1995) Cross- country data shows that relation between women’s work participation and GDP is U shaped. Declining part of curve is because of income effect, absence of a substitution effect (from higher relative wages of women), decline in home production and family stigma attached to married women working at manual labour (outside the home). These get reversed in the rising part of the curve – strong substitution effect (due to spread of education) and access to `stigma-free’ white collar jobs.
O THER COUNTRY EVIDENCE Goldin: Economic history of the United States in the early 20 th century. Humphries (1987): Women’s labor supply response to the factory system in England in the 19 th century. Cameron, Worswick and Dowling (2001): Using data from 5 developing countries, they show that education increases workforce participation in only some countries but not all. Is a stigma effect working here?
I NDIA EVIDENCE : S TIGMA AND S TATUS So is India in the falling segment of the U curve? Is stigma part of the story here? Das (2006): Workforce participation (dominated by agriculture) negatively correlated with education for women. Variation in participation rates across groups and regions. Castes ranked higher in the traditional heirarchy have lower participation rates. Sociological evidence: Status concerns serious for the `upper castes’.
I NDIA EVIDENCE …2 `Lower’ castes may not exhibit the same concerns but may imitate upper caste social norms when they acquire other markers of status (e.g., education, land, income) – Sanskritization hypothesis. Tribals outside the traditional caste hierarchy and status concerns are not expected to govern female labour supply. Female work participation higher in South and Central India which have higher proportions of lower castes and tribals.
B OSERUP (1970): O RIGINS OF S TATUS Lower castes and tribals have traditions of `female farming’. Wives from upper castes avoided work in fields to differentiate themselves from “the despised and hard-working female labourers even if this means that she must live in poverty.” Parallel in Rwanda,Burundi: Hutu (female farming tradition) and Tutsi (land-owning class)
O UR W ORK Set out a simple household model of family status Family status is a public good produced by married women. Derive comparative statics about how time devoted to status and time devoted to market work respond to wealth, education and membership of social group (caste). Examine these effects in two data sets The effect on female labour supply in a data set on work participation and employment. The effect on female labour supply and on status activities in a data set of time use.
O UR WORK A household comprises a couple. They consume 3 goods: a consumption good C, status good Z and leisure, R. Market good is jointly consumed. Leisure is private Status good is a household public good
S TONE -G EARY U TILITY F UNCTION
S TATUS C ONCERNS Beta is the `status’ parameter. How is status produced? It is the amount of market work (time) that the wife relinquishes. Consumption good is the numeraire Each person is endowed with one unit of time. There is non-labour income (wealth) denoted by A.
R ESULTS : H IGHER W EALTH Higher wealth increases consumption of all three goods: market, status and leisure But it decreases female labour supply more than male labour supply because status is a normal good and supplied only by women. Therefore, higher wealth decreases the ratio of female labour supply to male labour supply.
A PPLICATION OF THIS RESULT Main wealth variable in rural data set is land. Should we see the ratio of female to male labour supply decline for larger landowning households? Not necessarily: family labour is preferred to hired labour because of supervision costs associated with the latter. Therefore as land ownership increases, family labour will be first exploited. So this will lead female to male labour supply to increase with land. Net effect??
H IGHER B ETA (S TATUS C OEFFICIENT ) Higher `beta’ Increases female time to status Reduces her work and leisure time Increases male work time and reduces male leisure time. Therefore, reduces the ratio of female to male labour supply.
H YPOTHESES : C ASTE AND E DUCATION `Beta’ is a function of status markers – caste and female education. Then female labour supply (relative to male supply) should be lower for higher castes and for households with more educated women. Note in the absence of status effects, female education should increase relative wages and increase female labour supply (relative to males) – the substitution effect. Therefore, a contrary finding is strong evidence of status effects.
I NTERACTION AMONG STATUS VARIABLES Status effects would be magnified if status markers reinforce each other (wealth and caste, wealth and education, education and caste – interaction effects). For e.g., status effects of caste is greater in a household with more land and in a household with greater levels of female education. Competing hypothesis: `Sanskritization’ Households from lower castes that aspire to higher levels of social status emulate the status norms of higher castes. In this case, the interaction effects of caste would be in the opposite direction.
D ATA Test our hypotheses using two different cross- sectional data sets – (a) data set of employment and work participation and (b) data set of time allocation between work and non-work (including status activities).
E MPLOYMENT DATA REGRESSION, 2004/05 Dependent variable: Ratio of labor supplied by female members of household to labor supplied by male members of household. Status variables: Caste, land, female education and their interactions. Control variables: male education, children Village fixed effects: controls for relative wages Regression therefore uses only within village variation in status variables.
E MPIRICAL MODEL
T ABLE 3: R ESULTS FROM E MPLOYMENT D ATA – R URAL S ECTOR D EPENDENT V ARIABLE : R ATIO OF F EMALE TO M ALE L ABOR S UPPLY Excludes ST observationsIncludes ST observations CoefficientStd. Err.CoefficientStd. Err. ST *** OBC Other castes *** *** Proportion of females with at least primary education ** ** Land ST × Land OBC × Land Other Castes × Land ST × Proportion of females with at least primary education OBC × Proportion of females with at least primary education Other Castes × Proportion of females with at least primary education Land × Proportion of females with at least primary education -8.39E-06***2.89E E-06***2.77E-06 Proportion of males with at least primary education *** *** # of children below the age of # of children between 6 and *** *** Constant *** *** Religion dummiesYes Village Fixed EffectsYes Observations R 2 (within group) *** significant at 1%, ** significant at 5%, * significant at 10%,
F INDINGS (R URAL W OMEN ) Negative and significant at 5% (or 1%) level: caste and female education. Land is not significant. Is it the opposing effects of status and the superiority of family labour? Among interaction variables, the interaction of female education with land is significant. Others are not.
U RBAN SECTOR REGRESSIONS Urban sector data lacks a wealth variable. To include compensating controls, regression includes dummies for the occupation of head of the household – white collar jobs, service sector jobs (in trade, hotels and personal services) and blue collar jobs (manual labour in agriculture and industry).
T ABLE 4: R ESULTS FROM E MPLOYMENT D ATA – U RBAN S ECTOR D EPENDENT V ARIABLE : R ATIO OF F EMALE TO M ALE L ABOR S UPPLY CoefficientStd. Err.CoefficientStd. Err. OBC *** *** Other castes *** *** Proportion of females with at least primary education * Service sector dummy *** Manual labor dummy ** *** OBC × Proportion of females with at least primary education Other Castes × Proportion of females with at least primary education Servicesector dummy × Proportion of females with at least primary education *** Manual labor dummy × Proportion of females with at least primary education ** Proportion of males with at least primary education *** *** # of children below the age of *** *** # of children between 6 and *** *** Constant *** *** Religion dummiesYes Urban Block Fixed EffectsYes Observations24662 R 2 (within group)
F INDINGS Caste effects even stronger than rural regression. Female education has a negative effect Blue collar households have the highest female labor supply and the white collar households the least. The occupation dummies are likely to be correlated with wealth and caste and therefore these are also likely to be status effects. In the interaction specification, female education is not significant but the interaction with occupation dummies is negative and significant. The negative effect of female education is seen in blue collar and service sector households – not in white collar households – consistent with the Goldin hypothesis.
T IME U SE D ATA Time use survey of individuals above the age of 5 in 6 states of India. Our data set is for rural sector. From the data, we can compute the time devoted to economic activities and the time devoted to leisure. We also define status activities as time spent in social and cultural activities. This includes participation in social events, community functions, religious activities, socializing, arts & music, games & sports, reading and watching TV. Does not include household chores, learning, care of others, and personal care (including sleeping, doing nothing).
T IME USE DATA, 1998/99 Here we do three regressions. A regression with the ratio of female work time to male work time as dependent variable (with the same specification as earlier). A regression with female status time as dependent variable A regression with female leisure time as dependent variable
T ABLE 6: R ESULTS FROM T IME U SE D ATA Dependent Variable Ratio of women’s work time to men’s work time Proportion of women’s time spent in status activities Proportion of women’s time spent in status activities + personal care and maintenance CoefficientStd. Err.CoefficientStd. Err.CoefficientStd. Err. Castes other than SC *** *** *** Land *** *** Proportion of females with at least primary education ** *** Other castes × Land Other castes × Proportion of females with at least primary education * Land × Proportion of females with at least primary education Proportion of males with at least primary education ** House of permanent materials *** *** Children below the age of *** Constant *** *** *** Religion dummiesYes Village Fixed EffectsYes Observations R 2 (within group) *** significant at 1%, ** significant at 5%, * significant at 10%,
L ABOUR S UPPLY REGRESSIONS Caste and female education have negative and significant effects. Interaction between female education and caste is positive and significant – sign implies that the withdrawal of educated females from the work force is sizeable for the `low’ caste group – Sanskritization effect?
S TATUS AND L EISURE REGRESSIONS In status regression, wealth variables, `higher’ castes, female and male education have positive and significant impacts. In leisure regression, education variables (male and female) are not significant.
C ONCLUSIONS Female withdrawal from the workforce means that measured poverty does not fall as fast as it might in the absence of such responses – yet households would deem themselves considerably better off. Could it be that economic progress from low levels of income reinforces rather than undermine the traditional gendered division of labour? Female autonomy and family status may move in opposite directions. Current work: Looks at how status variables could be used as instruments for female labour supply and therefore examine the effects of exogenous variation in work participation on female and male wages.