1 The Objective of the Study
This is an exploratory study in search of suitable data and a methodology for investigating into the gender inequalities in household consumption. The objective of the study was to identify the data requirements and a suitable methodology, through a limited field investigation for bringing out explicitly the extent of gender disparity, if any, in household consumption in Kerala, one of the socially, economically, culturally and politically advanced states in India. Inter alia, the study was expected to provide some tentative estimates of the extent of
2 DATA AND METHODOLOGY
Because of the very nature and purpose of this study, no pre-decided and structured methodological framework was used in this study. Instead, an open-ended approach was used.We used a ‘Trial and Error’ approach to evolve an appropriate methodology and tools of analysis through the field investigation on the household consumption behaviour.
First, a household consumer expenditure survey in selected sample households on the lines of the NSS, was conducted. A specially prepared and pre-tested schedule was used to collect consumer expenditure data at the household level that would enable us to make suitable analysis to bring out the nature and extent of gender inequality in consumption. Further, to understand the extent to which the observed disparity could be attributed to discrimination, we supplemented the consumer expenditure schedules with two additional questionnaires, specially meant for a detailed probe into the nature and type of discrimination against girls or women , if any, compared with boys or men, by any of the family members in respect of allocation of food and expenditure on education, health and other consumption needs. The questionnaires were exhaustive with carefully framed questions and was tactfully administered during the survey without causing any embarrassment to the respondents.
The next step in the methodology was to analyse the quantitative data on consumption expenditure to examine the extent of gender- related inequality in consumption expenditure. This quantitative analysis and discussion on disparity was then followed by an analysis of the responses of the various respondents in the selected sample households on the questions put to them about their conscious habit in discriminating girls from boys with regard to consumption and other expenditures in the household.
A stratified three stage sampling design was adopted for the survey. District, Municipal/Panchayat ward and household constituted the three stages of sample selection. Six districts were selected for the survey in such a way that representation to different socio-economic and cultural groups is ensured. From each district one Municipal / Panchayat ward was selected at random and from each ward, a sample of 50 households with adult males, adult females, female children and male children was selected for the study.
Our first attempt was to examine how best the household consumption data collected through the small survey could be suitably tabulated and classified, to enable us to draw possible inferences on gender disparity in total consumption and item-wise consumption. Because we had only household level data and not individual-wise data, the data do not directly throw any light on gender disparity. Therefore, two approaches were adopted in this analysis.
The first approach was by tabulating and classifying the consumption data in such a way that we can understand the change in the per capita consumption - total and item-wise - at the household level due to a change in the sex composition of adults as well as children in the household, from where we can make certain inferences on gender disparity ( Of course, this is not on discrimination). Several groupings of the households, to make them more homogeneous, were experimented and the average per capita consumption expenditure for all these groupings worked out in this approach.
The household monthly consumption data collected from the 300 selected sample households were suitably classified and six sets of tables were prepared with a view to examine how best such an approach would be useful in detecting gender disparity in household consumption. The six sets of tables were subjected to detailed interpretation in order to draw inference on gender inequality. The scope for exhaustive interpretation of these tables was limited because of the lack of values in all the cells in these tables, due to the small size of the sample we used in the present exercise. We have data from only 300 households and these 300 households, when classified into various groups according to a number of characteristics, the number of households in each cell in the tables was very small and in some cases even zero. This, however, is a problem due to the small size of the sample, which could be overcome by increasing the sample size and, therefore, does not stand in the way of using the method. Also, when the number of households in each cell is very few, the monthly average per capita consumption figures in these cells would naturally be less reliable and representative. The interpretations based on these cell values, therefore, are subject to this limitation, which, of course, would also vanish when the sample size is large. These tables and their interpretations, in short, are to be viewed with caution and meant only for demonstration purpose. They are meant to demonstrate the methodology and not meant for drawing any firm conclusions about the magnitude of gender inequality.
In the second approach we made use of econometric method. The regression method is believed to be the most suitable technique to disentangle from the total, the likely contribution or share of individual factors which collectively contribute to the total. In other words, the technique enables us to split the total into different components, each of which is attributable to one of the several specified explanatory factors included in the regression model. Linear multiple regression technique was, therefore, used to examine the marginal change (increase or decrease) in the total household consumption expenditure due to the addition of one male adult, female adult, male child or female child. Several alternative sets of regression models with different explanatory variables, as explained below, were used in this exercise with a view to examining how best the gender disparity in consumption could be explained through regression analysis.
Six sets of regression models were estimated for demonstrating the application of regression technique in examining the gender inequality in consumption, using household level consumption expenditure data, collected through NSS type of surveys. Approach was found to be encouraging to throw light on gender inequality, provided data from a large sample survey like the NSS, was used.
The dependent variables (Y) in these regressions were respectively the total monthly household expenditure on all items taken in our survey, monthly expenditure on cereals and pulses, milk and milk products, fruits and vegetables, fish, meat and egg, tea, coffee and beverages, tobacco and liquor, other non-food items, clothing, footwear, medical, education, transportation and entertainment respectively. The explanatory variables were the number of male children in the age group of 0-4 (X1 ), number female children in the age group of 0-4 (X2 ), number of male children in the age group of 5-12 (X3 ), number of female children in the age group of 5-12 (X4 ), number of adult males in the age group of 13 –59 (X5 ), number of adult females in the age group of 13-59 (X6 ), number of adult males in the age group of 60 and above (X7 ) and number of adult females in the age group of 60 and above (X8 ). All the observations pertaining to the 300 households in the sample were used in the estimation by ordinary least square method. The estimated coefficients were tested for their significance at 5% or 1% probability levels.
The gender inequality was first examined by classifying and tabulating the consumption expenditure data from 300 households into six different set of tables. These tables were prepared so as to arrive at comparable figures on per capita total and intensive consumption expenditure of males and females and to examine the gender inequality. These exercises did not provide any conclusive evidence on gender inequality in the selected sample households, though we could observe some amount of disparity. The fact that this small amount of disparity observed was not consistently in favour of any sex. In some cases it was in favour of females. This mixed picture of inequality is perhaps indicative of the absence of inequality of any serious magnitude in Kerala. The inference drawn here may be taken as tentative, as it is based on a small sample study.
The regression exercises showed small ‘R’ values, indicating the weak explanatory or predictive power of these regressions. This, however does not discourage us from using them for discussing the likely marginal addition to the household consumption due to addition of a boy, girl, adult male or adult female, and the gender disparity or inequality in consumption. Out of the fourteen equations, only in eight equations the regression coefficients had the expected positive sign. In the other regressions, some coefficients were found to be negative. All the coefficients are statistically significant in the case of three regressions, relating to total expenditure on all items, expenditure on cereals and pulses and expenditure on fish, meat and egg. The regression relating to expenditure on tobacco and liquor had positive coefficients for adult males and females, and negative coefficients for boys and girls. Further, the only significant coefficient in this case is that of adult male, as expected. The negative sign of the coefficients of children is indicative of the possible influence of the presence of children on the adult male to reduce the expenditure on tobacco and liquor, perhaps leading to greater saving on this account.
The over all tentative inference that followed from the regression analysis is that there is some amount of gender inequality in household consumption expenditure, but the inequality is not very pronounced. It does not seem to be a serious issue, particularly because it is sometimes in favour of males and some other times in favour of females. Also, the gender inequality is more pronounced among children than among adults, as per the results. Needless to point out that firm conclusions are possible only if the study is conducted using data from a large sample.
In this study, we experimented with a methodology using NSS type of consumption expenditure data for examining the gender inequality in consumption expenditure. Two different approaches were adopted in the study. First, we examined how best the NSS type of consumption data could be suitably classified and tabulated to bring out the nature of gender disparity.Secondly, we tried to demonstrate the use of suitably specified regression models for discussing the nature of gender inequality. Both these approaches were found to be moderately appropriate and effective in studying gender inequality in consumption.
The tentative conclusion that follows from the different exercises in this study using data from a small sample of 300 households is that gender disparity in consumption exists in a limited way in varying degrees in Kerala. In some cases it is slightly in favour of boys and in some other cases slightly in favour of girls. This shows that gender inequality is not of any serious magnitude in Kerala. This conclusion is subject to the limitations due to the small size of the sample.
Since the gender disparity does not necessarily imply gender discrimination which, however, contributes to gender inequality, we tried to investigate the nature and extent of gender discrimination practised in the households by using data and information collected with the help of separate questionnaires. The tentative conclusion that emerges from this analysis is that gender discrimination in consumption is not a serious issue in Kerala, though, of course, because of the cultural background, traditions and other obvious reasons, girls are, by and large, treated differently from boys, giving them more protection, safety and other considerations. Here again, we reiterate the limitation imposed by the small size of the sample in arriving at this tentative conclusion. Further studies using data from large samples of say 3000 to 4000 households, is necessary for confirming the conclusions drawn from the present study.
From the experience of the present study, it can be concluded that the method of analysis using appropriate classification and tabulation of data as illustrated in Chapter 6, is capable of bringing out the inequalities or disparities which exist between males and females in the matter of household consumption, although such inequalities are relatively very much less in Kerala because of many factors. Kerala is completely covered by the public distribution system which provides cereals at a subsidised price substantially lower than the open market price. Though the average number of days of employment of an average labourer is much less in Kerala than in other parts of the country, the wage rate is considerably higher and even a few days of work in a month will fetch an income sufficient to buy at least the required cereals for the family. Near full literacy and a very high level of education has changed the attitudes and practices followed in respect of gender. Close interaction for centuries between different religious and social groups which are all, almost equally numerous in Kerala, has resulted in the adoption by society at large, of different desirable practices from the other sections. Although the per capita domestic product in Kerala is low compared with most other states in India, the per capita consumption expenditure in Kerala is second only to Punjab, mainly because the shortfall in GDP is more than made good by the inflow of large scale remittances from abroad especially from countries in the Middle East and also from other parts of India. Because of all these factors, gender inequality consciously practised will be rare in Kerala as a whole although such inequalities may be prevalent in some isolated pockets and among some orthodox sections.
What is needed now is to try out this method on the basis of data obtained from a much larger sample. The first requirement is to classify the sample households into a number of classes on the basis of some criterion which reflects the income level of the households. Since it is difficult to ascertain the family incomes with reasonable accuracy and since a part of the income of a large number of families will be remittances from elsewhere which may not be disclosed correctly, total household consumer expenditure will be a more reliable substitute. Since per capita expenditure on different items will be used as the discriminating criterion, it will be necessary to classify the households into a large number of expenditure classes, each sufficiently narrow so that all the families in a class can be considered to enjoy practically the same level of living.
The next possible characteristic of the population which can have a bearing on attitudes to gender issues is the social group of the families. Social group is a combination of religion , caste and tribe. There is an underlying uniformity in the culture and civilisation of Kerala. At the same time, members of the different religions and castes retain a number of distinguishing features in their traditions, social practices and culture. So also are the regional disparities in the traditions, customs and practices. Kerala is a long and narrow stretch of land which had three distinct administrative regions and different Governments for long periods of time. This would have brought about changes and differences in the practices and attitudes in relation to gender issues. Therefore, it is necessary to ensure adequate number of sample households to provide the minimum required sample size in each specified group of households so that the conclusions drawn are statistically precise and valid. These remarks are, mutatis mutandis, valid in the case of the analysis using regression methods also.
