Lessons from a survey

On the question of including data on the caste identities in Census enumeration, a survey conducted in Andhra Pradesh is instructive, especially in terms of the mechanics of its design and conduct. This sample survey, which covered 75,000 households in t he rural and urban areas of the State, is one of the largest such surveys undertaken in the country.

Dr. T.V. Hanurav, a former Dean of the Indian Statistical Institute (ISI), Calcutta, was the principal investigator of the rigorous survey pertaining to caste-related data. Conceived, conducted and analysed strictly on statistical principles, the survey has been hailed as a model that must be replicated across the country.

Hanurav was trained in the University of Sheffield, United Kingdom and at the Michigan State University, the Texas A&M and the University of Minnesota in the United States besides the University of Waterloo in Belgium. He was senior adviser to the United Nations Educational Scientific and Cultural Organisation (UNESCO) in Belgrade.

A special invitee to the seminar on 'Caste Enumeration in the Census' held in Mysore, Hanurav spoke to Asha Krishnakumar about the sample survey, the results of which were submitted to the State Backward Classes Commission in 1998.

Excerpts from the interview:

What is the background of the socio-economic and educational caste survey in Andhra Pradesh?

In 1994, a leader from the rich Kappu community went on a fast demanding the inclusion of his caste in the backward caste (B.C.) list. Chief Minister K. Vijayabhaskara Reddy approached the chairperson of the State Backward Classes Commission, Justice K.S . Puttaswamy, to go into the issue. But due to lack of data, the government decided to do a survey not only to enumerate the numbers but also understand the socio-economic and educational profile of the Kappus.

However, as there were likely to be similar demands from other castes and as none could be labelled forward or backward, except in comparison, I suggested that one omnibus survey be done to get an unbiased profile of various castes and their social, educ ational and economic parameters.

The survey was conducted and the data scrutinised and analysed in accordance with rigorous statistical principles. An advisory committee of sociologists, economists and statisticians was consulted in designing the schedules to make it as objective as pos sible. As caste is a sensitive issue, the respondents were assured that the primary data would not be given to the government and that they would not be forced to respond (statistical techniques were used to deal with non-response). Although Rs.40 lakhs was budgeted, only Rs.30.42 lakhs was spent.

A pilot survey was done in November 1994 and the formal survey was launched in January 1995. The report was submitted to the BC Commission in April 1998. It is, however, yet to see the light of day.

What was the survey design?

The survey was done scientifically. I adopted a stratified random sampling method. We adopted the sample survey (and not the Census) method because of its advantage as, apart from saving on time and cost, it has been proved that a rigorous and scientific survey is much better than even a Census. This was established by the late P.C. Mahalanobis more than 60 years ago, after which was set up the Central Statistical Organisation (CSO), the National Sample Survey Organisation (NSSO) and the State Statistic al Bureaux.

The survey covered all the 23 districts from which 3,600 urban enumerator blocks and 2,000 villages were statistically chosen. Out of these, 50,000 rural households and 25,000 urban households were statistically shortlisted. The choice of even the ultima te sampling unit, that is, a household, was made, and not left to field personnel. Care was taken to include, as far as possible, castes with small numbers.

Nearly 90 parameters on housing, education, occupation, income, expenditure, assets, family size and so on, were collected and scrutinised, and an Index of Forwardness (or Backwardness) worked out.

A novel concept to measure the coefficient of friendship (between castes) was also included. Both qualitative and quantitative data were collected. Each of the qualitative classification was graded. For instance, 'structure of the house' had five classif ications, each superior to its successor; 'location of the house', four and so on. All the qualitative variables were then indexed using well-proven statistical techniques with coefficients ranging from 0 (most backward) to 1 (most forward).

Broadly, four basic indices (of social, education and economic status and of political empowerment) were worked out to calculate the Index of Forwardness.

The Index of Social Status is a weighted average of such indices as quality of housing, ownership of agricultural land and equipment and gender-specific job security and job levels. The relative status of women measures the progress towards gender equali ty.

Each of these components has sub-components. For instance, the 'quality of housing' has nine components, eight qualitative, with 10 per cent weight each, and a quantitative parameter (structure of the house) with a weight of 20 per cent.

Finally, each component index is normalised over castes to make it comparable and then assigned weights to obtain the weighted average. This is the Index of Social Status, which gives the relative rank of each caste in the overall social status.

Similar is the calculation of the Index of Education Status (computed from six components, including adult male and female literacy and indices of school-going children). An Index of Economic Status was computed from five components including ownership o f assets, per capita income, value of consumer durables and proportion of expenditure on non-food items. An Index of Political Empowerment was computed from an 'auxiliary sample' of households of members of randomly chosen elected bodies, such as gram pa nchayats, municipalities and corporations, mandal praja parishads and zilla parishads, and the district cooperative central banks. A statistical measure using grades of elected bodies and designations gives the index.

What were the results of the survey?

The Index of Forwardness was calculated with 40 per cent weight for social status, 30 per cent for education status, 20 per cent for economic status and 10 per cent for political empowerment. This value, normalised for various castes, gives the composite Index of Forwardness, 'F', which like other indices, falls between 0 and 1. For each of the five indices, the castes are ranked in the descending order. Ranks are given bypassing the non-Telugu population.

The 'F' distribution is divided into five: Central 25 per cent average castes; 25 per cent on either side, forward and backward; and 12.5 per cent on either end, most forward and most backward castes. This is only a guide and if need be the scale can be divided into seven.

There were 369 castes listed in the households. In the sample there were no households for 63 castes, 104 had less than 10 households; and 30 castes had non-Telugu households. Ultimately, 172 castes/caste-groups were identified and studied in detail. The tables were prepared in alphabetical order and the castes ranked according to the index and population percentage. The index ranged between 0 and 1.

This index apart, we also worked out the coefficient of friendship. At the end of the schedule, we asked each respondent to name their five closest friends. Castes were grouped into 25 broad categories of single caste or homogeneous groups of castes and a 25x25 matrix was worked out, which was used to obtain the index coefficient which, if closer to 1 denotes high inter-caste friendship. Interestingly, the lower and higher castes were found to mix well.

Why is the report not being discussed or used?

The report, in all probability, has some embarrassing and difficult results for the Andhra Pradesh government. It may have to derecognise some castes that are now in the B.C. list. For instance, converted Christians are ranked 34th (out of 172) and have 81 per cent of the population below them. How can they be called backward? Similarly, the Kappu community, which has 77 per cent of the population below them, cannot be called backward. The B.C. Commission wants me to provide the primary data. I will not give them as I have made a commitment to my respondents.