Caste determines how long a person lives, shows study
A study shows that caste plays a role in determining life span and perception of good health even after accounting for differences in nature of work.india Updated: Apr 13, 2018 09:12 IST
An upper caste person is likely to have more land than a lower caste person in India. The former also has a better chance of having a decent job. In fact, caste is important in determining access to many kinds of privileges in India. Does it also play a role in determining how long people live?
Headline numbers from National Sample Survey Office’s (NSSO) Morbidity and Healthcare Surveys (henceforth survey) claim the answer to this question is yes. According to the 2014 survey, average age of death in Scheduled Tribe (ST) households was 43 years. This was 60 years in non-Muslim upper caste (NMUC) households. A similar pattern was observed in the 2004 survey as well. Statistics show that longevity increases as we go up the caste hierarchy. The surveys consider households which had reported at least one death in the preceding year.
Statisticians would object to such simplistic reasoning, though. After all, most ST households are poorer than their upper caste counterparts. This also means that they are likely to be living in unhealthy conditions.
How does one know whether the shorter life span of a member in ST household vis-à-vis that in an upper caste household is not just a function of their poverty? In other words, if all upper caste households were as poor as ST households, would there be any difference in longevity, a statistician would ask.
Statistical techniques actually allow us to apply such controls on a given data set and answer this question. An Economic and Political Weekly paper by Vani Kant Borooah, who teaches economics at University of Ulster in the United Kingdom, is based on such an exercise. His analysis shows that caste did play a role in determining longevity in the 2014 survey even after one takes into account the role of other factors. Borooah uses four other factors along with social group to estimate what he calls average age at death (AAD).
These factors are labourer or non-labourer nature of work, rural or urban domicile, living in a forward or backward state, and quality of the household’s toilet and cooking fuel. Each of these four factors plays a role in determining longevity. Chart 3 shows average age at death values estimated by Vani Kant Borooah for 2014 and 2004.
All the 2014 estimates are statistically significant at 1% level. This means that if one starts with the hypothesis that the average age at death for all other social groups are equal to that of Non-Muslim upper class households, then the probability of holding this hypothesis will be less than 1%.
Source for all charts: Caste, Religion and Health outcomes in India, 2004-2014, Vani Kant Borooah, EPW, March 2018
For the 2004 survey, SCs were the only social group whose average age at death was lower than non-Muslim upper castes at 1% significance level. For Muslim OBCs and Muslim upper class households the results are valid at 10% confidence level.
The survey also asks people aged 60 years or more whether their current state of health was excellent, fair or poor. It is a bit intriguing that even though longevity has increased between 2004 and 2014, a greater share of the 60-plus population felt that their health was poor. Across social groups, the share of those feeling their health was poor was the lowest among STs and non-Muslim upper caste persons in 2014 (Chart 2).
Boorah applies additional statistical controls — monthly per capita household expenditure, education level, gender; age and marital status and number of living children — to filter out the role of social background on self perception of health in addition to those considered for analysing average age at death.
The results show that people belonging to non-Muslim upper caste groups have a lower probability of poor self-perception of health than most other social groups in both rounds.
For 2014, these results are statistically significant at 1% level for STs, Muslim OBCs and Muslim Upper Class households. For 2004, they are statistically significant at 1% level for all households except Non-Muslim OBCs (Chart 4).
The paper does identify shortcomings in its method.
Health outcomes depend on a whole lot of other factors, such as smoking, diet, nature of work etc. which the study cannot account for due to lack of data.
Disproportionate presence of any of these attributes in a given social group could have contaminated the results Boorah has arrived at.
However, given the caveats, the findings give an important message. Belonging to a socially disadvantaged group can seriously limit a person’s physical ability to function in society, the paper argues.