In a recent study titled ‘Indian income inequality dynamics (1922-2014): From British Raj to Billionaire Raj?’ published on WID.world, we presented new estimates of the distribution of national income in India, from 1922, when the income tax was introduced, up to 2014. In this study, we systematically combine the best available data at hand and document a sharp rise in the level of inequality. According to our benchmark scenario, the share of national income accruing to the top 1 per cent Indians went up from 6 per cent in the early Eighties to nearly 22 per cent today, a level comparable to the 20th-century inequality peak observed in the 1930s, during the British Raj.
This paper has generated a fair amount of discussion in the media and in the research community. While several researchers pointed out that the new inequality estimates are consistent with other recent research findings, a few columnists questioned our results. We should stress at the onset that our conclusions on the extreme levels of income inequality reported in India, and on its rise since the early 1980s, remain unchanged. The new WID.world estimates appear as the most reliable source of historical income inequality data to date. At the same time, it is clear that much work still needs to be done by researchers, official statistical institutions and policymakers to better measure income and consumption inequality in India.
Official data to measure economic inequality in India and many other countries relies essentially on household surveys. This source of information has well-known limitations when it comes to measuring economic inequality: Surveys do not provide an accurate representation of income levels among the richest. This is particularly problematic since the dynamics of inequality were largely driven by the explosion of top incomes over the past decades, as is shown in many countries covered in the WID.world database (the most extensive historical database on income and wealth inequality).
Contrary to household survey data, tax data makes it possible to observe such evolutions with a much higher degree of precision. The reason is simple: Tax data covers all taxpayers at the very top of the income distribution, whereas household surveys generally do not cover at all the richest individuals. In addition, when asked for their income levels by survey interviewers, interviewees have no incentives to report their true income level. It was actually shown that the richest tend to largely underreport their incomes in surveys. On the contrary, tax data leads to more precise answers — at least, there is some degree of sanction against fraud. Indeed, non-compliance and tax evasion is a reality in India and other countries. This means that tax data should be treated as a lower bound to inequality levels, rather than an upper bound.
In a country like India, tax data covers only a small portion of individuals (the richest 7 per cent adults in 2013-14). To track inequality among the entire population (that is not only the top 1 per cent but also the bottom 50 per cent or the middle 40 per cent), it is thus necessary to cautiously combine tax data and household surveys. This is what we do in our recent paper and more generally in the WID.world project, which regroups more than a hundred economists over all continents. We seek to ensure that inequality estimates are consistent with national accounts, meaning that income growth rates for different groups of the population are consistent with average income growth rates discussed in public debates and comparable across countries.
In a recent article, Surjit Bhalla (‘Piketty has got it wrong’, IE, January 20) argues that these new income inequality estimates for India should not be trusted. First, the author argues that there is a relatively large difference between these numbers and previous estimates published by Abhijit V. Banerjee and Piketty in 2005. Our new methodology led to an upward revision of previous income inequality estimates largely because of the newly available tax data and because we combine tax and survey data in a more consistent manner.
Second, the author claims that new Indian income inequality numbers do not conform to any known model of savings behaviour and as such cannot be right. This is a peculiar statement. When facts do not conform to an abstract theoretical model, this does not necessarily mean that facts are wrong. It could also be that abstract theoretical models are what they are, theoretical models, and that they do not properly account for the complex dynamics of socio-economic inequality. Fortunately, tailors reason differently with their own models: When they find that the sewing pattern they have been working on does not fit the size of their client, they review their pattern rather than cut their client’s arms off. Let’s make it clear: Models are extremely useful to assess and generalise empirical research findings. But they should not be taken for granted.
Third, in his review, Bhalla mobilises data sources that cannot be compared as such, meaning that little can be inferred from his exercise. In order to measure savings rates for different groups of the population, one needs information on income and consumption levels of these groups. Bhalla uses income levels from WID.world, whereas for consumption, he relies on household surveys — those very same surveys which fail to capture income and consumption levels of rich individuals. By comparing artificially low consumption levels among the richest (obtained from household surveys), with more reliable income levels from WID.world, anyone would come up with savings rates for the richest that are much higher than what they actually are. By doing so, Bhalla unsurprisingly finds that the top 10 per cent richest make up 100 per cent of India’s savings in 2011.
At the moment, there is, unfortunately, no consistent source of information to obtain consumption levels among the richest in India. We hope to be able to provide such estimates in the near future on WID.world. Very simple estimations can, however, already help us to see more clearly here. Assuming that household surveys misreport consumption levels of the richest as they do for income, we can infer the true consumption levels of the richest on the basis of income levels measured in WID.world. When doing so, we find that the top 10 per cent make about three-quarters of total savings in 2011 and that the middle 40 per cent make up the rest. This is quite different from what Bhalla finds.
There is no doubt that measuring income and consumption inequality is a difficult enterprise, particularly in India where low data quality and availability raise a problem of democratic accountability, independently of the level of inequality observed. While our results are robust to a wide range of alternative scenarios accounting for data limitations and while they are consistent with other recent research findings, we certainly do not claim that they are perfect. We very much welcome all efforts to contribute to improving the quality of such estimates in the future. In this enterprise, the government should also play its role and release tax data for the 2000-2010 period. It is indeed surprising that the government stopped publishing such a crucial source of information to track inequality, precisely when the country entered its digital age. It is also critical that the government provides access to anonymous micro files of income tax declarations to researchers. Without increased transparency, it will be impossible for Indian society to have an informed democratic debate about rising inequality.