A Note on Studying Non-Agricultural Activity in Rural India
Introduction
A new set of claims on reduction in inequality in India has emerged following the publication of a World Bank report – one that has been highlighted by the Press Information Bureau (PIB). The evidence is based on the Gini index (a consumption-based inequality index, explained later) score that was estimated to be 25.5 for 2023, as compared to 28.8 for 2011–12. The Gini index ranges from 0 to 100, with lower numbers implying lower inequality. This projected improvement, as expected, has been hailed by certain sections.
The claim is that India has become the fourth most “equal” country in the world. As per the PIB report (2025), “India falls into the ‘moderately low’ inequality category, which includes Gini scores between 25 and 30, and is only a fraction away from joining the ‘low inequality’ group, which includes countries like the Slovak Republic with a score of 24.1, Slovenia at 24.3, and Belarus at 24.4. Apart from these three, India has a better score than all of the other 167 countries for which the World Bank has released data.” (Emphasis added.)
Economists, policymakers, and political experts have commented on this, and while the claims have been lauded by many platforms, there have been criticisms too. For one, on the level of inequality, it has been pointed out by Professor Santosh Mehrotra that consumption inequality (measured in India based on household consumer expenditure surveys (HCESs)) should not be compared with income inequality. The latter is reported by many countries and used by the World Bank in order to compare inter-country inequalities. Such a comparison is inaccurate, as income is defined as consumption plus savings. Poorer sections spend more on consumption than savings, while richer sections do the opposite. Therefore, a comparison of consumption inequality is not the same as income inequality, and likely to be lower.
A useful exercise was undertaken by Dr Surbhi Kesar on this, showing that if one compared India’s Gini index of pre-tax national income (using the national incomes from the World Inequality Database or WID) to that of other countries, India neither attained a high rank nor did the data show any improvement for India. First, according to her calculations, in the current period, India’s income Gini index was 0.61 in 2019 and 2023. She also found a steep increase in income inequality after 2009 – in 2009, India ranked 115 (out of 216 countries) in terms of income Gini, and in 2019, it ranked 176. So, contrary to the claims made by the PIB report, India’s rank in income inequality has fallen.
Second, returning to consumption, there are real concerns about a temporal comparison, and specifically, regarding comparing data from the National Sample Survey Office’s Consumer Expenditure Survey (NSSO CES) 2011–12 (68th round) and the Household Consumption Expenditure Survey (HCES) 2022–23. There are specific changes in sample design, recall periods, and questionnaire that pose important impediments in comparability (Kesar 2025).
Third, it has been argued that the claim of a decline in inequality does not match data from other indicators and reports that document rising wealth and income inequalities (Mehrotra 2025). With evident wage stagnation and irregular employment (decline in regular employment and rise in self-employment), any reduction in inequality needs to be scrutinised by social scientists. As argued by Das, Srujana, and Biswas, the aggregation of inequality through a single statistic, like in the case of the Gini coefficient, fails to capture exclusion by type of livelihood and across caste and gender.
In this Note, I offer a technical critique of the Gini coefficient and its limitations in understanding the level of and change in inequality. I do so by, first, providing an understanding of what is measured by the Gini Index of inequality and its limitations. Secondly, I attempt to show empirically why a decline in the Gini Index does not necessarily mean that existing inequalities (across socio-economic classes) are improving in rural India. The remainder of the Note (including the figures, numbers, and anecdotes) refers to rural India.
Gini Index in Application
First, let me explain what exactly is the Gini coefficient.
It is a measure of inequality based on a complete income or consumption distribution. The following equation presents a mathematical expression of the Gini coefficient:
Gini represents an average of “interpersonal income differences” relative to the mean income of the distribution (Ray 1998).
More specifically, in equation I, with n number of people in an economy: income (or consumption) of an individual is represented by X and the suffix alphabets (i or j) represent an individual among n number of persons in the whole economy. First, all n persons are arranged in ascending order of their incomes, so the i-th person has a lower income than the j-th person (or the 99th person has a lower income than the 100th person).
If we take the representative income difference between i-th person and j-th person, these differences should look like |Xi-Xj| and |Xj-Xi|. Since we are only interested in the absolute differences (hence the modulus sign), values of both will be the same.
Now, the first summation sign (where j is between 1 to n) means that all interpersonal income differences to be summed over j, that is, sum of all the differences: |Xj-Xi|. Then, the second summation would mean |Xi-Xj|, over i.
To avoid double counting of the cases in which i and j are repeated, we divide it by two. For each of these two summations, respectively, each (1/n) represents the average of the interpersonal differences (thus n squared in the denominator). Finally, x̄, represents the average income of all individuals.
The Gini index is therefore expressing average interpersonal differences relative to the mean income (or consumption) of the overall distribution.
Any acceptable inequality index is expected to rise with a regressive transfer (wherein the rich get richer and the poor get poorer) and fall with a progressive transfer (this is termed as the Pigou–Dalton Transfer Principle). The Gini coefficient satisfies this principle.
However, the Gini does not reflect upon the “magnitude” of such a transfer (even if progressive transfers take place) for different classes (income or consumption) within the same distribution over two different time periods. To illustrate, let us consider a four-person economy. In this economy, for year I, let the incomes would be 1, 2, 3, and 4 units. The Gini coefficient would be 0.25. Now, as an ad hoc policymaker, I would prepare policies that make everyone better in terms of their incomes, but at different rates.
So, the middle two people would experience an increase in income of 0.25 units, while the poorest and richest would experience an equal increase (in absolute units) – of 0.5 units in income. In year II, the income distribution would be: 1.5, 2.25, 3.25, and 4.5. Now, the Gini has declined (to 0.206 approximately). In the second period, total income increased from 10 to 11.5, thus an overall increase of 1.5 units. Therefore, the average per capita income changed from 2.5 units to 2.875 units, indicating an average 0.375 units of change in income.
The Gini declined in this case, although the middle two individuals experienced changes in income (0.25 units) that were lower than the average per capita increase in income (0.375 units). This implies that the middle two are not catching up with the top most or the richest, and are actually converging towards the poorest. Also, the difference between the richest and the poorest remained the same (3 units).
To put it in simple terms, a decline in the Gini cannot tell us if the gap between the bottom few income deciles is reducing, or the top few deciles is widening. The changing distribution in inequality cannot be captured by the Gini.
For understanding changes in inequality levels across households belonging to different classes, we need to look at changes in the overall distribution of incomes (Cowell 2000; Kolm 1976; Swaminathan and Rawal 2011). Despite this well-known fact, World Bank economists and PIB enthusiasts opt for the Gini index.
In the next section, I have attempted to show empirically how Gini in consumption expenditure declined in rural India between 2011 and 2024.
Real Changes in Levels of Consumption and Consumption Distribution
First, I compared average levels of consumption (in real terms) for each decile class (after sorting consumption in ascending order, and clubbing households into mutually exclusive groups) between 2011–12 and 2023–24.
It is important to look at real and not nominal consumption values, as inflation itself is a larger tax for the poorest (Bansal and Bansal 2025). However, before we compare the levels of Monthly Per Capita Consumption Expenditure (henceforth MPCE) between 2011–12 and 2023–24 (see Table 1), it should be noted that there have been important changes in the methodology (Sethu et al. (2024) have provided an excellent summary) (1) For the current analysis, I am not taking into account this problem.
Coming to MPCE, first, in real terms (2), the real consumption expenditure in rural India increased for all classes but at different rates for different decile groups of MPCE. Over the 12-year period, all classes (deciles of MPCE) showed an increase in the MPCE; thus, the overall average MPCE increased from Rs 1,427 in 2011–12 to Rs 2,082 in 2023–24.
However, if I make a simple assumption that in each year, this increase was uniform (which is, of course, inaccurate, as the COVID-19 pandemic affected the poor more than the rich), then the annual average increase would be about Rs. 54. Even with this uneven assumption, there is a class-differentiated story in the levels of progressive transfer.
Annually, the lowest decile witnessed an increase in MPCE of Rs 30.6, and the next decile (D2, i.e,. those with 10 per cent to 20 per cent of the MPCE distribution) had an increase of Rs 38.8. On the contrary, the top two deciles experienced an annual increase of real MPCE of Rs 75.3 (for D9) and Rs 65.9 (for D10), respectively.
Thus, in rural India, “progressive transfers” that led to a decline in the Gini were based on unequal changes in consumption across classes. The bottom 20 per cent population in the consumption distribution experienced a much lower average annual increase in consumption than those in higher deciles.
Table 1
Average Monthly Per Capita Consumption Expenditure, average annual changes in MPCE, by decile classes of MPCE, MPCE in constant 2011–12 prices and growth rates in per cent, rural India, 2011–2024
————————— table——–
A mere annual increase of Rs 30 and Rs 39 per annum for the bottom deciles and an almost double increase in the same for the top two deciles remains the crux of the story. The average annual increase in the per capita consumption expenditure remains class differentiated. Secondly, rural inequalities are not declining because the poorest are catching up at the same rate as the richest – but because the middle groups are “catching up” at a slower rate, their average consumption is getting closer to the averages of the bottom deciles (compared to before).
An economy in which the consumption of the middle 50 per cent has slowed down can have a perfect (nearly zero inequality) Gini index. However, the causes and consequences of this utter deprivation in rural India would not vanish.
To conclude, I have attempted to emphasise here certain technical limitations to using the Gini index. Without clarifying such innate characteristics of this inequality index, an overarching comment on declining inequality remains incomplete.
Acknowledgement
I thank Professor Madhura Swaminathan for her crucial inputs in the preparation of this note.
Correspondence: soham.bhattacharya@krea.edu.in
References
(1) A small clarification: in 2011–12, the NSSO consumption expenditure survey used the Modified Mixed Reference Period (MMRP), and the recall periods were different from earlier rounds for certain consumption items, such as food and perishables, durables, and other items, in one of their schedules. Current rounds of the consumption expenditure surveys (HCES 2023–24) can at best be compared with that recall period, and I have followed that here.
(2) I have used the Consumer Price Index of 1986 as the base year. Thus, in 1986, it was 100; in June 2012, it was 646; and in June 2024, it was 1280. So, to have 2024 prices at 2012 levels, all the values need to be divided by (1280/646), i.e., 1.98. This is called splicing.
