Income Inequality and Wealth Inequality in India

Section 1: Introduction

A recent recent working paper titled “Income And Wealth Inequality In India, 1922-2023: The Rise Of The Billionaire Raj” (Source: S1) (“Paper”) created quite a debate from both sides of the spectrum regarding inequality in India since 2014.
In this article, we shall look at some of my thoughts on the Paper. This shall cover the following aspects:

Problems with wealth inequality calculations
Growth trends in the incomes of the bottom 50% from 1993 to 2022
Problems with the savings rates assumptions
Problems with using IT returns
Poverty reduction post independence

It should be noted that this article does not attempt to carry out alternate calculations but rather points out some issues with the Paper.
The main data sources (not all inclusive) used by the authors of the Paper include (i) Income Tax Returns data, (ii) National income data, (iii) Indian Billionaire Rankings published by Forbes, (iv) Household consumption expenditure surveys (HCES) conducted by the Government, (v) Periodic Labour Force Surveys (PLFS) conducted by the Government, (vi) All India debt and investment surveys (AIDIS) conducted by the Government. The main assumption behind calculations of income inequality is that Government surveys like HCES, PLFS, AIDIS do not capture the wealth/ incomes/ consumption patterns of the Top 10% of the population and we should look into other data sources like the Tax returns data, Rich lists etc. to get an idea about the same. My article titled “Missing the Rich - Analysis of NSSO Data” (Source: S2) also independently analyses PLFS and HCES to conclude that:

The absolute income/ consumption (at an aggregate level) for the top 10% captured by PLFS/ HCES is most likely an underestimate compared to the actual numbers.
The growth the income/ consumption captured by PLFS (and its earlier avatar, Employment-Unemployment Survey (EUS))/ HCES from 2011-12 to 2017-18 is directionally wrong i.e., the per capita incomes/ consumption of the top 10% has supposedly declined as per EUS-PLFS/ CES while in all probability it would have increased.

Before we delve further, in an earlier paper titled “Top Indian Incomes, 1922-2000”, published in 2005 (Source: S3), the income share of Top 1% in India was ~9% for the year 2000. However, in the current paper (Source: S1), it increased to 15.1% for the same year. It would be good for the authors to provide more details on why this difference occurred since it is substantial.

Section 2: Key findings of the Paper

As per the Paper: “Our estimates suggest that inequality declined post-independence till the early 1980s, after which it began rising and has skyrocketed since the early 2000s. Trends of top income and wealth shares track each other over the entire period of our study. Between 2014-15 and 2022-23, the rise of top-end inequality has been particularly pronounced in terms of wealth concentration….. In other words, the ‘Billionaire Raj’ headed by India’s modern bourgeoisie is now more unequal than the British Raj headed by the colonialist forces. It is unclear how long such inequality levels can sustain without major social and political upheaval.”
Such strong conclusions especially on social/ political upheavals or more unequal than British Raj should be based on iron clad data. Is this the case?

Section 3: Problems with wealth inequality calculations

In case of calculating wealth of individuals, AIDIS is used to evaluate the same. As mentioned above, one of the key assumptions of the Paper (with valid reasons) is that the Top 10%ile or thereabouts of the population’s income/ wealth/ consumption patterns are not captured by the Government survey datasets. The Paper mentions “To put numbers to the issue at hand, consider this: the ratio of maximum wealth in the Forbes rich list to the maximum wealth observed in AIDIS was 3279 in 2012 and 7163 in 2018. In other words, there is enormous under-estimation of wealth at the top in surveys. Further, the issue of non-representativeness of the rich population appears to be worsening over time, especially with the last round of the survey conducted in 2018 - the total net wealth from the Forbes list as a percentage of total survey-based wealth increased from 1.26% in 2002 to 2.74% in 2012 to 6.01% in 2018 (based on 5, 46, and 117 individuals respectively).” This is in line with my conclusions on PLFS and HCES 2017-18 as mentioned earlier.
Nevertheless, there are clear breaks in terms of data usage for wealth inequality distribution as follows:

From 1961 to 2001, wealth inequality was just based on AIDIS data.
From 2002 onwards, it was based on AIDIS for around 90% of the population and Forbes’ Indian Billionaire Rankings for the richest set (basically billionaires) and the wealth of the remaining population was extrapolated.

Pic 1

Source: S1

As one can observe in the above Pic, till 1982 the wealth share of Top 1% was stagnant with a slight increase during 1992. However, with the introduction of Forbes Billionaires list to the data sets from 2002, there is a huge jump in 2002 compared to the earlier periods. It kept on increasing thereafter till about the times of Global Financial Crisis when stock market crash seems to have made India more “equal”. It further started meaningfully increasing after the increase in share market prices post COVID where it is around 40% as of 2022.
Hence, one should be careful in comparing the Wealth Inequality in India especially from 1961 onwards due to the additions of data sets. Further, even for Forbes database/ Hurun richlists whether it always covered all the billionaires (or individuals with incomes more than INR 1,000 cr) in the country or whether there has been an increase in data availability in the public domain over the years needs to be studied in detail before reaching any conclusion on Wealth Inequality.
Further, as per AIDIS (Source: S4), 49% of the individuals’ assets in urban areas and 69% of individuals’ assets rural areas is from land. Land was valued as per their normative/ guideline values. However, it is very clear that circle rates/ guideline values (as per Government) may not be equal to the actual market values. As per an article (Source: S5), even in metros the market values can be higher than guideline values anywhere between 6% to 75% depending on the location. This would indeed be the case in other locations in India as well. On the other hand, the net worth of the richest as per Forbes would be substantially based on values of companies (Listed/ unlisted) which is by default the market values for those companies. So, here is another comparability issue.
While it would be easy for ordinary people to offload their investments in stock markets without it causing a market crash, for many of the promoters (who will be included in the Rich list) to sell their shares in their companies would be quite difficult since any offloading of their shares (without a proper deal taking place) would most likely lead to a big crash in the market prices of those stocks. So, there is a big issue with actually realizing "wealth" for many of the rich while for ordinary people it is not that difficult. While technically not the same, it is akin to a holding company discount commonly referred to in valuation.
Thus, to conclude:

The wealth series from 2002 onwards is incomparable with the earlier period. So, at most one may say that India is the most unequal currently compared to the past 20 years.
Whether Forbes billionaire list/ Hurun list started getting more data in later periods compared to the earlier periods. E.g., the Forbes billionaire article for 2007 quotes an expert stating: friendlier tax laws, notably the 2004 elimination of capital gains tax on the sale of equity shares, encouraged entrepreneurs to list their companies. In the past, punitive taxes forced businessmen to stash away their personal wealth in a complex web of trusts and private holding companies…A lot of hidden value and assets have now come out of the closet" (Source: S6)
Comparability issues between land valuation as per AIDIS and market valuations of stocks.
Difficulty for billionaires for liquidating their potential net worth in their companies while it is easier for ordinary individuals

Section 4: Growth trends in the incomes of the bottom 50% from 1993 to 2022

Let’s take a simple example for this. In 2009, India’s GDP was ~INR 65 tn in 2009 and INR 87 tn in 2011 (Source: S7 and S8). The Paper says that bottom 50% earned 16.5% and 15.3% of the national income in those years respectively. This leads to bottom 50% earning INR 10.7 tn in 2009 and INR 13.4 tn in 2011 i.e., almost 25% growth in nominal incomes. However, the population increased by ~3% during the same period and the cumulative inflation was ~18% leaving just a growth of 0.6% CAGR in incomes for these 2 years. However, as per the HCES 2011-12, the growth in consumption expenditure for the overall population (which in turn is linked to incomes) during this period was at 7.1% CAGR from 2009-10. So, here’s another mismatch. I have done this analysis for a thirty-year period from 1993 to 2022.

Pic 2

Source: S1, S7, S8, S10, S11 and author’s calculations

Pic 3

Source: Pic 2 data and author’s calculations

Pic 4

Source: S1, S9 and author’s calculations

As can be seen in the pics above, except for the periods from FY05 to FY10 and FY12 to FY23, the growth in incomes of bottom 50% is at a significant variance with the growth in consumption.
Just as a cross-check, I also checked the change in Poverty rates over this period and compared with the growth in incomes of the bottom 50% as per the Paper. As can be seen from the pic below, poverty reduction (i.e., due to increase in consumption expenditure (and also income increase) of the poorer segment) happened at the fastest pace from 2009 to 2011 exactly when the growth in incomes as per the Paper is at the lowest.

Pic 5

Source: S1, S12, S13 and author’s calculations

Section 5: Problems with savings rate assumptions

As per the Paper, “While certainly a relevant proxy for incomes, consumption expenditures are likely to under-state incomes of the rich given their relatively lower marginal propensity to consume. To work around this issue, we follow Chancel and Piketty (2019) and rely on the India Human Development Survey (IHDS) which collected data on both incomes and consumption expenditure to estimate consumption-to-income scaling ratios at each percentile of the consumption distribution. Letting 𝑦 denote income, 𝑐 consumption, and 𝑝 percentiles, we begin by estimating ‘raw’ consumption-to-income scaling ratios as 𝛼𝑝 = 𝑦𝑝/𝑐𝑝. One issue when do this is that we are confronted with negative savings. We therefore define an alternate set of ratios where we winsorize the ‘raw’ ratios at 1 (from below) such that incomes are assumed to be at least as large as consumption. Our final choice that feeds into our benchmark series is an average of these two scenarios (which only affects the ratios at the bottom as ratios in the middle and top are not affected by the winsorization). Finally, since we have two rounds of IHDS, we estimate and average the ratios from the two round. These final scaling ratios are then applied to each percentile of the consumption distribution estimated from CES to arrive at a survey-based income distribution.”
It seems like a constant savings rate has been used for the whole period from 1922 to 2022 which is clearly incorrect. The Employment-Unemployment surveys (the earlier avatar of PLFS) (EUS) provided the incomes of regular wage earners and casual wage earners (who constitute almost 50-60% of the earning workforce). It did not give the incomes of Self-Employed workers. So, based on the weighted average earnings of regular and casual labourers and multiplying that with the workforce participation rates (after excluding unpaid family workers), one can arrive at a per capita income for the population (of course this does not include income from remittances and pension incomes). A more detailed explanation is of this methodology is provided in my article titled “Alternate methods of benchmarking consumption expenditure” (Source: S14). Similarly, we have the consumption expenditure of the population based on HCES. Since the 2017-18 survey was based on MMRP recall period, all the previous periods were also adjusted upwards to arrive at extrapolated MMRP based expenditures on per capita basis. With this, one could calculate the Savings rate for the population for the period from 1993-94. This has been compared with savings rates using IHDS and National accounts.

Pic 6

Source: S7, S8, S10, S11, S15, S16, S18, S19 and author’s calculations

As can be seen above, the savings rates as per EUS-HCES and even IHDS has been constantly increasing. Further, the savings rates as per National Accounts (referred to as Actual in the pic above) and EUS-HCES are much higher than the IHDS savings rates for FY05 and FY12. So, using a constant savings rate and applying it on consumption expenditure over time will effectively overstate the incomes in the earlier periods and understate incomes in the later periods.
Since the EUS incomes are based on earnings of Regular and Casual labourers, it is necessary to see what the incomes of Self-Employed could be. Based on EUS one can arrive at the Consumption expenditures of the Self-Employed, Regular and Casual employees for 2004-05 and 2011-12 for Rural and Urban areas. As can be seen from the Pics below, the consumption expenditures (and consequently the incomes) of Self-Employed class is somewhere in between Regular and Casual labourers for both periods and for both urban and rural areas. So, the savings rates can effectively be derived using EUS-HCES rather than using IHDS surveys.
Further, the PLFS provides incomes of the Self-Employed class as well. So, from 2017-18 onwards one can effectively use income distributions directly without converting expenditures into incomes.

Pic 7

Source: S17, S18 and author’s calculations

Pic 8

Source: S17, S18 and author’s calculations

Section 6: Problems with IT Returns

As per the Paper, “Our estimates suggest that inequality levels declined post-independence till the early 1980s, after which both top income and wealth shares began rising and have skyrocketed.”
As per an article (Source: S20),

The top income tax rates was at 82% in 1951-52 which got reduced to 65% in 1965-66 gradually.
It was increased to 98% by 1973-74 and gradually reduced thereafter to 72% in 1980-81
It was further reduced to 53% in 1986-87 and by 2000-01 it was at 33%

So, why on earth will one disclose their full incomes if the tax rates are at 70% or more. The actual tax rate reductions happened from the mid-80s and was brought down to the 30% levels only by 2000. So, isn’t it clear that the disclosures would have improved since the mid-80s and even more so since the 2000s? This is exactly the point at which the Paper said inequality started rising.

Section 7: A Word of Caution

Before we conclude, it should be acknowledged that inequality is a problem which must be dealt with using appropriate policy measures to the extent possible. However, since we are a poor country, one should also look at our efforts at poverty reduction since independence.
While full-fledged consumption expenditure surveys were carried out only since 1973-74, consumption expenditure surveys for smaller samples were conducted in the earlier periods as well and a paper called “Poverty In India And Indian States: An Update” by Gaurav Datt (Source: S21) gives India's poverty rates from 1951 onwards. We also look at poverty rates based on Tendulkar Poverty line from 1993-94 onwards. As can be seen below, there was barely any poverty reduction from the 1950s to the early 1980s even though inequality came down as per the Paper in this period. On the other hand, the period during which inequality increased substantially coincides with substantial decline in poverty.

Pic 9

Source: S12, S13, S21

Poverty as Multi-Dimensional Poverty published by UNDP also reduced substantially from 1998-99 onwards (the period when inequality skyrocketed). Multi-dimensional poverty basically covers non-monetary aspects of poverty like depravation in Health (mortality and nutrition), Education (years of schooling and attendance), Standard of living (electricity, drinking water, sanitation, housing, cooking fuel, owning assets like two-wheelers, ACs etc.).

Pic 10

Source: S22, S23 and author’s calculations (the estimates for 1998-99 was slightly different and had to be adjusted by the author to make it a continuous series)

Section 8: Conclusions

One cannot deny that inequality does not exist in India. Inequality is significant and has existed since ages and may have even increased since the 2000s as asserted by the Paper.
However, some issues with the Paper seems to be:

Growth in incomes do not properly match with growth in consumption as per HCES
Savings rates assumed does not look correct considering an almost constant savings rate has been assumed for decades together
Considering the extremely high tax rates till early 80s, usage of tax returns data from colonial times may have comparability issues with the later periods. This is mainly on account of people not disclosing their incomes adequately till the late 80s or even 2000s. Hence, any comparison of income inequality in the 2000s with the earlier periods (even with the 70s and 80s let alone the colonial period) seems to be fraught with issues.
Wealth inequality also suffers from comparability issues since a different data set is added from 2002 onwards and even for that data set the disclosures may have improved since 2004-2005 onwards which showed up as higher inequality.
Comparability issues between market values and guideline values

My issues are more with the long-term trends rather than the absolute levels of inequality. Even assuming the Paper correctly estimates India’s inequality trends and provided it increased substantially post the 80s and especially since 2000s, it should be contextualized with a substantially higher poverty reduction from the mid-80s and especially from 2004 onwards. One should also think what a potentially poor person would choose among the following options:

Increasing incomes but a declining share of the national income
Stagnant incomes but a higher share of national income

One would lead to lower poverty and the other would lead to lower inequality. India has supposedly had both these phases. Unfortunately, we may not have had a phase wherein substantial poverty reduction occurred along with inequality reduction.