The twist in the great Indian inequality debate

What's the inequality in India? As people who have experience working with Indian data acknowledge, it's difficult to get data to answer such simple questions, in India.

To be clear, when we talk about inequality, we are talking about the income inequality. An ideal way to measure the income inequality would then be to gather data on the income of all households and calculate the inequality. For reasons that we can't go into now, it's not possible to do this.

The alternate measure of incomes of people is to measure their consumption, i.e. consumption is used as a proxy for income. The more you consume, more should be your income. It's a reasonable assumption. India conducts NSS that gathers data on consumption of a representative sample of households in India. Understandably, one can't survey every household in India, so only a sample of households is surveyed.

Income inequality (consumption) can be calculated using the NSS data. Inequality is measured using an index called Gini Coefficient. Gini Coefficient is an index which ranges between 0 and 1. 1 is perfect inequality and 0 is perfect equality. India's inequality as measured by NSS's consumption data is around 0.3 and is fairly consistent across years. To give a context, US's (income) inequality is around 0.45, China around 0.5, and egalitarian countries have around 0.3.

This was a discomforting fact to many economists. Some have argued that India's 0.3 figure must be a lower estimate because NSS doesn't survey enough rich people (they are too small in number and hence get left out of sampling, they don't respond to surveys etc) and hence they get left out, and so on.

On the other hand, some economic commentators, generally categorized as the "right", used the 0.3 figure to dismiss all those who raise the question of inequality in India. 

For instance, in a TV show, Barkha Dutt interviewed Thomas Piketty, India's Chief Economic Advisor Arvind Subramanian and NITI Aayog's Bibek Debroy. Bibek Debroy almost snapped at Thomas Piketty for raising India's inequality (and other issues like public expenditure on education and health), calling Piketty as lacking knowledge of nature and context of India's data and its Constitution

Bibek Debroy went on to argue that India's inequality is only around 0.33 (Gini Coefficient). Even if it is understated, it can go to a maximum of 0.44, if actual incomes are measured. Responding to it, Thomas Piketty (rightly?) advised exercising caution on 0.33 figure because it is a "self-reported" figure and also it is about "consumption", not income. Piketty also pointed out that, in his experience of working with many countries' data, the income inequality as measured by income data is often way higher than the inequality measured using self-reported consumption data. Piketty hence cautioned not to dismiss India's inequality. But, Debroy bulldozed the soft-spoken Piketty.

What's the truth? Is India's income inequality 0.3 or 0.5?

Branko Milanovic has written an excellent primer on Indian equality in his blog, resolving this debate. Milanovic points out that a new survey, called the IHDS survey captured the incomes of people. The income inequality in India (as measured by the income data) is 0.51. It's almost 0.2 points higher than the inequality calculated through consumption data (0.33).

How do we explain this seeming inconsistency? On scrutinizing the microdata of both surveys, Milanovic finds that there's nothing wrong in both the surveys, NSS and IHDS. Both capture the data accurately. The twist lies somewhere else.

While using consumption as a proxy for income, we assumed that consumption is proportional to income. It turns out that this assumption isn't exactly true. On comparing percentile data of both surveys, Milanovic finds that people in lower percentile, consume much more than their income. In the bottom percentiles, people consume double their income. This trend is observed till 33rd percentile. On the higher end of the spectrums, it's the reverse. People consume half their income. So, when you measure the consumption alone, for the people at the lower end, you are overestimating their actual income by almost a factor of 2. It thus brings down the inequality when calculated using consumption data.

For instance, consider a person A, whose income is 1,000 but 1,000 is not sufficient to sustain. He borrows another 1,000 and consumes 2,000 in that month. Consider second person B, who earns 10,000 but consumes only 5,000 and saves 5,000.

We go and survey both A and B. A reports consumption of 2,000 and B reports consumption of 5,000. We calculate inequality as the distance between 2,000 and 5,000. Without going into indexes, let's say inequality is 3,000.


This measure of 3,000 as inequality is misleading because A's income was only 1,000 and B's was 10,000. The actual inequality is 9,000. The difference between consumption measures and income measures arises because the lower percentiles consume more than their income and higher percentiles save a significant part of their income. Consumption measures thus give lower estimates.



The question then is - how are the people at bottom percentiles consuming double their income? Debt! Borrowing. It takes us back to an earlier post on demonetisation on why it is incorrect to argue "poor don't have cash savings, hence they won't suffer". Poor don't use cash savings like us, as an instrument to protect against uncertainties. They smoothen their consumption using credit. So, their transaction intensity, the amount of money transacted in proportion to their wealth, is high. As seen here, they are essentially surviving on credit. They have negative wealth. When cash is sucked out, it impacts the instrument of the credit thereby hurting the poor adversely.

Returning to inequality, the final answer to 'what's India's inequality', as Milanovic points out is that it's around 0.5, putting it in the list of the top unequal countries in the world. It should hence not be dismissed using the NSS's misleading figure of 0.3. 

On a side note, as it turns out, the soft-spoken Piketty seems to have more understanding of the "nature and context of data" and was right on being cautious about using self-reported consumption data as a proxy for income.

4 comments:

  1. Good post. Thanks for highlighting this issue!

    ReplyDelete
  2. Nice one!

    I had a doubt: why is measuring income inequality better than measuring consumption inequality? Remember the Indian context — difficult to get income numbers etc

    And second, "NSS doesn't survey enough rich people" might be a result of India not having enough rich people, which would help in lowering the Gini coefficient.

    ReplyDelete
    Replies
    1. Well, that's the main argument of the post. :) I think I didn't explain properly in the first version. I later updated the post, explaining it clearly.

      By inequality, we are mainly concerned with income inequality. Consumption is just its proxy. It runs into problems some times.

      For instance, consider a person A, whose income is 1,000 but 1,000 is not sufficient to sustain. He borrows another 1,000 and consumes 2,000 in that month. Consider second person B, who earns 10,000 but consumes only 5,000 and saves 5,000.

      We go and survey both A and B. A reports consumption of 2,000 and B reports consumption of 5,000. We calculate inequality as the distance between 2,000 and 5,000. Without going into indexes, let's say inequality is 3,000.

      This measure of 3,000 as inequality is misleading because A's income was only 1,000 and B's was 10,000. The actual inequality is 9,000. The difference between consumption measures and income measures arises because the lower percentiles consume more than their income and higher percentiles save significant part of their income. Consumption measures thus give lower estimates.

      On NSS sampling, as I mentioned in the post, it was initially argued that the divergence is due to fewer rich people in India, who are left out of sampling and they are also difficult to survey. But as it turns out, it may not be the major factor.

      Agreed that income is difficult to capture in country like India without fully developed tax systems. In its absence, it was a matter of debate since we were relying on approximations. Some said it's high, some said it's low. As noted in the post, some denied inequality as a problem in India, saying that Gini won't be beyond 0.4 even if we measure actual incomes etc.. But, as it turns out, it's more than that, 0.5, which is high.

      You can read Prof. Branko Milanovic's post for full details. He's the one who scrutinized both NSS and IHDS and came up with this insight. http://glineq.blogspot.in/2016/05/the-question-of-indias-inequality.html

      Delete