Interview: Kaushik Krishnan on 'V-shaped recoveries' and how the lockdown affected Indian households

Kaushik Krishnan is chief economist at the Centre for Monitoring Indian Economy, an independent think tank that produces economic and business databases. Krishnan’s work focuses on CMIE’s flagship Consumer Pyramids Household Survey, which it touts as “the world’s largest household panel survey, the largest regular household survey in India, and the largest household survey to be conducted during the COVID-19 pandemic and lockdown.”

Based on data from more than 75,000 households over the course of this year, Krishnan co-authored an analysis that questioned assumptions about India seeing a “V-shaped” recovery – ie, a quick bounce-back following the remarkable contraction in the economy due to the lockdown earlier in the year – and explained how all was not what it seemed with the unemployment rate, which had seemed to recover almost fully.

Over email, I spoke to Krishnan about what indicators he’s focusing on, whether India is really on the path to recovery as many claim it is, and what we still need to understand about this unprecedented year.

Tell us a little bit about yourself and how you came to be chief economist at CMIE.
I wish I had a clear, linear story but reality is often messy! I actually started life as a lawyer, and went to law school in Kolkata. Somehow, I got it into my head that I should study economics.

While I was doing my masters in law, I took a course on microeconomic theory that really changed the way I saw the world and so I applied for a PhD in economics, and was mercifully admitted. After the PhD, I spent some time in the private sector as well as in academia.

When the opportunity to be involved with CPHS came along, it was too good to pass up. I have an almost religious belief in the utility of a high-frequency, high-quality household panel for economic research on India.

What were the trends regarding household well-being going into 2020? What were you expecting to look at over the year, before the pandemic changed everything?
There were signs of a slowing economy even in late 2019. Household incomes were declining and there were some hints of labour market distress. The pandemic made things much worse and makes the recovery more difficult.

In a way, the period right before the pandemic is quite crucial to study. Why were things already looking down at the beginning of this year? How bad could it have got? Are there lessons from that time about structural issues that require to be addressed, regardless of the pandemic?

So… are we dealing with a V-shaped recovery? What is your letter of choice?
Maybe the Kannada letter, ಈ? On a more serious note, [CMIE Managing Director and CEO] Mahesh Vyas has described the recovery as the greek letter nu (ν), where the uptick of the v is at a shallower slope than the downtick. This has broadly been the trend so far. The question really is how steep or shallow is the slope of the recovery side of the “v”.

The other day, a friend of mine at the IMF told me about “K-shaped” recoveries, where inequality is exacerbated as the recovery takes place. It’s interesting that our data doesn’t suggest that. If anything, the households with the highest per capita income have not recovered as fast as other groups.

Keep in mind, all of this is about households, which is what I focus on at CMIE.

What are the things you look for in trying to understand what is happening with the economy in an extraordinary situation like this year?
So the Consumer Pyramids Household Survey measures household well-being, to which there are many dimensions. Broadly, these are the “themes” that I would pay close attention to.

• Employment: This is what CMIE is best known for. Mahesh writes a weekly column on employment in our subscription service, Economic Outlook. We make our unemployment data freely available and do a lot more analysis in Economic Outlook. A recent piece with Marianne Bertrand, Rebecca Dizon-Ross and Heather Schofield also speaks about employment. The next research seminar that we will host is on different employment trajectories through Covid. So employment gets a lot of attention, as it should. One question that requires more investigation is whether young people joining the labour force, like students, are able to find adequate work.
• Incomes: Not all work is the same. The job you have now may be worse than your previous job. The same job may pay less than before. We find some evidence of this in our piece. Many occupations paid less money during the pandemic than before. Others have found that a large number of salaried workers moved into more informal arrangements. All of this has an impact on income earned, which continues to remain lower than it was a year ago.
• Household debt: As incomes start to dry up and consumption pressures remain, households may try to turn to debt. But the supply of formal debt may be low. In the next few months, we need to find out whether households with outstanding debt will be able to service their obligations, and what will happen if they can’t. Will they take on more debt just to meet consumption expenditure requirements? CPHS captures information on household debt from formal and informal sources for all kinds of purposes, including the accumulation of debt to repay prior debt, which could be particularly bad. There is already some work done on borrowing from friends and relatives during the pandemic. More work on household debt needs to be done.
• Non-food consumption: Many people say that the economy needs a revival in demand to fuel its recovery. We read news of a large bump in online sales and purchases around the holiday season but we don’t know how broad-based that is. The data from CPHS suggest that non-food expenditure, which was improving after April, still had a long way to go in June. It also started to flatten out in July. Compared to non-food expenditure in July 2019, households are still 35% below. Our data on consumption is lagged by a few months, so we don’t yet know how much the holiday season rallied consumption expenditure. We’ll need to watch this number as the months pass.
• Food consumption: Food consumption expenditure is remarkably stable for households before the pandemic. Covid hit food consumption, like it did, everything else. By July, overall food expenditure is about 12% lower [year-on-year]. Expenditure on cereals and pulses was 18% lower than it was a year ago while dairy, eggs and meat expenditure recovered to about 10% lower than its year-on-year average. Especially in households with children, food consumption patterns tell us about nutritional attainment. This depressed spending is something to watch out for.
• Education: Human capital formation was disrupted by Covid-19. It can play a large role in future earnings. We have not looked into this, but one needs to see whether learning is still on track. Are students dropping out of school or college? Are they spending less time on learning? CPHS tracks all this, but we’ve not yet seen anyone try to unpack this.
• Health: We capture some data on health. There is a growing literature that seeks to address the impact of negative health shocks on households long after the health episode has passed. Healthcare and medication can be expensive in time and money for many households. Researchers are now starting to analyse this information and CMIE is very keen on learning more about the physical and mental health of the households in our panel.

The primary focus for CMIE is employment, although clearly you work on other indicators as well (including collaborating on a Covid-19 survey). How trustworthy is the data that’s coming in? Are you engaging with it differently this year?
The primary focus of CMIE is not employment! Employment isn’t even the focus of our household survey. The survey began in 2014. Our work on employment came about in 2016 due to a partnership that CMIE had with BSE. Almost all of our work on employment comes out of one indicator that we collect in CPHS, employment status.

On employment alone, this is one of 9 or 10 indicators we collect. Employment in itself is a very small part of our entire survey. My answer above speaks to all the other areas for which we have surveyed households now for almost 7 years.

We rely heavily on our data collection and quality assurance processes while running the survey. I’ve given some talks on how we sample households as well as on how we execute, monitor and assemble the data from our survey. We survey 174,405 households across 28 states and UTs. Our survey covers over 3,900 villages and over 320 towns across the country. Each village and town is randomly selected. Within a village or town that we survey, each household is randomly selected. Surveying is done daily by a staff of over 200 people, whose work is monitored in real time by almost 100 supervisors.

All of the documentation on this is publicly and freely available on our website (you might need to sign in, but the CMIE user ID is free to create).

The lockdown didn’t fundamentally change any of this. At the worst times of the lockdown, we had to move from in-person surveying to surveying on the phone. This didn’t alter our coverage of households. Over 92% of our households have working phone numbers. We took a host of additional measures to ensure data quality remained high.

In fact, the representativeness of our survey depends crucially on two things – the geographic spread and the rural-urban balance of households surveyed every day. We did remarkably well on both measures.

The real issue during the lockdown was not as much of quality as it was of sample size. Moving to a phone survey, with our additional QA processes, meant that we could only collect data on approximately 76,000 households as opposed to our usual number of over 170,000. We are now back to in-person operations almost everywhere. Our response rate is now up to 70% and moving back to our historical average of 85%.

The lockdown has prepared us very well for surveying in the future. We now have the capability to rapidly switch from in-person to phone surveying. In-person surveying is preferred for a variety of reasons. However, if a natural disaster like a pandemic or a flood were to occur in the future, we are now set up to move seamlessly into phone-surveying until peaceful conditions prevail again.

At the end of the day, more data is always better. We try to increase our sample size each wave. The pandemic prevented us from doing this. We are hoping to increase our sample again by the middle of 2021.

Please explain for our readers the different ways of measuring unemployment, and your attempts to dig into what these actually tell us in your recent analysis.
Okay, so there are different ways to measure the health of the labour market. One way is to know how many people want jobs, but don’t have them. However, just that number is no good.

India and Switzerland might both have 100 people that don’t have jobs but are actively looking for them, but their populations are very different. So we could divide that number by the population, but that may not be a good idea either. The population of a country consists of everyone – young people, old people, infirm people, people who have no desire at all to work.

Instead of dividing by the total population, it might make sense to divide by the total strength of the labour force. Economists define the labour force as the number of people who are employed plus the number of people who are not employed but are actively looking for a job.

The unemployment rate is defined as the number of people who are not employed but are actively looking for a job divided by the total size of the labour force. That number skyrocketed to almost 24% at the height of the lockdown, but has now completely recovered to its pre-lockdown levels. In fact, the November unemployment rate is lower than it was last year.

The low unemployment rate is good, but it could be low because many people have stopped actively looking for a job. Remember, to be counted in the unemployment rate, you must be unemployed but also actively looking for a job! People might be discouraged from looking for a job at all.

Just like with your own health check-up, it is definitely good news that your blood pressure is normal, but that doesn’t guarantee that you have good cardiovascular health. You need to be healthy across a range of indicators. The same is true for the labour market. Another indicator of labour market health is the employment rate or employment-to-population ratio.

Employment rate is defined as the total number of employed people divided by the total working age population. For our purposes, we define working population as anyone above the age of 15. While the unemployment rate made a full recovery, the employment rate did not.

In January, employment rate was 39.8%. In April, it had fallen to 27.2%. As of November, it had recovered to 37.4%. A year ago, it was 39.2%. That might not sound like a large difference, but our current estimate of the working age population in India is over 1 billion. A 2 percentage point difference means that 20 million fewer people are working today compared to a year ago. This is worrying.

The last thing we do in our piece is to construct a slightly different measure of employment rate. The adjustment we make is to remove people that report as employed but when asked how many hours they worked on a representative day, said that they worked zero hours.

Usually, it’s very rare for people to say that they are employed but didn’t work at all. Indeed, before the lockdown, only 0.5% of employed people said this. During the lockdown, 8% of all employed people said this!

One reason this could be happening might be that people believe that they have a job with their employer, and that they are only temporarily not working. If we remove these people from our definition of employed, the employment rate falls even further to 33.8% as against 38% roughly a year before.

In May, you saw a big difference in household income loss as split by rural and urban. Has that trend held?
We saw that rural households were hit harder than urban households at the peak of the lockdown. The same seems true about the recovery in that urban households seem to have fared better. Many people were puzzled by our first finding. We haven’t even written about the continued rural/urban split in our second piece because we want to devote more time to understanding what’s going on.

The household income data that we have suggests a very different year [not recovering from the shock, as this recent piece points out] for those at the very top, which is Rs 10,801 per month and above, compared to everyone else. This might also tell us about how people understand the impact of this year on their finances...
First, it’s useful to point out that a monthly per-capita household income of Rs 10,800 puts a household in the top 10% of the income distribution! We often have an idea of the middle-class that is different from the actual middle of the income distribution.

It is true that these households have not seen the same recovery as others. This needs to be studied more. One idea is that people in these households have less labour market flexibility. So, when they lose a job, it takes longer for them to rejoin the labour force. Another possibility is that even before Covid hit India, there were some signs of a downward turn in the economy. These households in particular were seeing large reductions in per-capita income even before the lockdown.

As you say, it will be interesting to see how they manage their finances in the coming year. Many people predict that households will be saving much more than before, which might bear out to be true once the September-December data is released on January 1.

Middle-income households with monthly per-capita incomes between approximately Rs 4,000 and Rs 10,800 did not see as sharp a drop in incomes before the lockdown. They have recovered partially but are also struggling to reach the January levels. Low-income households earning less than Rs. 4,000 per month per-capita have almost caught up to their March levels.

Again, in May you said that “34% of all households report being able to survive for no more than one week without additional assistance”. What does this actually mean, and how has it evolved?
This is a little different from what we usually do in CPHS. From time to time, we partner with other institutions to ask questions that aren’t usually included in our survey. One such partnership was with the Rustandy Center at Chicago Booth. We did a very short survey at the height of the pandemic to ask two questions:

1. Has the lockdown caused a fall in household income?
2. How long can your household continue without borrowing or getting any help in cash or kind from anyone?

What you quoted comes from the answer to the second question. This question was not asked outside of the short period of the partnership, so it is hard to say how we are doing compared to the number in May.

What we can see though is that the patterns in income loss reported in May are consistent with the data on actual income levels that was subsequently collected. In our followup piece, we show modest increases in government transfers to households. While we do not focus on the importance of programs like MNREGS, other researchers will be presenting their findings on the stabilising role of these programmes in a few weeks. We also cannot say at this point to what degree the government’s intervention to provide free food helped.

You do find that spending on food is still substantially lower than one year ago. Could this be influenced by government interventions on providing food?
Most of the government’s interventions on providing food have been focussed on cereals and pulses. As per our estimates, cereals and pulses account for approximately 20% of total household spending on food items. In July, expenditure on cereals and pulses was 18% lower than it was a year ago. This might be because government intervention meant that households had to purchase less.

Even if not at the macro level, what are you expecting to see happening to households going forward – what indicators are you most interested in looking at closely?
My broad feeling is that there isn’t enough evidence of a sharp V recovery yet. Some early indicators don’t bode well. As of November, only 4.6% of households said that their incomes had increased compared to a year ago. Before the lockdown, this number was closer to 30%. The complete September-December data, when it is released in January, will help us know if things have improved.

Was there something that surprised you, from the data that has emerged since the pandemic hit? Something that puzzles you?
I think it was a real surprise to see how bad things got immediately after the lockdown. In late March, we saw the unemployment rate jump to over 20%. It was so high that we had to double and triple check that nothing had gone wrong in the survey. In May, we found that 84% of households reported income losses. These numbers were staggering but their impact is showing up in the data even today.

Equally surprising was the speed of the readjustment in the labour market. It is amazing that the unemployment rate dropped to pre-pandemic levels so quickly. In some ways, the economy seems very resilient. In other ways, it seems fragile. The big puzzle is to answer the question that you posed at the beginning – what will the true shape of the recovery be?

Is there a misconception about your area of expertise that you find yourself frequently having to correct? What do we – journalists, fellow-academics, the public – frequently get wrong?

1. The Consumer Pyramids Household Survey is about so much more than employment!
2. The meaning of the phrase “middle-class” as it is typically used is quite different from the actual middle of India’s income distribution.

What three recommendations (books, papers, podcasts, videos) do you have for folks who are interested in this subject?

I’ve linked to a lot of related literature in my answers above, so I will keep this a little more general:

1. I learn a lot by listening to the
   Suno India show.
2. One of the most important books I have read on labour economics is Monopsony in Motion by Alan Manning. (I mentioned it the last time we spoke, but I feel it is so underrated that it bears repeating).
3. A book that I’ve been waiting to begin reading is Fraternal Capital by Sharad Chari.