Debunking the K-shaped recovery theory
The Indian economy is supposed to have entered what is called a K-shaped 'recovery' after the COVID pandemic.
As per Investopedia, “A K-shaped recovery is one in which the performance of different parts of the economy diverges like the arms of the letter "K". In a K-shaped recovery, some parts of the economy may experience strong growth while others continue to decline.”
According to this theory, the top 10-20% of the Indian population has shown growth in incomes compared to the pre-pandemic levels while the rest has not yet recovered to the pre-pandemic levels even in FY23. Or to put it in other words, the rich became richer, and the poor became poorer.
The proponents of this view also quote some other data points:
1. As per a Pew Survey, “the number of people who are poor in India (with incomes of $2 or less a day) is estimated to have increased by 75 million because of the COVID-19 recession.” (Source: S3)
2. The above was largely substantiated by a World Bank study in 2022 (Source: S4) which saidthat over 56 million Indians were pushed into poverty in 2020.
In the following article, I'll try to explain why this is a less-than-accurate depiction of reality and is based on a superficial understanding of relevant data.
First, the Indian economy has grown by almost 10% cumulatively from FY20 to FY23 and this would be substantially driven by the top 10% of the population (with or without a K-shaped recovery). As per studies by Thomas Piketty et al. (Source: S5) the income shares of the top 10% of the population is ~55% in 2013.
Considering the above, leaving aside sales of two-wheelers, entry levels cars etc., the most critical aspect is to see how the bottom 50% of our population have fared in India. Are their conditions in 2022-23 worse than what it was in 2019-20? Let’s look at some data points in the subsequent sections.
I - Poverty Rates
In FY22, poverty came below the pre-pandemic levels in line with per capita incomes as per GDP calculations. The bottom 50% of the population has recovered from the pandemic at least by FY23. This shows that the K-shaped recovery when it comes to the poorest of the population does not stand.
The fact sheet of Household Consumption Expenditure Survey, 2022-23 (HCES) was officially released after a period of 11 years. However, it may be difficult to pin-point the levels of poverty just before the pandemic using the same with just the 2022-23 data. However, the HCES data shall be used in Section 7 to show that K-Shaped recovery is highly improbable.
So, we move to studies which estimated poverty rates just before and after the pandemic. There was an IMF working paper published in April 2022 which concluded that extreme poverty was as low as 0.8 percent in the pre-pandemic year 2019, and food transfers were instrumental in ensuring that it remained at a similar level in the pandemic year 2020 (Source: S6).
There was another paper co-authored by a former vice-chairman of NITI Aayog (Source: S63) which concluded that it is only during the strict lockdown period of April-June 2020 that rural poverty saw a “modest rise”. But it fell for the full year 2019-20, even if at a significantly lower rate. It witnessed a sharp decline in 2020-21 as in the pre-Covid year of 2018-19.
Both these papers show that even during the peak of COVID the poverty rates either remained stagnant or declined. These papers have borne many criticisms fromacademics.
Let's look at the main points for and against a fall in poverty during the pandemic year itself. Reasons for poverty reduction during the pandemic year:
1. While the rural population in India is almost 2/3rd of the overall population, the number of poor people in rural is almost 80% of the overall poor (Source: S8). In the pre-pandemic period, almost 60-65% (Source: S9 and author’s calculations)of the rural population were engaged in agriculture. A back of the envelope calculation would show that almost 50% of the poor people are engaged in agriculture. This would imply that agricultural growth would have a positive effect on reducing poverty. Our Periodic Labour Force Surveys (PLFS) are conducted using agricultural year as basis (June year ending). So, quarter ended June 2020 was during peak COVID, the calculations mentioned are from year ended June 2019 to year ended June 2021. From here onwards, fiscal years (year ending March) will be represented as FY (e.g., FY19 is year ended 31 March 2019) while agricultural years will be represented using the starting and ending years (e.g., 2018-19 represents year ended 30 June 2019). Agriculture gross value added (GVA) grew by almost 10% (Source: S10) from 2018-19 to 2020-21 (based on agricultural year) which would imply a reduction in poverty rates.
2. The Central Government’s food subsidy bill increased by almost 2.5% of GDP from FY20 to FY21 (Source: S11) on account of the free food grains and repayment of debtof Food Corporation of India. Further, MGNREGS expenses were also substantially increased. As per the study by Thomas Piketty et al. (Source: S3), the share of incomes of bottom 50% was around 15% in FY15. So, this additional expenditure was ~15% of the income of bottom 50%. PMGKAY would have benefitted bottom 50% the most and even assuming leakages and payment of FCI debt, it would have given an additional boost to the incomes of the bottom 50% andacted as a bulwark against the fall in incomes during the pandemic year.
II - Reasons against poverty reduction during the pandemic year:
1. The GDP fell by almost 6% in FY21 (Source: S12). The brunt of the fall would have been absorbed by the unorganized sector which means the negative growth would have beenmore for the unorganized sector (compared to organized sector) due to the pandemic shock. So, the incomes of the poor (since around 90% of our population works in the unorganized sector) would have fallen by a higher amount compared to the workers in the organized sector.
2. With reference to the agricultural growth, the calculations using PLFS surveys (Source: S9) show that while agricultural GVA increased by 10% from 2018-19 to 2020-21, agricultural employment increased from 20.5 cr to 25.8 cr (almost 27%) driven by the migrant labourers’ movement from urban to rural areas. So, on a per capita basis even agricultural incomes fell.
Considering these and the studies by Pew and World Bank, it would suggest that poverty rates would have increased during the pandemic year. Even our GDP growth rates were ~-6%, so, it is on expected lines. However, the more important question is whether the poverty rates remained higher than pre-pandemic years even in FY22, FY23 etc.
So, we go to a World Bank working paper published in April 2022 (Source: S7) which showed that India’s poverty rate reduced from around 22% in FY11-12 to 10% in FY19-20 (on the eve of the pandemic). Even this paper has come across criticism from many academics generally critical of the Government.However, its acceptance seems to be higher among them than the other 2 papers.
The main data used by them is the Centre for Monitoring Indian Economy’s (CMIE) Consumer Pyramids Household Survey (CPHS). However, as per the paper, since CPHS sampling methodology is not properly representative, this study made some adjustments to make it representative of the Indian population using various statistical methods.
The World Bank used CPHS data (Source: S13) and similar adjustments made by the authors of the aforesaid paper and have arrived at poverty rate of 12.7% in FY20 which increased to 14.7% in FY21 an increase by almost 29 mn (much lesser than the 56 mn in the 2022 World Bank study and the Pew Study). In FY22, it came down to 11.9% which is even lower than the FY20 figures and in FY23 it came down further to 11.3% (Source: S51). However,FY22 is based on actual data and FY23 is a nowcast and may get revised once later.
The recently released HCES data suggests that the poverty rates reduced to 4.5-5% in 2022-23 compared to 21.9% in 2011-12 (Source: S56). It seems the IMF Working Paper’sestimates of poverty came close to this.
Considering the larger impact of the pandemic on unorganized sector, there would have been some cushioning effect of PMGKAY and would have given a boost to the consumption of the poor. This effect would partly wear off in FY24 (till December 2022, the scheme was carried forward) but would be compensated by rising incomes.
In FY22 itself poverty came below the pre-pandemic levels.
For comparison purposes, from 1993-94 to 2004-05, poverty reduced at 0.74% per year from 45.3% to 37.2%. From 2004-05 to 2011-12, the reduction substantially increased to 2.18% per year from 37.2% to 21.9% due to our high GDP growth rates as well as high growth in real wages (refer Section 3 below). From 2011-12 to 2019-20 the per year reduction slowed down to 1.15% per year from 21.9% to 12.7%. From 2019-20 to 2022-23, the per year reduction was around 0.5% from 12.7% to 11.3% which broadly tracks the lower GDP growth rates during this period (CAGR of ~3.3%). The HCES 2022-23 data cited earlierhas not been used since we do not have comparable estimates for 2019-20.
The aforesaid poverty rate is at consumption levels of USD 2.15 in 2017 PPP which is very close to India's Tendulkar poverty line (i.e., INR 816 per month per person for rural areas and INR 1,000 per month per person for urban areas in 2011-12 prices). The World Bank has also given a Lower Middle Income poverty rate (USD 3.65 in 2017 PPP) which is much higher than the Tendulkar Poverty line. This was 45.9% in FY20 which increased to 49.7% in FY21 but fell below the pre-pandemic number to 45.1% in FY23.
Thus, broadly speaking the bottom 50% of the population has recovered from the pandemic at least by FY23. Within the bottom 50% also, the poorest section recovered faster.This will get clearer once we analyse the PLFS data on wages (Section 4 below).
Before we move forward, one would see headlines like “India’s Poverty Soared Pre-Pandemic, Eased in 2021 but Remained Above 2018 Levels” on the aforesaid World Bank study (Source: S14). These headlines seem to suggest that poverty rates did not come back to pre-pandemic levels in FY22. A careful reading of the World Bank study suggests that new poverty estimates are only from FY20 to FY22. So, the FY19 poverty estimates mentioned in the article quoted above are using a different methodology and not strictly comparable with the new FY20 poverty estimates. The FY20 poverty rates as per the earlier methodology was 10% while FY19 was 11.1%. However, the FY20 poverty rates as per revised methodology was 12.7% but we don’t have FY19 estimates of poverty using the revised methodology. So, the news headlines do not show the correct picture. Hence, our understanding that extreme poverty rates in FY22 moved lower than pre-pandemic levels still stand.
III - Wages
Data on wages also suggests that there has been a notable recovery in incomes in the lower segments. This data is taken from two sources, one is the Agricultural wages annual report for 2021-22 (the ones mentioned in AY in the Pic below or Agricultural year) (Source: S15) published by Ministry of Agriculture & Farmers Welfare and the other is from RBI’s annual publication titled Handbook of Statistics (Source: S16) on Indian States which are on financial year basis (the ones mentioned in FY in the Pic below) which in turn is based on data from Labour Bureau, Government of India. The Agricultural wages are mainly for agricultural labourers and the RBI data is used for non-agricultural labourers like horticulture, construction etc. The wage growth has been adjusted for CPI (rural) to arrive at real wage growth.
The wage growth seems to show that for agricultural and horticultural labourers, there has been real growth from the pre-pandemic years while for other rural labourers, the real growth is negative. So, for most rural labourers (may be two-thirds of them), there is positive real growth while for the others its negative as per this. K-Shaped recovery argument can find partial traction basis this data. It should be noted that both growth and degrowth mentioned above is not substantial, one can say it has largely remained stagnant.
If we look at the long-term trends also, as per an RBI Working Paper dated April 2018 (Source: S52), the rural wages for men followed 3 distinct stages:
• Phase I (from January 2002 to September 2007): Average real rural wages were slightly negative.
• Phase II (October 2007 to October 2013): Average real rural wages turned positive and were growing at 5-7% per year in real terms.
• Phase III (November 2014 onwards): Average real rural wages were slightly positive which basically became slightly negative since the onset of the pandemic.
Even during Phase I when real rural wage growth was negative, our poverty rate had reduced from 42.3% in 1999-00 to 37.2% 2004-05 which means incomes/ consumption of the poorest still increased (Source: S53). While in Phase 2 our poverty rate had reduced much from 37.2% in 2004-05 to 21.9% which is a much faster fall compared to Phase I which is an obvious conclusion.
The reasons for fall in poverty during Phase I, could be:
1. Same person moving from unskilled labour to skilled labour leading to higher wages for the same person over time. E.g., In the IT industry, the salary of an entry level employee may be INR 4 lakhs per annum, and it would have remained at a similar level for the past 10 years. So, the real wage growth for an entry level IT position would be negative. However, for a specific employee, over a 10-year span, that person would have received promotions and his/ her salary growth would have been higher than inflation rate.
2. If there is an increase in the workforce participation rates (as a percentage of population) over the period, for a specific household, there would be an increase in household incomes even if there is a fall in real incomes per person. E.g., If only one person was employed in a household who earned INR 200 per day in 1999-00 but in 2004-05 in the same household if there are 2 people earning INR 150 per day, while at a per person level there is a fall in income, however, at a household level the incomes have risen by 50%.
With regards to the first point, if we can observe an increase in real wages for Casual labourers during Phase I and post-COVID phase as per the Employment-Unemployment Surveys (EUS) and PLFS, it is indicative of the same.
• From 1999-00 to 2004-05, the wages of casual labourers in rural areas increased by 1.9% CAGR in real terms.
• From 2018-19 to 2012-23, the wages of casual labourers in rural areas increased by 4.0% CAGR in real terms (can also be observed in Section 4 below)
With regards to the second point, we can confirm the following (based on author's calculations from EUS, PLFS and Source: S54):
• In 1999-00, the workforce (as a % of overall population excluding unpaid household helpers) was 31.3% which increased to 31.9% in 2004-05
• In 2018-19, the workforce (as a % of overall population excluding unpaid household helpers) was 30.7% which increased to 33.6% in 2022-23
Considering both these points, we can corroborate the reduction in poverty in the periods from 1999-00 to 2004-05 as well as 2018-19 to 2022-23.
One can also carry-out a comparison of the wage growth as per Labour Bureau and as per EUS/ PLFS, it will give even better insights. From 1999 to 2004, EUS showed casual labour wage growth in rural areas at 0.6%, but it was negative 0.2% as per Labour Bureau (as per author’s calculations). From 2004 to 2009, it was 4.2% as per EUS while 0.3% as per Labour Bureau. From 2009 to 2011, it was 12.6% as per EUS while 8% as per Labour Bureau. So, one can see that Labour Bureau data has been consistently understating the annual wage growth.
IV - Wage data from PLFS
As per the PLFS, there are different classes of workers (Source: S9), which can be summarized as follows:
1. Self-employed: Persons who operated their own farm or non-farm enterprises or were engaged independently in a profession or trade on own-account or with one or a few partners were deemed to be self-employed in household enterprises.
2. Regular wage/salaried employee: These were persons who worked in others’ farm or non-farm enterprises (both household and non-household) and, in return, received salary or wages on a regular basis (i.e., not based on daily or periodic renewal of work contract).
3. Casual labour: A person who was casually engaged in others’ farm or non-farm enterprises (both household and non-household) and, in return, received wages according to the terms of the daily or periodic work contract, was considered as a casual labour.
As per India Labour and Employment Report 2014 (Source: S17), in 2011-12, the poverty rates among regular wage earners were at 9%, self-employed at 25% and casual workers at 36% while the overall poverty rate was at 22% (Source: S8).
Thus, if we track the movement in earnings of casual workers and self-employed, we should get a good idea on the post-pandemic recovery process among the bottom half of the population. In Pic 3 below, I have shown the Real indexed earnings (earnings of rural casual workers for 2018-19 indexed at 100) of these categories in rural and urban areas.
As one can see, the earnings of casual workers (rural and urban) is around 15% higher in 2022-23 compared to 2018-19. The earnings for rural self-employed is around 6% higher while the earnings for urban self-employed and regular wage workers is largely stagnant. This is in line with the slower poverty reduction for the USD 3.65 PPP poverty line vis-à-vis the USD 2.15 PPP poverty line which we saw above.
The bottom three lines have shown growth compared to 2018-19 while the top 3 lines have remained flat. So, it is among the top 50% earning category that have not yet recovered. On a lighter note, this is also K-shaped though the opposite of what one initially thought.
In an earlier article titled, “Missing the Rich - Analysis of NSSO Data” I had shown that PLFS data is not reliable to measure income growth in the top decile category. Hence, one can broadly assume that basis the below chart that the top 3 lines pertains towards the middle-income category (from 51st to 90th percentile) and seems like they have not fully recovered from the impact of the pandemic. There are some indications to this effect from some indirect indicators as we shall see below.
V : MGNREGS demand
As per an article in a prominent business newspaper (Source: S19), the number of person days of MGNREGS for the current fiscal year is 207 cr which is almost 10% higher than the previous fiscal (FY23). Further, the number of person days for full year FY23 was 293 cr (Source: S20) which itself was almost 10% higher than the person days generated in FY19 and FY20.
As per a study conducted by a professor at ISB (Source: S18), a rise in demand for MGNREGS in a district can serve as an early warning sign to red flag a region in distress. This shows that all is not well with the poorer sections of the country.
However, the following caveats should also be considered:
1. MGNREGS was rolled out across India from FY09 onwards and it generated 216 cr person days of labour which increased substantially to 284 cr in FY10 (Source: S21) (like FY23 numbers but with a 15% lower population base) due to drought. Even in FY11 which was a high growth year overall and where agriculture grew by 9% (Source: S10), it remained higher at 258 cr person days (almost 20% higher than FY09 levels) (Source: S22). I am sure no one will say that poverty rates remained stagnant or increased from FY09 to FY11 when our economy clocked 9% growth on average. It came down to 215 cr i.e., FY09 levels only in FY12 (Source: S22). But these were the years where we saw a drastic reduction in poverty and high economic growth.
2. Impact of unseasonal rains this year with an overall deficit of 6% vs normal while we had excess rains in FY23.
3. In FY17 and FY18, the number of person days generated was ~235 cr which increased to around 267 cr in FY19 and remained at a similar level in FY20 as well (even with the slowdown in growth that year). Is the increase in FY24 on account of election year like the trend observed in FY18 to FY19?
4. As per a recent article in a prominent business newspaper (Source: S55), acceleration in rural housing construction (Pradhan Mantri Away Yojana - Rural (PMAY)) in FY23 and FY24 may be the reason behind higher MGNREGS costs. It mentioned that in FY24, out of the INR 60k cr budgeted for MGNREGS, almost INR 10k cr (~17%) may have gone towards PMAY. Even in the earlier periods since FY17, this may have contributed partly to higher MGNREGS spends (like in FY19), but the increase in FY23 and FY24 seems to be much more substantial.
VI : Free Food Grains
Another oft repeated argument by the critics of the Government is that why do we need to provide free food grains to 80 crore people unless they can’t survive without it. The only counter argument to it is that the NFSA was passed in 2013. It was the period during which we got the poverty data that showed the fastest reduction in India’s poverty rates from 37% in 2004-05 to 22% in 2011-12 (Source: S8). The NDA Government just made the food grains free instead of heavily subsidizing the price. One can conclude its more of a political decision more than anything else.
VII : HCES Survey 2022-23
A quick analysis of HCES 2022-23shows that it poses a strong argument against the K-shaped recovery narrative. Let me explain.
Key assumptions:
1. There has been no recovery among the poorer sections since 2018-19 (consumption surveys are generally conducted from July to June and hence, taking 2019-20 would not be appropriate due to lockdowns in June 2020 quarter).
2. Consumption expenditure surveys of 2011-12 and 2022-23 are comparable and hence one can calculate growth rates in consumption from 2011-12 to 2022-23.
As per HCES 2022-23, the real growth in consumption expenditure (Source: S57) in rural areas is ~44% and in urban areas is ~35%.
Given the aforesaid assumptions, it would mean that in real terms the consumption expenditure would not have grown since 2018-19. So, we can assume all the real growth in consumption expenditure happened between 2011-12 and 2018-19 and inflate it using CPI (Source: S16) till 2018-19. This would give a CAGR (compounded annual growth rate) for per capita consumption from 2011-12 to 2018-19 in nominal terms.
This can be compared with the nominal per-capita PFCE growth (Private final consumption expenditure) as per National Account Statistics (used in GDP calculation) for the same period (Source: S10). And we get the chart shown below.
In the scenario where there is no growth in consumption since 2018-19, it is quite clear that for the bottom 50% of the population (both rural and urban) CAGR would have been higher than the growth in PFCE by ~1% on average. The same difference would carry forwardwhen we convert these into real terms. This means if the PFCE per capita grew by 5% in real terms, the consumption expenditure of the bottom 50% would have grown by around 6% (which is 20% higher).
Most people would not agree to this view including those who argued that the new system of National Accounts overestimates growth (that’s a different debate though). In a way, it is ironic that many of the same people argued for a K-shaped recovery post COVID.
Please note that for the purposes of this article, K-shaped recovery refers to the view that incomes/ consumption expenditures of the poorer segment have not recovered yet to pre-COVID levels (in real terms).
VIII : Comparability issues between HCES 2011-12 and 2022-23
Considering that it is the HCES 2022-23 offers a very strong argument against K-Shaped recovery narrative, it is important that some of the comparability issues raised by experts be investigated. While the comparability issues should be investigated in more detail, here are some preliminary thoughts. I shall look at some of the criticisms made by some prominent economists in an online news portal (Source: S58).
Quote: There seems to be a higher representation of the well-off groups in the HCES 2022-23 sampling approach, thereby resulting in higher consumption expenditure.
If we had used a similar sampling approach as HCES 2011-12, we may have gotten results like we got in HCES 2017-18 (which got junked due to bad data quality) (Source: S59). To quote a few anomalies in HCES 2017-18:
1. From my earlier article titled, “Missing the Rich - Analysis of NSSO Data”: “In the pic below, we have shown the monthly per capita consumption of the top decile (rural-urban combined) in FY12 and FY18 (both in 2011-12 prices) and the incomes of the protected regular wage earners for the same period. As one can note, the consumption expenditure fell by almost 12% in real terms and the earnings of protected regular wage earners also declined by 16%. Both seem to show a very similar trend.” This is opposite to the narrative of demonetization-GST which one used to hear.
2. Consumption of cereals, pulses fell by 15-20% per capita (in real terms)
3. Medical expenses fell by 26% and 44% in urban and rural areas respectively.
4. Expenditure on sugar, salt and spices fell by ~15%
Do those things represent something which is reflective of what was really happening?So, changing times may require changing the sampling methodology to stay relevant.
Quote: Moreover, 190 million workers (2021-22) in India are earning just up to Rs. 100 per day (in real terms at 2010 prices) which can be categorised as absolute poor, as compared to just 106.1 million workers in 2011-12. There has been a massive surge in the number of poor workers in recent times. There are 144.0 million workers (2021-22) that are earning between Rs 100 and Rs 200 per day which can be categorised as poor and vulnerable. Additionally, there are still 127.5 million workers (2021-22) who earn between Rs 200 and Rs 300 per day, which can be categorised as non-poor but definitely vulnerable.
This is a sensational claim. If true, one will have to conclude that the HCES 2022-23 is not comparable to earlier rounds and that poverty has gone up. To get more clues into this, I referred to one of the debates on a prominent news channel (Source: S60) wherein one of the co-authors of the aforesaid article mentions that in 2018-19, as per PLFS, the number of workers who earned less than INR 100 a day (I am sure he meant in Real prices with 2010 year as base) is at 15.2 cr.
To corroborate these, I was able to find a CSE working paper from Azim Premji University (Source: S61) whose Appendix gave important insights on 2018-19 PLFS data. I personally think they understate the monthly wages of casual labourers, but that's more of a quibble.
Anyway, after factoring inflation, INR 5,000 per month would roughly translate to INR 100 per day real wages in 2010 prices. So, as per the CSE working paper and author's simple extrapolations, around 11% of regular wage earners, 24% of the Self-Employed and 42% of casual wage earners earn less than INR 100 per day (in 2010 prices).
Considering Current Weekly Status based employment measurement, almost 25% or roughly 10 crore paid workers earn less than INR 100 per day (in 2010 prices). Where did the rest 5 crore go (since the aforesaid claim is ~15 cr)? Well, they are unpaid family workers who are not included in CSE working paper study as they don't earn by definition. So, we have broadly matched the 2018-19 numbers.
Now, coming to 2011-12 earnings, we did not have earnings of self-employed (who constitute almost 50% of the workforce) in the Employment unemployment surveys of 2011-12 and earlier. Plus, it seems like the article (my inference) did not add the crores of unpaid family workers in its calculation of 10.4 crores in 2011-12. So, the 10.4 crore number for 2011-12 would only include Regular plus Casual workers and does not include Self-Employed and unpaid family workers.
On the other hand, the number for 2018-19 includes all workers (including unpaid workers). So, the comparison is between apples and oranges.
Nevertheless, since the wages of regular employees in the informal sector have remained stagnant since 2011-12 to 2018-19 as per an ILO Study (Source: S62), one can assume that the proportion of regular wage earners earning below INR 100 per day (in 2010 prices) has remained the same at about 11%.
However, basis the methodology used by CSE working paper for calculating wages of casual labourers, one can say that more than 50% earned less than INR 100 per day (in 2010 prices) in 2011-12. Further, between 2011-12 and 2018-19, the number of casual workers reduced, and regular workers increased. So, a rough calculation after combining regular and casual workers suggests that around 36% earned less than INR 100 per day in 2011-12 which reduced to 27% in 2018-19.
This is a clear indication of increase in incomes/ consumption post 2011-12 (in real terms). The numbers for 2022-23 (as a proportion of total paid workforce) would also not be way different compared to 2018-19.
As a side note, as per the CSE working paper, the number of salaried workers earning more than INR 5.5 lakhs per annum as per PLFS would be ~65 lakhs while as per IT Return filing data (Source: S66), the number is close to 1.4 cr which is more than double. This shows that PLFS data understates the incomes of richer decile of the population (as also shown in detail in my article titled: "Missing the Rich - Analysis of NSSO Data").
Quote: "In fact, the National Survey Organisation (NSO), which conducts these surveys, has personally informed one of the authors that the HCES is not comparable with the earlier CES…. Additionally, it would be curious to argue that consumption expenditure is rising when real wages have been stagnating in recent years.
So, here the author is suggesting that poverty has increased from 2011-12 to 2022-23. Well, it means the real per capita consumption expenditure increase of ~40% from 2011-12 to 2022-23 is completely explained by the change in methodology alone and the actual increase is either 0 or negative.
Here’s some evidence to the contrary:
1. During 2011-12, Employment-Unemployment survey (EUS) also had a shorter questionnaire of ~40 questions on consumption expenditure vs the HCES 2011-12 of 400+ questions on consumption. Similar, was the case with EUS 2004-05 and HCES 2004-05 where the difference was around 5%. EUS 2004-05 also states: “The abridged worksheet that was used to reduce the respondent fatigue is known to understate the level of consumer expenditure in comparison with the detailed schedule.” Any NSO statistician would have also said that the consumption expenditure as per EUS and HCES are not comparable even though the difference is less than 10%.So, even with 10x increase in number of questions the consumption expenditure was higher by just 5-7% in HCES (Source: S64). This clearly suggests that the change in survey methodology / slightly increasing the questionnaire length / splitting the questionnaire would not lead to more than 40% increase in consumption expenditure.
2. Even the PLFS is using a different sampling methodology compared to the earlier Employment-Unemployment surveys. This was highlighted by NSO in the PLFS report just like it has been done in HCES 2022-23. But academics (critics and supporters alike) don't have many issues in comparing PLFS with the earlier Employment-Unemployment surveys.
3. Some prominent economists had even used consumption expenditure in the PLFS (with a single question on consumption vs 400+ in HCES) to estimate poverty rates with suitable adjustments (Source: S65) even though the PLFS report clearly says "Information on household Usual Monthly Consumer Expenditure (UMPCE) was collected in PLFS only to classify the households in different UMPCE classes and it cannot be used to estimate the household consumer expenditure which is generally estimated based on detailed survey". Here, PLFS makes it categorically clear that the consumption expenditure as per PLFS is not comparable with HCES.
4. HCES 2022-23 factsheet lists out the changes in the methodology and finally says "These are required to be noted while comparing the results of HCES:2022-23 with those of the previous surveys." It does not say they are not comparable with the earlier rounds.
These points would suggest even if there were an overestimation, it may not be much (lesser than 10% in all probability). Further, Dr. Pronab Sen, former Chief Statistician, in a recent interview with The Print said that even though there may be issues with comparability, the differences are not likely to be large.
IX : Other indicators
1. Vehicle sales: These are used frequently (as quoted in the beginning) to “prove” K-shaped recovery. However, as per National Family Health Survey V data, only around 50% of households have two wheelers and only 7% have cars. So, two-wheeler data is not representative of a K-shaped recovery for the poorest sections. Anyway, let’s look at the data now.
The data is shown for FY19 and FY23 since in FY20 there was a substantial degrowth except for the used car sales (which showed 10% growth). As mentioned in the beginning, two-wheeler sales are substantially lower and so is entry level cars.
With reference to entry level cars, one can see that used cars (which would be purchased at a similar rate as entry level car) has shown substantial growth compared to FY19. There is an article in a prominent business newspaper(Source: S27) which says that first time buyers moving from hatchbacks to SUVs. So, the entry level vehicle argument may be more of a choice than anything else and is not indicative of K-shaped recovery.
With reference to two-wheelers, it may show K-shaped recovery within the top 50% of the population and not for the bottom 50% since only 50% of households own two-wheelers as mentioned above. What that means is may be the middle 51st to 90th percentile of the population may not have recovered their incomes to their pre-pandemic levels.
But here also there is a caveat. As per a FADA press-release (Source: S28), “Over the past few years, the prices of various two-wheelers have risen significantly, impacting their affordability for consumers across India. This surge in prices can be attributed to multiple factors, including the rising cost of raw materials, stricter emission norms and higher taxes and levies. For instance, the price of the popular Honda Activa has escalated from Rs. 52,000 in 2016 to Rs. 88,000 in 2023. Similarly, the Bajaj Pulsar has witnessed a substantial increase from Rs. 72,000 in 2016 to Rs. 1,50,000 in 2023. The Hero Splendor and TVS Jupiter have also experienced considerable price hikes over the same period, with their prices soaring from Rs. 46,000 to Rs. 74,801 and Rs. 49,000 to Rs. 88,498, respectively. The continuous rise in two-wheeler prices has consequently led to a decline in sales, emphasizing the pressing need for intervention and GST rate reduction to restore the industry's growth trajectory.” It should be noted that out of the total 1.6 crore sales in FY23, around 45% of the overall 2-wheeler category comes from these 4 brands alone (Source: S45).
This means Honda Activa prices increased at 7.8% CAGR from 2016 to 2023 while CPI and WPI was 5% during this period. Similarly, Bajaj Pulsar prices increased by 11% CAGR and Splendor and Jupiter prices increased at a CAGR of 7.2% and 8.8% respectively. Further, the price increases from 2019 to 2022 would have been much higher than in the period from 2016 to 2019. Hence, from 2019 onwards, the CAGR of price hikes would be higher than the ones shown above.
However, if the earnings of the middle-income category have grown only around the CPI rates during 2019 to 2023 (i.e., around 6%) as shown in Pic 3 above, it is obvious that it would have a significant impact on demand. In the pre-pandemic period incomes would have shown faster growth than two-wheeler prices. So, one should just not look at the sales numbers and conclude K-shaped recovery etc. though it could be one of the reasons.
2. Residential real estate (Source: S46): Here,we shall look at the trend of residential sales in the top 8 cities (as per Knight Frank). The overall sales were 245,861 units in 2019 which increased to 310,592 for twelve months ended 30 June 2023. Out of this, the mid-segment (between INR 50 lacs to INR 1 cr per unit) sales increased from 81,409 to 117,183 while the premium segment (more than INR 1 cr per unit) almost doubled from 45,461 units to 89,846 units. This would imply the affordable segment sales fell from 118,991 units to 103,563 units. Another indication of a K-shaped recovery may be. Here also, it would be the middle-income segment who would be purchasing affordable housing and not the bottom 50%. Further, we should also factor the impact of inflation on the segmentation (i.e., an affordable segment home with INR 48 lacs in 2019 would be worth INR 60 lacs or so in 2023 and would fall in the mid-segment) and increase in interest rates before we conclude anything just based on this dataset.
3. Smart phone sales (Sources: S35 to S38): As per Annual Survey of Education Report 2022, in 2018 the number of rural households with smart phones were 36% which increased to 75% in 2022. With reference to number of smart phone sales, it declined by 1.7% in 2020 while it increased 11% in 2021 and a dip again in 2022 of 6% and a dip of 7-8% in 2023 up to September 2023. Does this imply K-shaped recovery or has market saturation got something to do with the dip. Here also, the K-shaped recovery, if at all would be implied to the middle section of the population (not the richest or the poorest).
4. Nielsen FMCG sales (Sources: S28 to S34): Here, we can look at sales growth from CY20 to CY23. So, FMCG sales clocked volume growth of -2% in CY20, 6.3% in CY21, -1.5% in CY22 and for the 3 quarters of CY23 it clocked 6.4% growth. Add all this together it will show that FMCG volumes may be 8% higher in CY23 compared to CY19. Though it is not much compared to historical FMCG growth rates, still being mass consumption products, it goes against a K-shaped recovery.
Another point is that the growth rates in rural areas is substantially lower compared to urban areas in 2022 and 2023. However, it should be noted that in the first year of the pandemic considering the migrant labour- movements and agricultural activity, rural areas did not slow down as much as urban. On average, overall FMCG value contracted 2% in CY20 while rural areas would have grown by 4%. Due to movement of the labourers back to urban areas and higher base effects (so to speak) FMCG volumes grew slower in rural areas in subsequent periods.
5. The Savings Rate Conundrum: As per Price ICE 360 income survey (Source: S47), out of the total savings, around 80% of our savings is coming from 1/3rd of the households and around 52% of total savings is coming from the middle class (who will basically form a large part of the middle 40% of the population). So, any income stagnation for this class will result in reduction in savings rate for this segment and a reduction in overall savings rate of the economy. Is this happening?
Basis a recent RBI publication, one would have seen headlines like “Household Savings Fall to Five-Decade Low in FY23, Debt Remains Sharply Elevated: Report”(Source: S48). If we read the article more carefully and see the RBI data, household net financial savings reduced from around 8% (pre-pandemic) of GDP to 5% in FY23.
This may indicate a huge distress at the first glance. However, household savings include financial savings and physical savings. With the National Accounts published in February 2024 (Source: S67), the overall household savings turned out to be close to 18.4% of the GDP. This is not way lower than the savings rates observed in the past. Even in FY16 and FY17, household savings rate was close to 18% and it was ~19% in FY20. The fall is only ~0.5-1%. There is a need for caution, but alarm bells need not start ringing yet.
There were articles in prominent business newspapers before the publication of national accounts which calculated household overall savings reduced from 19% of GDP in FY20 to around 16% GDP in FY23based on very erroneous assumptions.
Based on the growth in gross-fixed capital formation in the economy, there were sufficient data points available even at that time (Source: S50) wherein one could have reached a conclusion that the household savings rates would be closer to 18% in FY23.
6. Travel data (Sources: S39 to S41): Railway passenger traffic in FY23 was lower than FY20 by around 24%. Airline passenger movement in FY23 was lower than FY20 by around 4%. Does this indicate K-shaped recovery when both modes of passenger traffic have not even come to pre-pandemic levels? On the other hand, petrol (not petroleum) consumption, which is a mass consumption product used by individuals, increased from 30 mn MT in FY20 to 35 mn MT in FY23.
7. Page Industries (Innerwear sales) (Source: 42): Innerwear is another mass consumption item. If we look at the volume growth from FY20 to FY23 as per the Investor Call transcripts, volume growth was almost NIL in FY21, 29% in FY22 and 13% in FY23. It is in FY24 that we are seeing some degrowth, but still, it is quite above pre-pandemic levels.
Considering the data on earnings from Pic 3 above and data on two-wheeler sales, residential real estate, smart phones, savings and even to an extent FMCG sales, we seem to get a clearer idea that it is rather the middle 40% of the population (51st to 90th percentiles of income) whose incomes have remained stagnant/ slightly declined. That is not K-shaped,at least the way it is used in the Indian context.
Summary
In the manner shown above, one can pick and choose dozens of parameters to “prove” or “disprove” a K-shaped recovery. However, indirect indicators (like vehicle sales, MGNREGS demand) need not show the true picture and one must give priority to direct indicators (consumption/ earnings). After weighing all the data mentioned above, one can conclude that the argument for K-shaped recovery (at least when it comes to the bottom 50% of the population) is weak considering:
(i) HCES 2022-23 poses a very strong argument against K-shaped recovery considering it is highly implausible for the consumption expenditure to have grown in real terms by ~40% in just 7 years from 2011-12 to 2018-19 given the PFCE growth during this period.
(ii) Poverty has reduced in FY22 itself compared to FY20 showing the bottom 20% odd of the population has recovered in FY22thanks in part to schemes like PMGKAY. This is in-line with our GDP recovery from FY20 to FY22.
(iii) Increase in earnings of casual wage earners and rural self-employed shows that a large portion of the bottom 50% of the population would also have recovered. As a proxy, there was faster growth in the construction sector (CAGR of 6% from FY20 to FY23) vis-à-vis overall GDP growth. This was led mainly by revival in real estate sector and Government capex on infrastructure. Almost 45-50% of casual workers (Source: S9 and author’s calculations) work in construction and a large part of the remaining work in agriculture.
(iv) Rural wage data is a mixed bag since it largely shows that wages have remained stagnant but for various reasons mentioned above it consistently understates actual wages earned by the population on average and PLFS data on wages is more reliable.
(v) MGNREGS data supports the K-shaped recovery to an extent though the argument is indirect and has some flaws as highlighted above.
(vi) Food security argument would have political elements within it and is not relevant.
(vii) So, on an overall basis, the incomes of the bottom 50% would have recovered at rates somewhere closer to the GDP growth (may be slightly lower considering point (iii) above)with the poorest sections recovering even faster.
(viii) Recovery in GDP itself shows that top 10% has recovered and doing well. It is also seen in salary/ wage expenses of listed companies which increased from INR 9.5 tn in FY20 to INR 13.4 tn in FY23 (Source: S43). It seems that their incomes have grown even faster than the GDP growth rates.
(ix) Now, coming to the middle 40% of the population (51st to 90th percentile), there are indications that they may not have fully recovered yet, at least not seen growth in real incomes, especially considering stagnant earnings of regular wage earners and urban self-employed. This can be corroborated by indirect indicators like lower two-wheeler sales/ stagnant phone sales etc.
These would be the people who (i) would not have received benefits from the Government via food grains etc. (ii) would be mostly not paying income taxes and working in the informal sector (a good chunk of it would be in the urban areas which bore the brunt of the lockdown impact)that would have suffered more because of the pandemic (compared to the organized sector).
A lot of these people would be working in the Trade, hotels and transportation sector which employs 9-10 crore people. As per GDP calculations as well, this sector has recovered the least and the CAGR from FY20 to FY23 for this sector is just ~1%.
(x) These conclusions are somewhat covered by Dr. Raghuram Rajan in an interview dated December 2022 where he said (Source: 44): “Its not just about four or five industrialists, even the income of upper-middle class rose in the country….Extremely poor people get a lot of benefit from the government and the upper class was not affected by the pandemic, it was the lower-middle class who faced trouble to a larger extent.”
Section 11: Outlook
While there are views which suggest that we are in a full-blown crisis due to K-shaped recovery, the evidence for the same is weak.
For the bottom 50%, one must closely monitor the MGNREGS demand and the rural wage data. However, continued recovery in real estate sector and capex spending by Government (private capex recovery may still take time) will be able to keep this segment’s income growing. There would be some impact on real estate sector especially considering higher interest rates, the global growth slowdown, and the hiring freezes in IT sector for the next year or two. One also must look out for “Revdis” which may slowdown capex spending. However, they may be offset in the short-term due to higher revenue spending.
For the top 10%, CY24 may be tougher considering the hiring freeze in IT, global growth slowdown etc. Once we are out of that phase things would get better.
The middle 40% seems to have been the most affected though here also it’s not a full-blown crisis but rather a stagnation in incomes. However, with the continued recovery in the Trade, Hotels and Transportation sector and if the Government programs like National Logistics Policy have the intended effects, this segment will also start growing.
(The author is Associate Vice-President at one of the 'big four' accounting firms)
Sources
S1, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, S13, S14, S15, S16, S17, S18, S19, S20, S21, S22, S23, S24, S25, S26, S27, S28, S29, S30, S31, S32, S33, S34, S35, S36, S37, S38, S39, S40, S41, S42, S43, S44, S45, S46, S47, S48, S49, S50, S51, S52, S53, S54, S55, S56, S57, S58, S60, S61, S62, S63, S64, S65, S66, S67
S59: Household Consumption of Various Goods and Services in India, 2017-18 – Leaked report