by Ajay Shah.
The broad set of Indian listed companies have a high trailing P/E ratio. This suggests that the market believes there will be high earnings growth in the future.
Some finance practitioners back out an earnings time series as Nifty market capitalisation divided by Nifty P/E. This `Implied Nifty Earnings' series shows strong growth over long time horizons.
In this article, we show that this quick-and-dirty method has an upward bias in the estimation of aggregate earnings growth. In truth, earnings growth by Indian firms has been stalled for a decade.
The graph above shows the long time-series of the trailing P/E ratio of the CMIE Cospi index, which measures the broad market valuation. This shows that we are near some of the highest valuations in history.
These high P/E ratios would generally suggest that the stock market expects that a period of great earning growth is around the corner. It's important to look back at the recent history of earnings growth in order to evaluate this optimism.
The P/E ratio is market capitalisation divided by earnings. Hence earnings is market capitalisation divided by the P/E ratio. It's easy to obtain a time-series of the Nifty P/E (from NSE), and the Nifty market capitalisation (obtained by summing up the market capitalisation of all Nifty member firms as seen in the CMIE database). This gives the time-series:
As Nifty market capitalisation is measured in rupees, and the P/E ratio is dimensionless, the division yields an earnings value in rupees.
This shows pretty good growth in the earnings of the Nifty companies. In the latest few years, the growth is slow, but when compared with a decade ago, the earnings expansion is remarkable. Overall, it's a gain of 18$\times$ in 18 years, which is quite a performance. It is consistent with the common view that India is a high earnings-growth economy.
The set of firms that make up Nifty changes through time. From 1996 to 2017, there were 118 firms which have been a member of Nifty atleast once.
The Nifty components at time $t_1$ are often different from those prevalent at time $t_2$. Some firms are added and some are removed. We tend to think that these are a few random fluctuations which would tend to cancel out. However, the changes in the set are non-random, and they do not cancel out.
The management of Nifty uses a rule set that roughly summarises to this: (a) A pool of eligible firms is formed where the firms have adequate stock market liquidity based on the Impact Cost measure, and (b) If an eligible firm is over 2$\times$ larger (by market capitalisation) than the smallest incumbent, then a set change is effected where the smallest incumbent is removed and the large new liquid firm is brought in. The earnings of the new entrant will generally be higher than the earnings of the smallest incumbent who is removed, as the market value of the new entrant is over 2$\times$ higher.
Here is one example, from the April-May-June 2016 quarter. In this quarter, three firms were removed (Vedanta, Cairn India, Punjab National Bank) and three firms were added (Aurobindo Pharma, Bharti Infratel and Eicher Motors) to Nifty. The remaining 47 firms were unchanged. Let's pull together the information about earnings across these changes.
Table 1: Example of how the quick and dirty method over-estimates earnings growth
The best estimator of earnings growth is that which is made using the identical set of firms observed at two points in time. In the above example, there are 47 firms in Nifty who were present at both points in time. Their aggregate earnings declined from Rs.735B to Rs.635B, a decline of 14%.
Three firms were present in Q1 2016 -- Vedanta, Cairn, PNB -- and when their earnings data is used, the aggregate earnings of the 50 firms in Nifty at that point in time works out to Rs.717B. These were replaced by Aurobindo Pharma, Bharti Infratel, Eicher Motors in Q2 2016, and when their earnings data is used, the aggregate earnings of the 50 firms in Nifty at that point in time works out to Rs.657B. The earnings growth obtained by comparing these two inconsistent sets was -8%, which is a more optimistic picture when compared with the decline of 14% for the consistent set.
There is a big discrepancy, of 6 percentage points across one quarter, and the direction of the bias in in favour of greater optimism.
The wrong method (merely comparing the profits across inconsistent sets across time) does not just introduce random noise, it is biased. It systematically overstates earnings growth of the Nifty set.
How should we do this right? We exercise care with the following steps:
This yields the following picture of the index of nominal aggregate earnings of the Indian corporate sector:
This tells a story where the average earnings index grew from 126 in 2000 to 996 in 2010, but declined to 783.98 in the Oct-Dec 2016 quarter. Nominal earnings has stagnated in the last decade.
There is one problem of non-comparability in the above analysis. The time-series of the P/E that was shown in Figure 1 pertains to the 2500 odd firms in the CMIE Cospi index. The time-series of earnings that's shown in Figure 3 pertains to the performance of all listed non-finance non-oil firms. These two groups are slightly different. Could this difference be an important issue? In order to examine this, we apply the careful method (that was used for Figure 3) to the full universe of all listed firms. The two series are superposed here:
This shows that there are some differences between the two groups, but this difference is small. From Oct-Dec 2006 to Oct-Dec 2016, i.e. in the latest decade, we have three estimates of the compound growth rate of earnings:
The compound average growth rate of earnings of all listed firms is similar to that of non-finance non-oil firms. Both estimates are roughly 8 percentage points per year below the quick and dirty method.
Fine points in handling firm data matter! It appears easy and expedient to use the Nifty market capitalisation time series, and the Nifty P/E ratio time series, to back out an estimate of Nifty earnings time series. The use of the phrase `Implied Nifty Earnings' sets off an analogy with the genius of implied volatility. However, this procedure is highly misleading. Let's superpose the wrong and the correct index time-series on one graph:
The quick and dirty method suggests 18$\times$ earnings growth in 18 years. The correct method shows 8$\times$ earnings growth in the same period, and stagnation in the last decade.
The stock market believes that a great wave of earnings growth is around the corner, and India is generally considered a market with high earnings growth. However, earnings growth has been elusive for the last decade. More generally, in the past, the Indian stock market has done well on differentiating between firms -- in voting with a high P/E ratio for firms that will do well in the future -- but has fared poorly at macroeconomic thinking.
This R program when run using data from CMIE Prowess DX (Mar-2017 vintage) replicates Figure 3 above.
I thank Nilesh Shah and Mahesh Vyas for valuable discussions. Pramod Sinha wrote the code and it was audited by Dhananjay Ghei and Shekhar Hari Kumar.
The broad set of Indian listed companies have a high trailing P/E ratio. This suggests that the market believes there will be high earnings growth in the future.
Some finance practitioners back out an earnings time series as Nifty market capitalisation divided by Nifty P/E. This `Implied Nifty Earnings' series shows strong growth over long time horizons.
In this article, we show that this quick-and-dirty method has an upward bias in the estimation of aggregate earnings growth. In truth, earnings growth by Indian firms has been stalled for a decade.
The trillion dollar question
 |
| Figure 1: The long time-series of the CMIE Cospi P/E ratio |
The graph above shows the long time-series of the trailing P/E ratio of the CMIE Cospi index, which measures the broad market valuation. This shows that we are near some of the highest valuations in history.
These high P/E ratios would generally suggest that the stock market expects that a period of great earning growth is around the corner. It's important to look back at the recent history of earnings growth in order to evaluate this optimism.
Estimating aggregate earnings: a quick and dirty method
The P/E ratio is market capitalisation divided by earnings. Hence earnings is market capitalisation divided by the P/E ratio. It's easy to obtain a time-series of the Nifty P/E (from NSE), and the Nifty market capitalisation (obtained by summing up the market capitalisation of all Nifty member firms as seen in the CMIE database). This gives the time-series:
 |
| Figure 2: Time series of Nifty earnings (nominal rupees), quick and dirty method |
As Nifty market capitalisation is measured in rupees, and the P/E ratio is dimensionless, the division yields an earnings value in rupees.
This shows pretty good growth in the earnings of the Nifty companies. In the latest few years, the growth is slow, but when compared with a decade ago, the earnings expansion is remarkable. Overall, it's a gain of 18$\times$ in 18 years, which is quite a performance. It is consistent with the common view that India is a high earnings-growth economy.
The quick and dirty method over-estimates earnings growth
The set of firms that make up Nifty changes through time. From 1996 to 2017, there were 118 firms which have been a member of Nifty atleast once.
The Nifty components at time $t_1$ are often different from those prevalent at time $t_2$. Some firms are added and some are removed. We tend to think that these are a few random fluctuations which would tend to cancel out. However, the changes in the set are non-random, and they do not cancel out.
The management of Nifty uses a rule set that roughly summarises to this: (a) A pool of eligible firms is formed where the firms have adequate stock market liquidity based on the Impact Cost measure, and (b) If an eligible firm is over 2$\times$ larger (by market capitalisation) than the smallest incumbent, then a set change is effected where the smallest incumbent is removed and the large new liquid firm is brought in. The earnings of the new entrant will generally be higher than the earnings of the smallest incumbent who is removed, as the market value of the new entrant is over 2$\times$ higher.
Here is one example, from the April-May-June 2016 quarter. In this quarter, three firms were removed (Vedanta, Cairn India, Punjab National Bank) and three firms were added (Aurobindo Pharma, Bharti Infratel and Eicher Motors) to Nifty. The remaining 47 firms were unchanged. Let's pull together the information about earnings across these changes.
| Firm(s) | Earnings Q1 2016 (Rs. million) | Earnings Q2 2016 (Rs. million) | Change (Per cent) |
| The 47 common firms | 735,021 | 635,358 | -14 |
| Vedanta | 38,823 | ||
| Cairn India | -2,459 | ||
| Punjab National Bank | -53,671 | ||
| Aurobindo Pharma | 3,910 | ||
| Bharti Infratel | 14,769 | ||
| Eicher Motors | 3,371 | ||
| The full 50 at a point in time | 717,713 | 657,408 | -8 |
Table 1: Example of how the quick and dirty method over-estimates earnings growth
The best estimator of earnings growth is that which is made using the identical set of firms observed at two points in time. In the above example, there are 47 firms in Nifty who were present at both points in time. Their aggregate earnings declined from Rs.735B to Rs.635B, a decline of 14%.
Three firms were present in Q1 2016 -- Vedanta, Cairn, PNB -- and when their earnings data is used, the aggregate earnings of the 50 firms in Nifty at that point in time works out to Rs.717B. These were replaced by Aurobindo Pharma, Bharti Infratel, Eicher Motors in Q2 2016, and when their earnings data is used, the aggregate earnings of the 50 firms in Nifty at that point in time works out to Rs.657B. The earnings growth obtained by comparing these two inconsistent sets was -8%, which is a more optimistic picture when compared with the decline of 14% for the consistent set.
There is a big discrepancy, of 6 percentage points across one quarter, and the direction of the bias in in favour of greater optimism.
The wrong method (merely comparing the profits across inconsistent sets across time) does not just introduce random noise, it is biased. It systematically overstates earnings growth of the Nifty set.
What actually happened to earnings growth of Indian firms?
How should we do this right? We exercise care with the following steps:
- Oil companies have extreme earnings fluctuations based on fluctuations of global crude oil prices. Their profits do not describe what is going on in India. Finance companies have problems in earnings data, such as the concealment of bad assets by banks. Hence, we look at non-oil non-finance companies only. Aggregation of accounting data for this set of firms is an excellent source of insight into India's business cycle fluctuations.
- At every two consecutive quarters, we construct a set of listed firms which are observed in both quarters. We sum up the earnings of this set at each of the two quarters. These two summed earnings are comparable across time, as they pertain to the identical set of firms.
- This yields a nominal percentage growth of aggregate earnings from one quarter to the next.
- We start an index at 100 and cumulate it up through time using each of these carefully constructed estimates of earnings growth.
This yields the following picture of the index of nominal aggregate earnings of the Indian corporate sector:
 |
| Figure 3: Index of earnings of all listed non-finance non-oil companies |
This tells a story where the average earnings index grew from 126 in 2000 to 996 in 2010, but declined to 783.98 in the Oct-Dec 2016 quarter. Nominal earnings has stagnated in the last decade.
Ruling out an alternative explanation
There is one problem of non-comparability in the above analysis. The time-series of the P/E that was shown in Figure 1 pertains to the 2500 odd firms in the CMIE Cospi index. The time-series of earnings that's shown in Figure 3 pertains to the performance of all listed non-finance non-oil firms. These two groups are slightly different. Could this difference be an important issue? In order to examine this, we apply the careful method (that was used for Figure 3) to the full universe of all listed firms. The two series are superposed here:
 |
| Figure 4: Index of listed non-finance non-oil firms earnings and lll listed firm earnings (nominal) |
This shows that there are some differences between the two groups, but this difference is small. From Oct-Dec 2006 to Oct-Dec 2016, i.e. in the latest decade, we have three estimates of the compound growth rate of earnings:
| The quick and dirty method | 8.19% |
| All listed firms, the careful method | -1.11% |
| Non-finance non-oil listed firms, the careful method | -0.41% |
The compound average growth rate of earnings of all listed firms is similar to that of non-finance non-oil firms. Both estimates are roughly 8 percentage points per year below the quick and dirty method.
Conclusion
Fine points in handling firm data matter! It appears easy and expedient to use the Nifty market capitalisation time series, and the Nifty P/E ratio time series, to back out an estimate of Nifty earnings time series. The use of the phrase `Implied Nifty Earnings' sets off an analogy with the genius of implied volatility. However, this procedure is highly misleading. Let's superpose the wrong and the correct index time-series on one graph:
 |
| Figure 5: Superposing the quick and dirty earnings index and the careful index |
The quick and dirty method suggests 18$\times$ earnings growth in 18 years. The correct method shows 8$\times$ earnings growth in the same period, and stagnation in the last decade.
The stock market believes that a great wave of earnings growth is around the corner, and India is generally considered a market with high earnings growth. However, earnings growth has been elusive for the last decade. More generally, in the past, the Indian stock market has done well on differentiating between firms -- in voting with a high P/E ratio for firms that will do well in the future -- but has fared poorly at macroeconomic thinking.
Reproducible research
This R program when run using data from CMIE Prowess DX (Mar-2017 vintage) replicates Figure 3 above.
I thank Nilesh Shah and Mahesh Vyas for valuable discussions. Pramod Sinha wrote the code and it was audited by Dhananjay Ghei and Shekhar Hari Kumar.
hmm..interesting analysis,however i have two points to make.Eliminating oil and finance eliminates nearly half of the market,thats a huge chunk.Secondly it is misleading to think that high p/e ratio is indicating high future earnings.A market can have high p/e when the outlook on inflation is low.This is as the author would surely understand is due to low nominal return on debt hence the relative equity valuation moves high as the market accepts future lower return from equity.Lower inflation outlook -> lower debt return(nominal)->market accepts lower return on equity ->higher valuation -> Higher P/E.
ReplyDeletePlease clarify if the Data is to be discounted for Inflation
ReplyDelete