I downloaded some data from NSE and have been looking at ways to analyze and find patterns in them. I started by trying to find a way to identify if and what stage of a bull market we are at. While pure PE alone may not be a good indicator, I found (PBV - Dividend Yield) to do quite well. In the chart below, the left section shows the indicator as a heat map. In other words, I took (PBV - Div. Yield) over the entire data and broke it into 10 equal parts, mapping each part to a color between red and green. Deep Red indicates danger, while light green means the market is undervalued. In fact, whenever dividend yield has gone above PBV it has always been an excellent time to buy. The right section of the chart shows the individual ratios obtained from NSE.
For the last decade, while a PE of about 25 has been a good indicator of a top being reached I feel the level of madness in 2008 was way beyond anything seen anytime since. For 2011 you see the PBV being lower and the fall from that peak to be much less. The indicator bears this out with 2008 showing up almost as scarlet red, while 2011 is colored brownish.
If I am willing to trust this indicator, I can go one step further and ask when was the last time that the indicator was at the same level? I found all previous times when the indicator was in the same range and marked them in black and also connected all those points into a line. You see the results in the plot below.
For instance we saw similar levels around February 2004, and I hope we are in a similar situation this time. Looking at the ratios around that time, we see PE jumping rapidly from 12 to 22 and falling back again.
To dig a little deeper, I computed an approximate NIFTY EPS by dividing the closing price by PE. The right part of the chart shows EPS and you can see the rapid jump in EPS that led to the real bull market.
While the P/E, P/BV and Yield patterns/thresholds that you mention are all covered in earlier studies, you have brought up something new - strong buy signal when Dividend Yields crosses P/BV.
This will be useful in the aftermath of the Crash - when everyone is shit scared to get back in :-). Will be good to catch you then.
There seems to be something wrong in the last part - set of images for 2009-2013. While 2003-2007/8 were the best years for the economy 2011-2013 were some of the worst years for the economy.Please check the data and conclusions.
While this is a lot of effort, and I would compliment you on that, I would also caution against such data mining. If you take a bunch of parameters, and then try to see the best fit to data, it is pretty much certain that some parameter will come up, which will explain the pattern, purely by chance. The predictive or otherwise value of such a parameter needs to be taken with a pinch of salt.
@Donald: The NIFTY is an aggregate and does not indicate how badly individual sectors may have under-performed over any period. I feel 2009 was the worst we have seen. If you look at the EPS graph, you will see this is the only period where the NIFTY EPS fell, and significantly at that. We never really recovered from that fall, and have not matched the peak EPS growth since. For 2011 through 13, while cyclicals might have fared poorly, some sectors like IT, pharma and FMCG did reasonably well. But I think if I choose a better color scheme, what you say does show up where 2011 through 13 are average years. Even the PE graph shows it staying near the median line over that time frame.
@Samir: You are absolutely right. The data is too little to make even a statistically significant statement. I would have preferred atleast 4 or 5 complete boom-bust cycles. Even then, we all know that expensive stocks can keep getting more expensive and cheap stocks cheaper. That is why I prefer to use it as a heat map than a classifier that gives a sharp buy/sell signal. While nothing compares to experience and a level-head in the market, I think some sort of quantitative indicator atleast gives you pause to take stock and the confidence to go against the trend.
What I lack in experience I am trying to make up by analyzing and understanding the past using methods that I am comfortable with. I think the only sustainable way to invest is for each of us to create and fine-tune our own styles.
P/BV at around 2.5, is almost at historic low. PE ratio at around 18, which is at average level, looks high because Indian corporations are working at almost lowest profitability ever. Dividend yield at around 1.8% is very high- average DY has hovered around 1.4%, at market peaks it has reached to level of even less than 1%. Thus on various parameters, except PE, the market looks well below its average. If profitability increases in future, earning can go much higher bringing the PE parameter in sync with other parameters.
I have one more observation. 2008 was not something exceptional. Markets have gone to much crazy valuation almost every decade. Internet boom of 2000 was crazier than 2008, and Harshad Mehta boom of 1991 was craziest. More than 6 years have lapsed since January 2008, and may be we are moving towards crazy valuation of decade. However, it is never advisable to play greater fool game.
Another issue is valuation of boring quality defensive stocks. They are still commanding a crazy valuation, thus it is not possible to shift portfolio towards them. At market peaks these quality defensive plays goes extremely out of favour and are available at fair value. Once these stocks reach into fair valuation zone, that will be another indication of market approaching peak.
@Hitesh: I would be uncomfortable using this data for both sharp buy/sell signals. Even in late 2006 when markets were overheated, selling out would have meant missing the rest of the boom. When the market peaked out again in 2010, it was not as overheated as earlier. The PE was expensive but growth never picked up like it did in 2004. And future growth is the one thing that you cant really predict. Even now, we have all been waiting for the interest rate cycle to turn but it just keeps getting delayed.
I like the buy signal when PBV falls below Div. Yield because it suits my temperament as an investor, and is not necessarily very objective. Such instances are very very rare though.For instance, I have been waiting for a sharp correction this time around, and have not seen PBV and Dividend yield even come close.
I said earlier that corporate profitability is likely to rise. I would like to explain basis of this statement.
Historically Indian corporate have had 17% ROE on an average. That is the reason that market has given that return. Further prime lending rate of Indian Bank has hovered around 14% which is cost of debt for most the companies. Any figure of ROE, less than 14% for a prolonged period of time will result in corporate bankruptcy. 17-18% is sustainable, and necessary for stability in financial system.
We can find present ROE of the market by the formula- PBV/PE=E/BV. Presently it is 13.5%. Not sustainable. This is the prime reason of mounting NPA. It has to go up if the system has to survive. It cannot go up by any movement of price of the shares as price is out of the equation. It can go up only when rate of increase in earning is substantially more than rate of increase in book value. I expect that to happen… one caveat- Market ROE of substantially more than sustainable 18% is a danger signal. Most of the bear markets start when ROE is substantially higher than 18%.
Pretty good Random Walker… Very interesting study about market peaks and bottoms.
For your third chart, I would suggest you to use EPS growth over last year rather than EPS. Using plain EPS skews your graph and it will always shows recent time as great. Let us take following example for illustration.
2004 EPS Nifty - 50(only for illustration not actual)
2005 EPS Nifty - 60(20% growth)
2013 EPS Nifty-150(illustration only)
2014 EPS Nifty-165(10% growth)
In your graph, the EPS difference of 2013 and 14 will be seen as a larger amount (15) compared to EPS growth in 2004-05(10), even though the actual growth in EPS was 20% then compared to current 10. I believe Donald’s confusion are because of the same reason.
@Gyan: Thanks. I also got the same idea on EPS growth and I computed a finite derivative over the EPS values today. Unfortunately unless I find a good way to smooth the chart it is going to look pretty meaningless.
I am looking at ways to compare stocks based on their prices. What’s really of interest are the relative gains that a stock makes over a period of time. In other words, a movement from, say, 30 to 60 should matter the same as 60 to 120. For them to appear the same on a graph, plotting on the log scale seems the right way.
To compare different stocks they would need to be normalized so that they can be compared. One good way would be to divide each by its minimum value. The absolute minimum though is difficult for any investor to capture. To make the relative gains more realistic, I used the 10thpercentile of stock prices taken over the entire history as a normalizing individual stocks.
I plotted them on a log to the base 2, so each notch on the graph corresponds to a doubling in the stock price. For instance, I plotted TITAN versus PIDILITE on a log2 chart below. TITAN went up about 16 times (4 notches) between 2004 and 2007 while PIDILITE went up a little less than 8 times. Since then they have done about the same on a relative basis. The line at the bottom of the chart shows the (PBV a DivYield) heat map that I described earlier. I added an extra color blue to indicate undervaluation, and now green represents intermediate market conditions. The chart title also shows the correlation between the two charts. In this case its 98.5% over the log values.
I got the data from NSE and used a heuristic to automatically adjust for stock splits and bonuses, so the graphs may not be completely accurate. Please verify the conclusions on your own.
Penny stocks look very good on log charts. You can easily be misled into ranking poor quality companies better than the good ones. Log charts can atmost be another tool that you use to analyze stocks.
To find what stocks did well in the last big bull run, I looked at the maximum relative gains over the period between 1-Jan-2003 and 1-Jan-2009. The top 400 companies are listed in the file below:
Next to each company name is the log2 of the maximum relative gains made. For instance, UNITECH that tops the list went up 2^10.20 = 1176 times. As expected we see a lot of penny stocks. To filter the list further I added an extra condition that the stock should have closed yesterday (July 2) above Rs. 100 and tried again.
As expected this list is much more reasonable now. Anyone looking at a two or three year trading portfolio could use it as an initial filter. Some of the stocks have poor corporate governance, so buyer beware! Also, just because it went up last time does not mean it will happen again.
I dont like very deep data analysis, despite the fact that I graduated as Statistics Major. PBV for the market and PE ration of the market is available in every issue of Capital Market from where I picked those figures. ROE on shareholder money (BV) can simply be derived by using the formula- PBV/PE=E/BV=ROE.
I have done some simple study based on monthwise PE,PBV data which was available on BSE site- one can find monthwise data easily which can be copied and pasted on excel sheet and simple analysis can be done- link:-http://www.bseindia.com/markets/keystatics/Keystat_index.aspx?expandable=2 … I did analysis since 1991 for which data was available there. Unfortunately I cannot upload the excel worksheet readily as it was done on older laptop, and probably data could not be transferred on this one.
All that I am giving simple thumb statistics- average PE for the market, average DY, PE/DY when the market peaked in the past. Are these data something beyond anyone have access to?
I started only recently and don’t have data before 2000, because that’s the earliest that NSE provides. I am looking for more data sources hopefully going back earlier, but I think they are all from paid websites.
I took a look at risk-adjusted returns over the past decade, on a little prodding from Donald. For risk, I used drawdown which measures the fall from the peak. The higher the stock price falls the riskier it is. The maximum drawdown gives a single number which I can use to rank stocks by risk. For returns, an averaged version of the cumulative returns was used.
I focused on only those companies that were listed before 2003, narrowing it down to about 500 companies. Risk/returns were calculated between 1st Jan 2003 and yesterday (7th August 2014). Two interesting patterns stood out:
The fast and the furious
Some companies gave phenomenal returns in very short periods of time. You were lucky if you were in them and held on (against all reason). But you needed to get time your exits as well. Once the gains were made, they either fell off, consolidated, or slowed down. A few of the standouts were:
Eicher Motors
Havells
Kajaria Ceramics
Shree Cements
Sesa Sterlite
TTK Prestige
Unitech
The steady compounders (Perennials)
These stocks never fell much. You could buy them no matter what happened in the market, and not lose sleep. The top few were:
Colgate Palmolive
Hero Motor Co
ITC
Sun Pharma
P&G Hygiene and Healthcare
Apollo Hospital
Marico
GSK Consumer
Lupin
Cadila Healthcare
Asian Paints
Britannia
Glaxo
Dabur
Godrej Consumer Products
Bosch Limited
Hindustan Unilever
I noticed quite a few FMCG and Pharma companies. Most of them had some sort of foreign connection. Guess that’s the quality of professional management.
Notes:
There is no perfect way to measure risk. Modern Portfolio theory has spawned a number of statistical measures such as Gamma/Omega, Drawdowns, or SemiVariance. Each has its own flaws. I dont claim to understand the nitty-gritty, just enough to try them out.
For the perenenials is it possible to come up with some ranks - even quartiles will do. I am surprised not to see Nestle in the List!
By the way it will be better if we can make the data more represenatative that is access data from before the 2000 crash also and extract a common list of perennial dependables in both 2000 and 2008 crash.
Thanks Random Walker! Excellent analysis. Can you also share your spreadsheets so we can have a go at the data and hopefully try to come up with some more insights?
