zt: One-year Forward Returns based on P/E Level

Just FYI.

One-year Forward Returns based on P/E Level

 

This was inspired by Greenblatt's comment about one-year forward expected returns in the market based on P/E level. While bears are always telling us how overvalued the stock market is, Greeenblatt seems to always come up with a positive one-year forward expected return. He uses data from the past 30 years to calcluate this, but I was worried that this time period might be a little biased due to the large and persistent decline in interest rates for this period. So I decided to look at this myself over a longer time period.

Note: 1991.12 is December of 1991, the P/E is the P/E or CAPE10 at that time, and the return is the total return over the following 12 months.
Data source: Shiller


Since 1985:

P/E level of over: 20
Number of up years: 11
Total # years: 15
Percent up years: 73.33%
Average change: 5.5%
 

Date	P/E	return
1991.12	24.33	15.32%
1992.12	22.82	9.85%
1993.12	21.29	0.52%
1997.12	24.23	25.34%
1998.12	31.56	21.45%
1999.12	29.66	-5.70%
2000.12	26.62	-12.79%
2001.12	46.37	-20.06%
2002.12	32.59	22.11%
2003.12	22.17	12.77%
2004.12	20.48	7.09%
2007.12	22.35	-38.75%
2008.12	58.98	29.08%
2009.12	21.78	13.86%
2014.12	20.08	2.10%
2015.12	23.74

The data excludes total return for 2016, but we know it was more than 11%, so the results would be even stronger.
Since 1871:

P/E level of over: 20
Number of up years: 14
Total # years: 20
Percent up years: 70.0%
Average change: 5.9%
 

Date	P/E	return
1894.12	26.88	4.88%
1896.12	20.10	16.82%
1921.12	25.21	27.09%
1933.12	22.66	-2.61%
1961.12	22.49	-9.72%
1991.12	24.33	15.32%
1992.12	22.82	9.85%
1993.12	21.29	0.52%
1997.12	24.23	25.34%
1998.12	31.56	21.45%
1999.12	29.66	-5.70%
2000.12	26.62	-12.79%
2001.12	46.37	-20.06%
2002.12	32.59	22.11%
2003.12	22.17	12.77%
2004.12	20.48	7.09%
2007.12	22.35	-38.75%
2008.12	58.98	29.08%
2009.12	21.78	13.86%
2014.12	20.08	2.10%
2015.12	23.74

What happens if we do the above with a 25x P/E threshold?

Since 1871:

P/E level of over: 25
Number of up years: 5
Total # years: 8
Percent up years: 62.5%
Average change: 8.3%

Since 1985:

P/E level of over: 25
Number of up years: 3
Total # years: 6
Percent up years: 50.0%
Average change: 5.7%


All Months, not just year-end
Just to be thorough, I reran all of the above using all months, not just year-end.  I looked at all months where the P/E ratio was over 20x or 25x and what the total return was 12 months later.


Since 1871:
P/E level of over: 20
Number of up years: 139
Total # years: 223
Percent up years: 62.3%
Average change: 3.5%


Since 1985:
P/E level of over: 20
Number of up years: 111
Total # years: 162
Percent up years: 68.5%
Average change: 4.8%


Since 1871:
P/E level of over: 25
Number of up years: 58
Total # years: 96
Percent up years: 60.4%
Average change: 5.1%

Since 1985:
P/E level of over: 25
Number of up years: 54
Total # years: 90
Percent up years: 60.0%
Average change: 5.2%


Using CAPE10
Someone pointed out that CAPE10 might be more interesting to look at for this analysis as some years like 2008 show high P/E's due to depressed earnings (even though depressed earnings will push up CAPE10 too). Here are the results:

Since 1871
P/E level of over: 20
Number of up years: 23
Total # years: 33
Percent up years: 69.7%
Average change: 5.2%

Since 1985
P/E level of over: 20
Number of up years: 17
Total # years: 21
Percent up years: 80.95%
Average change: 8.5%

Since 1871
P/E level of over: 25
Number of up years: 9
Total # years: 14
Percent up years: 64.3%
Average change: 4.5%

Since 1985
P/E level of over: 25
Number of up years: 9
Total # years: 13
Percent up years: 69.2%
Average change: 5.2%

Using CAPE10, All Months

Since 1871
P/E level of over: 20
Number of up years: 281
Total # years: 410
Percent up years: 68.5%
Average change: 5.9%

Since 1985
P/E level of over: 20
Number of up years: 194
Total # years: 251
Percent up years: 77.2%
Average change: 9.4%

Since 1871
P/E level of over: 25
Number of up years: 93
Total # years: 153
Percent up years: 60.8%
Average change: 4.2%

Since 1985
P/E level of over: 25
Number of up years: 91
Total # years: 139
Percent up years: 65.5%
Average change: 5.96%

UncleFryFebruary 3, 2017 at 6:09 AM

Really enjoyed your analysis. I would love to see the results if you run the analysis on the basis of Shiller-PE or another "better" indicator for overvaluation. Because what do high PEs in 2002.12 or 2008.12 really mean?

ReplyDelete

Replies

 

• 
  kkFebruary 3, 2017 at 9:55 AM

  Good question. I tried to run it now but for some reason, the script is not reading in the correct csv; I will work on this later. I suspect, though, that the results may be similar. Why? Because the market has been 'overvalued' for a very long time and has done pretty well.
  
  Anyway, once I get it to run, I will probably put all of this on a single page in the 'scrapbook' section at brklninvestor.com. I will let you know when I put it up... I will try to do it within the next few days...

  Delete
   
   
• 
  kkFebruary 3, 2017 at 10:32 AM

  This is the same, annual version using CAPE10:
  P/E level of over: 20
  Number of up years: 23
  Total # years: 33
  Percent up years: 69.7%
  Average change: 5.2%
  
  ...and using 25x CAPE10 since 1871:
  P/E level of over: 25
  Number of up years: 9
  Total # years: 14
  Percent up years: 64.29%
  Average change: 4.5%
  
  so actually not that different. If you take it up to 30x CAPE10, then only late 1990's/2000 comes out, so becomes more or less a single data point...

  Delete
   
   
• 
  kkFebruary 3, 2017 at 3:14 PM

  OK, I did all the runs with CAPE10 and put it up at the website. Check it out here:
  http://brklninvestor.com/scrapbook/pe_fwd_returns.html

  Delete
   
   
• 
  UncleFryFebruary 7, 2017 at 7:24 AM

  Wow, I did not expect this outcome. Thanks a lot!
  
  But I am not fully convinced:
  - I think average change might be a bit too much simplification, because it underweights the effect of big loss months/years
  - maybe it would be interesting to calculate the return on a 100$ investment if only invested in the <20 CAPE10 month. That means leave the market if overvalued and put everything back in if overvaluation has past
  - Compare the return to the buy and hold strategy
  - I would do this on a monthly basis, to have the better approximation to being invested
  
  Also, and I think this is important, one should maybe look at this from a 2009 point of view. Of course returns look a lot better when they are calculated at all time highs. But I would want to be right wenn there is a major depression.
  
  This might become an interesting series where you could test some other overvaluation indicators, like Buffets MarketCap/GDP indicator.
  
  I might have to change my current strategy if you are correct. Currently that is hedging my stockpicks by shorting the market while the market is "overvalued" from my point of view.
  
  But as I said, I am not convinced yet ;-) Even though this is really suprising to me.

  Delete
   
   
• 
  kkFebruary 7, 2017 at 7:52 AM

  Hi,
  This is only one simple look at it, but it does show sufficiently to me that it is not so smart to short the market just because the P/E is high. There is no doubt about that. There may be other reasons to short the market, though.
  
  As for a simulation to see if you can improve on a buy-and-hold based on P/E ratio, Damodaran did that and showed pretty definitively that getting out and holding cash when P/E is high and getting back in is a losing strategy. You can google it and find it somewhere, maybe on his blog. But the results were pretty clear and unambiguous.
  
  My analysis above was deliberately simple; I didn't play with various P/E levels, time periods and did not isolate specific events/periods, because doing so would risk back-fitting / over-fitting. I didn't want to tweak the data.
  
  As for large losses/depressions, again, since we can't know when it will happen (if at all), we can't really exclude, and the above results include ALL events; great depression, great recession etc.
  
  But since those events and large losses in general are so rare, looking at them alone would result in insufficient sample size to make any determination about anything; this is where many economists went wrong. They tried to model the great recession on the great depression and expected the same or similar outcomes. But the problem with that is that they were comparing a single event to another single event for a statistically meaningless comparison, not to mention a sort of Heisenberg-like (well, again, that's a little off in terms of analogy, lol) situation where the central banks used the depression as a model, specifically, to prevent a recurrence thus throwing off the forecasts of the doom-predictors.
  
  But anyway, go look for the Damodaran simulation. It's pretty interesting.
  
  There is a reason why Buffett has done so well over the past half century and I have been digging into trying to figure out why on this blog since it started, and the above post, I think, gets to one of the main factors of his success. He knew it from the beginning. The rest of us are still too worried and afraid that we take self-destructive measures to shoot ourselves in the foot!