Post New

Grantham quote:
for 7 months of the Presidential Cycle, from October 1st of the second year (this cycle, 2022) through April 30th of the third year (2023), the returns, since 1932, equal those of the remaining 41 months of the cycle! This has a less than one-in-a-million probability of occurring by chance

During Grantham's 7 months the S&P was up from 3726 to 4146 = 11%. Grantham claims during the remaining 41 months the returns equal those +11%, resulting in around +22% ("around" because to know exactly one would need to know more details).

I am not familiar with US elections but google says the election is on 5.Nov 2024. So the cycle started 4 years before, on 5.Nov 2020, right?

S&P then: 3549. +22%: 4330. S&P now: 5158.

So from Grantham's PC statistics is practically "guaranteed" that at the end of the year the S&P will be around 15% lower than now.

Jeremy Grantham's work is of course backed by analysis - and this one is of the Statistical kind.

Specifically Mean/Median/Distributional variety of a long cohort ie timeframe of observations.

Said .... you can reproduce this yourself - its not too difficult to obtain daily prices for the DJ-30 from the 1930s. You just have to group the returns by Year and Month - and create another group called Leap ( literally that's what Pres Yrs are): Modulus Remainder( Year) %4 ie Grouping by Y0,Y1,Y2, Y3

What he has done is then break those 2 group MEANS into 2 separate periods and compare. The point is the MEAN is the average of multiple observations ie Years - for almost 100 yrs of observations every 4 year cycle - 25 observations ( 23 I think to be precise based on availability)

So its "NOT PRACTICALLY GUARANTEED" - its the Long term Average - so YES in the long run - the Hypothesis interpretation is as follows:

"Majority of the gains in the US markets are generated between the Fall of Year 2 to Summer of Year 3" ...

It DOESNT MEAN - that every 4 year cycle conforms to the Statistical Mean ie Average - that is contra to the concept of Statistics

Hope this helps!

Without arguing for the validity one way or the other, let me share with you a simple test of Grantham's hypothesis.

This test is based on being invested in the Dow30 from October 1st of the second year of every US President's term through April 30th of the next year. This is a period of seven months. I'm calling this PY Best Periods. I then compare this to what I call PY Worse Periods which is the remaining forty-one months. Keep in mind the ROI and CAGR values on the left side are based on holding cash (treasuries) the remaining time, while the CAGR on the right are based on just the CAGR during the holding period.

10/1/1902 to       3/7/2024   Dow                                                                 
Results           ROI         CAGR   GSD    Sharpe  DDD3    BBW   Win %  Drawdown  CAGR Invested  CAGR Cash
PY Worse Periods  5740.91%    3.41%  17.66  0.09    10.84%  0.34  69%    -84.49%   3.31%          3.99%
PY Best Periods   190794.44%  6.42%  6.24   0.51    2.82%   2.45  77%    -32.96%   20.16%         4.23%

The period is from 10/1/1902 to present. I'm able to test it from 1902 since the Dow30 has daily data back to 1885.

This next test is for GTR1's S&P 500 Dividend Adjusted which makes a big difference as you'll see. It is from 10/1/1926 to present.

10/1/1926 to       3/7/2024  ^S5T                                                                
Results           ROI        CAGR   GSD    Sharpe  DDD3    BBW   Win %  Drawdown  CAGR Invested  CAGR Cash
PY Worse Periods  66721.21%  6.90%  18.66  0.25    10.22%  0.73  85%    -80.41%   7.43%          3.78%
PY Best Periods   50420.78%  6.60%  6.31   0.47    2.98%   2.39  80%    -32.96%   22.99%         3.98%

Hope this helps!
Oh yes, immensely!!!

Because this you say: What he has done is then break those 2 group MEANS into 2 separate periods and compare.

means that I misinterpreted his: the returns, since 1932, equal those of the remaining 41 months of the cycle! This has a less than one-in-a-million probability of occurring by chance

Because of the probability mentioned by him I interpreted this as "was the case for EVERY SINGLE ONE of those 23 presidential cycles", as this surely would have an extremely low probability.

After this being clarified actually I think this wrong interpretation would be very worth a test! For how MANY of those 23 PC's was the as Zee calls it "best period" equal to it's "worst period"? A test like this:

1) compare each PC's "best period" with it's "worst period"
2a) set a criterion for them being equal, f.e. an interval of 20% (to start generous; can be tightened later if the test is positive) => If the returns in the "best period" deviate not more than 20% from it's "worst period" you treat them as being equal
3) count in how many of the 23 PC's the "best period" equals the "worst period"

2b) for investments even more useful: count in how many PC's the returns during the "best period" were "Not worse (not less than -20%) or better (>=0%)" than during the "worst period".

A test for which - see Zee's post - of course one would need "GTR1's S&P 500 Dividend Adjusted", to include dividends; not possible for me as I think to understand to run GTR1 one needs a Sip Pro subscription (which I don't have), correct?

QUOTE :A test for which - see Zee's post - of course one would need "GTR1's S&P 500 Dividend Adjusted", to include dividends; not possible for me as I think to understand to run GTR1 one needs a Sip Pro subscription (which I don't have), correct?

No dont think so - I have run GTR1 in the past and it works fine - as long as you need the actual underlying data on specific stocks or "current picks" on Screener. Index data testing should not be an issue

Said stated: A test for which - see Zee's post - of course one would need "GTR1's S&P 500 Dividend Adjusted", to include dividends; not possible for me as I think to understand to run GTR1 one needs a Sip Pro subscription (which I don't have), correct?

I believe anyone can obtain this data from GTR1.

Regarding the method, I've posted extensively on what I call the "Best Nine Months" research back in 2006 IIRC. In it I found the optimal start date to be October 6th, and then a hold for nine months. This is NOT based on Presidential Cycles per se, but rather, out of all periods of time in a four-year period what are the absolute best nine months, twelve months, etc.

So adjusting to this you end up with the following:

10/1/1926 to       3/7/2024  ^S5T                                                                
Results           ROI        CAGR   GSD    Sharpe  DDD3    BBW   Win %  Drawdown  CAGR Invested  CAGR Cash
PY Worse Periods  1674.30%   3.00%  17.84  0.03    10.87%  0.3   69%    -85.18%   2.81%          3.79%
PY Best Periods   59475.80%  6.78%  6.98   0.46    2.98%   2.44  88%    -40.23%   19.77%         3.98%

As you can see, there is a vast difference between the two - BTW - I had to correct an error in what I previously posted. This is now correct to the best of my ability.

As to the actual periods compared:

Date       Signal         % G/L      Date       Signal         % G/L
6/30/1927  Worse 39 Mths  22.00%     10/6/1926  Best 9 Months  7.95%
6/30/1931  Worse 39 Mths  -38.18%    10/6/1930  Best 9 Months  -25.94%
7/1/1935   Worse 39 Mths  24.64%     10/6/1934  Best 9 Months  27.97%
6/30/1939  Worse 39 Mths  -14.62%    10/6/1938  Best 9 Months  -11.80%
6/30/1943  Worse 39 Mths  17.78%     10/6/1942  Best 9 Months  28.56%
6/30/1947  Worse 39 Mths  30.71%     10/7/1946  Best 9 Months  4.99%
7/2/1951   Worse 39 Mths  49.37%     10/6/1950  Best 9 Months  5.28%
6/30/1955  Worse 39 Mths  18.81%     10/6/1954  Best 9 Months  23.86%
6/30/1959  Worse 39 Mths  -8.94%     10/6/1958  Best 9 Months  20.01%
7/1/1963   Worse 39 Mths  6.88%      10/8/1962  Best 9 Months  19.67%
6/30/1967  Worse 39 Mths  -9.04%     10/6/1966  Best 9 Months  14.76%
6/30/1971  Worse 39 Mths  -31.82%    10/6/1970  Best 9 Months  13.89%
6/30/1975  Worse 39 Mths  0.12%      10/7/1974  Best 9 Months  44.68%
7/2/1979   Worse 39 Mths  13.21%     10/6/1978  Best 9 Months  -5.22%
6/30/1983  Worse 39 Mths  46.04%     10/6/1982  Best 9 Months  29.41%
6/30/1987  Worse 39 Mths  4.35%      10/6/1986  Best 9 Months  35.53%
7/1/1991   Worse 39 Mths  27.62%     10/8/1990  Best 9 Months  17.22%
6/30/1995  Worse 39 Mths  69.95%     10/6/1994  Best 9 Months  20.67%
6/30/1999  Worse 39 Mths  -32.34%    10/6/1998  Best 9 Months  41.69%
6/30/2003  Worse 39 Mths  31.88%     10/7/2002  Best 9 Months  21.05%
7/2/2007   Worse 39 Mths  -18.97%    10/6/2006  Best 9 Months  14.22%
6/30/2011  Worse 39 Mths  36.87%     10/6/2010  Best 9 Months  13.19%
6/30/2015  Worse 39 Mths  50.33%     10/6/2014  Best 9 Months  3.69%
7/1/2019   Worse 39 Mths  12.01%     10/8/2018  Best 9 Months  0.87%
6/30/2023  Worse 39 Mths  12.74%     10/6/2022  Best 9 Months  14.97%
Average                   12.86%                               15.25%
Median                    13.21%                               14.97%
Percent    Positive       72.00%                               88.00%

Keep in mind that the table on the left is invested 39 months out of the four year period, while the one on the right is invested just 9 months and holds cash the remainder of the time. So you cannot just compare the one to the other since the periods are so vastly different!

As this "Presidential Cycle" is fascinating for me today I've spend the whole day (too much time on my hands) to read practically everything Zee and others in 2005/6/7 had to say about it (Zee, thanks for pointing to your 2006 posts, but actually yours from 2005 are even more illuminating). It seems to come down to this:

Jim (in 2007): I believe the "presidential cycle" effect to be a crock. For my reasons, see post 185404. The bottoms in 1949-1953-1957 tend to indicate to me that a cycle exists that is a few years long, often four, but that it doesn't necessarily align with years divisible by four.

Zee (in his reply): Whether you consider the cycle to be a Presidential Cycle or just a four year cycle, there is little doubt that a cycle exists and it is approximately four years in duration.

I think the evidence (Zee, Grantham etc.) is overwhelming and that it's not "just a four year cycle", but definitely a "Presidential Cycle". Why? Because a cycle doesn't come out of thin air. It's an effect, so there must be a cause. And the only plausible argument I found why such a cycle exists (which even Jim admits, though he's not convinced about it being exactly 4 years nor about it being presidential) that makes sense is what WendyBG said (MI-178033, 11/16/2005):

http://datahelper.com/mi/search.phtml?nofool=youBe...

Again: Even Jim admits to a cycle. A cycle does not come out of nowhere, there must be causes, (psychological) reasons. WendyBG presents them. This for me is convincing, more than "Ok, there is a cycle; no idea why".

One can be of different opinions though whether Zee's follow-up work in which he tried to differentiate between months and sectors might be too extreme for one's personal taste or not, but the evidence data for the general principle is so convincing that it's extremely tempting to use this more than anything else as "primary timing indicator" (eventually in conjunction with Zee's 9 months Oct-Jul instead of Oct-May plus "Superior 8").

Of course it's not perfect. 2007/8 dramatically contradicted it: 2007 (3rd PC year) was a down year, 2008 (4th year) even a crash year. During the PC bullish period from Oct 2006 (2nd year) to the end of the cycle Nov 2008 (4th year) the S&P went down a full 35% (would be interesting how often this happened, how often the bullish around 2 year period actually was a Desaster). And during the current and the previous cycle the 1st year also contradicts it --- but the other 2x 3 years are fully (as far as one can say that yet for 2024) in line with it.

One more thing:
historical returns for 3rd Year Presidential cycle very interesting... The last down year (-3%) was 1939. 24 of the 29 third years were up...
Source: http://datahelper.com/mi/search.phtml?nofool=youBe...

I don't see that as significant as over time indices go up - - - but it leads to an idea for a more interesting test: In how many out of those 29 was the 1st year as "predicted" (hey, that's in Apostrophs!) actually going DOWN?

Post New

Mechanical Investing FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.