8/9/2015 The Tortoise and the Hare

Sector Model	XLE	-2.10%
Full Model	Date	Return	Days
PWR	3/9/2015	-16.55%	153
CBI	4/2/2015	9.36%	129
MTZ	4/9/2015	-14.60%	122
DRQ	5/15/2015	-23.49%	86
RES	5/19/2015	-18.65%	82
SPN	5/28/2015	-27.85%	73
NOV	6/23/2015	-19.18%	47
INT	7/7/2015	-14.67%	33
AHC	7/28/2015	12.42%	12
FFIC	8/3/2015	-1.73%	6
(Since 5/31/2011)
S&P	Annualized	10.93%
Sector Model	Annualized	20.98%
Full Model	Annualized	13.54%
S&P	Total	54.44%
Sector Model	Total	122.15%
Full Model	Total	70.26%
Sector Model	Advantage	10.05%
Full Model	Advantage	2.61%
	Previous	2015
S&P	53.06%	0.91%
Sector Model	142.84%	-8.52%
Full Model	101.13%	-15.35%

Now we come to the time for wondering just what the point can be when after four years of effort the Full Model has collapsed so spectacularly that it is nearly as bad as the S&P.  To make matters worse, taxes will eat away any alpha that was left, leaving a simple holding of SPY superior to many hours of effort.

The Sector Model, of course, continues to tick away without much care in the world, even after itself losing over 8% so far this year:

An annoying year, but most profitable overall.

And yet, even here there is a problem: the Sector Model trades short term. With capital gains rates at 43.80%, the NET return rate for the Sector Model would be 20.98% * (1-43.8%) = 11.79%.

In other words, you might as well have just held SPY and forgot about the entire exercise.

That’s what IRAs are designed to solve (and yes, I run this in an IRA). The delayed capital gains tax is collected at long term rates after many years of compounded returns.

IRA accounts are fine for these short term versions of the model. But neither will work for a taxable account.

Nevertheless, there is a factor that can be used to full advantage if measured correctly: an adaptive long term holding period.

That is, the Full Model has a secondary holding period based on the collapse of Effective Annualized Capital Gains rates.

Here’s a table to show how that works:

< Year	Annualized Rate	Base Rate
1	43.80%	43.80%
2	11.27%	23.80%
3	7.38%	23.80%
4	5.48%	23.80%
5	4.36%	23.80%
6	3.62%	23.80%
7	3.10%	23.80%
8	2.70%	23.80%
9	2.40%	23.80%
10	2.16%	23.80%
11	1.96%	23.80%
12	1.80%	23.80%
13	1.66%	23.80%
14	1.54%	23.80%
15	1.43%	23.80%
16	1.34%	23.80%
17	1.26%	23.80%
18	1.19%	23.80%
19	1.13%	23.80%
20	1.07%	23.80%

We all know that the short term rate is higher than the long term rate.  What we do NOT usually calculate is the fact that the effect of a capital gain tax can be greatly reduced against compound returns if we can hold onto a stock for a number of years.

This becomes an Effective Tax Rate as seen in the following chart:

That’s all well and good, if one can pull it off.  But who can pick long term outperforming stocks?

Well, Warren Buffett, for one.  He’s not much of a trader.  He waits for a stock to create significant long term value and then holds on – often for decades at a time.

Once we reach a holding period of greater than 5 years, the Effective Tax rate is less than 5%.

I’ve not only tracked the returns of the Full Model from the time I’ve bought and sold each stock – but I’ve also continued to track those sold stocks as if I were still holding them.

The annualized return RATES in the following graph are calculated in two ways: first without tax, and second with tax.

The top line is the return rate in an IRA account of each stock since I began tracking live trades on 5/31/2011; all 189 of them.

As time lengthens, the Effective Tax Rate falls closer and closer to zero, until the taxed account is almost as good as an IRA account.

The ideal holding period for a stock in an IRA account is about 99 calendar days.

The ideal holding period for a stock in a taxed account is about 1525 days (so far).  I’ve not yet found the maximum point.

Now, this is an interesting graph, since we can see that the return rate gradually decays from 99 days until a bottom around 886 days.

After that two things happen:

1)      About 10% of the stocks cease to exist (either through purchase or bankruptcy).

2)      The remaining stocks recover aggressively.

In Haugen’s studies, “value” companies tend to under-perform “growth” companies for about five years, but the stocks associated with those companies have been too aggressively discounted.

The market is very accurate in identifying which companies will do better than others in the next few years, but the market is NOT very accurate at pricing those discrepancies beyond three years.

Investors tend to hyperbolically discount time as it fades further into the future. The difference between two and four years is treated the same as the distance between one and two years.

Logarithmically, those two are equal, but time does not progress logarithmically – it progresses linearly. 

To put this simply, investors will price correctly from one to two years, but price two to four years as if it were two to three years. I’ve shown this kind of perception error in the following table.

Reality	Perception	$100.00
1	1	$105.00
2	2	$110.25
3	4	$115.76
4	8	$121.55
5	16	$127.63

The actual “value” of a company after 5 years is discounted as if it were 16 years in the future. What may seem to be a moderate 5% growth rate is perceived as if it were closer to 1% because of the logarithmic perception of time.

This is how value stocks can become growth stocks. Earnings will begin to surprise and traders will over react in buying the stock as if it had miraculously made 16 years’ worth of growth in only 5 years.

We are always trying to pull a rabbit out of the hat, but if Aesop’s Fable is to be believed, we’d be better off trying to pull a tortoise out of our hat instead.

By the end of this year I will either convert the Full Model into a long term holding period, or else add a long term version.  I haven’t decided yet.  But four and a half years of data has confirmed something that Benjamin Graham argued many decades ago: in the short term the market is a voting machine; but in the long term it is a weighing machine. Good companies are worth buying and holding for the long term.

The key is to find out which companies, and for how long.

We aren’t exactly there yet – but we are close enough to begin the final version of the Full Model – a version that is infinitely patient, and infinitely scalable.

Tim