# January as a Predictor (or: Fun with Numbers), from Alston Mabry

February 1, 2016 |

Question: Does January serve as a predictor of volatility over the next 11 months?

For fun, I dug up an old measure of volatility I used to use, where I calculate the "minimal path" that a price would have to traverse through the normal Open-High-Low-Close data, i.e., either C-O-H-L-C or C-O-L-H-C. This can be over any given time period. (For many calculations, the High-Low range gives similar results. Note at bottom on calculation.)

Using SPY, calculate the minimal path for January, as a % of the December Close, along with the minimal path for February-December, as a % of the January Close. The results:

Date / Jan minpath / Feb-Dec minpath
Jan-15   5.1%   29.9%
Jan-14   6.0%   28.7%
Jan-13   7.4%   25.9%
Jan-12   10.1%   25.0%
Jan-11   6.6%   43.8%
Jan-10   10.6%   31.0%
Jan-09   23.7%   76.4%
Jan-08   22.7%   67.5%
Jan-07   4.9%   27.7%
Jan-06   6.8%   22.2%
Jan-05   6.7%   19.4%
Jan-04   9.2%   20.4%
Jan-03   19.6%   45.6%
Jan-02   15.8%   50.3%
Jan-01   13.8%   47.9%
Jan-00   13.1%   37.3%
Jan-99   9.8%   27.7%
Jan-98   17.1%   44.0%
Jan-97   14.1%   42.5%
Jan-96   9.6%   35.3%
Jan-95   4.0%   36.8%
Jan-94   4.7%   15.2%

Correlation between the two series:  +0.843
R square:  +0.711

Given the strong R square, I ran a regression with "Jan minpath" as the independent variable, and got the equation:

Feb-Dec minpath = 0.123062 + 2.197213779 * Jan minpath

The minimal path for January 2016 is 18.8%.  Plugging that into the equation:

Feb-Dec minpath = 0.123062 + 2.197213779 * 0.188208

= 53.7%

Multiply that by SPY December Close of 193.72

= 103.95

Or approximately 1040 points on the S&P, i.e., the equation predicts that the minimal path for the S&P for Feb-Dec 2016 will be 1040 points.

Even assuming this estimate is accurate, it doesn't tell you what
*shape* the market will have over the next 11 months. You can, however, model some scenarios.

For example, I saw a recent collection of big-bank predictions for where the S&P (cash) would end 2016, the highest of which was 2350. If you plug 2350 in as the Close and also assume it is the High, then the minimal path for the next 11 months looks like this:

Jan Close:  1940
2016 High:  2350
2016 Low:  1623
2016 Close:  2350

In that scenario you get a predicted Low of 1623.

Assume that the Close is 2350, and raise the High to 2450, and you
also raise the low by 100 points, but still have to get down to 1723:

Jan Close:  1940
2016 High:  2450
2016 Low:  1723
2016 Close:  2350

Assume a Close of 2100 and a High of 2200:

Jan Close:  1940
2016 High:  2200
2016 Low:  1598
2016 Close:  2100

Et cetera…

For me, this analysis works as a mental exercise to help me with a severe shortcoming: My idea of what is possible is always much narrower than the market's version. For example, if from here we drop to 1600, it will be hard for me to think, "We could easily pop back up to 2200 and then finish at 2100".

_____

The minimal path calculation looks like this:

abs(Open - prevClose)
+ (High - Low)
+ min(
[ (H-O)+(C-L) ],
[ (O-L)+(H-C) ]
)

Divide the result by the previous Close to get the %.

`SELECT * FROM wp_comments WHERE comment_post_ID = '10871' AND comment_approved = '1' ORDER BY comment_date`