Some things are just plain hard to explain. I did the following test using all 17 of the iShares single country funds that started trading 1996:
I tested the performance of each single country ETF on each specific trading day of the month (i.e., the 1st trading day of any month is TDM 1, the next day is TDM 2, etc.)
I also examined the last 7 trading days of the month counting backwards (i.e., the last trading day of the month is TDM -1, the day before that is TDM -2, etc.)
(See also Wild-Eyed T-Bond Speculation (Sort of))
The basic idea was to see if there were any consistently favorable or unfavorable days of the month. I wasn’t necessarily expecting much given that the stock markets of each of 17 different countries could rightly be expected to “walk to the beat of their own drum”, given that the fundamentals underlying the stock market in any given country may be unique from that of other countries.
Or maybe not so much.
For what it is worth, the results from 3/25/1996 through 6/20/2016 appear in Figure 1.
(click to enlarge)
Figure 1 – Trading Day of Month Results for 17 iShares single country ETFs; 3/25/96 through 6/20/16
Each column displays the cumulative % return for each individual single country ETF if we held a long position in that ETF on only that particular trading day of the month.
As you can see – and what are the odds of this, I am not even sure how to calculate that – there are 4 days (highlighted in green in Figure 1) that were uniformly “favorable”, i.e., all 17 ETFs showed a net gain on that particular trading day of the month (for the record, there are three trading days – TDM #7 and TDMs -7 and -6 – during which all 17 ETFs showed a net loss (highlighted in yellow in Figure 1. Go figure).
Using as a Strategy
I am not recommended the following strategy, but wanted to test it out for arguments sake. Figure 2 displays the results of the following test:
*Buy and hold an equally weighted position in all 17 single country ETFs only on TDM 1, 9, 13 and -4 (i.e., if Friday is the last trading day of the month then – barring a holiday – TDM -4 would be the Tuesday of that week), earn annualized interest of 1% per month while out of ETFs.
*Versus simply buying and holding all 17 single country ETFs from 3/25/1996 through 6/20/2016.
Figure 2 – Cumulative % return for holding all 17 single country ETF only 4 days a month (blue) versus buying and holding (red); 3/25/96 through 6/20/16
Summary
Yes, this model can probably be accused of being “curve fit”. Also, does anybody really want to trade in and out of 17 single country ETFs 4 times a month? I don’t know. But the bigger points are:
*Why would all 17 single country ETFs all be up (or down) on a particular day of the month over a 20 year period?
*If I were considering buying and holding a cross section of individual country ETF’s, um, well, I can’t help but think that I would be haunted by the thought that there might be a better way.
Jay Kaeppel
You used 20 and a part years of data, but appear to have only reported 17 years of results. EG, on TD1, you have 17 up years and 0 down years. What happened to the other 3 years?
Monday, June 27 is a favorable day (TDM – 4). Let’s see what happens :>)
TDM-4 would not have worked at all this month. TDM-3 was much better.
Jay, have you looked at a long entry the close of the day before the favourable dates? Get the run up in the overnight and then hold during the day session.
Ryan, for the record the results do involve buying at the close the trading day before the favorable day in order to capture the entire day’s worth of price action. Jay
My follow up question is the split between the night and the day session. This may in fact represent a 4 nights seasonable edge as opposed to 4 days?
Ryan
You curve-fit your comment responses (Don asked a good question) like you curve-fit your data, Jay. I think one could possibly make an argument that all of your stuff is curve-fit. The sample sizes is too small and you have done no validation at all (pretty-looking graphs are nothing of the sort).