No. of Recommendations: 10
The annual report and annual meeting are a big event every year. I wondered if that might cause some buying, either leading up to it, or afterwards based on things learned.
So I got the meeting dates and pricing data for 1997 through 2022 (26 years worth) and aligned the prices to the meeting date for each year. I adjusted Berkshire's returns against SPY's daily returns to remove market movement, then used a 20 day moving average to further smooth the data. I averaged those 26 time series and was hoping to see some outperformance in the months leading up to the annual meeting. Unfortunately I didn't find anything useful.
I graphed average daily returns for 6 months on both sides of the annual meeting. They stayed within roughly +/- 0.1% the entire time. (There was a small bias in the positive direction which you'd expect because Berkshire has a positive trend.) A daily 0.1% return seemed too small to be of interest but I wasn't sure what my expected error bars should be. So to check it, I generated a bunch of random numbers with a distribution that matched the Berkshire data inputs and ran them through the same program multiple times. The random data generated similar graphs that seemed to reach up to about the same 0.1% daily return.
My conclusion is I found nothing and am not going to pursue it any further. The only glimmer of hope was the actual data had average daily returns of roughly 0.1% that lasted longer (~2 months) than what was typical for the random data. I would look further at that except it is not where I would expect to see such an effect (months 2 and 3 preceding the annual meeting). I would expect it to immediately precede the meeting, and possibly extend past the meeting date. But that's not I saw. So, I've tossed it into my failure pile.
Comments/suggestions about any errors or improvements are welcome. If anyone wants a copy of the data or software, I'm happy to send them, although the software is a quick and dirty one off.
No. of Recommendations: 8
Being a numbers geek, I did the same analysis not long ago.
I did find a modest effect.
I found that there is on average a small rally which straddles the release of the annual report, starting a few days before and ending a few days after.
The low point at the start of the rally is on average 4-5 trading days before the release, and the high point is on average 7 days (say, 5-8 days ) after the release.
The "typical" bump the way I averaged it is around +2.4% from lowest point to highest point.
The price was on average weak in the several days before and several days after that bullish window, so the net result from roughly day -14 to day +19 is pretty much flat.
It should be noted that if you average any 18 random numbers you will also see a trend : )
An effect which I noticed long ago does seem to keep happening.
Berkshire's stock price tends to have a little rally around each month end, and a weak stretch just before that.
This has also been true of the broad market on average over the decades, but it seems to be quite strong for Berkshire compared to other single stocks.
The speculation is that some people get paid around month end and have automatic buying going on, causing the rally.
e.g., consider dividing time into the really bad stretches and the rest of the time.
Define bad stretch as being from market close on the 9th-last trading day of the month until market close on the 6th-last day of the month: four trading days per month.
The CAGR of BRK's stock price in those days since 1999 has been -23.9%/year, and the CAGR the rest of the days has been +18.5%/year.
(the 9th-last day can be replaced with the 10th-last or 11th-last without changing the result much).
The total return in all "bad" days was negative in 16 of the 22 calendar years 2000-2021. Average "total result of bad days in the calendar year" was -4.5%, and five of those were double digit total percent drops.
Do I try to make use of this? Heck no.
Well, I suppose that if you were about to buy some stock around the 11th-last to 9th-last trading day of the month anyway, it statistically might not hurt to wait just a few days more.
Jim
