Hacker News new | past | comments | ask | show | jobs | submit login
Sell in May, and go away? (agoraopus.com)
61 points by theocs on April 9, 2015 | hide | past | favorite | 35 comments



This analysis is only looking at the price of the S&P 500, not the total return. An investor that sits out half of the year will miss out on about half of the dividends paid, which are always positive.

EDIT: Here is a graph highlighting how important including dividends is: https://i.imgur.com/YZSq6K3.png

Another consideration is taxes. The short-term gains produced by selling after 6 months are taxed at normal income rates (or slightly higher), as is the interest from the "risk free" interest-paying investment held the other 6 months. Long-term capital gains and dividends are taxed at favorable rates.


Thank You!

Its ridiculous how many of the timing/tactical strategies ignore these factors because of the complexity, even though they introduce massive drag vs buy & hold.

With the amount of data available, we should be able to do these sorts of backtests fairly accurately. At some point I hope I can compile a bunch of open prices/distribution data sets for people to use. You can easily get, for instance, daily close and distributions for VFINX (Vanguard's S&P 500 fund) back to 1980, but its not neatly compiled anywhere. Trickier is classifying distributions (dividend/LCG/SCG), but again all the required data exists (Sadly it means manually trawling through Edgar)


In case you're unaware, or someone reading this is, have a trawl through the MSCI [1] and FT websites [2] for detailed pricing data.

[1] http://www.msci.com/ [2] http://www.ftse.com/

If you're interested in getting hold of historic tick data you could go to a data provider (for a price) or perhaps more convenient would be to see if you could get access to a Bloomberg terminal.

Academics that research this area, especially questions of policy and market efficiency of both market pricing and funds, can be extremely open about their work, and are often keen to share.


Aren't dividends figured in to the historical prices anyway, though? Via adjusted closing prices?


There are several versions of the S&P 500. The version used here is the price index, which doesn't include dividends. The total return index, ^SPXTR[1], does account for dividends.

https://ycharts.com/indices/%5ESPXTR/level


In theory, but look at the dataset (https://github.com/AgoraOpus/Sell-in-May/blob/master/resourc...) and the adjusted close is the same.


The taxes issue could be solved in a fund that implements the strategy.

There is another advantage: during 6 months of the year you have 0 risk.


0 equity risk. You're still running currency risk with whatever currency you're holding in.


But that difference is diminished when you also take into account that you would be earning interest on your capital in the period you were out of the markets (e.g. through a savings account or short-term bonds).


They are already factoring that in, using a 5% risk-free rate.


> The question then is: What do you do with your cash when it’s not invested in the stock market? Assuming we put them in high caliber (“risk-free”) interest bearing fixed income instrument they would still give us a return on our investment while we’re out of the stock market. Based on data from various historical sources I’ll stick with an average annual geometric risk-free rate of 5%.

I must be misunderstanding this, otherwise WTF??? Tell me where I can get this risk-free 5% rate.


The author is claiming 5% is an historical average, which is plausible because in past decades, both interest rates and inflation were much higher.


About a decade ago in Australia, at-call online savings accounts in Australia were getting a 6% interest rate. Nowadays that's down to 3.75% with an extra requirement to deposit $1000 every month (but still at call).


You're not. If you find it, sign me up.


He fit some time periods that happen to avoid the 87 crash and most of the 2008 crash. It's funny, when stats tell you this story, it feels more compelling than some instruction to just avoid the two largest market crashes since 1929, but in reality neither are telling you anything.

If you chop up financial data sets enough, you can always find some generally defined subsets that perform better than the whole set.


From the graph it looks to me like the entirety of the gain comes from skipping some of the 2008 financial crisis. http://www.agoraopus.com/wp-content/uploads/2015/04/Sell_in_...


Mark Twain allegedly quipped:

   OCTOBER: This is one of the peculiarly dangerous
   months to speculate in stocks in. The other are
   July, January, September, April, November, May,
   March, June, December, August, and February.
Still relevant 100+ years later.


Shouldn't he test his model on another dataset besides the one used to generate it? Maybe it's overfitted. It would have been nicer to see the whole thing developed with only a decade or two of data, then shown to work for all the other decades too.


Correct. This is quantitative analysis 101. A training set and testing set.


I moved a large portion of my IRA to cash towards the end of last year, there were just too many red flags and unknowns, and seemingly unexplainable reactions to the daily news cycle, and the market had performed just so well over the last several years, it seemed like a peak would have to be near.

Of course, what actually happened is I managed to miss a 15% bump in IJT, a 5% bump in VT, a 6% bump in IYY, and about flat in GLD. The two stocks I held directly however did go down about 10% since selling.

So yeah, the two most basic pieces of advice; you can't time the market, and don't hold individual stocks (without spending the time to actively manage your portfolio), both rang quite true for me at least the last 6 months. I just haven't gotten back in, because I'm sure the day I decide to do that will prove to be the actual peak. :-/


You can do dollar cost averaging[0]: buy chucks over time instead of all at once. Of course this increases commissions and other transactions costs.

[0]: http://en.wikipedia.org/wiki/Dollar_cost_averaging


Sounds like you aren't listening to your own advice about timing the market. Just put the money back in and know that 20 years from now it will(more likely than not) be worth more than it is today.


A few months back, I performed a two part analysis of this very subject.

http://www.westonbeckett.com/posts/sell-in-may.html

http://www.westonbeckett.com/posts/sell-in-may-redux.html

In short, while returns for the May to October period may on average be less, they are positive and if anything, the returns for this period have been increasing. The one notable exception being the incredible selloff in 2008. Taxes are a more consistently important issue to consider.


I heard this many years ago. Asking a fund manager colleague, she mentioned it was mostly a fear of liquidity, not a question of HFTs, but should bad news strike, the markets she tended to invest in didn't hold up well (or were not perceived to hold up well, perception causing reality) and price falls were irrationally large, due to this lack of liquidity.

Now, a fund manager that tends to hold onto a stock when its falling by definition sees the falls as irrational, but this was confirmed by her colleague that walked into the room. He added "what's really important is when retail investors start buying, then that's really the time to sell [as after the stock gets popular with the public, who's left to pump it up]."


I don't think this factors in the cost of selling and buying all your stocks twice a year.


To ignore taxes and to assume that you're only holding an S&P500 index, it's 2yearsfee (let's say $10 * 2 * 20==$400). Kind of insignificant.

To not ignore taxes though it becomes more clear. If you literally sell every May, you're never holding for more than a year. In the US at least you're always paying short term capitol gains. This is the real killer here.


Given it's an analysis of the S&P500 you just buy/sell SPY and follow the index. Not that I recommend that, for reasons other people here have pointed out.


By far, the worst month for me has always been March. Now that we've made it to April, I'm planning to sit tight till after the Santa Claus rally.


Seasonal stock market tendencies are real. I created a website (Seasonalysis.com) which quantifies these tendencies. Past performance is no guarantee of future returns but the information gleaned can be incredibly useful if combined with other indicators/analysis.


Several problems: once you start looking at monthly data, you don't have enough Mays to say with any statistical significance that May means anything

another problem is, transaction costs and taxes from selling and buying would probably eat up your gains


I sold in early March. To me it seems like more of a gamble to be in than out at this point.


Depends on your time horizon.


Not really... It depends on the basis of his intuition. If you knew some details about the market that others did not it would make sense to act on that information regardless of if you needed the money tomorrow or forty years from now.


Can you elaborate on why you think its more of gamble to be in than out?


I want to know what null hypothesis was used to generate those significance results.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: