There are a whole bunch of numbers on the internet being thrown around and proclaimed as the long term return of the U.S. stock market. Anywhere from 7%-11% depending on who you believe. This number is the basis for a lot of folks' investment decisions and its accuracy is not something to be taken lightly... especially when a couple of little problems become apparent when you stop to think about the methodology most commonly used.
So I've decided to go into the historical data for the S&P 500 index and try to calculate it for myself.
The most common way I've seen it done is by picking some point in the 1920's and then saying that the compound annual return is X% through today.
"It's almost 100 years of data. Surely any imperfections would be averaged out, wouldn't they?"
Not necessarily.... The first problem here is that starting and ending point matter.
Let's look at the returns of the S&P 500 from June 1932 through October 2018. That's around 86 years of data. According to this online calculator, you get a nominal 7.7% annualized return (11.7% if you reinvest the dividends). Not bad, huh?
Now let's change the dates by a only few years. Let's do September 1929 through March 2009. That's still close to 80 years.
This time, you get a nominal annual return of just 4.1% (8.3% with dividends reinvested).
That's quite a difference. As you can see, cherry-picking start and end dates can be a big problem.*
*For those who don't see what I did, I chose the peak right before the Great Depression through the low point of the 2000's housing bubble for the first scenario. For the second scenario, I chose the low of the Great Depression as the start point and the current bull market as the end.
The other problem is that nobody who started investing in the 1920s is still around. Most people's investment horizon is probably closer to 35 years. Sure the average return over a 100 year period may be X%, but there are occasional 10-20 year periods where stocks go nowhere.
The solution would therefore be to:
Below, I've done just that using monthly data from Robert Shiller (I hope it's accurate, since that could be another issue) and ran it for all the rolling 420 month periods (35 years) from January 1926 to September 2018, with dividends reinvested. I chose 1926 as the start because that is when better data for dividends became available. I am including the dividends because if you are buying and holding, you will be getting them.
Returns here are nominal (not accounting for inflation).
Here is the distribution...
Here is the same chart using real returns (adjusted for inflation), for what it's worth.
Looking at a distribution chart may be a more realistic way of looking at returns than cherry-picking dates, but still take it with a grain of salt...
Forget the expected 11% annualized return. In the real world, you could lose money one year and make bank the next. And to make things scarier, returns at the end matter more than returns in the beginning. If you lose 40% on your $5,000 portfolio when you are only 20 years old, you will be okay. You can't say the same thing to a 50 year old with a $500,000+ portfolio.
Also, nothing grows at 11% forever. The Roman Empire was once the be all end all... and it went away. U.S. equity returns will be no different. Eventually these outsized growth rates will go away. Will it happen in our lifetime? Who knows?