He's missing dividends. The average S&P 500 dividend yield is 1.37% according to something I just found on Motley Fool. If that's accurate, that's pretty substantial.
Too bad they didn't have SPDRs in 1950, he could have been much more accurate with the same amount of effort.
Good point. I would like to analyze returns assuming re-invested dividends (TSR), which I will try to do in my next analysis. However, as I search around the web I'm having trouble finding enough data (SPDRs only go back to '93, and even the economagic data only starts in 1970, which isn't really enough data to meaningfully look at 'average' 30 year investments)
Does anyone have access to data going further back?
I will continue looking myself, maybe also looking at DJI data.
In the spirit of hypothesis lead analysis, perhaps we (or I) should lay out hypotheses for how the TSR analysis will differ (Total Shareholder Return). Clearly the curves will be shifted to the right, probably by the ~1.37% that Matt mentions. But what about the variance? (which is more interesting IMO). I need to think about that..
The dividend yield used to be much higher, the long term average is 4-5%. The current yield of 2% is pitiful. Dividends are really how you make money investing in stocks, the price appreciation is really just a proxy for money supply growth, see http://www.cashflowanalytics.com/news.php?articleID=172. You can get that same kind of appreciation investing in gold, art, 1950's baseball cards - anything that people value and find hard to dilute.
I've wondered if the appreciation of the S&P index takes into account the constant reweighting that needs to be done. For instance, there's a lag as the company falls out of the index and when a mutual fund can actually sell it. I've also wondered if it takes into account the dilution rate of stocks, which traditionally has been around 2% a year.
There are lots of interesting things to consider here. Thanks for putting this together, it set me thinking.
The analysis doesn't consider investment over time - as far as I can tell, this graph shows the probability distribution of returns on a one-time investment after a period of years. The goal of investing long-term is to dampen the effect of investing at a (in retrospect) very good or very bad time. (In addition, of course, you need to put your money somewhere to get returns at all).
Random oscillations can be dampened further by investing at regular intervals, disregarding what the stock price is at the moment: for instance, investing a smaller amount every month for five years (or even every month, period). It would be interesting to see how the probabilities of such an investment scheme fares against a savings account.
You could even make two probability distributions: one for stock market returns and one for savings account returns, considering taxes where they apply. (In my part of the world, income from savings accounts is taxed annually while returns on stock is only taxed when the gains are realized, which affects returns over time substantially). Inflation affects all investments equally, and could be disregarded.
Now that I think of it, maybe I should do this myself. This is exactly the kind of investment scheme I am betting on for the next 20 years, so it seems a bit irresponsible not to do this kind of thing beforehand.
Investing a lump sum by spreading it out over a period of time ('dollar cost averaging') has been shown to produce lower returns than making one large initial investment. Most experts regard it as a marketing gimmick used to ease nervous customers into investments.
That assumes that you have all the money in the beginning, correct? Makes perfect sense in that case - the market goes up on average, so if you invest early on, you get a better price.
But most people don't have a large pile of money sitting around, and earn it over time. AFAICT, dollar-cost-averaging gives you a better return than the alternatives of buying a certain number of shares monthly, buying the best performing stocks, or holding all the cash and investing it when you think it's a good time to invest.
Yes, the term 'dollar-cost averaging' tends to be overloaded. Strictly speaking, it refers to a strategy where you start with a lump sum and invest it in chunks over a period of time, rather than at once. Some people have extended the concept to periodic investments, like you might make in a 401k. In that case, there's really no reason to sit on your money rather than investing it as you go.
I'm pretty sure he meant "on average." And it's hardly surprising, since the market on average goes up. Dollar cost averaging decreases your expected value while also decreasing the standard deviation (risk.) I don't know the result gamble was referring to, but I'm guessing he meant that if you wanted to decrease your risk, then the better way to do so is by adjusting your stock/bond ratio.
For completeness I'll mention that there's also a strategy called "value cost averaging", which gives better returns than dollar cost averaging but is less straightforward to implement, you need to keep a spreadsheet and do some calculations--the basic idea is that instead of investing a fixed amount of dollars every period of time, you invest the amount needed to keep the value of your portfolio growing at a certain rate (i.e., you buy more when the stocks go down and less when they go up.)
Good thoughts.
As you say, this is for a 'lump sum' investment, not investment over time, which has the benefit of dollar-cost-averaging. As I mentioned in my post, I may look at that next, probably in v3.
The tax is is good point too, but I'll leave that off the table for the next version as people's tax situations vary, and it gets complex - maybe v5..
One of the flaws in this is that since he's looking back in time he doesn't include potential investments in companies that go bankrupt. This would shift the whole thing somewhat to the left.
OK, here is the explanation. Maybe it isn't as obvious as I thought.
Let's say that you invest in 100 random stocks in 1978 and intend to keep them as a long term investment. In 2008 when you want to sell your stock it has on average risen by x%. But some of the companies you invested in have gone bankrupt, and thus these shares are worth nothing. This pulls your entire portfolio down by quite a bit.
Now if you look back from 2008 instead of looking forward from 1978 you will see a different picture. If you pick 100 random stocks and see what their stockprice was in 1978 (which is what this guy seems to have done) you might expect to get the same result, but you don't. A lot of companies have gone bankrupt in those 30 years, and you don't include these in your back-looking portfolio. Yet, as we see these days, it is a real scenario that the company you have invested in will simply tank and your shares will be worth nothing. Over thirty years my guess is that 3 out of 100 companies will go bankrupt, meaning that bankruptcies alone diminishes your portfolios value by 3%.
This is an interesting point. The analysis was done on the 'index value' of the S&P 500 (from Yahoo Finance), not on any individual equities. I'm trying to discover how long a 'long term investment' need be, if one invests in an S&P 500 index / ETF / iShare. (I'm no expert on the subtleties of those investment vehicles). My understanding is that they track (as best they can) the value of the index by investing in the stocks that compose the index. Therefore I guess that when company in the S&P 500 goes bust the effect that has on one's investment closely matches the effect it has on index value. I'll research that. Assuming this is true, I think we needn’t be concerned by the bankruptcy issue that you make. But please correct me if I'm wrong about that - I'm interested.
ahh, yes you're tracking S & P 500. Sorry, I probably read the aricle a little quickly... I don't know for sure, but I would think you're right in your assertion that the index incorporates the risk of companies failing.
Also bankruptcies count into your average return of x% if you're tracking the market as a whole. 3 companies will go bankrupt (-100%) but 1 that grossly outperforms the market (+1000% in a few years, for example) more than covers it.
I still don't understand. He's tracking the S&P 500 which changes over time. Have any companies gone bankrupt from that before being delisted? If so, they'd be factored in already. If not, then someone with SPDRs (had they existed) wouldn't have cared.
Too bad they didn't have SPDRs in 1950, he could have been much more accurate with the same amount of effort.