Dow historical chart: reassuring or unsettling?

[Latestarter]

http://graphics8.nytimes.com/images/2007/08/11/business/11charts.650.gif

This appeared in the NY Times today (Saturday) to illustrate an article about the Dow's (and certain other indexes') amazing run for the last 25 years. But I can't stop looking at what preceded this period. Forget the 3-year tech bust in the early part of this decade, or even the recession of the '70s. An investment in the Dow index at about 1957 would not have broken even for more than forty years.

The accompanying article acknowledges that this excludes dividends, but points out that it also excludes all transaction costs, annual fees, and taxes. And of course it's true that we're talking just about the Dow here, not a diversified portfolio. But still.

Because few of us have been investing for longer than a quarter century, our idea of risk may be framed by 2000-2002 or maybe 1987. But who's to say that the market won't enter another period of losses (or at least no gains) that extends as long as an individual's working lifetime? This isn't just a theoretical possibility; it happened immediately before the current 25-year period.

Please tell me I'm overlooking something here.

[EmergDoc]

No, you're not overlooking something here. Investing in equities is risky and sometimes that risk even shows up.

[cudaman]

But still.

Seems to me you places your bets and you takes your chances. It's one interesting game for sure, albeit a high stakes game. It's only money though in the grand scheme of things. Even the GEICO cavemen seem to survive. I mean how much can a caveman's portfolio be worth?

[Not trying to be non-responsive but I make it a point to make one smartaleck comment per week just to make sure I don't take life too seriously :-) ]

[goosecat]

An investment in the Dow index at about 1957 would not have broken even for more than forty years.

These "price" graphs are always flawed because they don't include reinvested dividends.

You need to use a "Growth of $10,000" type graph to get some useful data. Here is one source of such graphs - SmartMoney. It doesn't allow you to see historical Dow data but I imagine the Dow probably broke even in less than 20 years from 1957.

[bob90245]

Those dividends were not chump change, ya know. They averaged around 5% historically -- nearly half the total return. Even today, you can get a nice dividend yield at around 3% with something like an exchange traded fund such as DVY.

Speaking of cool charts. You can check out the Excel charts at my website.

http://bobsfiles.home.att.net/download.html#Growth

[Taylor Larimore]

"An investment in the Dow index at about 1957 would not have broken even for more than forty years."

The facts:

The Dow Jones Industrial Average closed at 436 on December 31, 1957

The Dow Jones Industrial Average closed at 7,908 on December 31, 1997.

Best wishes
Taylor

[Latestarter]

"An investment in the Dow index at about 1957 would not have broken even for more than forty years."

The facts:

The Dow Jones Industrial Average closed at 436 on December 31, 1957

The Dow Jones Industrial Average closed at 7,908 on December 31, 1997.

Best wishes
Taylor

Comforting, but also puzzling! Since those figures obviously don't include reinvestment of dividends, how can we explain what appears to be an extraordinary discrepancy with the graph?

[MossySF]

It's amazing everytime I see an analysis in the news about the Dow or S&P, they always use just the index numbers but ignore the dividend yield. Here's what I get for the S&P500+dividends running the same scenario as the NY times article.



Compare against:

[peter71]

this is just the message in which i thought the NYT data was one of the lines on mossy's graph . . .

[bob90245]

Comforting, but also puzzling! Since those figures obviously don't include reinvestment of dividends, how can we explain what appears to be an extraordinary discrepancy with the graph?

I, too, find the NY Times graph to be puzzling. I cannot reconcile two conflicting items:

1) The title says "Compound annual percent change in Dow Jones Industrial Average for 25-year periods"

Yet,

2) It shows 25-year periods starting in 1926 through 2007.

Unless they have a clear crystal ball, how do they know the 25-year compound annual percent change for the Dow Jones Industrial Average starting in 2007?

[nisiprius]

The facts:

The Dow Jones Industrial Average closed at 436 on December 31, 1957

The Dow Jones Industrial Average closed at 7,908 on December 31, 1997.

Best wishes
Taylor

Fact: What cost $436 in 1957 would cost $2514.35 in 1997.

http://www.westegg.com/inflation/infl.cgi

So, in constant dollars, the Dow multiplied by a factor of 3.145 over forty years, for a "real" interest rate of 2.9% above inflation.

Probably not terribly different from a passbook bank account or a United States savings bond.

But, as others have observed, the dividends were not chump change.

The most interesting thing to notice about these charts, in my opinion, is how long the big ups and downs lasted. In order to smooth them out, you can't just average over ten or twenty or thirty years. They are very long compared so, say, the length of time during which I have been working and making enough money to be seriously saving for retirement. In planning one's financial future one can't just plug in a number for historical returns, one had better seriously plug in numbers for the wide range of returns that could credibly apply even over a thirty-year period.

The next twenty years absolutely could be like 1960-1980 and they absolutely could be like 1985-2005, and nobody knows which.

[peter71]

sorry, i thought the two lines on mossy's own chart were comparing returns with and without dividends reinvested -- i assume they're instead real vs. nominal returns. i'll leave this post up in case others make the same error but delete my previous one.

mossy, if you can quickly produce a single chart that allows for direct comparison that'd be great!

pete

[MossySF]

interesting graph, though it seems a little counterintuitive that the role of dividend reinvestment would increase over time (i.e., given the CW that dividends were a bigger deal in the past than they are today). do you think it's that the CW is wrong or some more complicated story about "reinvestment on the dips" in the context of higher volatility and returns over time?

I guess I should have labelled the lines. The 2 lines I did are nominal and real return. I did not graph against index-only since that number was already there in the NYT graph.

To compare against index-only, my eyeball on the NYT graph shows:

1980: 2.5% versus 8.5%
1985: 2.5% versus 9%
1990: 5% versus 9.5%
1995: 7% versus 10%
2000: 10% versus 15%
2005: 10% versus 12%

You can see the gradual creepup for index-only versus total return. What was 6% in 1980 is now 2% in 2005.

Unless they have a clear crystal ball, how do they know the 25-year compound annual percent change for the Dow Jones Industrial Average starting in 2007?

It probably means the 25-year period ending 2007. The Dow started in the late 1800s so it is possible to run numbers starting from 1900.

[nisiprius]

Standard and Poor's website has a way-too-fancy company history timeline at http://www2.standardandpoors.com/portal/site/sp/en/us/page.topic/aboutsp_ch/4,2,2,0,0,0,0,0,0,0,0,0,0,0,0,0.html . If you click on that, and when the picture of the steam engine comes up click on S&P 500 Index near the upper right, it gives you a pretty picture of a graph which I frankly can't interpret, and a little table of average annual total return including reinvested dividends.

1928 1929 17.60%
1930 1939 5.30%
1940 1949 10.30%
1950 1959 20.80%
1960 1969 8.70%
1970 1979 7.50%
1980 1989 18.20%
1990 1999 19%
2000 2000 -9.10%
2001 2001 -11.90%
2002 2002 -22.10%
2003 2003 28.70%
2004 2004 10.90%

Here's the same data with a little spreadsheet work, in which the fourth column is the CPI and the fifth column is annual interest rate over the stated time period adjusted for inflation, i.e. the interest rate in constant dollars, and I've also combined 2000 through 2004:

1928 1929 17.60% 17.1 18.63%
1930 1939 5.30% 13.9 7.50%
1940 1949 10.30% 23.8 4.52%
1950 1959 20.80% 29.1 18.40%
1960 1969 8.70% 36.7 6.21%
1970 1979 7.50% 72.6 0.41%
1980 1989 18.20% 124 12.04%
1990 1999 19.00% 166.6 15.54%
2000 2004 -2.29% 172.2 -2.94%

Remember, the last column is in constant dollars, i.e. relative to inflation. Oh, for the entire period 1928-2004, S&P says the return including reinvested dividends was 10.03%, and when I adjust for inflation I get a constant-dollar return of 6.64%.

So it does appear that if you reinvested dividends, even during 1970 through 1979 you would not have lost real value. What (say) Treasury bills or bank CDs did during the same period, I don't know.

This is my own rather hasty spreadsheet work and it could be wrong... and I used the December CPI and I'm too lazy to figure if that created any off-by-one-year errors... wouldn't affect the results much.

[Latestarter]

So it does appear that if you reinvested dividends, even during 1970 through 1979 you would not have lost real value.

What was the dividend rate then, as compared with now?

[bob90245]

Unless they have a clear crystal ball, how do they know the 25-year compound annual percent change for the Dow Jones Industrial Average starting in 2007?

It probably means the 25-year period ending 2007. The Dow started in the late 1800s so it is possible to run numbers starting from 1900.

Thanks, that makes more sense. I see that the first year on their chart is 1926. This would make the the first data point the 25-year period from 1901-1926.

[EyeDee]

I could have easily made a calculation error, as I just did a quick calculation to bring the numbers up through 08/10/2007, but I get an estimated annual return for 2000-08/10/2007 of +1.51%. We have a long way to go to even get near to the 5.30% of the 1930’s.

I used the -2.29% nisiprius listed for 2000-2004, 4.9% for 2005 and 15.8% for 2006 for S&P 500 Index from the Annual Report for the Vanguard 500 Index fund (the actual fund numbers were slightly lower), and 3.61% for 2007 from year-to-date listed on Vanguard’s web site for 500 Index Admiral shares.

[MossySF]

Grabbed data from the Schiller numbers to do 100+ rolling 25-year periods. Shows nominal index, total return and real return. Real return is slightly different from my previous graph since I'm using CPI here versus 85% of Treasuries previously as a proxy.

[Taylor Larimore]

Hi Bogleheads:
Most of us have an investing lifetime of at least 50 years. It is important for us to think "long-term."

The link below is to a chart that shows nearly 200 years of U.S. stock market real (after inflation) returns.

During all our wars, riots, depressions, 9-11 and other calamities, our capitalistic system has survived intact.

It is important to understand that after every bear market, and stock market correction, U.S. stocks have always come back higher than before.

Stay-the-course

http://img404.imageshack.us/my.php?image=20060709goldreturnsve9.jpg

Best wishes.
Taylor

[MossySF]

Simple graph of annual dividend yields 1871-2006.

[Latestarter]

Simple graph of annual dividend yields 1871-2006.

Very helpful, thanks.
As long as we're speculating here (no pun intended): Suppose the stock market is about to enter another period of returns that are as dismal as those from the early '60s to the early '80s. Which scenario is more likely?
(a) Dividends will spike back up to where they were during that last period, providing decent adjusted returns and making us glad we were still invested; or
(b) Dividends will remain at 1-2%, where they've been for the last decade.
If (a), what's the mechanism that would lead that to happen?

[thehumangame]

Assuming reinvested dividends, and using the same 1871-2004 data other posters above me are using. Shiller's ie_data.xls included dividend yields but he didn't compute a total return series, so I did the computation myself.

By the way:

(Annualized, continuously compounded)
Mean 6.72%, Standard Deviation 14.31%.

People here like CAGR instead of continuously compounded return, so CAGR = exp(.0672) - 1 = 6.95%.

[bob90245]

Assuming reinvested dividends, and using the same 1871-2004 data other posters above me are using. Shiller's ie_data.xls included dividend yields but he didn't compute a total return series, so I did the computation myself.

Nice chart. Does anyone else notice the stair-step pattern? The $100 and $1,000 levels represented heavy resistance (to use the jargon of technical analysts). Time will tell, but it appears that the same pattern of heavy resistance is currently holding stocks below the $10,000 level. If this plays out like the two previous times, it could try the patience of long-term buy-and-hold investors.

[peter71]

By the way:

(Annualized, continuously compounded)
Mean 6.72%, Standard Deviation 14.31%.

People here like CAGR instead of continuously compounded return, so CAGR = exp(.0672) - 1 = 6.95%.

Hi Humangame:

The difference between the "continuously compounded" and CAGR returns is kind of what I was getting at the other day . . . care to elaborate on which you yourself think is the more accurate predictor of expected fund returns? Any sympathy for those who defend the arithmetic mean?

All best,
Pete

[matt]

Historical total return charts substantially overstate returns for everything but the most recent past. The costs of trading were substantially higher throughout most of the prior century. The simplicity of mutual fund dividend reinvestment was not around in 1871.

[thehumangame]

They're different ways of measuring the same thing. Academics tend to favor continuously compounded returns because they make it easier to apply calculus.

Suppose I can invest in something that returns 0.5% of my invested amount each month. I want to know how much I will have in five years if I invest $1,000.

If I didn't reinvest interest, I'd have $1,300 at the end of five years. $1,000 + 60 months*($1,000*0.5%) = $1,300.

But in the real world we reinvest interest and dividends. If I think in terms of discrete time, in chunks of one year, I can measure my true annual gain using CAGR. (0.5%+1)^12 months - 1 = 6.17% CAGR. This means that at the start of every year I have 6.17% more than at the start of last year. After five years I will have $1,000*(6.17% + 1)^5 years = $1348.85.

If I think in terms of continuous time, like academics, I can use the Pe^rt formula we learn in calculus class to compute my return. This formula expects r to be a continuously compounded return, so I take log($1 / $1+1*0.5%) = log($1*0.5%) = 0.499% as my continuously compounded return given a one-month time unit. Then I apply the Pe^rt formula: $1,000*e^(0.499%*60 months) = $1,348.85.

I can also compute my annualized continously compounded return: 0.499*12 months = 6.00%, and apply the Pe^rt formula: $1,000*e^(6.00%*5 years) = $1348.85.

There's no difference between the two so long as you always use the correct formulas to manipulate them. I like continuously compounded returns because in the random walk model of stock market returns they are normally distributed.

[peter71]

thanks humangame,

'd certainly assume that academics would prefer the most mathematically-complicated solution, but, if interested, check out this actuary's attack on Fama-French in particular . . .

http://www.actuaries.org.uk/files/pdf/l ... erbert.pdf

and Mehra and Prescott's "proof" of the superiority of arithmetic means for estimating equity premia in appendix A of the following paper:

http://www.academicwebpages.com/preview ... ospect.pdf

i have neither the ability nor the willingness to figure out what the latter in particular are talking about (and all i can think to argue against the former paper is that, as it concedes, it's just an observed vs. expected thing) but whether or not you're able to reckon the two i do think you'll find them interesting!

all best,
pete

p.s. assuming you have j-stor you can also do an interesting search for the relevant keywords -- tons of stuff in the 60's and 70's but unclear that the debate got dropped because it got resolved . . .

[nisiprius]

"Inflation-adjusted growth of $1 in the U. S. stock market, 1871-2004." That is a beautiful chart. Resisting the urge to pick out patterns in it... the take-home message I get from it is that the stock market is indeed a low-risk way to earn a steady real rate of 6.5% to 7% (above inflation) but only if you average over periods of about fifty years or so!

Over the periods of one's serious earnings (when one is actively saving significant amounts of money for retirement), and over retirement itself, the stock market still carries a significant degree of risk.

There are three twenty-year periods on that chart during which the stock market was essentially flat. If one of those flat periods happened to occur during one's first twenty years of retirement, it would have a serious impact on finances.

I believe that trying to guess when those periods will occur is about as productive as trying to guess short-term patterns.

[MossySF]

Some more eye candy. This time comparing real returns of 5,10,15,20,25,30 year rolling periods.



So for the past 100 years, 25 year holding period is the minimum to ensure some real return at any given point. 20 year periods have had 3 different periods of 0% real return -- 20s, mid-40's, late 70s-early 80s.