Wednesday, May 02, 2007

Cornflakes Without the Milk

“You without me is like cornflakes without the milk”

-Oran “Juice” Jones

Macro Man was recently involved in a discussion about the relative merits of equities and gold. One of the participants observed that gold and other commodities had handily outperformed US and broad global equities in the current millennium, a fact which is clearly indisputable. However, he also quoted a passage from a financial blog [a famously unreliable source of analysis ;) ] discussing the historical relative performance of equities and gold. The author of the original piece noted that at the Dow peak in 1929, the Dow divided by the gold price was 19, the same ratio as today. Therefore, he concluded, equities have failed to outperform gold over a timespan encompassing nearly eighty years.

Can you spot the mistake in the analysis? That’s right, he conveniently forgot to include the income from dividends. Simply put, measuring stock market performance without dividends is like eating cornflakes without milk- it’s just not complete.

Most readers will probably be aware that dividend income represents a substantial portion of stock market return over time. However, it is easy to forget how powerful the impact is over long periods. Unfortunately, Macro Man does not subscribe to Ibbotson Associates or similar providers of historical data, so he can only go as far back as the end of 1969 in his analysis. However, even going back just 37 and a bit years, the impact of the compounding of dividend reinvestment is very striking indeed.

First off, let’s consider the na├»ve analysis referenced above: measuring equity indices not in dollars (or any other currency), but in ounces of gold. Now, it is certainly debatable whether gold is an appropriate measure of “absolute value”, but enough people seem to think it is that Macro Man is prepared to use it for the purposes of this example. The chart below shows the price of MSCI World and MSCI USA priced in ounces of gold. On December 31, 1969, both indices could “purchase” 2.84 ounces of gold. Today, on the other hand, MSCI World can only purchase 2.32 ounces of gold, while MSCI USA only buys 2.05 ounces of gold. So according to the author of the blog referenced above, gold has “won” and was a better investment than equities.
Not so fast, my friend. When we add dividends back in, equities win in a landslide. Even with the recent dramatic outperformance of gold, $100 invested in either MSCI World or MSCI USA in 1969, with dividends reinvested, would still be worth loads more than an equivalent amount invested in gold. Whereas one unit of either index could buy 2.84 ounces of gold in the Age of Aquarius, MSCI World with dividends reinvested now buys 7.03 ounces of gold, and MSCI USA buys 6.97 ounces. In other words, dividends reinvested take the equities from underperforming gold to easily more than doubling in value relative to the yellow metal. Small wonder, then, that no less an authority than Albert Einstein called compounding “the greatest mathematical discovery of all time.”

Since 1969, the dividend yield of MSCI World has been about 3% and the annual price increase has been about 7.4%. Now 10.4% per year sounds better than 7.4% per year, but over the course of an investing lifetime one might think that it would allow one to retire a year or two earlier at most. Right? Wrong. $100 invested in MSCI World at the foundation of the index by someone who took the dividends and bought bell bottoms with the proceeds every time they were paid would currently be worth $157,785. Not too shabby, to be sure. However, had our fashion victim instead plowed the dividends back into the index, that same $100 would now be worth $478,731. The latter figure would fund a substantially nicer retirement pad...
A few comments/queries/suggestions about this study:

* This analysis does not include the tax costs of dividends. Total returns of equities would surely be lower if dividend taxes are paid out of dividend income. Macro Man has not seen any total return analysis that shows the post-tax compound returns of dividend reinvestment, and would be curious to know if any readers have seen such a study.

* Similarly, Macro Man would be curious to know how the numbers break down for longer data sets, such as those of Ibbotson. He’d imagine that the effect would, if anything, be more pronounced the earlier one makes the starting point, given the relatively high dividend yields that prevailed before the 1950’s.

* There is a school of thought in behavioural finance that dividends should be irrelevant to total returns, given that stock prices decline by the amount of any dividend payment. In other words, future returns of a $100 stock should be identical as those of a $100 stock that pays a $1 dividend and drops the price to $99. This school of thought pooh-poohs the notion that one might wish to purchase dividend-paying stocks as a source of income, suggesting instead that one could replicate the cash flow by selling tiny amounts on non-dividend paying stock (ignoring transaction costs, of course.) Intuitively, this makes sense, and does not appear to jive with real-world observation. Perhaps it is because as companies mature, it becomes more and more difficult for retained earnings to generate the same marginal profitability via fresh investment. Dilutive investment would therefore reduce the overall return on capital, therefore dampening future stock returns. Presumably the future return of a company that distributed as dividends any earnings that could not be invested to at least match existing marginal profitability would be superior. Microsoft and Intel come to mind as companies that held on to too much cash for too long, though they have belatedly seen the error of their ways.

* In that vein, Macro Man would be curious to know if any readers have ever seen studies performed on relative total returns generated by companies of similar size and profitability, but with one persistently offering a substantially higher dividend yield than the other. In other words, do companies with a market cap of $10 to $20 billion , earnings growth of 10%-15%, and a dividend yield greater than 3% generate superior, inferior, or identical total performance to companies with a similar size and profit growth but who pay a dividend yield of less than 1.5%. Macro Man reckons that an exhaustive study across countries, market cap, and time would make for a really neat thesis project, if there are any would-be academics out there.


wcw said...

Not to be too flip, but I think you have just rediscovered the value effect.

Yes, over the (sometimes very) long term and in the aggregate, dividend payers outperform. The question is, why? I incline towards the mechanical explanation: as you segregate by dividend (or book multiple, the other favored 'value' measure) you include many more beaten-down names than any lower-yield/higher-multiple set. Random chance will deliver some as unfairly beaten down, while the residual set receives a dollop which fails to be beaten down fairly. This artifact of chance alone will boost your chances at long-term outperformance.

I have some further sympathy for the value-as-risk assertion. Lower-yielding bonds, ceteris paribus, are lower risk.

As for your data questions, I would ignore that price-weighted artifact the DJIA and focus on the S&P 500. If you had bought at the 1929 peak, held, and reinvested all dividends, you would now have multiplied your initial investment by.. thousand.

I do not think gold's pissant little thirty-times rise can really compete.

Now, I grant, buying at the 1929 peak hurt. In May 1932, you were down 85%. Indeed, it took you until 1979 to hit 30x your 1929 investment, at which point gold really had been the better investment. Still, if your interlocutor is so much of a gold bug that he ignores dividends and storage costs, he deserves your casual note that the S&P beat gold not by a few percent but by a factor of 31.6 from the 1929 peak to the end of April.

Macro Man said...

OK, I see what you are saying re div yield as value, as price is a vital component of yield. But what if we further narrow the focus to div payers versus non-div payers? Does the outperformance continue to hold? And is it persistent over time?

I concur on the DJIA. For the life of me I cannot figure out why people who are ostensibly equity professionals even bother make reference to such a stupidly constructed index.

"Cassandra" said...


First, unless you've a vault in your cellar, free from government knowledge, returns from gold are overstated since it COSTS money to store physical bullion, for security and safekeeping - not just from privateering thieves, but also from many governments who from time to time sour upon private ownership of the metal. These costs are rarely if ever debited from time Gold Price series. And of course there is occasionally transport and insurance costs if one ever desired to move it from say UBS to say Credit Suisse, or even Scotia Bank (owner of what was once Mocatta's vault under the WTC).

Second, almost all micro-level equity research should be done in "return space", rather than in price space. There are some occasions where it doesn't really matter (e.g. Japanese stocks 88-98 where yields are paltry to non-existent, and if one in fact finds themselves sensitive thereto, one should be exc avating for alpha elsewhere, for they were objectively "rounding errors" to most numerical gymnastics.

Rob Arnott and Cliff Asness wrote a paper that demonstrated higher dividends (i.e. stocks with highest dividend yields proxying for those with lower growth) translated into higher total returns (as compared to stocks with the lowest dividend yields (implication being those with the highest growth), ostensibly as a riposte to lingering growth-humping managers. But perhaps that is because higher dividend stocks are less volatile as suggested by in this paper Blitz & van Vliet.

I think both Shiller and Fama/French have Ibbotson-like data on their website, though I haven't visited there recently.

Macro Man said...

Yeah, I'm aware of the storage cost issue for gold, though of course no time series exists for such costs over time. So yes, the return series on gold is actually lower than the price series. I was surprised, however, at how modest storage costs are currently: UBS quote me 0.20% in a communal vault, whereas I had supposed it was 1% or so. Compounded, of course, it makes a bigger differnce.

Thanks for the links to the papers. I'll check them out. What shocks me is how much mainstream equity analysis focuses on

a) indices like the Dow that are, frankly, crap

b) price, rather than total return

c) even when focus is on total return, how little appears to be on risk adjusted return

I know Arnott's a smart cookie...I'll check his paper out.

Anonymous said...

Have to be a bit careful about interpreting dividend yield since it is also a proxy for the value effect (as mentioned in a earlier comment), thus the historical outperformance of high dividend stocks is tainted by the fact that high book-to-market stocks have outperformed. I believe that once you control for the value effect (i.e. sort performance by dividend yield controlling for book-to-market levels) high dividend stocks have the same or worse performance.

"Cassandra" said...

Dr Asness is no slouch either (and has a better sense of humour). It was a joint undertaking, though I think the AQR guys (Asness' bunch) did the legwork (as well as the jokes and the sarcasm)

flipper said...

s'n'p500 is much better measure than the dja ofcourse.

i think if you take some very broad an equal weighted market index like wilsire5000 you'll get even beteer results.

the only problem with all that calculation is that an average investor cann't follow index...

index tracking funds are only recent fenomena, and are available not for all indices.

they still charge managment fees, etc.

and there are taxes on dividends, etc...

still if think equites will outperform gold even if we take all that into account

Anonymous said...

Shouldn't we measere the S&P against an index of gold stocks? The HUI for example?

Anonymous said...

a) High dividend stocks are better performers because dividends are a discipline/constraint on capital management. Managers face a higher bar for capital projects, and make better decisions as a result. Dividends involve the shareholder in the overall capital budgeting process.

b) From a capital accumulation perspective, dividend reinvestment is a form of effective price averaging. It’s dull but it works. The long run investor in a non-dividend paying stock gets paid absolutely nothing for market volatility. The dividend reinvestor gets paid for it. Incremental stock added to the position is cheaper, and the investor buys more of it, when downside volatility is high, and vice versa. The long-term investor is getting paid for a risk assumed by short-term investors. There’s more liquidity and more return in such a process over the long run, than when managers wreck their firms by overspending on capital projects just before a down cycle. The resulting cumulative return, while not approaching the speed of light, is impressive over the long term.

The market at large essentially confirmed the thesis with the macro move to dividend payers and higher payouts following the tech crash.

Anonymous said...

Are those indices nominal or adjusted for inflation?

Macro Man said...

Nominal....just like the price of gold.