Common Sense on Social Security
A Centrist Strategy for Social Security Reform
Those who would reform Social Security by mandating a system of Personal Retirement Accounts (PRA's) have long pointed to the magic of compound interest as a source of superior returns for future retirees.
According to PRA advocates, a 20th Century career investor who deposited a regular percentage of his paycheck in an index fund based on the Standard & Poor's 500 would have compounded his savings at an inflation adjusted rate of 7 percent a year, on average. We should, therefore, expect much the same returns in the future. Social Security's Office of the Actuary essentially concurs, and forecasts real returns of 6.5% in decades ahead.
Are these glowing forecasts right? Are they reasonable? Or are they dangerously misleading? Let's find out.
For starters, it is sensible to focus on index fund returns. If 150 million American workers were required to invest in PRA's, all at the same time, almost all their funds would have to flow into very broad index funds. And Social Security would have to base all its forecasts on the average returns earned by all 150 million PRA investors at the same time. It's not Lake Wobegon, it isn't true that every investor will be above average. Quite the contrary. As we're analyzing an issue that deals with millions of investors, there's no alternative to index funds as the main vehicle for stock market investing.
Were a PRA program ever to be adopted by the Congress, an interesting debate would break out on which index fund is the best. Perhaps the Standard & Poor's 500 is too narrow. Perhaps a much broader index fund, such as theWilshire 5000, would be more appropriate. As the Wilshire 5000 is a relatively new index, only thirty years old or so, it doesn't have a track record that goes back nearly as far as the S&P 500, so for our purposes in this essay, I will focus on the S&P 500 and its return record.
1. Historical Returns
Most commentators who refer to stock returns of 7% refer to the experience of the Standard & Poor's 500 from 1925 to 1995 or so. And for the time period in question, they're absolutely right. The S&P 500 data series does show an average annual return of 7%, once inflation is taken out.
But what elements feed into that seven percent? What does it consist of? How often does one read commentary that breaks it out into its component parts?
Not often enough. The 7% track record has two components, capital growth and reinvested dividends. Capital growth accounts for only a third of the 7% average. Reinvested dividends account for two-thirds of the long run growth rate. Capital growth in the S&P 500 averaged 2.3% a year from the mid-1920's to the mid-1990's. Dividend yields average 4.6% a year, for a combined gain of 7% a year. (1.023 x 1.046 ~ 1.07)
2. What Drives Capital Growth? Is it GDP Growth?
As I first began to research this issue, I expected to see capital growth occurring at roughly the same rate as GDP growth. After all, if the economy grows at an average rate of 3.3% a year, isn't it reasonable to assume than capital growth in a broad index fund would more or less match the GDP growth rate?
3. Total Market Capitalization versus GDP
For my opening hypothesis, I assumed that I'd find a regular relationship between total stock market capitalization, on the one hand, and total Gross Domestic Product (GDP), on the other. Market capitalization is an asset concept, much like the Balance Sheet side of an annual report. GDP is a revenue concept, much like the top line in an annual Income Statement. And normally there's a fairly predictable relationship between total assets and total revenues.
In the corporate world, annual Revenues generally have a predictable relationship to total assets. In my consulting work, I was for a time deeply immersed in telecom issues. At one point I looked at revenue to asset ratios for AT&T and for Bell Atlantic (now Verizon). AT&T produced one dollar of revenue for each dollar of assets on its Balance Sheet. Verizon produced one dollar of revenue for every two dollars of assets on its balance sheet. In other words, AT&T's long distance assets were cheaper than Bell Atlantic's local loop assets. Different companies, different ratios, but in both cases, the ratios were predictable and fairly constant.
Wouldn't the same sort of relationship hold for the economy as a whole? Wouldn't there be a predictable relationship between the American economy's total assets, as partly reflected in stock market capitalization, and that same economy's total revenue stream, as reflected in Gross Domestic Product? The answer is Yes. Over a several decade time period, it's clear that stock market capitalization roughly tracks Gross Domestic Product.
In the chart below, with a semilog scale on the left reflecting current dollar numbers both for stock market capitalization and GDP, note the hundredfold growth (in current dollars) from the 1920's to the late 1990's. The American economy has grown from a GDP of $100 billion a year to a GDP of $10 trillion a year. The GDP growth path was a bit rocky from 1929 through 1945, through the Great Depression and World War II, but followed a steady track thereafter.
Total stock market capitalization fluctuated a bit more, but it's still quite clear, over a quite a lengthy period of time, that the size of the asset base bears a regular relationship to the size of the annual revenue stream.
4. But there's a lag. Capital growth lags GDP growth.
Yet something didn't quite make sense. The long-run GDP growth rate was roughly 3.3% a year. The long-run capital growth rate in the S&P 500 index was only 2.3% a year.
Wharton's Jeremy Siegel, a long-time student of returns to capital, explains the lag this way. As new firms are added to the stock market, the total capitalization of the market grows. On the other hand, adding new firms to the market does nothing for the value of the firms represented in the S&P 500 index.
Other factors may be at work as well. Corporations that are in the S&P index may issue stock bonuses to officers, which over time will dilute - if slightly - the growth rates of their pre-existing shares. Secondary offerings - the sale of new stocks by companies already listed on the stock exchange - can add capitalization to the market as a whole without adding anything to the value of stocks already represented in the index. Firms not in the S&P index may be in faster growing sectors of the economy. And so on. In other words, some sort of lag is a natural occurrence.
So - using the 1925 to 1995 period as the yardstick - it appeared that capital growth in the S&P 500 lagged growth in GDP by roughly one percentage point.
5. How valid is the Lag Factor estimate of one percent?
As I analyzed the relationship between GDP growth, on the one hand, and stock market capitalization growth, on the other, I thought I'd test the ratio between the two. What's the "normal" ratio between stock market capitalization and GDP? I developed a graph showing the Market Capitalization to GDP ratio.
I was surprised, initially, by how small the stock market seemed to be, relative to GDP. Total capitalization dropped as low as thirty-five percent of GDP, rose as high as 105% of GDP, and fluctuated substantially within that very broad band. Over the seventy years from the mid-twenties to the mid-ninetites, the Market Capitalization to GDP Ratio averaged approximately 65%.
In the middle of this analysis, I soon came to realize that my Lag Factor estimate of 1% was based on a time period that has a significant upward bias built in. Were I to follow Siegel's lead, and measure capital growth in the S&P 500 from 1925 to 1995, I'd find myself using a Start Year a year with a relatively low Market Capitalization to GDP Ratio, and an End Year a year with a somewhat higher ratio. Maybe the Lag Factor for the s&P 500 is greater than 1 percent.
6. What if the Lag Factor were to be measured differently?
The follow-on question is a natural one. What if I were to construct a "Peak to Peak" measurement of capital growth versus GDP growth? Or a "Trough to Trough" measurement?
I chose two time periods with very similar Start Years and End Years. And conducted just such a test.
7. The Peak to Peak measure showed a larger difference
I used three data series to run this test. Total Market Capitalization, S&P 500 capital value, and capital growth in the Dow Jones Industrial Average. I normalized all the data series at 100 for the year 1965, and then plotted the results for each series from the same starting point.
Total market capitalization grew at a compounded rate of 3.0 percent a year. (Real GDP grew 2.9% a year over the same time period.) Capital growth in the S&P 500 averaged only 1.5 percent a year. Capital growth in the Dow averaged only 0.7% a year. S&P 500 growth lagged GDP growth by 1.4 percentage points. [NOTE: CAGR = "Compound Annual Growth Rate"]
8. The Trough to Trough measure also showed a larger difference
When I applied the same approach to the 1954 - 1990 time period, I got very similar results. Capital growth in the S&P 500 lagged the Compound Annual Growth Rate in the GDP by 1.5 percentage points. The Dow Jones Industrial series lagged GDP growth by 2.5 percentage points.
9. GDP Growth - What drives GDP growth?
In filling in the historical chart (see below), I use 1% as the Lag Factor, because it's the right number for the 1925 - 1995 time period.
As my next step, I asked: What drives GDP growth?
At a very high level, the answer is quite simple. More people, times more output per person, yields more GDP.
(Or, to give it greater precision, growth in the number of hours worked, times productivity per hour, yields growth in GDP. As population grows, the workforce grows, though not necessarily at the same exact rate. As the workforce grows, the number of hours worked every week also grows, though, again, not necessarily at the same exact rate.)
Be that as it may. It's more convenient to refer to Population Growth as one of the two primary drivers of GDP growth. For much of the Twentieth Century, annual population growth averaged 1.3% a year.
And annual productivity growth averaged about 2.0% a year. Those two growth rates combined to produce a GDP growth rate of 3.3% a year.
Now we see the main cause-and-effect linkages on the Capital Growth side. Population growth plus productivity growth produce GDP growth. GDP growth, minus a Lag Factor, generates index fund capital growth.
10. Dividend Yields: What drives dividend yields?
Turn now to the right side of the Equity Return diagram. Dividend yields averaged 4.6% a year in the 1925 - 1995 time period. What do those yields consist of?
With an individual stock, we know the answer instantly. Yield equals total dividends paid on the stock over the course of the year, divided by the total value of the stock.
In an index fund, the answer is similar. And, since we're looking at this from a tax-free retirement perspective, we also assume that all dividends are reinvested in purchasing more stock, with no tax charges (and - in this analysis - no management fees!), in order to generate maximum compound growth over time.
Again, at the level of the total economy, total dividend yields come from much the same forces. Total dividend payments, as a percent of GDP, represent the cash payment numerator. Total stock market capitalization, also as a percent of GDP, represents the stock value denominator. Cash payment divided by value equals yield. Reinvested yields produce compound growth.
11. What do we know about Dividend Payments over time?
Let's start with dividend payouts. Dividend payments over time are fairly steady. Any corporation that pays dividends likes to show its shareholders a steady upward rise in quarterly dividend payments, even if the upward increases are modest. Since this is such a basic feature of corporate psychology, this philosophy into a fairly steady pattern of dividend payments across the economy as a whole.
The New York Stock Exchange provides a data series showing annual dividend payments for all firms listed on the Exchange. I translated these payments into GDP terms, and charted them over a two decade period.
The results, as you can see in the chart below, are reasonably steady. You see much less fluctuation in dividends as a percent of GDP than you'll see in stock market capitalization as a percent of GDP. I take it as a fair rule of thumb that regular cash dividends for the market as a whole are roughly equal to 2% of GDP. (This estimate might be a slight overstatement, but it's not a bad rule of thumb.)
Dividend yields in the S&P 500 family of corporations are, I suspect, a bit higher than they are for the stock market as a whole. The sort of firm that makes it onto the index is larger, older, more established, and, therefore, somewhat more likely to pay dividends. 4.6% dividend yields in the S&P 500 probably overstate the dividend yield pattern for the market as a whole.
12. What do we know about Stock Market Capitalization over time?
Now we come back to the Stock Market Capitalization, which is best thought of in terms of the Market Capitalization to GDP Ratio.
Dividend payments as a percent of GDP may be relatively stable. Stock market capitalization is anything but stable. When I first did this analysis, the stock market had just begun to inch above its historic highs. In 1998, I had no idea how much higher it would go, but I suspected a bubble and doubted that the capital growth runup would be able to continue indefinitely. And, of course, it didn't. When the bubble burst, there was a huge collapse.
On the other hand, a long-run shift appears to have taken place as well. For many decades, the stock market wasn't the sort of place that average Americans normally thought of when they wanted to invest a buck or two. Investing in the stock market was the province of an elite, not of a mass market.
This seems to have changed. The stock market is more popular than it was half a century ago, and capital flows into the market more readily than it used to. In the 1942 - 1953 trough phase, the Market Cap to GDP Ratio found itself in the 35% to 50% range. In the 1974 - 1984 trough phase, the Market Cap to GDP Ratio ranged from 40% to 55%. I think it most unlikely that we'll see the Market Cap to GDP Ratio ever dip into those ranges again, let alone spend an entire decade at a time at those levels.
In fact, it wouldn't be too surprising, given the experience of the last decade (1995 - 2005) to see the Market Capitalization to GDP Ratio of the future fluctuate between 100% and 150%, with few if any excursions into the "below 100%" range.
13. How will these trends affect Dividend Yields?
We now have two pieces of evidence well in hand. Dividend Payments - the numerator in the dividend yield equation - tend to be fairly steady. Stock Market Capitalization - the denominator - fluctuates much more. It also looks like its overall level will be somewhat higher than it has been in the past.
Is there a relationship between Dividend Yields and the Market Capitalization to GDP Ratio? How might these factors affect dividend yields in an S&P 500 index fund in years to come?
One way to answer this question is with logic. The Dividend Payment numerator is likely to be fairly constant. The Market Capitalization denominator might well be twice as large in the future as it was in the past. Dividend yields, therefore, are likely to be cut in half.
The other way to answer the question is with data. The following chart maps the Market Capitalization to GDP Ratio in red. And it maps Dividend Yields (in the S&P 500) in blue.
From 1940 onward, the pattern is absolutely clear. When Market Cap to GDP Ratios are low, dividend yields in the S&P 500 tend to be high. When Market Cap to GDP Ratios are high, dividend yields tend to be low.
Either way we slice it, with logic or with data, the evidence points in the same direction. Unless the stock market of the future loses its appeal, its Market Cap to GDP Ratios are likely to be much higher than they were in the past, while its Dividend Yield percentages will correspondingly be somewhat lower than in the past.
There is one caveat, not shown on this chart. Stock buybacks have become a popular alternative to the cash dividend. If a firm offers to buy stock back, shareholders have two choices. Sell it back, and pay taxes at a lower capital gains tax rate. Or hang onto the stock, and watch its value rise as the total number of shares outstanding declines just a bit.
Stock buybacks thus constitute a hidden form of dividend. Experts who've studied this think that stock buybacks add about one percentage point to the market's overall dividend yield.
14. Recap: Historic Returns to the S&P 500
One reads this chart by starting at the bottom. Dividends divided by stock values (represented by Market Capitalization) equal dividend yields. The S&P 500 had quite an admirable dividend yield track record of 4.6%.
Populations growth (i.e. workforce growth) plus productivity growth equals GDP growth. GDP growth, minus the Lag Factor, equals index fund capital growth.
And capital growth, combined with reinvested dividends, equals long run index fund returns.
It's not enough to know the top level number - seven percent. It's essential to understand the structure that underlies that number. Because, after all, it's the underlying structure that will determine whether the seven percent returns of yesteryear will be able to repeat in the decades to come.
15. Forecasting Future Returns
Given what we know about the past, and the present, what sort of guess can we make about the future?
I will start with four caveats.
1) Social Security forecasts productivity gains of only 1.6% a year, down from 2% during the twentieth century. They base this on America's ongoing shift from manufacturing to services and assume future productivity gains in the service sector won't be as rapid. To me, the "slow productivity in the service sector story" overlooks the importance of a new sector, the digital sector, and its likely impact on productivity gains. The new digital sector is essentially a zero marginal cost sector, an attribute which translates into very high productivity for products with mass market appeal. The digital sector is one that's also capable of automating many portions of the service sector, thus raising productivity in the service sector more rapidly than we once thought possible.
So my estimate marks up Social Security's. Social Security estimates future productivity gains of 1.6%, I estimate 2.1%.
2. The Lag Factor. The S&P 500 index is too small to support Personal Retirement Accounts for 150 million American workers. A comprehensive PRA system will require a market-wide index, along the lines of the Wilshire 5000. For a market wide index, it's my guess that the Lag Factor won't be as high. Instead of a Lag Factor of 1.5%, I'm guessing the Lag Factor will come in at 0.7%. (This estimate needs further research.)
3. Dividends as a percent of GDP. To take account of stock buybacks, I have boosted the estimate for Dividends as a percent of GDP from 2% to 3%.
4. Market Cap to GDP Ratio. The historic average was roughly 65%. I peg the future average at about 120%. Who knows if this will turn out to be right, but it's a reasonable estimate.
Now add up all the pieces.
Workforce growth slows to 0.2%, according to Social Security's actuaries. Add productivity growth of 2.1%. This suggests long-run GDP growth of 2.3%. Subtract a Lag Factor of 0.7%. It's not unreasonable to assume long-run Capital Growth of 1.6%.
Meanwhile, dividends at 3% of GDP, divided by market capitalization equal to 120% of GDP, translate into an average dividend yield of 2.5%.
Capital growth at 1.6%, compounded by Dividend Yields of 2.5%, suggests long run index fund equity returns of 4.1%.
This is a far cry from the 7% returns of yesteryear. Slower population growth slows GDP. Higher Market Cap to GDP Ratios reduce Dividend Yields. Even with some favorable caveats, it's hard to make a case for long run returns much above four percent.
The argument on behalf of personal accounts is a reasonably sunny one if we believe the stock market will produce real ROI's of 6.5% in the future. It's much less sunny if we take the prudent investor view I've just outlined and assume real ROI's of around 4 percent.
17. A Note on Sources.
GDP figures come from the Commerce Department. (National Income and Product Account data, known as NIPA data). GDP estimates for the 1920's were constructed from a discontinuous GDP series for that decades. They're reasonable estimates, but they may be slightly off.
Stock Market Capitalization estimates from the 1950's onward come from the Federal Reserve's Flow of Funds report, the Z series reports. Estimates prior to 1950 were constructed using NYSE capitalization figures. The gap between the NYSE capitalization numbers, and the Fed's numbers, during the 1950s, was applied to the NYSE figures from the earlier time period. It is undoubtedly off, by an unknown margin, but what's important are the directional trends, and the directional trends shown in my estimates are almost surely the same as we'd see if the dataset were more robust.
S&P 500 capital growth and dividend yield data series come from Ibbotson.
NYSE dividend payment data comes from the New York Stock Exchange.
Copyright 2005, by Simcivic.org. You are welcome to quote from this essay, or use its charts, provided you let us know ahead of time, and give us proper credit.
A Non-Profit Successor to the Collaborative Democracy Project
Page Version 1.01
Revision Date April 13, 2006