"The issue is no longer whether the currnet market resembles those preceding the 1929, 1969-70, 1973-74, and 1987 crashes. The issue is only - are conditions like October of 1929, or more like April? Like October of 1987, or more like July? If the latter, then over the short-term, arrogant imprudence will continue to be mistaken for enlightened genius, while studied restraint will be mistaken for stubborn foolishness. We can't rule out further short-term gains, but those gains will turn bitter... Let's not be shy: regardless of short-term action, we ultimately expect the S&P 500 to fall by more than half, and the Nasdaq by two-thirds. Don't scoff without reviewing history first."
- John P. Hussman, Ph.D., Hussman Econometrics, February 9, 2000
"On Wall Street, urgent stupidity has one terminal symptom, and it is the belief that money is free. Investors have turned the market into a carnival, where everybody 'knows' that the new rides are the good rides, and the old rides just don't work. Where the carnival barkers seem to hand out free money just for showing up. Unfortunately, this business is not that kind - it has always been true that in every pyramid, in every easy-money sure-thing, the first ones to get out are the only ones to get out... Over time, price/revenue ratios come back in line. Currently, that would require an 83% plunge in tech stocks (recall the 1969-70 tech massacre). The plunge may be muted to about 65% given several years of revenue growth. If you understand values and market history, you know we're not joking."
- John P. Hussman, Ph.D., Hussman Econometrics, March 7, 2000
On Wednesday, the consensus of the most reliable equity market valuation measures we identify (those most tightly correlated with actual subsequent S&P 500 total returns in market cycles across history) advanced within 5% of the extreme registered in March 2000. Recall that following that peak, the S&P 500 did indeed lose half of its value, the Nasdaq Composite lost 80% of its value, and the tech-heavy Nasdaq 100 Index lost an oddly precise 83% of its value. With historically reliable valuation measures beyond those of 1929 and lesser peaks, capitalization-weighted measures are essentially tied with the most offensive levels in history. Meanwhile, the valuation of the median component of the S&P 500 is already far beyond the median valuations observed at the peaks of 2000, 2007 and prior market cycles, while our estimate for 10-12 year returns on a conventional 60/30/10 mix of stocks, bonds, and T-bills fell to a record low last week, making this the most broadly overvalued instant in market history.
There is a quick, knee-jerk response floating around these days, which asserts that “stocks are still cheap relative to interest rates.” This argument is quite popular with investors who haven’t spent much time getting their hands dirty with historical data, satisfied to repeat verbal arguments they’ve heard elsewhere as a substitute for analysis. It’s even an argument we recently heard, almost inexplicably, from one investor we’ve regularly agreed with at market extremes over several decades (more on that below). In 2007, as the market was peaking just before the global financial crisis, precisely the same misguided assertions prompted me to write Long-Term Evidence on the Fed Model and Forward Operating P/E Ratios. See also How Much Do Interest Rates Affect the Fair Value of Stocks? from May of that year. Let’s address this argument once again, in additional detail.
Valuations and interest rates
There’s no question that interest rates are relevant to the fair valuation of stocks. After all, a security is nothing but a claim to some future stream of cash flows that will be delivered into the hands of investors over time. The higher the price an investor pays for a given stream of future cash flows, the lower the long-term return the investor can expect to earn as those cash flows are received. Conversely, the lower the long-term return an investor can tolerate, the higher the price they will agree to pay for that stream of future cash flows. If interest rates are low, it’s not unreasonable to expect that investors would accept a lower expected future return on stocks. If rates are high, it’s not unreasonable to expect that investors would demand a higher expected future return on stocks.
The problem is that investors often misinterpret the form of this relationship, and become confused about when interest rate information is needed and when it is not. Specifically, given a set of expected future cash flows and the current price of the security, one does not need any information about interest rates at all to estimate the long-term return on that security. The price of the security and the cash flows are sufficient statistics to calculate that expected return. For example, if a security that promises to deliver a $100 cash flow in 10 years is priced at $82 today, we immediately know that the expected 10-year return is (100/82)^(1/10)-1 = 2%. Having estimated that 2% return, we can now compare it with competing returns on bonds, to judge whether we think it’s adequate, but no knowledge of interest rates is required to “adjust” the arithmetic.
There are three objects of interest here: the current price, the future stream of expected cash flows, and the long-term rate of return that converts one to the other. Given any two of these, one can estimate the third. For example, given a set of expected future cash flows and some “justified” return of the investor’s choosing, one can use those two pieces of information to calculate the price that will deliver that desired expected return. If I want a $100 future payment to give me a 5% future return over 10 years, I should be willing to pay no more than $100/(1.05)^10 = $61.39.
So when you want to convert a set of expected cash flows into an acceptable price today, interest rates may very well affect the “justified” rate of return you choose. But if you already know the current price, and the expected cash flows, you don’t need any information about prevailing interest rates in order to estimate the expected rate of return. One does not have to “factor in” the level of interest rates when observable valuations are used to estimate prospective long-term market returns, because interest rates are irrelevant to that calculation. The only thing that interest rates do at that point is to allow a comparison of the expected return that’s already baked in the cake with alternative returns available in the bond market.
The Fed Model is an artifact of just 16 years of history
There’s an additional problem. While it’s compelling to believe that the expected return on stocks and bonds should have a one-to-one relationship, history doesn’t bear that out at all. Indeed, over the past century, the correlation between bond and stock yields has historically gone in the entirely wrong direction except during the inflation-disinflation cycle from about 1970 to 1998. What investors may not realize is that the correlation between interest rates and earnings yields (as well as dividend yields) has been negative since 1998. Investors across history have not been consistent at all in treating stocks and bonds as closely competing substitutes.
As I noted during the bubbles that ended in 2000 and 2007, the problem with the Fed Model (which compares the S&P 500 forward operating earnings yield with the 10-year Treasury yield) is that this presumed one-to-one relationship between stock and bond yields is wholly an artifact of the disinflationary period from 1982 to 1998. The stock market advance from 1982 to 1998 represented one of the steepest movements from deep secular undervaluation to extreme secular overvaluation in stock market history. Concurrently, bond yields declined as inflation retreated from high levels of the 1970’s. What the Fed Model does is to overlay those two outcomes and treat them as if stocks were “fairly valued” the entire time.
The chart below shows the S&P 500 forward operating earnings yield alongside the 10-year Treasury bond yield. The inset of the chart is the chart that appeared in Alan Greenspan’s 1997 Humphrey Hawkins testimony, and is the entire basis upon which the Fed Model rests. The same segment of history is highlighted in the yellow block. Notice that this is the only segment of history in which the presumed one-to-one relationship actually held.
The Fed Model is not a fair-value relationship, but an artifact of a specific disinflationary segment of market history. It is descriptive of yield behavior during that limited period, but it has a very poor predictive record with regard to actual subsequent market returns.
When investors assert that stocks are “fairly valued relative to interest rates,” they are essentially invoking the Fed Model. What they seem to have in mind is that regardless of absolute valuation levels, stocks can be expected to achieve acceptably high returns as long as the S&P 500 forward operating earnings yield is higher than the 10-year Treasury yield.
No, no. That’s not how any of this works, and we have a century of evidence to show it. The deep undervaluation of stocks in 1982 was followed by glorious subsequent returns. The steep overvaluation of stocks in 1998 was followed by one crash, then another, which left S&P 500 total returns negative for more than a decade. I fully expect that current valuations, which are within a breath of 2000 extremes on the most historically reliable measures, will again result in zero or negative returns over the coming 10-12 years. Let’s dig into some data to detail the basis for those expectations.
First, a quick note on historically reliable valuation measures. The value of any security is based on the long-term stream of cash flows that it can be expected to deliver over decades and decades. While corporate earnings are certainly required to generate future cash flows, current earnings (or even forward earnings) are very poor “sufficient statistics” for that stream of cash flows. That’s true not only because of fluctuations in profit margins over the economic cycle, but also due to very long-term competitive forces that exert themselves over multiple economic cycles. From the standpoint of historical reliability, valuation measures that dampen or mute the impact of fluctuating profit margins dramatically outperform measures based on current earnings. Indeed, even the Shiller CAPE, which uses a 10-year average of inflation-adjusted earnings, provides substantially better results when one also adjusts for the embedded profit margin (the denominator of the CAPE / S&P 500 revenues). For a brief primer on the importance of implied profit margins in evaluating market valuations, see Two Point Three Sigmas Above the Norm and Margins, Multiples, and the Iron Law of Valuation.
The chart below shows the ratio of nonfinancial market capitalization to corporate gross value-added, including estimated foreign revenues. I created this measure, MarketCap/GVA, as an apples-to-apples alternative to market capitalization/GDP that matches the object in the numerator with the object in the denominator, and also takes foreign revenues into account. We find this measure to be better correlated with actual subsequent S&P 500 total returns than any other measure we’ve studied in market cycles across history, including price/earnings, price/forward earnings, price/book, price/dividends, enterprise value/EBITDA, the Fed Model, Tobin’s Q, market cap/GDP, the NIPA profits cyclically-adjusted P/E (CAPE), and the Shiller CAPE.
MarketCap/GVA is shown below on an inverted log scale (blue line, left scale), along with the actual subsequent 12-year total return of the S&P 500 (red line, right scale). From current valuations, which now rival the most extreme levels in U.S. history, we estimate likely S&P 500 nominal total returns averaging less than 1% annually over the coming 12-year horizon. As a side note, we tend to prefer a 12-year horizon because that is the point where the autocorrelation profile of valuations drops to zero, and is therefore the horizon over which mean reversion is most reliable (see Valuations Not Only Mean-Revert, They Mean-Invert).
I’m often asked why we don’t “adjust” MarketCap/GVA for the level of interest rates. The answer, as detailed at the beginning of this comment, is that given both the price of a security, and the expected stream of future expected cash flows (or a sufficient statistic for those cash flows), one does not need any information at all about interest rates in order to estimate the expected long-term return on that security. Each point in the chart below shows the actual 12-year subsequent total return of the S&P 500 index, along with two fitted values, one using MarketCap/GVA alone, and the other including the 10-year Treasury bond yield as an additional explanatory variable. That additional variable adds absolutely no incremental explanatory power. Both fitted values have a 93% correlation with actual subsequent 12-year S&P 500 total returns.
We’re now in a position to say something very precise about current valuations and interest rates. Given the present level of interest rates, investors who are willing to accept likely prospective nominal total returns on the S&P 500 of less than 1% over the coming 12-year period are entirely welcome to judge stocks as “fairly valued relative to interest rates.” But understand that this is precisely what that phrase implies here.
Moreover, as one can see from the foregoing charts, there’s not a single market cycle in history, neither in the period before the 1970’s (when interest rates regularly hovered near current levels), nor in recent decades, that has failed to raise prospective 10-12 year S&P 500 total returns to the 8-10% range or beyond over the completion of that cycle. So even if investors are willing to accept 10-12 year total returns of next to nothing, they should also be fully prepared for an interim market loss on the order of 50-60%, because that is the decline that would now be required to restore those 8-10% return expectations, without even breaking below historical valuation norms.
- John P. Hussman, Ph.D., Hussman Econometrics, February 9, 2000
"On Wall Street, urgent stupidity has one terminal symptom, and it is the belief that money is free. Investors have turned the market into a carnival, where everybody 'knows' that the new rides are the good rides, and the old rides just don't work. Where the carnival barkers seem to hand out free money just for showing up. Unfortunately, this business is not that kind - it has always been true that in every pyramid, in every easy-money sure-thing, the first ones to get out are the only ones to get out... Over time, price/revenue ratios come back in line. Currently, that would require an 83% plunge in tech stocks (recall the 1969-70 tech massacre). The plunge may be muted to about 65% given several years of revenue growth. If you understand values and market history, you know we're not joking."
- John P. Hussman, Ph.D., Hussman Econometrics, March 7, 2000
On Wednesday, the consensus of the most reliable equity market valuation measures we identify (those most tightly correlated with actual subsequent S&P 500 total returns in market cycles across history) advanced within 5% of the extreme registered in March 2000. Recall that following that peak, the S&P 500 did indeed lose half of its value, the Nasdaq Composite lost 80% of its value, and the tech-heavy Nasdaq 100 Index lost an oddly precise 83% of its value. With historically reliable valuation measures beyond those of 1929 and lesser peaks, capitalization-weighted measures are essentially tied with the most offensive levels in history. Meanwhile, the valuation of the median component of the S&P 500 is already far beyond the median valuations observed at the peaks of 2000, 2007 and prior market cycles, while our estimate for 10-12 year returns on a conventional 60/30/10 mix of stocks, bonds, and T-bills fell to a record low last week, making this the most broadly overvalued instant in market history.
There is a quick, knee-jerk response floating around these days, which asserts that “stocks are still cheap relative to interest rates.” This argument is quite popular with investors who haven’t spent much time getting their hands dirty with historical data, satisfied to repeat verbal arguments they’ve heard elsewhere as a substitute for analysis. It’s even an argument we recently heard, almost inexplicably, from one investor we’ve regularly agreed with at market extremes over several decades (more on that below). In 2007, as the market was peaking just before the global financial crisis, precisely the same misguided assertions prompted me to write Long-Term Evidence on the Fed Model and Forward Operating P/E Ratios. See also How Much Do Interest Rates Affect the Fair Value of Stocks? from May of that year. Let’s address this argument once again, in additional detail.
Valuations and interest rates
There’s no question that interest rates are relevant to the fair valuation of stocks. After all, a security is nothing but a claim to some future stream of cash flows that will be delivered into the hands of investors over time. The higher the price an investor pays for a given stream of future cash flows, the lower the long-term return the investor can expect to earn as those cash flows are received. Conversely, the lower the long-term return an investor can tolerate, the higher the price they will agree to pay for that stream of future cash flows. If interest rates are low, it’s not unreasonable to expect that investors would accept a lower expected future return on stocks. If rates are high, it’s not unreasonable to expect that investors would demand a higher expected future return on stocks.
The problem is that investors often misinterpret the form of this relationship, and become confused about when interest rate information is needed and when it is not. Specifically, given a set of expected future cash flows and the current price of the security, one does not need any information about interest rates at all to estimate the long-term return on that security. The price of the security and the cash flows are sufficient statistics to calculate that expected return. For example, if a security that promises to deliver a $100 cash flow in 10 years is priced at $82 today, we immediately know that the expected 10-year return is (100/82)^(1/10)-1 = 2%. Having estimated that 2% return, we can now compare it with competing returns on bonds, to judge whether we think it’s adequate, but no knowledge of interest rates is required to “adjust” the arithmetic.
There are three objects of interest here: the current price, the future stream of expected cash flows, and the long-term rate of return that converts one to the other. Given any two of these, one can estimate the third. For example, given a set of expected future cash flows and some “justified” return of the investor’s choosing, one can use those two pieces of information to calculate the price that will deliver that desired expected return. If I want a $100 future payment to give me a 5% future return over 10 years, I should be willing to pay no more than $100/(1.05)^10 = $61.39.
So when you want to convert a set of expected cash flows into an acceptable price today, interest rates may very well affect the “justified” rate of return you choose. But if you already know the current price, and the expected cash flows, you don’t need any information about prevailing interest rates in order to estimate the expected rate of return. One does not have to “factor in” the level of interest rates when observable valuations are used to estimate prospective long-term market returns, because interest rates are irrelevant to that calculation. The only thing that interest rates do at that point is to allow a comparison of the expected return that’s already baked in the cake with alternative returns available in the bond market.
The Fed Model is an artifact of just 16 years of history
There’s an additional problem. While it’s compelling to believe that the expected return on stocks and bonds should have a one-to-one relationship, history doesn’t bear that out at all. Indeed, over the past century, the correlation between bond and stock yields has historically gone in the entirely wrong direction except during the inflation-disinflation cycle from about 1970 to 1998. What investors may not realize is that the correlation between interest rates and earnings yields (as well as dividend yields) has been negative since 1998. Investors across history have not been consistent at all in treating stocks and bonds as closely competing substitutes.
As I noted during the bubbles that ended in 2000 and 2007, the problem with the Fed Model (which compares the S&P 500 forward operating earnings yield with the 10-year Treasury yield) is that this presumed one-to-one relationship between stock and bond yields is wholly an artifact of the disinflationary period from 1982 to 1998. The stock market advance from 1982 to 1998 represented one of the steepest movements from deep secular undervaluation to extreme secular overvaluation in stock market history. Concurrently, bond yields declined as inflation retreated from high levels of the 1970’s. What the Fed Model does is to overlay those two outcomes and treat them as if stocks were “fairly valued” the entire time.
The chart below shows the S&P 500 forward operating earnings yield alongside the 10-year Treasury bond yield. The inset of the chart is the chart that appeared in Alan Greenspan’s 1997 Humphrey Hawkins testimony, and is the entire basis upon which the Fed Model rests. The same segment of history is highlighted in the yellow block. Notice that this is the only segment of history in which the presumed one-to-one relationship actually held.
The Fed Model is not a fair-value relationship, but an artifact of a specific disinflationary segment of market history. It is descriptive of yield behavior during that limited period, but it has a very poor predictive record with regard to actual subsequent market returns.
When investors assert that stocks are “fairly valued relative to interest rates,” they are essentially invoking the Fed Model. What they seem to have in mind is that regardless of absolute valuation levels, stocks can be expected to achieve acceptably high returns as long as the S&P 500 forward operating earnings yield is higher than the 10-year Treasury yield.
No, no. That’s not how any of this works, and we have a century of evidence to show it. The deep undervaluation of stocks in 1982 was followed by glorious subsequent returns. The steep overvaluation of stocks in 1998 was followed by one crash, then another, which left S&P 500 total returns negative for more than a decade. I fully expect that current valuations, which are within a breath of 2000 extremes on the most historically reliable measures, will again result in zero or negative returns over the coming 10-12 years. Let’s dig into some data to detail the basis for those expectations.
First, a quick note on historically reliable valuation measures. The value of any security is based on the long-term stream of cash flows that it can be expected to deliver over decades and decades. While corporate earnings are certainly required to generate future cash flows, current earnings (or even forward earnings) are very poor “sufficient statistics” for that stream of cash flows. That’s true not only because of fluctuations in profit margins over the economic cycle, but also due to very long-term competitive forces that exert themselves over multiple economic cycles. From the standpoint of historical reliability, valuation measures that dampen or mute the impact of fluctuating profit margins dramatically outperform measures based on current earnings. Indeed, even the Shiller CAPE, which uses a 10-year average of inflation-adjusted earnings, provides substantially better results when one also adjusts for the embedded profit margin (the denominator of the CAPE / S&P 500 revenues). For a brief primer on the importance of implied profit margins in evaluating market valuations, see Two Point Three Sigmas Above the Norm and Margins, Multiples, and the Iron Law of Valuation.
The chart below shows the ratio of nonfinancial market capitalization to corporate gross value-added, including estimated foreign revenues. I created this measure, MarketCap/GVA, as an apples-to-apples alternative to market capitalization/GDP that matches the object in the numerator with the object in the denominator, and also takes foreign revenues into account. We find this measure to be better correlated with actual subsequent S&P 500 total returns than any other measure we’ve studied in market cycles across history, including price/earnings, price/forward earnings, price/book, price/dividends, enterprise value/EBITDA, the Fed Model, Tobin’s Q, market cap/GDP, the NIPA profits cyclically-adjusted P/E (CAPE), and the Shiller CAPE.
MarketCap/GVA is shown below on an inverted log scale (blue line, left scale), along with the actual subsequent 12-year total return of the S&P 500 (red line, right scale). From current valuations, which now rival the most extreme levels in U.S. history, we estimate likely S&P 500 nominal total returns averaging less than 1% annually over the coming 12-year horizon. As a side note, we tend to prefer a 12-year horizon because that is the point where the autocorrelation profile of valuations drops to zero, and is therefore the horizon over which mean reversion is most reliable (see Valuations Not Only Mean-Revert, They Mean-Invert).
I’m often asked why we don’t “adjust” MarketCap/GVA for the level of interest rates. The answer, as detailed at the beginning of this comment, is that given both the price of a security, and the expected stream of future expected cash flows (or a sufficient statistic for those cash flows), one does not need any information at all about interest rates in order to estimate the expected long-term return on that security. Each point in the chart below shows the actual 12-year subsequent total return of the S&P 500 index, along with two fitted values, one using MarketCap/GVA alone, and the other including the 10-year Treasury bond yield as an additional explanatory variable. That additional variable adds absolutely no incremental explanatory power. Both fitted values have a 93% correlation with actual subsequent 12-year S&P 500 total returns.
We’re now in a position to say something very precise about current valuations and interest rates. Given the present level of interest rates, investors who are willing to accept likely prospective nominal total returns on the S&P 500 of less than 1% over the coming 12-year period are entirely welcome to judge stocks as “fairly valued relative to interest rates.” But understand that this is precisely what that phrase implies here.
Moreover, as one can see from the foregoing charts, there’s not a single market cycle in history, neither in the period before the 1970’s (when interest rates regularly hovered near current levels), nor in recent decades, that has failed to raise prospective 10-12 year S&P 500 total returns to the 8-10% range or beyond over the completion of that cycle. So even if investors are willing to accept 10-12 year total returns of next to nothing, they should also be fully prepared for an interim market loss on the order of 50-60%, because that is the decline that would now be required to restore those 8-10% return expectations, without even breaking below historical valuation norms.
