A Better Way to Intrinsic Value
Focus on expected returns from conservatively forecast cash flows and not on the false precision of academic models, writes Morningstar’s Sam Lee.
A version of this article was published in the May 2014 issue of Morningstar ETFInvestor. Download a complimentary copy here.
In the first part of this two-part series, "Understanding Intrinsic Value," I briefly described intrinsic value and its relationship with time and risk. In this second and final part, I describe a more practical way to apply the intrinsic-value framework.
Intrinsic value may be the central idea of investing, but it has a big problem: It collapses two key estimates--future cash flows and the discount rate--into a single value. Even minor changes in future cash flows or discount rates can result in huge swings in intrinsic value.
Consider the Gordon dividend growth equation, which provides the present value of a cash flow that grows at a constant rate in perpetuity:
P = D/(r - g),
where P is present value, D is the cash flow one year from the present, r is the required rate of return (or the discount rate), and g is the annualized growth in the cash flow.
The table below shows the present values of cash flow streams where D is $100, g is 1%, and r ranges from 3% to 10%. Even though I'm only changing r, raising g by 1 percentage point has the same effect as lowering r by 1 percentage point. Note that as the discount rate falls, present value becomes increasingly sensitive to changes in the discount rate.
Many investors use historical returns and the capital asset pricing model, or CAPM, to estimate appropriate discount rates. This, unfortunately, is a fatally flawed approach that relies on two bad assumptions. First, it assumes historical returns provide an unbiased estimate of future returns--that is, the market prices assets to always provide a constant return in excess of the risk-free rate, or a constant risk premium. Academics used to believe that the efficient-market hypothesis implied constant risk premiums. This is not true at all. In good times, the market tends to demand low risk premiums and in bad times high premiums. Second, the CAPM, which assumes that an asset's expected return is linearly related to its market beta, simply does not work: High-beta and low-beta assets tend to have similar returns, an anomaly that exists in almost every asset studied, whether stocks, junk bonds, Treasury bonds, or commodities.
The use of CAPM persists despite its widely known empirical failures because investors want to make valuation look scientific. In doing so, they exchange knowledge on their limitations for false knowledge. A better way, in my opinion, is to simply focus on the expected returns from conservatively forecast cash flows and figure out whether that provides an adequate return for the perceived risk taken. This is Ben Graham's margin of safety: Estimate expected cash flow yield on a conservative basis and compare that with the yield on high-quality corporate or government bonds. What this method gives up in false precision, it gains in robustness.
Note that this practice reverses the role of the risk premium. In the academic intrinsic-value framework, the risk premium is the "fair" market return for an asset of a given level of riskiness (however that's defined)--it's handed to the investor from some model or historical data set. In the margin of safety approach, the risk premium is transformed into your private estimate of expected excess returns.
For asset allocators, the Gordon dividend growth equation is a good starting point for estimating expected returns. We can rearrange and reinterpret some of the variables to make it more useful to us:
r = D/P + g,
where r is the expected return, D/P is dividend divided by price, or the yield, and g is the growth in per-share dividends or cash flows. R is no longer the discount rate but becomes our expected return. This equation assumes that over the long run, valuation changes cancel out and yield plus cash flow growth dominate returns.
A valuation change occurs when price changes but cash flow doesn't. For example, if a bond is yielding 10%, and its price quadruples so that it yields 2.5%, the realized 300% return on the bond is much higher than what the 10% yield projected. The reverse can happen. Unfortunately, while valuation changes dominate returns in the short run, they are also the least predictable element of expected returns.
The full equation for expected returns, then, can be written:
r = y + g + s,
where r is expected return, y is the current cash flow yield (sometimes called carry), g is per-share cash flow growth, and s is the return attributable to valuation changes (what Jack Bogle calls the speculative return).
We are finally equipped with the tools necessary to make informed judgments about expected returns for a wide variety of assets. This framework, unfortunately, isn't terribly useful for individual securities, but it is useful for broad asset classes that have fairly predictable cash flows and long-run growth rates.
An Example: High-Yield Bonds
The U.S. high-yield bond market is often represented by the Bank of America Merrill Lynch U.S. High Yield Master II Index. As of the end of April, its yield was 5.5%. In an ideal world with no defaults and no yield-curve changes, you can expect to earn about 5.5% annualized.
However, junk bonds have historically lost about 2.7% annualized to defaults. Year to year, however, losses are either much smaller or much bigger. During boom times, when financing is easy and the corporate earnings are robust, defaults tend to be low. During bad times, defaults spike. If you want to be clever, you can calculate expected credit losses based on your forecast of the economy during the next few years. That's a hard game to play. A better bet for most investors is to assume something close to the historical average.
Now we have the two biggest components for our expected return forecast: We have y, yield, and we have g, the expected per-share growth rate of that cash flow. Plugging into our equation, we get:
r = y + g = 5.5% - 2.7% = 2.8%.
That's downright horrific when you consider the fact that 1) high-yield bond funds have struggled even before fees to match the junk-bond index, and 2) you can get 2% risk-free return by socking your money in FDIC-insured five-year CDs.
So why are investors willing to pay such high prices for junk bonds today? There are two plausible and non-mutually exclusive explanations. First, the market might be forecasting much lower annual credit losses. PIMCO, BlackRock, and many other managers believe the next year will be benign for credit, thanks to a recovering economy, ample liquidity, and low interest rates. Second, investors' desperate search for income has pushed high-yield bond valuations near bubble territory.
We've so far omitted the third term, "s"--the return from valuation changes. Junk bonds can do better than 5.5% annualized if their yields fall. However, s, as I mentioned, is the hardest component to forecast. I think it's better to think of "s" in directional terms--what is its plausible contribution over the long run?
Interest rates and spreads are at all-time lows. Further positive surprises will likely have only modest impact. However, negative surprises could send yields and spreads spiking. The payoff is asymmetrical. Valuation changes are likely to hurt more than help.
Keep in mind that expected returns are just that--expectations, not guarantees. That's investing. It's a probabilistic exercise. Because the future is difficult to forecast, one should use conservative estimates as a cushion against one's errors. That's margin of safety in a nutshell.
| ||ETFInvestor Newsletter|
|Want to hear more from Sam Lee? Subscribe to Morningstar ETFInvestor for fresh ideas for income and total return plus a bird's-eye view of valuations around the globe, portfolio construction advice, and data on the biggest and most popular ETFs.|| One-Year Digital Subscription |
12 Issues | $189
Premium Members: $179