As Bernstein admit, it has always been know that DCF models put a larger than ideal weight on uncertain cash flows that run far into the future. It’s just that the problem is now more intense. They put together a simple model to make the point involving two companies, the first of which “grows at 10% for 5 years, then linearly fades to a terminal growth rate of 3.5% at year 10 and the second is a more cyclical company that on average manages 5% growth for the first 5 years and then undergoes a similar fade.” They then construct DCFs of these two companies where the discount rate varies from 10% down to 3.6%. The point they’re making here, which was already flagged at the start, is that when “the discount rate is 5%, roughly 90% of the NPV rests in the terminal value and 5% in the 6-10 year segment.” And that “any framework that ascribes anything approaching 90% of current value to cash flows that are necessarily unknowable far in the future has to be flawed. We have never had discount rates at such low levels as we have today and hence have never had this problem before.”
What should investors do? We cannot reject discounting and there is no choice but to use it anyway. So we will keep using DCFs. We just have to be aware that a by-product of the low rate world is a scale of forecast error that is outside the bounds of what has been previously seen and it is likely that those forecast errors may swamp any other differences between stocks. What would definitively break such a model? By construction, if the overall discount rate fell below the growth rate then the NPV becomes undefined, though in a semi-permanent low rate environment presumably expected growth rates fall too. We have already seen European companies issue debt at negative yields and in Japan 10 year rates are set at 0. Any move lower in discount rates would, we suggest, cause a significant methodological problem for financial analysis.