John Hussman. Still very bearish on stocks.
In his latest weekly note he discusses the last lost decade (0% returns for 10 years) and predicts a nearly-as-dismal result for the next 10 years.
When starting valuations are elevated, investors require similarly elevated terminal valuations – 3 years, 5 years, 7 years, 10 years and further into the future – in order for stocks to achieve acceptable long-term returns. Investors and analysts entirely miss the point when they propose that stocks are “fairly valued” based on a short-term condition, whether it is the prevailing level of 10-year bond yields (which can change significantly over periods of much less than 10 years), the current inflation rate, or the expected level of next quarter’s profits. Once valuations become elevated, particularly on profit margins that are also already elevated, investors require terminal valuations to be stretched to the limit, years and years into the future, in order for their speculation to be bailed out. At present, stock valuations are elevated on a variety of smooth metrics. In contrast, stocks appear reasonably valued only on metrics which place excessive weight on short-term factors, and which can therefore be shown to perform poorly in historical data.
Based on our standard methodology (see The Likely Range of Market Returns in the Coming Decade for the basic approach) we estimate that the S&P 500 is priced to achieve a 10-year total return of just 5.05% annually. Using our forward operating earnings methodology (see Valuing the S&P 500 Using Forward Operating Earnings), the projected 10-year total return is just 4.69% annually. With the S&P 500 dividend yield at 1.96%, the 10-year projection from dividend-based models is even lower, at about 2.30% annually (though prospects are good that faster growth of index-level dividends will bring that estimate closer to earnings-based projections, see No Margin of Safety, No Room for Error).
Overall, the projected returns for the S&P 500 are now lower than at any time in U.S. history prior to the bubble period since the late-1990’s (which has resulted in predictably dismal returns for investors). At present, investors rely on a continuation of this bubble to achieve further returns. With respect to the bubble period, the current projected returns match those we observed at the April 2010 high, and are at about the same level as we observed before prices collapsed in 2008. Valuations were even more extreme, of course, at the bubble peak of 2000, which was predictably followed by a decade of zero returns.
Keep in mind that near-term returns much higher than 5% annually would essentially be shifted from future years, meaning that higher returns today simply imply even lower long-term return prospects tomorrow. For very long-term investors, say 15 years or more, such variations hardly matter. By our estimates, investors are looking at prospective 15-year total returns in the range 5.8-6.5% annually, with enormous volatility in the interim, no matter how you cut it.
I should note that while I clearly underestimated the extent to which investors would concentrate their 10-year return prospects into an 8-month span from March to November of 2009, this was no fault of the valuation methods. Rather, I refused to discard lessons from prior historical credit crises in the U.S. and internationally. Except for the relatively contained S&L crisis, no major credit crises were observed in post-1940 U.S. data (which is what we use most heavily in our investment analysis). It is unfortunate that we would have actually performed better if I had assumed that the recent downturn was nothing but a typical post-1940 recession and recovery, but I am still convinced that this would not have been appropriate. When you develop a model using some set of data that includes valuations, market action, economic conditions, and other variables, you can’t reasonably apply it to data that is clearly “out of sample” and unrepresentative of what you observed in the data underlying the model. Rather, it’s essential to examine additional data that is as representative as possible of conditions you actually observe. In the case of 2009, we could not rule out other post-credit crisis evidence (such as pre-1940 data), and (unfortunately, as it turned out) that data implied the need for much more stringent valuation criteria than post-war data did.
I remain unconvinced even now that we should view the current economic climate as a standard post-war economic cycle. Still, the experience of the past few years has clearly added to our post-war dataset in that it now contains a full-blown credit-crisis. Every new data point adds information, either in creating new distinctions, or increasing confidence in the distinctions one has already learned. The past three years have confirmed much of what we already knew about valuation, while the enormous volatility of prices, despite an overall market loss, has contributed significantly to our analysis of market action.
In any event, whether or not one believes the current economic cycle should be viewed as a “typical post-war recovery,” it is clear that our valuation methods have been accurate measures of likely market returns over time – even during the recent crisis. At present, the long-term outlook for equities is unfavorable on the basis of valuation, so regardless of shorter-tem influences, I expect that long-term investors are likely to be ill-served by investing in stocks at current levels.