Excerpt from the Hussman Funds' Weekly Market Comment (4/26/10):
What is most striking about Depression-era data between 1929-1935 (and post-credit crisis data more generally) is that when we examine periods when one would generally grade market action as favorable on the basis of major trends and market internals, we find that taking positive exposures would actually have resulted in a net loss. This is due to the abruptness of trend reversals, so even periodic gains of 30-50% would have been largely erased through a combination of abrupt initial losses and subsequent whipsaws. While the compound net loss from periods of "trend following" over the full period was tolerable (about a -25% drawdown), that figure represents a combination of large individual gains and losses - substantial volatility, with a negative overall contribution to returns.
Partitioning the data provides a clearer picture. When valuations exceeded even 12 times normalized earnings (on our most comprehensive measure discussed above), seemingly "favorable" market action was followed by profound losses averaging -69.8% on an annualized basis (generally reflecting a few weeks of vertical losses until enough damage was done to kick the market action measures negative). Once the initial damage was done coming off of the uptrend, valuations over about 12 were still hostile, but were associated with slightly less profound losses averaging -37.7% annualized.
In contrast, only when valuations became quite depressed did the combination of favorable valuations and market action produce positive subsequent returns. Multiples below 12, coupled with favorable market action, were associated with annualized returns of 12.5%, while multiples below 12 coupled with unfavorable market action were associated with further mild losses averaging -4.5% annualized.
In 2009, we observed only a few weeks in March when the S&P 500 was priced at less than 12 times normalized earnings (again, on our best measure). At that time, indicators of market action were still negative. Faced with two possible data sets, one assuming further credit strains and one assuming that the problems had been solved, I noted "even giving the two possibilities equal weight is harsh, because as I've repeatedly noted, post-crash markets have included advances as large, and larger, than we've observed since March, but with devastating follow-through." Needless to say, sharply negative return figures don't "average in" very well.