Generally the CXO Advisory blog does great work. Their digests of the latest finance papers from SSRN are especially valuable. But in this particular case they got it wrong.
Exhibit A. Fortunately, Hussman actually publishes what his fund return would be if no hedging strategy were employed. For the period 21 Nov 2000 (inception) to 30 June 2009:
| $10000 becomes | Annualized return % | |
| HSGFX | 21346 | 9.21 |
| Without hedging | 15137 | 4.93 |
Subtracting the reported return without hedging (4.93%) from the total return (9.21%) gives an annualized return due to hedging of 4.07%. (Arithmetic on annualized returns is done geometrically. In this case, 1.0921/1.0493 = 1.0407)
Comparing HSGFX with an S&P 500 index fund, Vanguard Index Trust 500 (VFINX):
| $10000 becomes | Annualized return % | |
| VFINX | 7939 | -2.65 |
| HSGFX outperformance | 12.18 | |
| Due to hedging | 4.07 | |
| Due to stock selection | 7.79 |
HSGFX (9.21%) outperformed VFINX (-2.65%) by an annualized 12.18%. Of this outperformance, as we saw above, 4.07% was due to hedging. This leaves the remaining outperformance of 7.79% presumably due to superior stock selection.
Exhibit B. Let’s look at the data another way. I calculated a 20-day rolling beta for HSGFX, using VFINX total return as a proxy for the market. It looks like this:
Hussman turned bullish early in 2003, then gradually scaled back as the market moved higher, reaching maximum bearishness in mid-2007. What’s not to like about that? In 2008 he did turn bullish too early, just before the October crash, which cost him dearly (though his fund holders still came out well ahead of anyone fully invested in stocks).
Now associate each of the betas calculated above with the market return that occurred during the corresponding 20-day period, sort the data by market return, then divide the data into deciles. Down market periods to the left, up market periods to the right. Within each decile, what was the average beta exhibited by HSGFX?
Very clearly Hussman, on balance, was positioned with lower beta during down periods and higher beta during up periods—exactly as he intended.
Why did CXO come up with a different answer? Primarily because CXO put too much faith in statistical regression. The least squares regression formula minimizes squared error, whereas the market rewards or penalizes you in proportion to actual (unsquared) error. In least squares regression, outlying points end up getting disproportionately large weighting. Dropping only two of the oddball data points from CXO’s regression (crash week, 10 Oct 2008, and week horribilis, 8 May 2009, when HSGFX was down 5% while the market was up 6%) gives very different results. Rather than leave out those two data points, however, I leave them in but redo the regressions using a nonstandard method: minimizing the sum of the absolute errors. This method of regression gives the result we would expect if the hedging were successful: a steeper slope (beta) during up weeks than during down weeks. The new regressions are shown by the black lines in the chart below; the original regression is shown in red and green (click on the image for a larger version).
Conclusion: Hussman’s performance has been stellar, with estimated outperformance of 4.07% due to market timing and of 7.79% due to stock selection. Either of the two is beyond most fund managers. For one manager to combine both is truly unusual. I would be surprised to find any mutual fund farther above the security market line over the past 8 1/2 years than Hussman Strategic Growth.
Happily for us, Hussman is very open about his methods. He publishes his current stance on the market every Monday morning in his weekly market comment.
