For those that might not be familiar with the the relationship between arithmetic (simple average), standard deviation and annualised (geometric), have a look at Gummy's article

http://www.gummy-stuff.org/AM-vs-GM.htmNear the bottom of that page is a calculator, next to this image

Have a play and see how a higher standard deviation (volatility) results in smaller arithmetic results (and at the end of the day its the arithmetic (geometric) that investors actually get to spend).

Now consider that since 1927 (up to 2011) the rolling 10 year arithmetic average standard deviation for small cap value (SCV) was 30% compared to 17% for the S&P i.e. the volatility of SCV were a lot more volatile than the S&P. Plot those rolling 10 year averages and a degree of inverse correlation was also apparent i.e. they weren't both highly volatile at the same times.

Assuming a 2% inflation rate (Fed/BoE targets), Gummy's calculator indicates you'd have to have a 6.2% arithmetic average with a 30 stdev to get a 2% annualised. For the S&P's 17% stdev you only need a 3.4% arithmetic with 17% stdev to achieve the same 2% annualised.

If you hold only 57% exposure to an asset that had a 6.2% average, 30% standard deviation and stuffed the rest under a mattress, you'd have a combined portfolio that averaged (0.57 x 6.2) average = 3.54% average, with a (0.57x30) = 17% standard deviation. Which is similar in this example to having held 100% S&P.

Instead of a 50% weighting in S&P, you might hold 0.57x50 = 28.5% in SCV to similar effect and endure similar overall volatility. If the 21.5% remainder of funds not invested (rest of the 50% that would otherwise have been invested in the S&P) earn a reasonable return, or move counter direction to stocks, that might further bolster rewards/reduce risk.

You might say that around 30% SCV, is similar to holding 50% S&P. But those two don't track each other closely, so its more like two different stocks, that aren't highly correlated. In Japan post 1990 for example SCV have performed well, whilst their TSM has been dismal.

If 30% SCV provides similar reward to 50% S&P, then rather than a 50-50 portfolio of S&P with perhaps 5 year Treasury (bonds) you held 15% SCV, 25% S&P, 60% 5 year Treasury then according to SImba's backtest spreadsheet (which TrevH was a contributor) you achieved similar reward, but did so with significantly less portfolio risk (volatility) - something like 9.6% annualised with 8.4% stdev compared to 9.8% annualised with 18.2% stdev.

Small stocks may outperform the market at times. However, the inherent risk is great. The AAII has a small stock model portfolio and it was down over 50% in 2008.

Small stocks are too volatile for me. I'd rather go with smooth and steady.

I don't know many folks who could stay with a portfolio down 50%.

But if you only held 57% exposure to a -50% decline = 0.57 x -50 = -28.5%. And the remainder might have moved counter direction to reduce (make smaller) those losses further. 2008 50% TSM, 50% 5 year T endured a -11% in 2008, compared to 25% TSM, 15% SCV, 60% 5 year T declining -6%.

Have a look at this comparison - broadly similar results despite holding (some) SCV exposure.

One investor might opt to hold 30% SCV, 70% 5 year T as a proxy for 50-50 TSM/5 Year T. Another might opt to invest the 20% surplus from holding 30% SCV instead of 50% TSM in gold (20% gold). Yet another investor might opt to hold a STT/LTT barbell instead of 5 year T. Another might opt to hold 15% in a 2x (leveraged) SCV instead of 30% SCV....etc...etc. Which broadly are all subsets of Ben Graham's 50-50, with different names (Larry's Fat Tail Minimisation, Permanent Portfolio, Nassim Taleb's/Zvi Bodie's mostly safe, some highly speculative ...etc.). Just pick and stick with whichever is the most comfortable for you - according to which might be the most cost/tax efficient for you, and/or the better valued at the time of investing.