Page 3 of 3
In a practical sense, we need to take the data another step. How might an actual (for purposes of this article all data is in fact historical and hypothetical) portfolio of sector indexes be constructed and expected to perform? It has been our experience that in an effort to most honestly look at backtested material, the allocations within those tests must be of equal dollar weight. With 20/20 hindsight, it would be easy to allocate more assets to better-performing sectors and skew the results. It is also how we manage assets in the real world, since no one has 20/20 foresight.
The data in the Adding Sectors Chart holds no real surprises. Adding sectors will lower the overall return, but has the advantage of alleviating volatility (as measured by standard deviation). There is clearly added benefit to eliminating sectors from a portfolio on an annual basis, as it is clear that the diminished volatility of owning all sectors in the sample is too costly (in our opinion) a price to pay for the diminished reutrn.
Where along the arc of return and volatility an investor would be most comfortable is an individual choice. Our experience in the cauldron of actual portfolio management and client relations suggests that some level of sector diversification is called for, and full disclosure of expected returns and volatility is a must. There is no "right" allocation, only an allocation with which investor and advisor can align expectations.
Implications And Summary
The most apparent implication of our work here is that, once again, the stability and market sensitivity of indexes, be they grouped in a Three Factor Model priority or, as we have attempted to show here, in a sector model, will best serve investors. This superior service is also evident if a long-term momentum discipline is used to create and maintain a portfolio of sector indexes. I restate that our data here can be recreated in real portfolios using iShare/Dow Jones sectors or using other products to an individual investor's liking, but that our results and data are reflections of the iShare/Dow Jones product.
Another implication that I think is very significant is the proof that diversification in and of itself has limited benefit to a portfolio designed with long-term momentum as a basis. As we see in our second exhibit, the more diversification, the lower the return. There is a point along the arc that will suffice for most investors that is well short of the need to own all sectors at a single time. Diversification up to a point is beneficial, in particular as it eases investor discomfort associated with volatility, but beyond that it would seem to have negative implications.
We give no weight or significance to any individual sector. Each is allocated to our model exclusively based on its prior year ranking in the sample.
Perhaps the implication that is the most unnerving for many investors is one that we draw from our first exhibit of the Ranked Sectors. It is our belief that this data suggests that the old saw of "buy low, sell high," needs a little revising. Buying anything low is of course ideal; determining when "low" is "low" is a real problem. What we think our data may best suggest is that an investor should hold high - it will serve them better - until the sector (in this model) for whatever reason reverses its course from positive to negative momentum.
In summary, back in 2001 we missed the power of ETFs to propel a sector strategy based on our unique momentum discipline. The fact that there are index-based ETFs in the sector space, coupled with ETFs' inherent cost and tax efficiency, make these asset classes prime targets for momentum-based strategies. Because ETFs lend themselves so well to a prudent momentum program, we have now entered the space with enthusiasm.