Multifactor model for enhanced indexing

CFA Level III Fixed Income
#1

The curriculum differentiates between the cell-matching technique (or stratified sampling) and the multifactor model technique in order to construct an effective indexed portfolio.

I don’t really see the difference because I can’t really see how you could use the multifactor model technique without dividing the portfolio into cells as well, according to the various risk factors.

For exemple if you want to match the following risk factors (to make it easy): sector and quality percent and sector duration contribution, assuming there are only two sectors (industry and service), two qualities (A and B) and two durations (1 and 2), then there would be 8 possible cells (I, A, 1; I, A, 2, I, B,1… S, B, 2). You look at the bonds from the index that fit in each cell and at the weight of each cell and you invest accordingly.

Or is a computerized technique? And if so, isn’t the computer just doing exactly the same thing except you don’t see it because it is running in the background?

Very practical question again… Hope you guys don’t mind: And I hope it does not seem too stupid but I am not at all working in portfolio management.

#2

If you look deeper into indexes they tend to have large biases.

e.g. the dow has a large GS weight. emerging market bonds mostly argentina.

with cell matching, you might choose to invest in some other banks than just GS.or choose other south american countries for your bond portfolio.

with mutifactor you using more technical attributes. large/small cap. cyclicality, value/growth.

I seem to remember that cell matching is used to increase diversification. but multifactor is is to reduce the number of instruments.

the topic is covered for both FI and EQ, so if it’s not clear read the other one.

#3

oh ok interesting.

I thought that will cell-matching, you ended up classifying each bond of the index in one cell, and regarding your portfolio, you were only taking bonds of the index to invest in. For example, considering the index, you would end up with 6 bonds in one cell and 10 bonds in another, and regarding your portfolio you would invest in 3 and 5 bonds thereof respectively.

But if I understand you right, the above is not entirely correct as you may use totally different bonds for your portfolio. You would only try to match the weighting per cell of the index, the cell having given caracteristics, but not being necessarily related to one bond or another.

Is it correct? And sorry, writing my thoughts down in English is a bit challenging for me here. I hope it does not look like Chinese…

#4

Cell matching is simply match each weight of the benchmark cells in the bond index portfolio regardless of the number of bonds within the cells. If financial sector has 30 stock that represent 30% of S&P 500, for example, you should match this weight with One stock or 10 or whatever as long as the weight is matched at 30%. Multi factor model has different approach, it ignores the weights of any cell or sectors. So, as long as the bonds in the benchmark matches the factores it should be included in the bond index portfolio.