2
$\begingroup$

I am trying to do some performance attribution for a few portfolios we manage. What I am trying to examine are three different sources of returns:

  1. The general asset allocation
  2. Security Selection
  3. The performance of our tactical allocation

I am quite familiar with the textbook way of performance attribution, categorizing assets and then selecting category benchmarks (easier said than done sometimes) and then comparing alpha and beta. But when you are constantly making portfolio security and weighting changes then this simplistic approach doesn't hold (unless I could calculate returns on a daily basis, which I don't have data for.)

I was thinking of making a replica portfolio and then entering trades every time a real trade is done. That way the replica portfolio will have an updated asset allocation (benchmark) to compare with my actual portfolio and the difference should be my security selection alpha.

Ideas or knowledge of papers/books/research in this area would be greatly appreciated.

Improve this question
$\endgroup$

2 Answers 2

Reset to default
2
$\begingroup$

There are a number of issues here. First, there are a number of methodologies called “performance attribution” each providing answers to different questions. So I am not sure what type of question you wish to address. I will here assume that you wish to evaluate the effects of investment decisions as opposed to the effects of market factors. I will also assuming that your distinction between “general asset allocation” (as different than Security Selection) and “tactical allocation” means that you are not simply nesting asset allocations but incorporating “tactical” decisions that are not instantiated in the actual weights implemented by the fund. Then the “textbook way of (decision) performance attribution” needs to be expanded/adjusted in order to incorporate “tactical allocation”. Similarly, if one wishes to evaluate the effects ON beta, since “textbook” decision mythologies only meaningfully address active return. Alternately, if you want to evaluate the effects OF beta then this would have to be correctly formulated as a decision or one might consider other types of attribution, such as market factor attribution, instead of decision attribution. Finally, all methodologies need to be tailored to the content of the data you provide. Since you say that you do not have daily data, to properly answer your question, I would need a clearer understanding of what data you do have. The level of seriousness that you seem to have regarding getting “performance attribution” correct indicates that you need an industrial strength solution. Such are not buildable as a small project, but takes years of careful thought and implementation. This suggests that you look into buying a robust solution rather than trying to build one yourself.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Tks for the feedback any software suggestions? or models listed somewhere I could look at to give me a clue as to what you are referring to $\endgroup$
    – Avi
    Oct 10, 2012 at 16:10
0
$\begingroup$

" ...making a replica portfolio and then entering trades every time a real trade is done."

You can do this in excel with free historical data off yahoo finance. I'm giving you a link to some example data analysis with excel, http://matdays.blogspot.com/2012/11/excel-data-analysis-10-basic.html

Improve this answer
$\endgroup$
2
  • $\begingroup$ Since this is a link to your own blog, could you summarize the contents here? $\endgroup$
    – chrisaycock
    Dec 17, 2012 at 3:52
  • $\begingroup$ The gist of the idea is to create spreadsheets with prices of products you're trading. Then using various excel functions, you could simulate trades, and review associated stats. :) $\endgroup$
    – Rock
    Dec 18, 2012 at 1:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.