1
$\begingroup$

I have been reading about Sharpe's return-based style analysis, which tries to determine the manager's exposure/effective asset mix to changes in the values of the asset classes. It does so by using quadratic optimization (minimizing the variance of the manager return minus coefficient times returns of chosen asset classes).

My question is: 1) Does this analysis give the same/different interpretation as/from the multivariate linear regression model? 2) I want to answer this question: which sector had the biggest return and risk contribution to the market from t = t1 to t= t2? What percentage of total return and risk is coming from the TMT sector from t = t1 to t = t2? I have daily returns of MSCI ACWI (the world equity market index) and MSCI ACWI for all the sectors (tech, financials, health care, etc). Would Sharpe's return-based style analysis be a good method to answer those questions?

$\endgroup$
2
$\begingroup$

1) Return-based style analysis (RBSA) generally uses multivariate linear regression over rolling windows to get results. Sometimes Kalman Filters are employed instead.

2) Not the best, honestly. returns-based style analysis tries to guess the dependence of your fund (ACWI) and it struggles with resolution for large numbers of variables. The number of sectors is not too large but large enough to cause problems.

You can likely can get the weights for the ACWI sectors over time. So you should be able to just do performance attribution and risk attribution which is much easier.

$\endgroup$
  • $\begingroup$ Thank you so much. How can I get the weights for the ACWI sectors over time? $\endgroup$ – Jun Jang Jun 12 '18 at 20:20
  • $\begingroup$ mmm... I would get it through Bloomberg, but that is stupid expensive and you may not have access. I don't know a free source. Maybe Quandl? You might have to ask quant.se. $\endgroup$ – rhaskett Jun 12 '18 at 20:32
  • $\begingroup$ I have access to Bloomberg! I will research and see if I can download the weights over time. Thanks! So besides that, I will also have to use big data (~100 independent variables and equity long/short hedge fund returns) to explain the fund returns. As you mentioned in your own post (Robust Returns-Based Style Analysis), using RBSA is not the best method for long/short funds or funds that change strategies over time. What method should I use? Did you end up using Kalmal Filter? I found the Markov (2006) paper too difficult to read.. lol $\endgroup$ – Jun Jang Jun 12 '18 at 20:35
  • $\begingroup$ I looked into the risk attribution link. I think it's very helpful, but the problem is that weights (w_i) change over time, but the equation on the website assumes that the weights are constant over time.. $\endgroup$ – Jun Jang Jun 12 '18 at 21:15
  • $\begingroup$ Sector weights change slowly. So the common easy solution is just to average the weights if you have a short to medium time period. Otherwise, I know it is possible with weights as a time series, but I've never looked into it myself. $\endgroup$ – rhaskett Jun 12 '18 at 21:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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