Wilmott Forums - time-linking performance measurement

CQF

Forum Navigation:

login

FORUMS > Student Forum

Topic Title: time-linking performance measurement
Created On Thu Mar 10, 05 03:40 AM

Topic View:

View thread in raw text format

tigerbill
Senior Member

Posts: 338
Joined: Apr 2004

Thu Mar 10, 05 03:40 AM

User is offline

I notice at the book 'Investment performance attribution' by David Spaulding, the author highly recommends a linking across time method--Carino model or logarithmetic model, while i also found a paper 'Multiperiod Attribution Calculations' by Damien Laker, who is against this model, proves Carino model is just an approximation, not an exact method.

I m a little bit confused, not sure which one is better.

whoever has experience on this field, could you give me some suggestion?

thanks a lot in advance.

-------------------------
Quantitative Finance Collector

Aaron
Senior Member

Posts: 6354
Joined: Jul 2001

Fri Mar 11, 05 06:38 PM

User is offline

There are two distinct issues.

(1) What is the best way to predict future performance, given past history? The rule of thumb is if the manager choses the amount of investment, dollar weighted returns, but if the manager has no control over the amount of investment, time weighted returns. However, there are nuances in all practical cases.

(2) What is the best way to estimate past performance, given limited data? This is the more controversial point. It's not nearly as important as it used to be. When people had to compute performance with pencil and paper, given published periodic data, an easy approximation was valuable. Now that everyone downloads data and can do exact computations by computer, approximations are not used as much. In any case, there's no correct answer, everything is an approximation. The trick is to pick the best approximation for the data you have and what you're using it for.

-------------------------
Aaron Brown

Apati
Member

Posts: 26
Joined: Mar 2003

Tue Mar 22, 05 11:30 PM

User is offline

Quote
Originally posted by: Aaron
There are two distinct issues.

(1) What is the best way to predict future performance, given past history? The rule of thumb is if the manager choses the amount of investment, dollar weighted returns, but if the manager has no control over the amount of investment, time weighted returns. However, there are nuances in all practical cases.

(2) What is the best way to estimate past performance, given limited data? This is the more controversial point. It's not nearly as important as it used to be. When people had to compute performance with pencil and paper, given published periodic data, an easy approximation was valuable. Now that everyone downloads data and can do exact computations by computer, approximations are not used as much. In any case, there's no correct answer, everything is an approximation. The trick is to pick the best approximation for the data you have and what you're using it for.

Hi Aaron,

Since you seem to have a lot of experience on these topics I'd like to ask any suggestion on the areas/topics that you'd suggest someone to research/work on as part of a PhD in Performance Attribution? What would you say would be the most hottest areas to play with that will contribute to academia and mainly in the market? i know it's not an easy question but since you got some experience and definetely have faced much more issues u might have sth in mind. thanks in advance!

-------------------------
Apati

Aaron
Senior Member

Posts: 6354
Joined: Jul 2001

Thu Mar 24, 05 10:44 PM

User is offline

I say this a lot, but it's never a good idea to get a topic this way. Pick an area of interest and start researching it. Questions will occur naturally, and you'll have the context to understand them. Plus you'll have a lot of your research done, and you'll know who to contact for more information.

-------------------------
Aaron Brown

Apati
Member

Posts: 26
Joined: Mar 2003

Fri Mar 25, 05 08:53 AM

User is offline

Quote
Originally posted by: Aaron
I say this a lot, but it's never a good idea to get a topic this way. Pick an area of interest and start researching it. Questions will occur naturally, and you'll have the context to understand them. Plus you'll have a lot of your research done, and you'll know who to contact for more information.

thanks for the sincere reply..i think you are fair at least L

it just sometimes seems to me that academia is way behind reality and thought i could get some "inside" tips..

anyway, i am trying to research on relevant subjects and constrain myself from widening the areas of interest

actually the more you read the more you understand that you are ignorant...which at the end makes things more complex. thank anyway!

-------------------------
Apati

Aaron
Senior Member

Posts: 6354
Joined: Jul 2001

Fri Mar 25, 05 04:00 PM

User is offline

You're right that academia sometimes drifts far from reality. The reason is that academics sometimes search for topics rather than genuinely learning about an area. If you take the latter approach, topics are easy to find, and you will stay firmly anchored in reality.

This is not a question of theory versus practice, it's a question of meaningful theory versus silliness.

-------------------------
Aaron Brown

Apati
Member

Posts: 26
Joined: Mar 2003

Fri Mar 25, 05 10:18 PM

User is offline

So, are you implying that I'd better give priority to applied issues and in a way put aside for a while the pure academic/theoritical models? in other words where would you suggest to study from for issues relating on asset allocation issues concerning mutual funds, index funds, retirement portfolios(i think this one will gain popularity as the whole retirement system seems to collapse). any material to look into?

-------------------------
Apati

Apati
Member

Posts: 26
Joined: Mar 2003

Fri Mar 25, 05 10:18 PM

User is offline

So, are you implying that I'd better give priority to applied issues and in a way put aside for a while the pure academic/theoritical models? in other words where would you suggest to study from for issues relating on asset allocation issues concerning mutual funds, index funds, retirement portfolios(i think this one will gain popularity as the whole retirement system seems to collapse). any material to look into? thanks in advance for your time and help.

-------------------------
Apati

Aaron
Senior Member

Posts: 6354
Joined: Jul 2001

Sat Mar 26, 05 04:42 PM

User is offline

No, not at all. As I said, it's not the distinction between theory and practice that matters. The point is to ground your research in something meaningful. That means, learning about something, then following up the questions that arise naturally, rather than starting with a question suggested by someone else.

You are interested in performance attribution. Why? Do you want to separate skillful managers from lucky ones? Do you want to refine a hedge fund's strategy dynamically? Are you interested in risk management of multi-manager portfolios? There are other applications of PA, each will suggest different questions. All these questions can be considered usefully from a theoretical or practical perspective.

Let's say you want to use PA to separate skillful managers from lucky ones. If you read the literature, professional and academic, you will find a major concern is factor overlap. A good starting point is to collect all published methodologies for eliminating or correcting for overlap and see if you can replicate them on a standard set of data. In the process of this, you will undoubtedly communicate with other researchers in the field, because you'll need to get details that are missing from published sources.

At the end of this you have a good section for your paper (a standard comparison of the literature to date) and will probably have dozens of topic ideas suggested as you try to implement other people's work; and speak to people. You will know that your topics have not already been addressed, and that they interest other people in the field. If you plan to look for a job in finance, you will be able to drop a lot of names and offer impressive references. Some of your sources may offer you jobs themselves.

If you skip all this and decide to think up a method for eliminating factor overlap on your own you will likely find that it's unrelated to real concerns, or reinventing the wheel, or suffers from well-known problems. Even if none of that is true, you won't cast your research into familiar terms, and no one will know you; thus no one will read your paper. If you plan to use it to look for a job, no one will be impressed, because you won't be able to describe it in terms of existing knowledge base.

-------------------------
Aaron Brown

Edited: Sat Mar 26, 05 at 04:43 PM by Aaron

Apati
Member

Posts: 26
Joined: Mar 2003

Sun Mar 27, 05 08:41 AM

User is offline

Thank you once more for sharing your thoughts and experience overall. I'd certainly try to do the "networked" approach and see how it goes..I actually have like 80% of the literature in hand and have done a first scan (a year ago!) so I will start pointing the critical ones and gathering more and then move on with the rest. It seems to me that there are a lot of things to be done but that i knew..i just don't want doing useless efforts..thanks for sharing!

-------------------------
Apati

View thread in raw text format

FORUMS > Student Forum

Forum Navigation:

© All material, including contents and design, copyright Wilmott Electronic Media Limited - FuseTalk 4.01 © 1999-2010 FuseTalk Inc.