There has been a lot of confusion about the difference between the attribution formulas in the micro attribution section (reading 47) and the global attribution section (reading 48), myself included. Thus, I have done some extra-curricular reading to figure out what’s going on, and thought I would post it here in case other people are still confused. The attribution model in reading 47 is known as Brinson-Fachler attribution. It uses the following formulas: allocation = (portfolio sector weight - benchmark sector weight) * (benchmark sector return - benchmark return) selection = (benchmark sector weight) * (portfolio sector return - benchmark sector return) interaction = (portfolio sector weight - benchmark sector weight) * (portfolio sector return - benchmark sector return) Active return = allocation + selection + interaction The attribution model in reading 48 is known as Brinson-Hood-Beebower attribution. It uses the following formulas (note that these don’t match the formulas in the curriculum for reasons I will explain later): allocation = (portfolio sector weight - benchmark sector weight) * (benchmark sector return) selection = (benchmark sector weight) * (portfolio sector return - benchmark sector return) interaction = (portfolio sector weight - benchmark sector weight) * (portfolio sector return - benchmark sector return) Active return = allocation + selection + interaction I have removed the yield and currency component, for simplicity’s sake. The allocation term is different between the two models (this is why this term is different between readings 47 and 48). Some people don’t want to include the interaction term in their attribution reports. If so, the interaction term is combined into the selection term (which is the other reason that the formulas presented in readings 47 and 48 are different). You can remove the interaction term from either the model in reading 47 or 48 in the same manner. However, whichever formula you choose (the one in reading 47 or 48), you will end up with the same overall result. The curriculum and the Schweser material does not do a very good job of fully explaining the relationship between the two approaches. Oh, and Fachler’s first name is “Nimrod”.
Maybe the greatest name ever - Nimrod Fachler
awesome…this helps…thanks a ton…now my fried brains can cool down… nimrod fachler…that is classic…is that even worse than gaylord focker!
Thanks a lot. It is a great job!
bump nimrod fachler. i’m having some trouble figuring out when i’m supposed to use what…
Alright, this is how our Schweser instructor put it and told us we should use on the exam: ATTRIBUTION, when not comparing to a benchmark, but assigning attribution WITHIN a portfolio: (shown below for individual weights so no “j” subscript; W=weight, R=return, p=portfolio, b=benchmark, l=local ccy, d=domestic ccy) Mkt Return Effect: Wp ( Rbl ) Security Selection Effect: Wp ( Rpl - Rbl ) Currency Effect: Wp ( Rpd - Rpl ) Add in yield for total attribution. Note you’re only using PORTFOLIO WEIGHTS here. Also note, our instructor felt that the interaction effect, since it’s generally so small anyway, is unimportant. ALLOCATION, when comparing to a benchmark: Mkt Allocation Effect: ( Wp - Wb ) Rbl Security Selection Effect: Wp ( Rpl - Rbl ) * Note this is the same as above Currency Allocation Effect: Wp ( Rpd - Rpl ) - Wb ( Rbd - Rbl ) Add in the benchmark DOMESTIC return and yield for total allocation effects. Not only should you know the calculations, but know what each measure implies about the investment manager. Any thoughts on the above? Let me know!
darkstar Wrote: ------------------------------------------------------- > There has been a lot of confusion about the > difference between the attribution formulas in the > micro attribution section (reading 47) and the > global attribution section (reading 48), myself > included. > > > The attribution model in reading 47 is known as > Brinson-Fachler attribution. > > selection = (benchmark sector weight) * (portfolio > sector return - benchmark sector return) > > The attribution model in reading 48 is known as > Brinson-Hood-Beebower attribution. > > > selection = (benchmark sector weight) * (portfolio > sector return - benchmark sector return) In reading 48, the selection effect (from the book, is in fact ) = (PORTFOLIO sector weight) *(portfolio sector return - benchmark sector return) Why use PORTFOLIO weight in the Global performance attribution, but BENCHMARK weight in Micro Attribution???
“Why use PORTFOLIO weight in the Global performance attribution, but BENCHMARK weight in Micro Attribution???” Because you’ve combined the interaction term into the selection term.
darkstar, if you answered the previous post, then there is a mistake in ur formula: the selection should be (PORTFOLIO SECTOR weight) * (portfolio sector return - benchmark sector return) if we are asked to evaluate the selection attribution in a global context. Am I right?
sunnycfa1 Wrote: ------------------------------------------------------- > darkstar, if you answered the previous post, then > there is a mistake in ur formula: the selection > should be (PORTFOLIO SECTOR weight) * (portfolio > sector return - benchmark sector return) if we are > asked to evaluate the selection attribution in a > global context. Am I right? Yes, you are right. I’m not sure which post you’re talking about…I just reread all my posts above and they all look correct. I did present the selection term in the Global section differently than they presented it in the source readings and Schweser (I DID NOT collapse the interaction term into selection, which is what they did and this is why the two selection terms are different) since my point was to highlight that the attribution models in reading 47 and 48 really aren’t that different from each other. From reading 47 (domestic): Selection = (portfolio sector return - benchmark sector return) * PORTFOLIO sector weight From reading 48 (global): Selection = (portfolio sector return - benchmark sector return) * BENCHMARK sector weight The reason for the difference between the two formulas is that you’ve added the interaction term into the selection term.
isn’t just supposed to the opposite? in the global context, we are using portfolio weight and in the domestic context, we are using benchmark weight?
sterling76 Wrote: ------------------------------------------------------- > Alright, this is how our Schweser instructor put > it and told us we should use on the exam: > > ATTRIBUTION, when not comparing to a benchmark, > but assigning attribution WITHIN a portfolio: > > (shown below for individual weights so no “j” > subscript; W=weight, R=return, p=portfolio, > b=benchmark, l=local ccy, d=domestic ccy) > > Mkt Return Effect: Wp ( Rbl ) > Security Selection Effect: Wp ( Rpl - Rbl ) > Currency Effect: Wp ( Rpd - Rpl ) > > Add in yield for total attribution. Note you’re > only using PORTFOLIO WEIGHTS here. Also note, our > instructor felt that the interaction effect, since > it’s generally so small anyway, is unimportant. > > ALLOCATION, when comparing to a benchmark: > > Mkt Allocation Effect: ( Wp - Wb ) Rbl > Security Selection Effect: Wp ( Rpl - Rbl ) * > Note this is the same as above > Currency Allocation Effect: Wp ( Rpd - Rpl ) - Wb > ( Rbd - Rbl ) > > Add in the benchmark DOMESTIC return and yield for > total allocation effects. > > Not only should you know the calculations, but > know what each measure implies about the > investment manager. > > Any thoughts on the above? Let me know! Sorry, I just saw your post! That looks right, although I would have used slightly different terminology. Contribution or Return Decomposition: This determines the sources of your TOTAL RETURN (this is what you labeled attribution when not comparing to a benchmark) Attribution: This determines the sources of your ACTIVE RETURN (this is what you labeled allocation when comparing to the benchmark) Technically, I don’t think that your instructor is necessarily correct about the interaction term. You can’t just “ignore” interaction - if you don’t want to see it (which many people don’t), you need to add it into the selection term like they did in the global performance attribution section. See my post above. If we get asked an attribution question on the exam, I’m going to follow the formulas EXACTLY as they provide them in the source readings…I’d hate to lose some points because I ignored a term because it was small!
darkstar Wrote: ------------------------------------------------------- > sunnycfa1 Wrote: > -------------------------------------------------- > ----- > > darkstar, if you answered the previous post, > then > > there is a mistake in ur formula: the selection > > should be (PORTFOLIO SECTOR weight) * > (portfolio > > sector return - benchmark sector return) if we > are > > asked to evaluate the selection attribution in > a > > global context. Am I right? > > Yes, you are right. > > I’m not sure which post you’re talking about…I > just reread all my posts above and they all > look correct. I did present the selection term in > the Global section differently than they presented > it in the source readings and Schweser (I DID NOT > collapse the interaction term into selection, > which is what they did and this is why the two > selection terms are different) since my point was > to highlight that the attribution models in > reading 47 and 48 really aren’t that different > from each other. > > From reading 47 (domestic): > Selection = (portfolio sector return - benchmark > sector return) * PORTFOLIO sector weight > > From reading 48 (global): > Selection = (portfolio sector return - benchmark > sector return) * BENCHMARK sector weight > > The reason for the difference between the two > formulas is that you’ve added the interaction term > into the selection term. AAAAHHHHH!!! I mixed up the formulas from reading 47 and 48. 47 --> should be BENCHMARK sector weight 48 --> should be PORTFOLIO sector weight Sorry!
Good job anyway darkstar. The crux of the matter is that interaction is combined in selection. BTW, I’ve met Nimrod
I’m just gonna take my chances that this isn’t on the test.
I can almost bet that either fixed income, global or micro attribution will be on the test this year!
skillionaire, what’s your issue with attribution? It’s mostly rote memorization (unfortunately). Is there something you don’t get?
inbead Wrote: ------------------------------------------------------- > I can almost bet that either fixed income, global > or micro attribution will be on the test this > year! Fixed Income has to be on the test as it is mandatory.
pupdawg82 Wrote: ------------------------------------------------------- > inbead Wrote: > -------------------------------------------------- > ----- > > I can almost bet that either fixed income, > global > > or micro attribution will be on the test this > > year! > > > Fixed Income has to be on the test as it is > mandatory. Fixed income will definitely be on the exam, but I think inbead meant that it might show up in an attribution-related question.
darkstar Wrote: ------------------------------------------------------- > skillionaire, what’s your issue with attribution? > It’s mostly rote memorization (unfortunately). > > Is there something you don’t get? I thought there was when I looked at (what I perceived to be yesterday) the complicated formulas and concepts above; then I got it on the Schweser 1am practice test and got almost a 100% for the three questions - it’s intuitive as hell, honestly. The combination of my not knowing what “attribution” (as it relates to the curriculum) really was (I call it the “return chart thingy”) and your long-winded (although I’m sure helpful) post to start off this thread kinda threw me for a loop. All better now, though 