# Performance Attribution - in the real world

hi guys, so i’m trying to do some performance attribution for one of our accounts and i am calculating all three effects, allocation, selection and interaction. however, it doesn’t seem to add up to the difference between the one year return and the one year benchmark’s return. Can someone help me see why?

Global portfolio and is exchange rate a factor here?

what application are you using? did you look at the cash, other, and unassigned buckets?

there might be excess cash that’s not somehow not accounted for. you should give that cash to me. problem solved.

numi Wrote: ------------------------------------------------------- > there might be excess cash that’s not somehow not > accounted for. you should give that cash to me. > problem solved. hhaaha

jeks Wrote: ------------------------------------------------------- > hi guys, > so i’m trying to do some performance attribution > for one of our accounts and i am calculating all > three effects, allocation, selection and > interaction. however, it doesn’t seem to add up > to the difference between the one year return and > the one year benchmark’s return. Can someone help > me see why? compounding might be part of the issue

janet, ill help you if you want: you can email me at mooch22 at gmail.com i created an attribution excel file that does this… it’s fairly simple, so you can just compare my sheet w/ yours.

It’s not quite clear what type of returns you’re working with, but if you’re using e.g. monthly values, then you can’t just simply add or compound each attribute over the year to arrive at the yearly contribution. You have to compound each period’s attribute contribution with the benchmark’s return in the following period and you have to compound the following period’s attribute with the portfolio’s return from the previous period…or something like that. However, usually when I see attribution reports “in the real world”, I find that they are quite different and far more complex from the basic theory taught in the curriculum.

Janet, For multi-period attribution, you’ll need to apply some algorithm in order for the total effect and benchmark excess return to match. Here’s a paper that delves into the issue more. http://www.edhec-risk.com/research_news/choice/RISKReview1095940690711932922

i basically used an excel spreadsheet. there’s no exchange rate factors to worry about. i’m actually just an assistant PM right now so seriously don’t usually delve into this kind of stuff but i did just finish the Level 3 so thought it would be a piece of cake. The only numbers i used were the annual returns of the portfolio as a whole (ex. portfolio returned 11% at 12/31), the benchmark as a whole (ex benchmark return which consisted of 5% Dex t-bill, 10% Dex Universe, etc. etc. returned 9.4%), and then the annual returns of each component and the benchmark components and of course the percentages. Cape Cobra, i think that might be my problem, maybe i have to do multi period analysis…maybe I was over simplifying things. i read somewhere about timing effect. i wonder if that comes into play as the difference between selection/allocation/interaction and the total value added. i will see if i might be able to paste the excel sheet into here tomorrow morning. i am overwhelmed at the response though. Thanks everybody.