I would normally think that the transaction costs and implementation costs of an index replication strategy would legitmately be considered contributors to tracking error of a strategy. Not so the CFAI text, it seems (unless someone can find part of the text that disagrees with me). Questions 16 and 17 (page 259 of Volume 4, Reading 33): 16. Which of the following index strategies [for replicating the MSCI All Countries World Index] is MOST appropriate given the investment committee’s goal [of minimizing] tracking error? [size of portfolio = $2mm] A. Optimization B. Style Indexing C. Full Replication Now… at first glance I went with “C. Full Replication”, but then I considered the fact that full replication of the all countries index would be pretty darn inefficient, especially given the portfolio size of $2mm. These inefficiencies would of course contribute greatly to tracking error. Therefore, I went with “A. Optimization” However, the text goes with “C. Full replication” I probably would have just moved on (and certainly wouldn’t have posted here), but take a look at the immediately following question in the text: Which of the following is MOST LIKELY to have the highest investment cost associated with meeting the investment committee’s tracking error goal? A. MSCI ACWI ETF B. MSCI ACWI mutual fund C. Full replication of the index And, naturally, the text indicates that “C” would present the greatest investment cost. Does anyone find this just slightly contradictory? Or am I just being picky because I woke up at 2:30 AM in order to study before work?
well, your thinking is correct, full replication is difficult in this case and shouldn’t be done but as CFAI puts it , client constraints comes first in any given circumstances no matter what you think even if its not cost effective. so i would have chosen A but CFA answer is CFA answer, hence i mean we could argue with the answer. and given only constraint of minimizing tracking error, answer comes to C
bdeora Wrote: > > and given only constraint of minimizing tracking > error, answer comes to C Well, I guess my point was that if transaction costs significantly and negatively impact tracking error, then “C” would NOT be the right answer. However, in the clarity of day-light, I realize (I think) that the answer to my question in the thread title is “No.” Tracking error is measured as the variation of the investment strategy about the index or benchmark. Transaction costs are a consistent drag on the strategy that introduce a consistent negative bias, but would probably not introduce much actual variation (up and down) versus the benchmark. So I guess, strictly speaking, the transaction costs are not an appropriate consideration if minimization of tracking error is your only goal. Is this how I should be thinking about this?
plyon: I would always choose C in both cases. Here is why? for the 1st one, the goal is to minimize the tracking error. You have to have a good definition of what tracking error is. It simply arrive because you cannot replicate the index 100%. If you could 100% replicate the index, your tackink error is zero. Please take a look at exibit 3 of the reading of " Passive Equity Investing" and you willget it. THe second question has to do with cost. The only reason that you do not want to practice full replication when the universe of selection is large is because it is very costly in money, time and liquidity. I think the reason that you did not get them right away is because you are a deep thinker . This is what confuse me about this exam, sometime, i feel if you put all of your knwoledge, you may loss a question and sometime if you do not put everything, you lose. Therefore, I have mix deeling about this test. I hate CFAI!!!
plyon Wrote: > However, in the clarity of day-light, I realize (I > think) that the answer to my question in the > thread title is “No.” Tracking error is measured > as the variation of the investment strategy about > the index or benchmark. Transaction costs are a > consistent drag on the strategy that introduce a > consistent negative bias, but would probably not > introduce much actual variation (up and down) > versus the benchmark. So I guess, strictly > speaking, the transaction costs are not an > appropriate consideration if minimization of > tracking error is your only goal. > > Is this how I should be thinking about this? this is how I am thinking about it, plyon. btw, no doing questions at 2:30 in the morning in February, man. 
good explanation, plyon!
plyon Wrote: ------------------------------------------------------- > bdeora Wrote: > > > > > and given only constraint of minimizing > tracking > > error, answer comes to C > > Well, I guess my point was that if transaction > costs significantly and negatively impact tracking > error, then “C” would NOT be the right answer. > > However, in the clarity of day-light, I realize (I > think) that the answer to my question in the > thread title is “No.” Tracking error is measured > as the variation of the investment strategy about > the index or benchmark. Transaction costs are a > consistent drag on the strategy that introduce a > consistent negative bias, but would probably not > introduce much actual variation (up and down) > versus the benchmark. So I guess, strictly > speaking, the transaction costs are not an > appropriate consideration if minimization of > tracking error is your only goal. > > Is this how I should be thinking about this? I told you that either had some vodka and too smart.
plyon – tracking error is measured traditionally with GROSS returns not NET. so they wouldnt take into account the transaction costs.
mike0021 Wrote: ------------------------------------------------------- > plyon – tracking error is measured traditionally > with GROSS returns not NET. so they wouldnt take > into account the transaction costs. I’ve never seen an asset manager who can (or would) break out transaction costs. Tracking error is measured gross of FEES (and maybe gross of commissions… I don’t know about that). But not gross of all transaction costs.