Are those two the same thing? I want to calculate ex ante tracking error of a fund, however I believe that I need the variance-covariance matrices of all the stocks in the portfolio. Let's assume that I do this now with the current portfolio, a week later stocks will be different. I'm quite confused about ex-ante tracking error. For example I calculate an ex-ante TE of 5% for a fund, then put this as a limit for the ex-post TE. How reliable is that and would that mean the same thing as Relative VaR?Any help on the subject would be appreciated.I couldn't find any concrete information about that, excel files or so on in the internet...Thanks a lot in advance

could you explain where you came across the term "relative VaR"? it reminds of hours of senseless discussions with textbook educated auditors and financial regulators....are you managing insurance portfolios and are subject to some kind of regulatory reporting requirement?in any case, relative var could also mean VaR(Portfolio)/VaR(Index). This says nothing about the TE, but (supposedly) gives some kind of indication of how much more volatility you are taking vs your benchmark. it is a stupid measure that only regulators can want (please tell me)about ex ante and ex port TE, first of all, the portfolios have to be static, meaning no change in the positions. if you reshuffle the portfolio, any ex-ante/ ex-post comparison becomes meaningless.ex-post TE is problematic for other reasons, for example, if you take daily obersvations you absolutely must make sure that the index and the portfolio are revalued at the same time. this is not always the case!correlations can change etc etc etcforecasting TE is only useful within a factor based risk framework that is simultaneously used in the investment decision making process. because then the TE can be broken down into its constituents and give the portfolio manager a sense of where he is taking the most risk versus his index.

First of all thanks a lot for your reply.I work at asset management and manage mutual funds, however I am not very experienced about risk. I wanted to do a study for our funds and calculate their ex ante tracking errors, however since we change positions almost daily and even change our portfolio almost weekly it seemed a bit meaningless to me to do that. Nevetheless, to put some tracking error limits for our funds I guess an ex ante tracking error study could be useful. I just didn't know where to start.http://www.financial-risk-manager.com/r ... var_b.html , in this link I could find some information. Since our funds have benchmarks, Relative VaR seemed a good measure to me. Do you think it is still irrelevant even for funds with benchmarks?I am not subject to regulatory requirement to calculate Relative Var but during a recent visit to our group headquarters Head Of Investment Risk told me that Relative VaR would be a better measure than VaR for funds with benchmarks and that I need to calculate ex ante tracking error. We didn't discuss a lot and maybe I got him totally wrong but I started to think that those two were the same idea.As far as the ex post TE is concerned you are totally right, we tend to have high TEs since the stocks in the funds are valued with weighted average prices while the index is valued with close prices. TE tends to be higher than it should be but since ex post TE is a group requirement we still calculate it..I think I'm going to start valuing the index with weighted average prices to sort this out...You wrote "forecasting TE is only useful within a factor based risk framework that is simultaneously used in the investment decision making process. because then the TE can be broken down into its constituents and give the portfolio manager a sense of where he is taking the most risk versus his index"Could explain this a bit more if you don't mind? And also could you recommend any practical sources about how I can make an ex ante TE study?Thanks a lot.QuoteOriginally posted by: Gmike2000could you explain where you came across the term "relative VaR"? it reminds of hours of senseless discussions with textbook educated auditors and financial regulators....are you managing insurance portfolios and are subject to some kind of regulatory reporting requirement?in any case, relative var could also mean VaR(Portfolio)/VaR(Index). This says nothing about the TE, but (supposedly) gives some kind of indication of how much more volatility you are taking vs your benchmark. it is a stupid measure that only regulators can want (please tell me)about ex ante and ex port TE, first of all, the portfolios have to be static, meaning no change in the positions. if you reshuffle the portfolio, any ex-ante/ ex-post comparison becomes meaningless.ex-post TE is problematic for other reasons, for example, if you take daily obersvations you absolutely must make sure that the index and the portfolio are revalued at the same time. this is not always the case!correlations can change etc etc etcforecasting TE is only useful within a factor based risk framework that is simultaneously used in the investment decision making process. because then the TE can be broken down into its constituents and give the portfolio manager a sense of where he is taking the most risk versus his index.

The software available for computing TE of equity portfolios (from vendors such as B@rra or Northfield Information Services) works a little bit like this, in my experience:They have a "risk model" which is basically a table that gives for each stock ticker, the exposure of that stock to a few "factors" and the residual risk. Let's say, for each ticker they have a vector of 5 or 6 numbers; and they have 1500 tickers. You pay them $$$ and they re-estimate this table every once in a while (weekly?) and send the update to you so it is always fresh.(In fact they have a choice of risk models, usually, some with macroeconomic factors, some with industry factors, US only factors or global and you choose the risk model you want).They have an application that lets you enter the composition of your portfolio every day and computes the factor exposures of your portfolio, of the benchmark, and the tracking error vs the benchmark. The TE if I recall correctly is usually expressed as a standard deviation, rather than a VaR. Even more useful than the TE is the tools that allow you to diagnose a TE problem (i.e. "your tracking error is too high because you have too much exposure to oil stocks", "if you sold blabla shares of XOM your TE would go down to xyx", etc.).As you can see, the idea is pretty simple, but to develop such a thing from scratch yourself could be a lot of work (which is what keeps the above mentioned companies in business).