[R-SIG-Finance] making sense of 100's of funds

Brian G. Peterson brian at braverock.com
Thu Aug 16 14:35:02 CEST 2007


CalculateReturns works from price series.  That's what it's for.  The 
original poster (Paul) was looking for a calculation to take a price 
series of a mutual fund and turn it into a geometric return series.

In most cases, it is not easy to get historical interest, dividend, and 
capital gains distributions information on mutual funds in a form that 
is useful for analysis (i.e. not on a paper statement), especially of 
hundreds of funds.

It would not be too difficult to add distribution calculations to the 
CalculateReturns function, but users would need to be educated that when 
distributions are included, the price series must be turned into a 
position value (NAV) series or the distributions must be scaled to a 
per-share basis.  In general, I think it is better in practice to use 
this kind of analysis with positions actually held, and to apply a time 
and cash flow weighted return calculation as you've suggested.  In 
practice, because of data availability, the closest proxy for the 
distributions you mention is often to use a total return series on the 
index, which are generally more easily obtainable from 
Ibbotson/Morningstar or similar locations.


   - Brian

Tobias Muhlhofer wrote:
> Paul,
> Sorry for the delay in my reply.
> Yes you are looking for funds that produce intercepts that are as
> strongly positive as possible when their fund returns are regressed on
> the returns to the benchmark. Remember to consider standard errors!
> Since we are looking at a risk-adjusted return measure, only a fund
> which has a *statistically significantly* positive intercept has a
> positive intercept. Sorry if this is obvious to you, but many people
> miss it.
> For choice of benchmark, the ASX SP200 strikes me as a very good
> starting point.
> In terms of calculating returns, CalculateReturns (which I have never
> used, but just looked at the manual) does not seem to take into account
> distributions, which is imprecise. The total one-period return at time t is:
> r_t = [Price_(t) + Dividends_(t) + CapGainsDist_(t) - Price_(t-1)] /
> Price_(t-1)
> Here Price_(t) is the price at time t, Dividends_(t) and
> CapGainsDist_(t) are respectively the dividends and capital gains
> distributions made by the fund between t-1 and t. The distributions may
> be substantial, so they will make a difference. If you use "adjusted
> closing" prices from Yahoo Finance, you are OK, as these are adjusted
> for exactly that (and splits).
> Toby
> paul sorenson wrote:
>> Toby,
>> Thanks for the tips.  I am somewhat of a cynic also. When it is my own
>> retirement fund, academic meets real-life in a fairly personal way!  I
>> have heard figures that something like 98% (from memory) of Australian's
>> "don't understand superannuation".
>> I am trying to get myself well into the 2%.  Writing R code leveraging
>> some of the great packages out there is just a method of learning which
>> I find works for me.  It can be slow going at times though.
>> I have picked out 8 funds to crunch through and using the ASX SP200 as
>> the benchmark for exercising my code.  My "practice" data set is at
>> http://www.metrak.com/tmp/exitprices.csv and I retrieved the SP200 using
>> get.hist.quote("^AXJO", start="2002-01-01", quote="Close", retclass="zoo").
>> I have used CalculateReturns from PerformanceAnalytics to create returns
>> as zoo objects so hopefully I will be able to calculate the alphas.  If
>> I understand your comment below, I am looking for a more positive
>> intercept on my choice of fund compared with the benchmark?
>> cheers
>> Tobias Muhlhofer wrote:
>>> Paul,
>>> Unless you are looking at index funds, you need to see whether your
>>> funds produce alpha. To do this, pick a set of benchmarks according to
>>> your fund's style and investment strategy, like Morningstar category
>>> index or something like that (or perhaps just the general stock market
>>> plus the two Fama-French factors), regress the fund's returns on the
>>> benchmark returns, and see whether you have a significantly positive
>>> intercept after fees. This is the best way of measuring systematic-risk
>>> adjusted returns.
>>> Being a finance academic (and therefore a cynic), and judging from my
>>> own research, if benchmarked correctly, very few fund managers generate
>>> positively significant alphas, and so I personally buy index funds for
>>> whatever style I want to invest in, and there I choose the one with the
>>> lowest expense ratio.
>>> Best,
>>>     Toby

More information about the R-SIG-Finance mailing list