Adding transaction fees to Return.portfolio

Thanks, Ilya.  I assume then I would have to subtract the custom
transaction costs from the  relevant ?return? list object entries to
generate a cost-adjusted portfolio return for each period.  Is that right?



*Robert C Wages*

USA Mobile: +1 717 618 2828

robert at rwages.com



*From:* Ilya Kipnis [mailto:ilya.kipnis at gmail.com]
*Sent:* Tuesday, May 10, 2016 3:53 PM
*To:* Robert Wages <robert at rwages.com>
*Cc:* r-sig-finance at r-project.org
*Subject:* Re: [R-SIG-Finance] Adding transaction fees to Return.portfolio



Essentially, given a change in weights over any period, you can apply your
own custom transaction costs to the difference in weights.



On Tue, May 10, 2016 at 3:52 PM, Ilya Kipnis <ilya.kipnis at gmail.com>
wrote:

Robert,

There's actually a way to compute turnover with Return.portfolio, with
verbose = TRUE. It returns end of period weights, and beginning of period
weights, so the turnover can be computed as beginWeights - lag(endWeights).

-Ilya



On Tue, May 10, 2016 at 3:50 PM, Robert Wages <robert at rwages.com> wrote:

Return.portfolio is an easy function to use to calculate the aggregate
performance of a portfolio that is rebalanced on a regular basis or that
has
a custom weighting scheme on the same periodicity of returns.  I cannot
find
any provision in Return.portfolio, however, for incorporating the
transaction costs for the rebalancing in the portfolio.  The resulting
portfolio performance would  seem to be high to the extent of the
transaction costs.



On the other hand, running strategies in Quantstrat on individual
securities
(or groups of securities) allows one to take into account estimated trading
costs to generate somewhat realistic performance estimates.  Quantstrat,
however, does not seem to have any readily accessible functions to balance
a
portfolio of securities within a portfolio limit.  Previous related
discussions seem to suggest the only solution is to write custom order
sizing functions to include in the strategy's rules.



I may be able to figure how to write custom functions in Quantstrat to do
this after a lot of brain damage, or alternatively my own version of
Return.portfolio, but I thought I should check with the community to see if
there are any better suggestions for this.  Given that many investment
strategies rely on portfolio rebalancing under some set of rules, as
opposed
to buying and selling individual securities on the basis of rules, I would
be surprised if I were the first person to wonder about this.



Robert C Wages

USA Mobile: +1 717 618 2828

robert at rwages.com <mailto:robert at rwages.com>




        [[alternative HTML version deleted]]