[R-SIG-Finance] R-SIG-Finance Digest, Vol 190, Issue 1

[Vivek Rao]

Suppose an actively-managed fund is benchmarked to the S&P 500. I would expect that
the weights of stocks gradually converge to the market cap weights. So you could 
simulate deviations of individual weights from a benchmark as following an AR(1) process
with an AR(1) coefficient being slightly less than 1. For a multivariate model of all stock
weight deviations, you could simulate from a VAR(1) model where the matrix of autoregressive
coefficients is diagonal. This model does ignore the fact that many portfolios have exactly
zero weights in many stocks and do not short stocks. It predicts that managers trade small
amounts of all stocks every day, but in reality, active discretionary managers make large
trades in a few stocks. If managers tend to trade in the same direction over several days, 
a higher-order VAR model could be used. As others said, the model really depends on
the investment process used.

Vivek Rao


On Sunday, March 1, 2020, 08:45:51 AM EST, G Mac <mcewan.gareth using gmail.com> wrote: 





Hi Brian, All

Forecasting or a way to simulate next period's weights, based on a series
of past period's weights, is more accurate of what I'd like to do. I
realize the tough ask, but curious if there is a package that could be used
for this.

Any directions, in the world of R packages or statistical methods would be
very much appreciated.

Thanks
Gareth

On Sun, Mar 1, 2020, 6:00 AM <r-sig-finance-request using r-project.org> wrote:

> Send R-SIG-Finance mailing list submissions to
>        r-sig-finance using r-project.org
>
> To subscribe or unsubscribe via the World Wide Web, visit
>        https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> or, via email, send a message with subject or body 'help' to
>        r-sig-finance-request using r-project.org
>
> You can reach the person managing the list at
>        r-sig-finance-owner using r-project.org
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of R-SIG-Finance digest..."
>
>
> Today's Topics:
>
>    1. Portfolio Composition Forecasting (G Mac)
>    2. Re: Portfolio Composition Forecasting (Brian G. Peterson)
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sat, 29 Feb 2020 09:06:03 -0500
> From: G Mac <mcewan.gareth using gmail.com>
> To: r-sig-finance using r-project.org
> Subject: [R-SIG-Finance] Portfolio Composition Forecasting
> Message-ID:
>        <
> CALGfFMYGMnKfz4rJMsVOLm6HmbNcn72O+Kie32HbELYKoVAkQA using mail.gmail.com>
> Content-Type: text/plain; charset="utf-8"
>
> Hi All
>
> I was hoping someone could point me in the right direction.
>
> Is there an R-package (or other software) that can be used to forecast the
> next period's portfolio composition?  There are many portfolio optimization
> packages, but this is not the same question.  Say I take the past* x*
> periods, each period holds the percentage composition of an investment
> portfolio (sums to 1); the composition of assets will contain key assets
> held (or increased/decreased) through periods, but new assets will be added
> to the portfolio over time, while some holdings will be dropped, so we will
> have nuisance here.  I would like to model the past* x* periods, accept
> this mentioned error, and forecast or simulate for* x+1* period.
>
> Does anyone have any experience with this, or have any pointers within the
> broader domain of statistics?
>
> Many thanks in advance,
> Gareth
>
>        [[alternative HTML version deleted]]
>
>
>
>
> ------------------------------
>
> Message: 2
> Date: Sat, 29 Feb 2020 17:41:02 -0600
> From: "Brian G. Peterson" <brian using braverock.com>
> To: G Mac <mcewan.gareth using gmail.com>, r-sig-finance using r-project.org
> Subject: Re: [R-SIG-Finance] Portfolio Composition Forecasting
> Message-ID:
>        <0cd5fe99a1dde7d3a427fbf4e5e487b280f05ae6.camel using braverock.com>
> Content-Type: text/plain; charset="utf-8"
>
> On Sat, 2020-02-29 at 09:06 -0500, G Mac wrote:
> > Is there an R-package (or other software) that can be used to
> > forecast the next period's portfolio composition?  There are many
> > portfolio optimization packages, but this is not the same
> > question.  Say I take the past* x* periods, each period holds the
> > percentage composition of an investment portfolio (sums to 1); the
> > composition of assets will contain key assets held (or
> > increased/decreased) through periods, but new assets will be
> > added to the portfolio over time, while some holdings will be
> > dropped, so we will have nuisance here.  I would like to model the
> > past* x* periods, accept this mentioned error, and forecast or
> > simulate for* x+1* period.
> >
> > Does anyone have any experience with this, or have any pointers
> > within the broader domain of statistics?
> >
>
> Itr seems to me tomorrows portfolio is the same as today's portfolio
> except for organic change in weights caused by market price
> fluctuations, or by a rebalancing event.  The 'forecast' is the
> standard naive forecast: today's portfolio will still be held tomorrow,
> unless you rebalance.
>
> I don't see any value in a simulation from the prior holdings.
> Portfolios are rebalanced for some business reason, and those reasons
> are usually pretty well understood, and not the result of a random draw
> from some distribution of prior holdings.
>
> What am I missing?
>
> Regards,
>
> Brian
>
>
>
>
> ------------------------------
>
> Subject: Digest Footer
>
> _______________________________________________
> R-SIG-Finance mailing list
> R-SIG-Finance using r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>
>
> ------------------------------
>
> End of R-SIG-Finance Digest, Vol 190, Issue 1
> *********************************************
>

    [[alternative HTML version deleted]]

_______________________________________________
R-SIG-Finance using r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.