Monthly Rebalancing Strategies with Fees

Question

I have a question regarding PerformanceAnalytics and Quantstrat packages. I want to test a monthly rebalancing portfolio strategy, but I want to incorporate the effect of annual fees and buy&sell fees for rebalancing.

Is there any practical way of doing this? In PerformanceAnalytics, I can easily test the monthly rebalancing strategy, but fees are problematic. In Quantstrat, I can easily incorporate fees, but is there a way to construct a monthly rebalance structure?

Thank you very much,

Kind regards.

Answer 1

In case an alternative package is also acceptable: here is how it could be done with the PMwR package.

Start with example data: 30 industry series from Kenneth French's website.

library("NMOF")
library("PMwR")
library("datetimeutils")
library("zoo")
P <- French(tempdir(),
            "30_Industry_Portfolios_daily_CSV.zip",
            price.series = TRUE, na.rm = TRUE)
P <- zoo(P, as.Date(row.names(P)))
P <- window(P, start = as.Date("2000-1-1"))

Now we compute the dates at which transactions take place.

eoy <- nth_day(index(P), period = "year", n = "last")
## [1] "2000-12-29" "2001-12-31" "2002-12-31" "2003-12-31"
## [5] "2004-12-31" "2005-12-30" "2006-12-29" "2007-12-31"
## [9] "2008-12-31" "2009-12-31" "2010-12-31" "2011-12-30"
## ....

eom <- nth_day(index(P), period = "month", n = "last")
## [1] "2000-01-31" "2000-02-29" "2000-03-31" "2000-04-28"
## [5] "2000-05-31" "2000-06-30" "2000-07-31" "2000-08-31"
## [9] "2000-09-29" "2000-10-31" "2000-11-30" "2000-12-29"
## ....

For the backtest, I'll use function btest in the PMwR package. It takes as input a function signal which evaluates to the target portfolio. Suppose we wanted an equal-weights portfolio; then signal could look as follows:

signal <- function() {
    k <- ncol(Close())
    rep(1/k, k)
}

btest also accepts an argument cashflow that can be used to add/subtract arbitrary amounts to the cash balance. The following function deducts 10% of the total wealth at every end of year (it's just an example).

cf <- function() {
    if (Timestamp(0) %in% eoy)
        -Wealth()*0.1
    else
        0
}

This would take care of the annual fee. The easiest way to include trading fees proportional to the amount traded is via the separate argument tc. It remains to run the backtest:

bt <- btest(list(coredata(P)),
            timestamp = index(P),
            signal = signal,
            do.signal = eom,
            cashflow = cf,
            tc = 0.0025,  ## 0.25% transaction fees 
            convert.weights = TRUE,
            initial.cash = 100)
summary(bt)
plot(bt)

## ---------------------------------------------------------
## 03 Jan 2000 ==> 30 Oct 2020   (5242 data points, 0 NAs)
##         100         55.2453
## ---------------------------------------------------------
## High                 110.97  (28 Dec 2000)
## Low                   35.32  (09 Mar 2009)
## ---------------------------------------------------------
## Return (%)             -2.8  (annualised)
## ---------------------------------------------------------
## Max. drawdown (%)      68.2
## _ peak               110.97  (28 Dec 2000)
## _ trough              35.32  (09 Mar 2009)
## _ recovery                   (NA)
## _ underwater now (%)   50.2
## ---------------------------------------------------------
## Volatility (%)         19.1  (annualised)
## _ upside               12.1
## _ downside             14.7
## ---------------------------------------------------------
## 
## Monthly returns  ▁▁▃▇█▂▁ 
## 
##        Jan   Feb   Mar  Apr  May  Jun   Jul  Aug   Sep   Oct  Nov   Dec   YTD
## 2000   0.0  -3.6   8.0 -0.2 -0.8  0.0   0.3  6.0  -2.3   1.4 -3.8  -4.7  -0.7
## 2001   2.8  -3.5  -4.8  7.5  1.8 -3.1  -1.3 -3.0 -11.1   3.6  6.6  -7.1 -12.5

(Disclosure: I am the maintainer of packages PMwR, NMOF and datetimeutils.)