Qplum Fractional Rebalancing

Fract ional
Rebalancing
www.qplum.co
Page
1620 Plaza 5 Jersey City, NJ 07311

(201) 377-2302
contact@qplum.co
(201) 604-5422
Fract ional Rebalancing

Target risk without timing the market
July 1st 2016
Rebalancing is a simple yet powerful portfolio management trick. In this paper, we show that it is better
than static portfolios in some cases, and examine the situations under which it can hurt performance.
We expand on the reasoning behind it and highlight better ways of doing it. We also look at the pitfalls
of using the traditional method of rebalancing, as observed in both back-testing as well as in actual
performance. We then propose fractional rebalancing as an alternative to periodic or threshold based
rebalancing, and demonstrate why it is much better at eliminating biases.
It is beyond doubt that thoughtful decision making leads to sound plans, and sticking to the plans leads to better results.
Planning in advance shifts the burden of decision making from the future to now. This helps because in future, one might be
prone to error in judgment in the face of crisis, whereas now one can make rational decisions and choose to stick to it no
matter what happens.
The most import ant part of port folio

management st icking t o t he plan. That
means we must remember t o rebalance
t he port folio.
Managing a portfolio is no different. It is also about making plans

and sticking to them. Planning here is coming up with a risk
mandate, which results in a portfolio allocation. The second and an
equally important part is sticking to the plan, which in this case
corresponds to rebalancing the portfolio.
Rebalancing your portfolio means bringing the allocation back to the percentages you decided to have in your plan. For this
you would need to sell the assets that have grown in value, again and again no matter how luring the prospect of holding on to
the winners. This ensures that you stick to the risk mandate you settled on at the start and the the portfolio doesn?t tilt too
much towards one type of asset.
But does rebalancing make more money?

If you look at two popular indexes, S&P500 and the Volatility Index, using the famous Mean Variance Optimization would
have only earned 7% over thirty years, while a daily-rebalanced portfolio would have earned 19%. The risk of the portfolio
would have decreased by a third too.
Does it always help?

No. One situation where it could hurt is when we are rebalancing between stocks which can have vastly different
performance. For instance if we were to rebalance a 50-50 portfolio on Lehman Brothers and Goldman Sachs daily, we would
have lost all our money.
In the rest of the paper we will tell you exactly which situations to use rebalancing in. We will also talk about a novel method
of rebalancing that in data-science terms is much better than the naive method.
Broadly speaking, rebalancing is a simple, wonderful and at the same time a very intuitive trick. Rebalancing is a must since
we absolutely need to stick to our investment plan. Quantitatively, the benefits of rebalancing are:
1.
2.
3.
4.
It helps in closer risk targeting

It makes for steadier risk taking that leads to better real returns
It prevents exposure to unexpected risks
It ensures that if we can?t make up for lost capital later on, we don?t lose it in the first place
All investments carry risk. This material is for informational purposes and should not be considered specific investment advice or recommendation to any
person or organization. Past performance is not indicative of future performance. Please visit our website for full disclaimer and terms of use.
(201) 377-2302
contact@qplum.co
(201) 604-54222

To illustrate the points above, let us look at a traditional 60/40 portfolio (60% in stocks and 40% in bonds). Anyone invested
in this portfolio would expect a 10% drop in stocks to hurt them by only 6%, given the allocation to stocks is 60%. But let us
say we never rebalance this portfolio. For someone who invested in such a strategy in October 2003, the share of equities in
her portfolio would have steadily increased till 2007, owing to the spectacular growth of US stocks in that period ( as can be
seen in exhibit-1).
Since equities are inherently riskier than bonds, the portfolio

risk was also growing along with the allocation to equities. We Rebalancing can help avoid wild swings
and keep t he allocat ion st eady, making it
would never realize the risk till right before the 2008 crisis,
when the portfolio would have been almost 80% in stocks and a bet t er choice t o a prudent invest or.
at its highest risk level till then. Now a 10% drop in stocks will
lead to a decline of 8% in portfolio value rather than an initially planned 6%. Imagine how much this would have hurt when
the stock market had lost more than 56% by March 2009.
Exhibit-1 shows the growth of stocks in the years leading to the crisis and it?s phenomenal. At such times an investor
would have felt their gut instinct tell them that stocks go up and go up a lot. There is no logical reason to take money out of
stocks and put it somewhere else!
Prices of both stocks and bonds move a lot. Often they are moving in opposite directions. In tech-speak - ?they are not
positively correlated to each other?. One of them may grow faster than other, but later face some market correction before
rising again. So if we start with 60/40 allocation and hold it without rebalancing, it is possible to find ourselves at 40/60 or
80/20 after a while, and back to 60/40 again. On the other hand, constant rebalancing will keep the allocation steady at
60/40. Both of these approaches might even result in similar net growth over a complete boom-bust cycle. But since the
former is prone to wild swings in the portfolio value, the latter makes a better choice to a prudent investor.

This cyclic nature of asset-class returns makes rebalancing a great tool for risk management. This is especially true over
longer investment horizons. In general, rebalancing works for a strategic portfolio when the assets involved satisfy these 2
important conditions.
Condition I. Assetsare uncorrelated to each other.

When the securities in a portfolio are highly correlated to each other, rebalancing just serves to shift the asset allocation
from one product to another very similar product. This way it ceases to be an effective risk management tool and only leads
to higher turnover. Rebalancing works best for a well diversified portfolio.
Condition II. There isa reasonable expectation of their survival.
In fact , when asset s exhibit mean

revert ing behavior, not only does
rebalancing help in risk t arget ing but also
in improving risk adjust ed ret urns.
Rebalancing a portfolio which had some allocation to Lehman

Brothers? stock during 2007-09 crisis would have meant
allocating more and more to it while it kept bleeding money till it
was a penny stock. Rebalancing is not ideal for portfolios which
have toxic assets. It is put to better use when the securities
involved tend to revert to their mean path in the long run.
Tradit ional approaches t o rebalancing

There are two commonly used methods for rebalancing passive portfolios - fixed period rebalancing and threshold based
rebalancing.
Fixed Period Rebalancing

Investors choose a fixed interval after which they look at their portfolio again and rebalance it to the targeted allocation.
Rebalancing period can be daily,, monthly,, or annually . Most asset managers tend to use this method, but there is a big
problem with this method, that we will illustrate next.

Threshold Based Rebalancing
In this, the investor monitors allocation regularly. Rebalancing is done whenever the portfolio diverges beyond the desired
allocation by a margin wider than the threshold. For example, if the threshold is 5% and we start with 60/40 we will rebalance
only if it crosses 55/45 or 65/35.
One can also take the hybrid approach and do both of the above.
Lying wit h st at ist ics

An investment strategy is nothing but a set of trading rules and parameters. We can try to learn optimal values of these
parameters from data. However, all we can learn is the optimal value in past data. As any quant hedge fund researcher would
testify, parameter robustness is very important. So while drawing inferences from the statistics observed, we try to ensure
that these statistics are not prone to external biases and our conclusions aren?t very sensitive to minor variations in inputs. If
proper caution is not exercised, one might end up mistaking a tree for the forest. Our only real yardstick is the performance in
future.
One such bias is the market-timing bias. In the context of analyzing a
strategy, we are referring to the dates over which we back-test the strategy.
Take the S&P500 as a strategy for example. Investments made in it in 2007
had barely recovered their principle till 2013, while the same investments
made in 2009 more than doubled.
The same st rat egy can look very

good or bad depending upon
when you st art t he back-t est .
Exhibit-3 shows the performance of a trend following futures
It is quit e surprising, how a minor

change in st art dat e can change
performance st at ist ics so drast ically.
strategy that starts from different dates in Q1 of 1995. This strategy

rebalances positions once every quarter. One would expect the final
cumulative returns of these strategies to be very close, since a period of
15 days is insignificant when compared to a testing period of 20 years.
But what we see is nowhere close to the expectation.
Int roduct
Fract
ionalion
Rebalancing
This presents a big challenge for a data-scientist in finance. If we believe that what we see as the average line here is the true
unbiased estimate of the strategy performance, the extremes give us us an idea of the extent to which statistics can be
biased. Imagine how wrong would we have been had we not conducted this experiment and just expected the strategy to
perform as good as our back-tests starting from March 16th, 1995 or Jan 1st , 1995 did.
The reason that fifteen days make so much difference lies not in what happened in those fifteen days, but rather in how the
portfolio looked on certain dates in the following 20 years. And this is where rebalancing played a major role. The choice of
the start date and the rebalancing frequency dictates in what time periods will the portfolio be most divergent from the
mandates. And chances are that market sees some major events in these periods. What this means is that a significant
variation in performance is introduced by sheer chance.
As we noted above, an average of performance seen from a range of such start dates can give us an estimate which is closest
to being unbiased. An alternative way to do that would be to assume daily rebalancing. This way the portfolio will look the
same every day and it wouldn?t make a difference when we started the simulation.
But in real life, daily rebalancing is not practical for most portfolios. While it offers the benefit of better returns through
smaller divergence from the mandate, it can also drive transaction costs through the roof. In some strategies it might even be
desirable to hold on to trades for a longer time because the trading signal arrives late.
To solve this problem, we propose an alternative approach of
portfolio rebalancing that nullifies extraneous factors like start
date. The central idea in this approach is that a portfolio can be
thought of as a combination of many smaller portfolios, each of
them a fraction of the original portfolio. We call this fractional
rebalancing and we will talk about this at length in the next
section.
So should we st ay resigned t o t he fact

t hat st at ist ics will have larger uncert aint y
or can we do somet hing bet t er?

Fractional rebalancing is a portfolio rebalancing style in which a portfolio is assumed to be made up of multiple smaller
portfolios. The number of smaller portfolios depends on the desired rebalancing frequency. Rather than rebalancing an
entire portfolio at the same time, we rebalance these fractions alternatively and in a cyclical manner.
For instance, the average portfolio in exhibit-3 can be assumed to be simply a combination of the other 6 portfolios, with
allocation of 1/6th to each. Since each of these 6 portfolios is quarterly rebalanced, its corresponding share in the average
portfolio is quarterly rebalanced too. This way the average portfolio is effectively being rebalanced every quarter as well.
While a portion of rebalancing occurs every 15 days, the average rebalancing is still equivalent to a quarterly rebalanced
portfolio.
So how does this help?It ensures that even for two different start dates, the allocation stays more or less the same on any
given day. This way:
-
we make sure that two investors who invest in the same strategy on different days of the week converge to the same
portfolio everyday
we de-couple the strategy's performance from the spurious parameter of the start date and eliminate the possibility
of over-fitting on it
Int roduct
Fract
ionalion
Rebalancing
Exhibit-4 further illustrates this through a portfolio represented as a pie chart. On Monday, the green shaded slice which is
one-fifth of the portfolio is rebalanced. This portion is expected to remain untouched on the next 4 days. Similarly, each slice
is rebalanced only on its designated days. Overall, all 5 portions are rebalanced over a week. This is an example of a
fractionally rebalancing portfolio, which is practically an average portfolio of 5 simple periodically rebalanced portfolios. Each
of these 5 portfolios can be said to have begun on one of five successive days, and then rebalanced every fifth day.
Exh ibit 4: Sch em at ic r epr esen t at ion of f r act ion al r ebalan cin g
It is obvious from this schematic that the portfolio looks equally close to the mandate on every day of the week. If this were a
simple quarterly balanced portfolio it would have been closest to the mandate on the day of rebalancing, and furthest from
the mandate on the day just before it.
The same concept can be expanded to create a fractional portfolio for each day. Combination of all these fractional portfolios
is the total portfolio. Every day, a small part of the total portfolio will be rebalanced and each of the fractional portfolios will
be rebalanced once a cycle on its designated day. This way the entire portfolio is rebalanced over a cycle
Int roduct
Fract
ionalion
Rebalancing
Conclusion
Rebalancing works wonders when it is used in the right way. A portfolio of ETFs or safe stocks, or one that employs risk
management aimed at cutting huge losses, is much better off with rebalancing than without it.
Fractional rebalancing is a good alternative to traditional time based rebalancing methods. It effectively mitigates problem of
extraneous biases like date of rebalancing that creep in when we use periodic rebalancing. As robust back-testing is essential
for strategy development, it is definitely helpful when comparing statistics. Further, it allows us to place smaller orders
spread out over the rebalancing cycle. This helps us in avoiding the costs associated with price movements caused by huge
orders.
It is worth noting that implementation of fractional rebalancing is only possible with some advanced, execution sensitive,
algorithms and that is biggest reason why many asset managers are unable to take advantage of this powerful method.
The scope of this paper was intentionally limited to passive investment strategies. In our future work we will talk about the
best way to rebalance alpha strategies.
References
1. Colleen M. Jaconetti, Francis M. Kinniry Jr., Yan Zilbering, 2010, Best practices for portfolio rebalancing, Vanguard
Research
2. Pliska, Stanley R., and Kiyoshi Suzuki, 2004, Optimal Tracking for Asset Allocation With Fixed and Proportional
Transaction Costs. Quantitative Finance 4(2): 233?43
3. Grinold, R., and Kahn, R., 1995, Active Portfolio Management: Quantitative Theory and Applications
4. Nick Granger, Douglas Greenig, Campbell Harvey, Sandy Rattray, David Zou, 2014, The unexpected costs of
rebalancing and how to address them
5. Sheng Wang, Yin Luo, Miguel-A Alvarez, Javed Jussa, Allen Wang, Gaurav Rohal, 2014, Seven Sins of Quantitative
Investing

Qplum Fractional Rebalancing

Uploaded by

Copyright:

Available Formats

Qplum Fractional Rebalancing

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Qplum Fractional Rebalancing

Uploaded by

Copyright:

Available Formats

Fract ional

1620 Plaza 5 Jersey City, NJ 07311

Fract ional Rebalancing

July 1st 2016

The most import ant part of port folio

Managing a portfolio is no different. It is also about making plans

But does rebalancing make more money?

Does it always help?

It helps in closer risk targeting

Fract ional Rebalancing

Since equities are inherently riskier than bonds, the portfolio

Fract ional Rebalancing

Condition I. Assetsare uncorrelated to each other.

Condition II. There isa reasonable expectation of their survival.

In fact , when asset s exhibit mean

Rebalancing a portfolio which had some allocation to Lehman

Tradit ional approaches t o rebalancing

Fixed Period Rebalancing

Fract ional Rebalancing

Lying wit h st at ist ics

The same st rat egy can look very

Exhibit-3 shows the performance of a trend following futures

It is quit e surprising, how a minor

strategy that starts from different dates in Q1 of 1995. This strategy

So should we st ay resigned t o t he fact

Fract ional Rebalancing

You might also like