See discussions, stats, and author profiles for this publication at: [Link]

net/publication/228258829

Fundamental Indexation: Rebalancing Assumptions and Performance

Article in The Journal of Index Investing · August 2010

DOI: 10.3905/jii.2010.1.2.082

CITATIONS READS

16 793

3 authors, including:

David Blitz Pim Van Vliet

89 PUBLICATIONS 1,574 CITATIONS

Erasmus University Rotterdam
55 PUBLICATIONS 1,242 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Pim Van Vliet on 16 May 2014.

The user has requested enhancement of the downloaded file.

 Fundamental Indexation:
Rebalancing Assumptions and Performance

David Blitz, Bart van der Grient, Pim van Vliet *

ABSTRACT

We show that the performance of a fundamental index with annual rebalancing,

as proposed by Arnott, Hsu and Moore (2005), can be highly sensitive to the
subjective choice of when to rebalance. For the year 2009, for example, we find
that a fundamental index rebalanced every March outperformed the
capitalization-weighted index by over 10%, whereas a fundamental index
rebalanced every September underperformed. We provide intuitive and statistical
evidence in support of the hypothesis that if two fundamental indexes diverge,
they do not tend to mean-revert subsequently, i.e. the gap is likely to be
permanent. This performance ambiguity is an undesirable feature for an index
which is used for benchmarking purposes. We introduce the idea of blending
multiple underlying fundamental indexes, each one rebalanced annually, but at
different dates, as an example of how to construct a more robust fundamental
index without increasing turnover.

First version: March 2010

This version: August 2010

* Blitz, van der Grient and van Vliet are at the Quantitative Strategies team of Robeco
Asset Management. E-mail addresses are: [Link]@[Link] (corresponding
author), [Link]@[Link] and [Link]@[Link].

Electronic copy available at: [Link]

1. INTRODUCTION
In a traditional stock market index, the weight of each stock is proportional to its
market capitalization. As an alternative to this conventional approach, Arnott, Hsu
and Moore (2005) introduce the concept of a fundamental index, in which the
weight of each stock is set in proportion to its fundamentals, such as book value,
sales, cash-flow and dividends. These authors empirically show that fundamental
indexes outperform the capitalization-weighted index by around 2% per annum,
while risk, in terms of for example volatility and maximum drawdowns, is similar.
Their proposed fundamental index is rebalanced once a year at the end of
December. The authors note that more frequent rebalancing increases turnover
but not average returns. The real-life FTSE/RAFI indexes, which are based on
the fundamental indexation concept as outlined in the paper, are also rebalanced
once per year. The date chosen in this instance, however, is the third Friday of
March.1
The existing literature is silent on whether rebalancing assumptions
matter, thereby implicitly suggesting that these are not of major concern. We
contribute to the literature by showing that the performance of fundamental
indexes in a given year can, in fact, be highly sensitive to the choice of the
annual rebalancing moment. For the year 2009, for example, we find return
differences exceeding 10% for fundamental indexes with different annual
rebalancing dates. Furthermore, performance differences between fundamental
indexes do not exhibit mean-reversion, implying that divergences are effectively
permanent wealth shocks for long-term investors. Based on these findings we
conclude that the choice of the annual rebalancing date of a fundamental index
introduces a significant degree of arbitrariness that is undesirable for an objective
benchmark index. This arbitrariness arises because fundamental index weights

1
Quoting from sections 5.1 and 5.2 from the 1.17 version (Nov 2009) of the
methodology for the management of FTSE RAFI index series: “The FTSE RAFI Index
Series will be reviewed annually based on data as at the close of business on the last
trading day of February, taking into account any additions and deletions planned in the
underlying indices. Changes arising from the annual review will be implemented after the
close of the index calculation on the third Friday of March each year.” Source:
[Link]

Electronic copy available at: [Link]

are allowed to deviate from fundamental values in between rebalancing
moments. We introduce the idea of blending multiple underlying fundamental
indexes, each one rebalanced annually, but at different dates, as an example of
an index which is more representative for the general concept of fundamental
indexing, without resulting in increased turnover.
Various papers have shown that the performance of fundamental
indexation is robust and similar when applied to international equity markets.2
Much of the literature on fundamental indexation focuses on the source of
outperformance. The proponents of fundamental indexation argue that
fundamental indexes are a new breed of passive indexes, which by design invest
less in ‘overvalued’ stocks and more in ‘undervalued’ stocks.3 Others have
argued that capitalization weighting does not, by itself, create a performance
drag4 and that fundamental indexation is essentially an active strategy, which
mainly profits from loading on the classic value premium5. The results in this
paper can be interpreted as supportive for the view that fundamental indexation
is an active value strategy. The sensitivity of fundamental index returns to the
subjectively chosen rebalancing moment arises because rebalancing can
dramatically change the value exposure of a fundamental index. Although a
fundamental index is passive in the sense of being entirely mechanical (i.e. rules-
based), it behaves more like an active strategy in the way subjective timing
decisions can alter the profile of the strategy and thereby make or break the
return in a given year.

2. DATA AND METHODOLOGY

In order to construct a fundamental index we consider the same four
fundamentals as in the FTSE/RAFI indexes, namely book value, sales, cash-flow
and dividends. We follow Arnott, Hsu and Moore (2005) by taking the past five-

2
See for example Chen, Chen and Basset (2007), Estrada (2008), Hemminki and
Puttonen (2008) and Houwer and Plantinga (2009).
3
See for example Arnott, Hsu and Moore (2005), Treynor (2005, 2008), Arnott and
Markowitz (2008), Hsu (2006, 2008).
4
See for example Perold [2007, 2008].
5
See for example Asness [2006], Kaplan [2008], Blitz and Swinkels [2008].

2
year average for these variables, with the exception of book value. We compare
an index consisting of the 1,000 largest U.S. stocks in terms of fundamentals with
an index consisting of the 1,000 largest U.S. stocks in terms of market
capitalization. Fundamental data was obtained from Compustat for the period
January 1986 through December 2009. As a fundamental index requires five
years of historical data, this allows us to construct fundamental indexes over the
period 1991-2009. The frequency of the data is monthly.
Similar to the FTSE/RAFI indexes and Arnott, Hsu and Moore (2005), we
rebalance once a year. Instead of choosing one particular rebalancing moment,
however, we distinguish between four fundamental indexes, rebalanced at
month-end March, June, September and December respectively. Later on, we
also consider an alternative fundamental index which invests one-fourth in each
of these four indexes. This ‘blended’ index is effectively rebalanced by one-fourth
every quarter. For example, at the end of March, we rebalance the sub-index that
was created at the end of March one year ago, but leave the other three sub-
indexes (that were last rebalanced at the end of June, September and December
in the previous year) untouched. Contrary to a fundamental index that would be
fully rebalanced every quarter (or even more frequently), the blended index
exhibits a turnover level that is comparable to the annually rebalanced
fundamental indexes.

3. MAIN RESULTS
Our base-case results are similar to those reported by Arnott, Hsu and Moore
(2005). The first two columns in Panel A of Table 1 show that fundamental
indexation with annual rebalancing at the end of December outperforms the
market-capitalization weighted index by approximately 200 basis points per
annum. The next three columns show that the fundamental indexes rebalanced
in March and September outperform the cap-weighted index by almost the same
amount, but the June index only outperforms by about 140 basis points per
annum. The absolute risk of each fundamental index is very close to that of the
capitalization-weighted index.

3
 INSERT TABLE 1 ABOUT HERE

The annualized tracking errors of each index with respect to the others are
shown in Panel B of Table 1. We observe that all four annually rebalanced
fundamental indexes exhibit a tracking error of approximately 5% against the
cap-weighted index. Although this might suggest that the four fundamental
indexes are close to identical, an inspection of their cross-tracking errors shows
that this is not the case. Varying between 1.4-2.1%, the tracking error of one
fundamental index versus another is not negligible, but in a range which is
typically observed for low-risk active management strategies, such as enhanced
indexing.
In the short run, the performance difference between fundamental indexes
rebalanced at different dates can be even larger than their long-term tracking
errors might suggest. Figure 1 shows the annual outperformance of the four
fundamental indexes over the past eight years. For example, for 2009 we
observe that the March fundamental index outperforms the cap-weighted index
by over 10%. Over the same year, the official FTSE/RAFI U.S. 1000 index, which
was rebalanced on 20 March 2009, even outperforms the Russell 1000 index by
13.6%.6 Our three other fundamental indexes exhibit a much weaker 2009
performance though, with the September index even showing a slight
underperformance compared with its cap-weighted counterpart. In other words,
the choice of rebalancing date turns out to be crucial for the performance of
fundamental indexes in 2009, with March appearing to be the best choice that
year. Figure 1 also shows that 2009 was not the only year in which the
rebalancing date had a large impact. For 2003 we also observe a significant
return spread of 5% between the best- and worst-performing fundamental
indexes (the December and June ones respectively, in this instance).

INSERT FIGURE 1 ABOUT HERE

6
Source: Thomson Financial Datastream.

4
 Why can the choice of a specific annual rebalancing date have such a
large impact on the performance of fundamental indexes? In order to answer this
question we take a closer look at what happened in 2009. In the beginning of
2009, equity markets continued their downward trend in the wake of the credit
crisis. But after bottoming out on 9 March 2009, the remainder of the year was
characterized by a strong rally. This rally was particularly strong for beaten down
stocks with relatively high historic fundamentals, such as Bank of America (BoA).
Figure 2 shows the weights of BoA during 2008-2009 for the cap-weighted index
as well as for the fundamental indexes rebalanced in March and September.
From October 2008 until February 2009, the fundamental indexes suffer from
their overweight in BoA, which exhibits a string of large negative returns. As a
result of these negative returns, the weight of BoA in both fundamental indexes
drops from around 2% to 0.4%. At the end of March, however, this weight is
brought back to around 2% in the March fundamental index, whereas the weight
in the September index is still only 0.6% at that point. As a result, the March
index is suddenly ideally positioned to take advantage of the rally in BoA (and
similar stocks) from April onwards. As the FTSE/RAFI index is already
rebalanced on 20 March 2009, it can additionally benefit from part of the March
rally, which explains why it posted an even higher return. In retrospect, we can
conclude that the timing of the rebalancing of the FTSE/RAFI index in 2009 was
near-perfect. However, it would be naive to expect that rebalancing on the third
Friday of March will again prove to be spot-on if a similar scenario were to unfold
in the future. In all fairness, the performance of the FTSE/RAFI index in 2009
contains a significant amount of luck as opposed to repeatable investment skill.

INSERT FIGURE 2 ABOUT HERE

What happens after a large performance discrepancy between

fundamental indexes with different rebalancing dates, such as in 2009, has
arisen? One possibility is that the gap disappears by itself. For this to happen, the

5
fundamental indexes should continue to exhibit large differences in stock
weights, the stocks concerned should again exhibit significant price movements
and these price movements should be in the ‘right’ direction (as otherwise the
gap would only widen further). Intuitively, this scenario is not very likely.
Therefore, we expect that large performance discrepancies, such as in 2009,
tend to be permanent shocks, which affect the terminal wealth of long-term
investors. In other words, our hypothesis is that when two fundamental indexes
diverge, they do not exhibit mean-reversion subsequently. We perform
Augmented Dickey-Fuller (ADF) tests in order to formally analyze whether this
intuition is correct. The ADF tests are applied to the difference between the
cumulative return series of two fundamental indexes, e.g. December and March
(six pairs in total).7 The ADF null hypothesis is that there is a unit root, so a
rejection of the null hypothesis can be interpreted as evidence for mean
reversion. Table 2 shows that even at a mild 10% significance level, we cannot
reject the null hypothesis of a unit root for every index pair. Hence, there is no
evidence of mean-reverting behavior. This supports our intuition that divergences
in the performance of one fundamental index versus another tend to have a
permanent nature.

INSERT TABLE 2 ABOUT HERE

4. A POSSIBLE SOLUTION
The underlying cause of performance differences between fundamental indexes
is a time-varying exposure to value. First, fundamental indexes deviate more
from the cap-weighted index when dispersion in valuations is large; and, second,
the exposure to value changes over time, because fundamental index weights
are allowed to deviate from fundamental values in between rebalancing
moments. The first effect is one of the attractive, or at least intentional features of
the index. The second effect is not particularly desirable though. One solution to

7
We choose the optimal number of lags using the AIC criterion, which results in the
inclusion of one or two lags.

6
this could be to rebalance a fundamental index more frequently, but this goes at
the expense of increased turnover.8 This is why we consider a blended
fundamental index, as described in the methodology section. A blended index
can mitigate the sensitivity of a fundamental index to rebalancing assumptions,
while maintaining a low level of turnover. If one fundamental index, rebalanced
annually at some fixed date, can deviate substantially from another fundamental
index, rebalanced at another date, then, by definition, averaging will give a more
balanced view of the general concept of fundamental indexing. Compared to the
underlying individual fundamental indexes, a blended index has less extreme
weights in individual stocks, and adjustments to these weights are made more
gradually and more frequently over time. For the Bank of America example
discussed in the previous section this is graphically illustrated in Figure 2.
In the last column of Table 1 we show the characteristics of the blended
index. Most characteristics (e.g. average return, volatility, tracking error with the
capitalization-weighted index and turnover) hardly change compared to the four
underlying fundamental indexes, which is not surprising in light of the fact that the
blended index is defined as the average of these individual indexes. The only
statistic which does change materially is the tracking error with the individual
underlying fundamental indexes. Whereas the individual indexes exhibit cross-
tracking errors of 1.4-2.1%, their tracking errors with the blended index are only
around 1%. Recognizing that there is no such thing as “the” performance of
fundamental indexing, this statistically illustrates that the blended index provides
the most nuanced (or least ambiguous) view on the performance of fundamental
indexing. Put differently, the underlying indexes represent specific examples of
fundamental indexing implementations, but the blended index is better able to
capture the performance of the general concept of fundamental indexing.

8
Turnover increases from around 13% to 25% (40%) in case of fully rebalancing the
fundamental index on a quarterly (monthly) basis.

7
5. CONCLUSION
We have shown that the performance of a fundamental index with annual
rebalancing, as proposed by Arnott, Hsu and Moore (2005), can be highly
sensitive to the subjective choice of when to rebalance. In 2009, a fundamental
index rebalanced in March, such as the official FTSE/RAFI index, outperformed
its market-capitalization-weighted counterpart by over 10%, while, over the same
year, a similar index rebalanced in September exhibits a slight
underperformance. Using Bank of America as an example, we show that the
FTSE/RAFI index benefited from almost perfect timing by rebalancing back to
fundamental weights at the start of a huge rally in value stocks. Based on our
detailed examination of the causes behind the performance differences in 2009,
we hypothesize that if two fundamental indexes diverge, they are not likely to
mean-revert subsequently. We find that a formal statistical test cannot reject this
hypothesis, which supports the notion that performance divergences between
fundamental indexes tend to be permanent shocks. Large dispersions in the
performance of fundamental indexes with different rebalancing assumptions such
as in 2009 are an undesirable feature for an objective benchmark index. An
example of a simple yet effective solution to this problem is to create a more
robust fundamental index by blending multiple underlying fundamental indexes,
each one rebalanced annually, but at different dates. A blended index exhibits
similar long-term return, risk and turnover characteristics, but can resolve the
ambiguity regarding the performance of fundamental indexing in the short-run.

8
REFERENCES

Arnott, R.D., J. Hsu and P. Moore (2005), Fundamental indexation, Financial

Analysts’ Journal, vol. 61, no. 2 (March/April), pp. 83-99.

Arnott, R.D. and H.M. Markowitz (2008), Fundamentally flawed indexing:

comments, Financial Analysts’ Journal, vol. 64, no. 2 (March/April), pp.12-14.

Asness, C. (2006), The value of fundamental indexation, Institutional Investor

(October), pp. 94-99.

Blitz, D.C. and L.A.P. Swinkels (2008), Fundamental indexation: an active value
strategy in disguise, Journal of Asset Management, vol. 9, no. 4, pp. 264-269.

Chen, C., R. Chen and G.W. Bassett (2007), Fundamental indexation via
smoothed cap weights, Journal of Banking and Finance, vol. 31, no. 11, pp.
3486–3502.

Estrada, J. (2008), Fundamental indexation and international diversification,

Journal of Portfolio Management, vol. 34, no. 4 (Spring), pp. 93-109.

Hemminki, J. and V. Puttonen (2008), Fundamental indexation in Europe, Journal

of Asset Management, vol. 8, no. 6 (February), pp. 401-405.

Houwer, R. and A. Plantinga (2009), Fundamental indexing: an analysis of the

returns, risks and costs of applying the strategy, SSRN working paper abstract
no. 1343879.

9
Hsu, J. (2006), Cap-weighted portfolios are sub-optimal portfolios, Journal of
Investment Management, vol. 4, no. 3 (Third Quarter), pp. 44–53.

Hsu, J. (2008), Why fundamental indexation might – or might not – work: a

comment, Financial Analysts’ Journal, vol. 64, no. 2 (March/April), pp. 17-18.

Kaplan, P.D. (2008), Why fundamental indexation might – or might not – work,
Financial Analysts’ Journal, vol. 64, no. 1 (January/February), pp. 32-39.

Perold, A.F. (2007), Fundamentally flawed indexing, Financial Analysts’ Journal,

vol. 63, no. 6 (November/December), pp. 31-37.

Perold, A.F. (2008), Fundamentally flawed indexing: author response, Financial

Analysts’ Journal, vol. 64, no. 2 (March/April), pp.14-17.

Treynor, J. (2005), Why market-valuation-indifferent indexing works, Financial

Analysts’ Journal, vol. 61, no. 5 (September/October), pp. 65–69.

Treynor, J. (2008), Fundamentally flawed indexing: comments, Financial

Analysts’ Journal, vol. 64, no. 2 (March/April), pp. 14.

10
Table 1: Long-term index characteristics. We show return and liquidity characteristics of various indexes. ‘Cap-
weighted’ refers to a capitalization-weighted index consisting of the 1,000 largest U.S. stocks in terms of market
capitalization, which we compare to fundamental indexes consisting of the 1,000 largest U.S. stocks in terms of
fundamentals, based on the methodology of Arnott, Hsu and Moore (2005). We consider fundamental indexes with annual
rebalancing at the end of December (Dec), March (Mar), June (Jun) and September (Sep). ‘Blend’ refers to the equally
weighted average of these four indexes. Panel A shows the long-term return, risk and turnover characteristics of the
individual indexes. Panel B shows the tracking error of each index against the other indexes. The sample period is
January 1991 to December 2009.

Cap-weighted Fund. Dec Fund. Mar Fund. Jun Fund. Sep Fund. Blend
Panel A: characteristics (annualized)
Geometric return 10.00% 12.08% 11.86% 11.42% 11.98% 11.84%
Excess return 5.50% 7.50% 7.29% 6.86% 7.41% 7.27%
Standard devation 14.43% 14.68% 14.53% 14.30% 14.25% 14.40%
Sharpe ratio 0.38 0.51 0.50 0.48 0.52 0.51
Turnover 7% 14% 13% 13% 13% 13%
Panel B: cross tracking errors
Cap-weighted 5.2% 5.5% 5.0% 4.7% 5.0%
Fund. Dec 1.8% 1.9% 1.8% 1.1%
Fund. Mar 1.7% 2.1% 1.2%
Fund. Jun 1.4% 0.9%
Fund. Sep 1.1%
Table 2: ADF tests. For all six possible index pairs, we calculate the cumulative
difference series and test for the presence of a unit root using an Augmented
Dickey-Fuller (ADF) test. The 10% critical value for the reported test-statistics is
-1.63. We use the AIC to choose the optimal number of included lags
Fund. Mar Fund. Jun Fund. Sep
Fund. Dec -0.38 1.25 -1.40
Fund. Mar 0.08 -0.76
Fund. Jun 0.63
Figure 1: Calendar year index returns. We show calendar year returns over the 2002-2009 period for fundamental
indexes with annual rebalancing at the end of December (Dec), March (Mar), June (Jun) and September (Sep).

12%

10%

8%

6%

4%
Dec
Mar
2%
Jun
Sep
0%

-2%

-4%

-6%

-8%
2002 2003 2004 2005 2006 2007 2008 2009
Figure 2: Bank of America in 2008-2009. In the top part of the graph we show the weight of
Bank of America in the capitalization weighted index, the fundamental index with annual
rebalancing in March, the fundamental index with annual rebalancing in September and the
blended fundamental index. In the bottom part of the graph we show the monthly returns of Bank
of America. The period considered is January 2008 through December 2009.

Capitalization-weighted index Fundamental index rebalanced in March

Fundamental index rebalanced in September Fundamental index blended

4.0%

3.5%
Index weight Bank of America

3.0%

2.5%

2.0%

1.5%

1.0%

0.5%

0.0%

80%
Return Bank of America

60%

40%

20%

0%
-20%

-40%

-60%
Jan-08
Feb-08
Mar-08
Apr-08
May-08

Jun-08
Jul-08
Aug-08
Sep-08
Oct-08
Nov-08
Dec-08
Jan-09

Feb-09
Mar-09
Apr-09
May-09
Jun-09
Jul-09
Aug-09
Sep-09

Oct-09
Nov-09
Dec-09

View publication stats