See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309138994

Timing the Market with a Combination of Moving Averages: Combination of

Moving Averages

Article  in  International Review of Finance · October 2016

DOI: 10.1111/irfi.12107

CITATIONS READS

5 2,025

1 author:

Paskalis Glabadanidis
University of Adelaide
46 PUBLICATIONS   262 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

mean-variance spanning tests and portfolios with zero-cost spreads View project

All content following this page was uploaded by Paskalis Glabadanidis on 16 November 2017.

The user has requested enhancement of the downloaded file.

International Review of Finance, 2016
DOI: 10.1111/irfi.12107

Timing the Market with a

Combination of Moving Averages*
PASKALIS GLABADANIDIS
Accounting and Finance Business School, University of Adelaide, Adelaide, Australia

ABSTRACT

A combination of simple moving average trading strategies with several win-

dow lengths delivers a greater average return and skewness as well as a lower
variance and kurtosis compared with buying and holding the underlying asset
using daily returns of value-weighted US decile portfolios sorted by market
size, book-to-market, momentum, and standard deviation as well as more than
1000 individual US stocks. The combination moving average (CMA) strategy
generates risk-adjusted returns of 2% to 16% per year before transaction costs.
The performance of the CMA strategy is driven largely by the volatility of stock
returns and resembles the payoffs of an at-the-money protective put on the
underlying buy-and-hold return. Conditional factor models with macroeco-
nomic variables, especially the market dividend yield, short-term interest
rates, and market conditions, can explain some of the abnormal returns. Stan-
dard market timing tests reveal ample evidence regarding the timing ability of
the CMA strategy.

JEL Codes: G11; G12; G14

I. INTRODUCTION

Technical analysis involves the use of past and current market price, trading vol-
ume, and potentially, other publicly available information to predict future mar-
ket prices. It is highly popular in practice with plentiful financial trading advice
that is based largely, if not exclusively, on technical indicators. From the stand-
point of classical economic and finance theory, it is not at all clear that technical
analysis in general and moving averages in particular will have any role or power
in predicting the returns of individual stocks as well as portfolios of stocks. Sev-
eral potential reasons come to mind in terms of justifying the use of moving av-
erages. First, investor heterogeneity as well as information asymmetry may lead
to the persistent manifestation of behavioral biases in stock market prices. Prior
studies that have touched upon these issues include Treynor and Ferguson
(1985); Brown and Jennings (1989), and Hong and Stein (1999) among many
others. Furthermore, the theoretical model in Wang (1993) shows explicitly

* I would like to thank the editor, Hong Yan, and one anonymous referee for their very detailed and
thoughtful comments. Any remaining errors are my own responsibility.
© 2016 International Review of Finance Ltd. 2016
 International Review of Finance

how a rational economic agent inhabiting a classical model of choice under un-
certainty and differential information will find signals based on average past
prices quite useful, informative, and revealing of other agents’ private informa-
tion. Secondly, active investors in practice very often follow price trends which
may lead to the continued persistence of trends, both upward as well as down-
ward. These trends present other investors with the ability to follow them at least
in the short-term. Academic work in this area is perhaps best exemplified by Fung
and Hsieh (2001), and their construction of trend following indicators based on
the returns of look back straddle options. Thirdly, the study by Brock et al. (1992)
document the performance of various implementations of the moving average
and conclude that it is the most popular strategy followed by investors who use
technical analysis. More formally, Brock et al. (1992) find evidence that some
technical indicators do have a significant predictive ability. Fourthly, Blume
et al. (1994) present a theoretical framework using trading volume and price data
leading to technical analysis being a part of a trader’s learning process. A more
thorough study of a large set of technical indicators by Lo et al. (2000) also found
some predictive ability especially when moving averages are concerned. Zhu and
Zhou (2009) provide a solid theoretical reason why technical indicators could be
a potentially useful state variable in an environment where investors need to
learn over time the fundamental value of the risky asset they invest in. More re-
cently, Neely et al. (2010, 2011) find that technical analysis has as much forecast-
ing power over the equity risk premium as the information provided by
economic fundamentals. The practitioner’s literature also includes Faber (2007)
and Kilgallen (2012) who thoroughly document the risk-adjusted returns to the
moving average strategy using various portfolios, commodities, and currencies.
In addition, Huang and Zhou (2013) use the moving average indicator to predict
the return on the US stock market while Goh et al. (2012) apply the same idea to
government bond yields and risk premia. Motivated in part by the predictive
power of the moving average indicator, Han et al. (2016) and Jiang (2013) con-
struct a trend factor with considerable cross-sectional explanatory power and
substantial historical performance. In a similar vein, Glabadanidis (2014,
2015a, 2015b) investigates and documents the performance of the simple mov-
ing average strategy with various US and international portfolios as well as indi-
vidual US stocks.
The contribution of this paper is three-fold. First, I propose a novel strategy
which is an equal-weighted average of simple moving average. The novelty here
is that the investment or disinvestment in the underlying risky asset is propor-
tional to the number of moving average windows that have generated a buy or
sell signal, respectively. This is in stark contrast with trading signals generated
by a single moving average window which involve either being completely
invested in the risky asset or completely invested in the risk-free asset. Secondly,
I report on the performance of the combination moving average strategy with a
large number of portfolios and individual stocks. Finally, I provide a link between
technical indicators and fundamental indicators by presenting evidence that the
performance of the combination moving average strategy can be partially

2 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

explained by a conditional asset pricing model with the market’s dividend yield,
short-term interest rates, and a recession indicator.
This paper is similar in spirit to Glabadanidis (2014, 2015a, 2015b) and Han
et al. (2013). However, several important differences stand out. First, I use daily
value-weighted returns of decile portfolios constructed by various characteristics
like size, book-to-market, momentum, and standard deviation of return. Value-
weighted portfolios at a daily frequency have a much smaller amount of trading
going on inside the portfolio compared with the daily equal-weighted portfolios
investigated by Han et al. (2013). Secondly, the cross-sectional results in this
study are just an artefact of the decile portfolios and not the main focus of this
paper, while Han et al. (2013) is mostly concerned with the inability of stan-
dard empirical tests to account for the moving average strategy average returns
differences across portfolios. I argue that this is largely due to using the wrong
benchmark pricing model. Using a dynamic market-timing tests and condi-
tional asset pricing models with macroeconomic state variables leads to mostly
negative or statistically insignificant risk-adjusted returns for the moving aver-
age strategy. In light of this, my take on the performance of the combination
moving average strategy is that it is not an anomaly, but instead a dynamic
trading strategy that exposes investors to potential upside returns derived from
risky assets via its market timing ability. Similarly, the combination moving av-
erage strategy manages to avoid substantial market downturns more often than
not, thus, insulating investors from periods of sustained bear markets. This per-
formance is more pronounced the more volatile the returns of the underlying
risky assets are. A final caveat is that I assume the moving average trading has
no price impact. Large investors using this strategy will necessarily experience
an inferior performance. This is largely due to the adverse price impact of liqui-
dating and initiating large positions, especially for less liquid assets with lower
trading volumes.
The highlights of this study are the superior performance of the combina-
tion moving average portfolios relative to buying and holding the underlying
portfolios, the fact that the switching strategy returns resemble an imperfect
at-the-money protective put, and that cross-sectional differences are not a
new anomaly as maintained in Han et al. (2013), but are due to volatility dif-
ferences in the underlying portfolios and stocks as well as factor exposure
differences to a few macroeconomicstate variables. The returns of the combina-
tion moving average strategy relative to the buy-and-hold strategy are quite
convex with respect to the return of the buy-and-hold strategy and, hence, will
be hard to explain using standard linear asset pricing models. The anomalous
risk-adjusted performance relative to standard linear asset pricing models ap-
pears to be largely due to omitting market timing factors in a simple piece-wise
linear framework that captures the moving average strategy’s convexity.
Furthermore, the moving average strategy appears to be antifragile in the sense
of Taleb (2012) meaning that for securities with more volatile returns there is a
greater improvement of the moving average returns relative to buy-and-hold
returns.

© 2016 International Review of Finance Ltd. 2016 3

 International Review of Finance

II. A COMBINATION OF SIMPLE MOVING AVERAGES

I use daily value-weighted1 returns of sets of 10 portfolios sorted by market cap-

italization, book-to-market, momentum, and standard deviation. The data is
readily available from Ken French Data Library. The sample period starts on Jan-
uary 4, 1960 and ends on December 31, 2013.
The following exposition of the moving average strategy follows closely the
presentation in Han et al. (2013). Let Rjt be the return on portfolio j at the
end of month t and let Pjt be the respective price level of that portfolio. De-
fine the moving average of portfolio j Ājt,L at time t with length L periods as
follows:

PjtLþ1 þ PjtLþ2 þ ⋯ þ Pjt1 þ Pjt

Ajt;L ¼ (1)
L

Throughout the paper, I use a combination moving average comprised of an

equal-weighted combination of simple moving averages of length L = 5, L = 10,
L = 20, L = 50, L = 100, and L = 200 days. The way I implement the simple moving
average strategy in this paper is to compare the closing price Pjt at the end of ev-
ery day to the running moving average Ājt,L. If the price is above the moving av-
erage this triggers a signal to invest (or stay invested if already invested at t - 1) in
the portfolio in the next day t + 1. If the price is below the moving average this
triggers a signal to leave the risky portfolio (or stay invested in cash if not
invested at t - 1) in the following day t + 1.2 As a proxy for the risk-free rate, I
use the daily return on the 30-day US Treasury Bill.
More formally, the returns of the moving average switching strategy can be
expressed as follows:
(
Rjt ; if Pjt1 > Ajt1;L
ejt;L ¼
R (2)
rf t ; otherwise;

in the absence of any transaction costs imposed on the switches. For the rest of
the paper and in all of the empirical results quoted, I consider returns after the
imposition of a one-way transaction cost of τ. Mathematically, this leads to the
following four cases in the post-transaction cost returns:

1 I use value-weighted portfolio returns to control for the amount of rebalancing trading inside
the various portfolios. The empirical results in this paper are much stronger when equal-
weighted portfolios are used. However, this may understate the break-even transaction costs
as equal weighted portfolios require a lot of trading to be replicated.
2 An alternative version of the switching strategy involves investing in the market portfolio in-
stead of the risk-free asset. This version of the switching strategy has a somewhat inferior per-
formance compared with the baseline case investigated in the article. Nevertheless, it is an
interesting case to consider, and I am grateful to an anonymous referee for suggesting this idea
to me.

4 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

8
>
> Rjt ; if Pjt1 > Ajt1;L and Pjt2 > Ajt2;L ;
>
>
< R  τ; if Pjt1 > Ajt1;L and Pjt2 < Ajt2;L ;
ejt;L ¼
R
jt
(3)
>
> rf t ; if Pjt1 < Ajt1;L and Pjt2 < Ajt2;L ;
>
>
:
rf t  τ; if Pjt1 < Ajt1;L and Pjt2 > Ajt2;L :
depending on whether the investor switches or not. Note that this imposes a cost
on selling and buying the risky portfolio, but no cost is imposed on buying and
selling the Treasury bill. This is consistent with prior studies like Balduzzi and
Lynch (1999); Lynch and Balduzzi (2000), and Han (2006), among others. Re-
garding the appropriate size of the transaction cost, Balduzzi and Lynch (1999)
propose using a value between 1 and 50 basis points. Lynch and Balduzzi
(2000) use a mid-point value of 25-basis point. Instead of choosing a controver-
sial value for the one-way transaction cost, I use τ = 0 and report the break-event
one-way transaction cost that will completely eliminate any outperformance of
the combination moving average strategy relative to the buy-and-hold strategy.

e e e e e e
f jt ¼ Rjt;5 þ Rjt;10 þ Rjt;20 þ Rjt;50 þ Rjt;100 þ Rjt;200
CR (4)
6
I construct excess returns as zero-cost portfolios that are long the combination
moving average (CMA) switching strategy and short the underlying portfolio to
determine the relative performance of the moving average strategy against the
buy-and-hold strategy. Denote the resulting difference between the return of
the CMA strategy for portfolio j at the end of month t, CR f jt  Rjt , and the return
of portfolio j at the end of month t, Rjt, as follows:

f jt  Rjt ;
CMAPjt ¼ CR j ¼ 1; …; N (5)
The presence of significant abnormal returns can be interpreted as evidence in
favor of superiority of the moving average switching strategy over the buy-and-
hold strategy of the underlying portfolio. Naturally, the moving average
switching strategy is a dynamic trading strategy, so it is perhaps unfair to com-
pare its returns to the buy-and-hold returns of being long the underlying
portfolio.
Table 1 presents the first four moments and the Sharpe ratio of the buy-and-
hold (BH) strategy, the CMA strategy and the combination moving average port-
folio (CMAP) strategy for decile portfolio sorted by market capitalization, book-
to-market, momentum, and standard deviation of return. The first strong finding
that emerges for all portfolios is that the standard deviation of return is reduced
by the CMA strategy relative to the BH strategy. Secondly, the risk-return trade-off
is improved for all portfolios as evidenced by the increased Sharpe ratios for all
portfolios. Thirdly, in the vast majority of cases the average return of the CMA
strategy exceeds the average return of the BH strategy. The only exception are
decile high sorted by size and decile eight sorted by momentum. Fourthly, the
kurtosis of almost all portfolios is reduced as well with the exception of

© 2016 International Review of Finance Ltd. 2016 5

6
Table 1 Summary statistics
Panel A: size sorted portfolios
p μ σ s k SR μ σ s k SR μ σ s k SR
BH portfolios CMA portfolios CMAP portfolios
Low 11.88 13.58 0.90 14.82 0.52 23.61 7.43 0.34 9.57 2.53 11.73 9.47 1.68 42.24 1.24
2 11.82 16.71 0.46 13.86 0.42 20.25 9.07 0.50 10.56 1.71 8.43 11.65 0.52 36.55 0.72
3 12.90 16.59 0.48 12.03 0.49 19.70 9.19 0.41 9.00 1.62 6.80 11.38 0.62 32.86 0.60
4 12.10 16.32 0.48 12.49 0.45 18.36 8.94 0.44 8.77 1.52 6.25 11.29 0.66 35.91 0.55
5 12.62 16.22 0.47 11.97 0.48 18.10 8.99 0.45 8.87 1.48 5.48 11.19 0.62 34.38 0.49
6 12.16 15.32 0.53 13.43 0.48 17.76 8.45 0.40 8.11 1.54 5.60 10.61 0.87 41.62 0.53
7 12.19 15.47 0.55 15.55 0.48 17.48 8.51 0.34 8.70 1.49 5.28 10.72 0.95 48.95 0.49
8 11.91 15.69 0.51 16.63 0.45 16.02 8.68 0.30 8.37 1.30 4.11 10.82 0.85 54.89 0.38
9 11.48 15.45 0.59 20.91 0.43 14.02 8.55 0.23 7.97 1.08 2.54 10.68 1.04 72.40 0.24
High 10.09 15.99 0.49 21.64 0.33 9.74 8.88 0.18 7.74 0.56 0.34 11.01 0.94 78.14 0.03
Panel B: book-to-market sorted portfolios
p μ σ s k SR μ σ s k SR μ σ s k SR
BH portfolios CMA portfolios CMAP portfolios
Low 9.32 17.52 0.17 13.47 0.26 12.27 9.58 0.17 7.42 0.78 2.95 12.21 0.05 42.84 0.24
International Review of Finance

2 10.89 16.06 0.39 16.38 0.38 12.44 9.01 0.13 7.93 0.85 1.55 11.04 0.77 57.39 0.14
3 11.01 15.42 0.51 20.71 0.40 13.03 8.75 0.04 7.28 0.94 2.02 10.47 1.06 78.27 0.19
4 10.82 15.79 0.71 20.94 0.38 13.25 8.87 0.20 8.86 0.96 2.43 10.81 1.54 73.64 0.22
5 11.10 15.61 0.45 23.03 0.41 12.34 8.73 0.30 9.13 0.87 1.24 10.75 0.59 80.18 0.12
6 11.77 15.14 0.44 16.96 0.46 13.44 8.55 0.23 8.10 1.01 1.67 10.26 0.72 59.03 0.16
7 12.43 14.89 0.64 22.73 0.51 14.09 8.55 0.20 8.60 1.09 1.65 10.02 1.24 87.83 0.16
8 12.89 15.81 0.60 30.17 0.51 14.15 8.87 1.20 25.12 1.06 1.26 10.93 0.81 92.42 0.12
9 13.82 15.99 0.68 20.85 0.57 15.45 9.26 0.46 11.05 1.15 1.64 10.74 1.04 72.16 0.15
High 14.90 17.77 0.46 15.87 0.57 18.43 10.32 0.27 11.15 1.32 3.53 11.87 0.97 53.66 0.30

© 2016 International Review of Finance Ltd. 2016

 Table 1 (continued)
Panel C: momentum sorted portfolios
p μ σ s k SR μ σ s k SR μ σ s k SR
BH portfolios CMA portfolios CMAP portfolios
Low 2.51 24.92 0.41 25.73 0.09 16.23 12.33 0.51 41.68 0.93 13.72 18.28 0.58 46.59 0.75
2 7.92 20.34 0.16 22.19 0.15 14.78 10.98 0.37 34.60 0.91 6.86 14.39 0.74 51.81 0.48
3 10.21 17.58 0.04 17.99 0.31 13.80 9.70 0.09 14.68 0.93 3.59 12.17 0.29 43.70 0.29
4 10.15 16.60 0.15 18.82 0.32 13.08 9.15 0.06 11.36 0.91 2.92 11.43 0.27 56.09 0.26
5 10.00 15.89 0.01 21.66 0.33 11.63 8.86 0.10 13.88 0.77 1.63 10.93 0.29 68.11 0.15
6 10.84 15.35 0.58 26.69 0.39 12.19 8.65 0.26 10.04 0.86 1.35 10.48 1.21 99.19 0.13
7 10.76 15.19 0.55 22.87 0.39 12.51 8.51 0.25 7.45 0.91 1.76 10.41 1.08 86.63 0.17
8 13.42 15.49 0.73 21.20 0.56 13.27 9.16 0.27 7.75 0.93 0.15 10.14 1.71 93.02 0.01
9 12.66 16.46 0.53 15.10 0.48 13.92 9.46 0.29 7.35 0.97 1.26 11.05 0.90 56.48 0.11
High 17.71 20.36 0.50 12.74 0.63 19.69 11.69 0.44 7.57 1.28 1.98 13.64 0.64 45.42 0.15

© 2016 International Review of Finance Ltd. 2016

Panel D: standard deviation sorted portfolios
p μ σ s k SR μ σ s k SR μ σ s k SR
BH portfolios CMA portfolios CMAP portfolios
High 40.03 20.26 0.10 14.39 1.74 49.35 13.03 1.16 15.15 3.42 9.32 12.64 0.33 50.77 0.74
2 17.70 19.29 0.31 14.53 0.67 28.52 10.74 0.32 13.08 2.21 10.82 13.37 0.80 41.93 0.81
3 15.14 18.25 0.38 16.40 0.57 24.73 9.93 0.04 11.51 2.01 9.59 12.82 0.78 47.10 0.75
4 15.67 16.88 0.47 18.78 0.64 23.10 9.19 0.12 10.99 1.99 7.43 11.88 0.88 54.03 0.63
Combination of Moving Averages

5 15.60 15.51 0.55 20.48 0.70 22.41 8.45 0.20 10.74 2.09 6.81 10.88 0.90 60.56 0.63
6 14.74 14.38 0.67 24.46 0.69 20.91 7.81 0.27 10.44 2.07 6.18 10.11 1.08 73.82 0.61
7 14.30 12.77 0.66 28.32 0.75 19.94 7.01 0.16 9.13 2.16 5.64 8.97 1.03 94.82 0.63
8 13.60 11.15 0.78 36.33 0.79 18.81 6.12 0.00 7.95 2.29 5.21 7.90 1.53 127.66 0.66
9 12.47 9.27 0.58 54.74 0.83 17.54 5.04 0.17 8.69 2.53 5.07 6.68 1.00 187.69 0.76
Low 10.77 6.86 0.47 71.63 0.87 16.03 3.72 0.51 10.76 3.03 5.25 4.94 0.73 245.31 1.06

This table reports summary statistics for the respective buy and hold (BH) portfolio returns, the combination moving average (CMA) switching strat-
egy portfolio returns and the excess return of CMA over BH (CMAP) using sets of 10 portfolios sorted by size, book-to-market, momentum and stan-
dard deviation of return. The sample period covers January 4, 1960 until December 31, 2013 with value-weighted portfolio returns. μ is the
annualized average return, σ is annualized standarddeviation of returns, s is the annualized skewness, k is the annualized kurtosis, and SR is the an-
nualized Sharpe ratio. The lengths of the moving average windows are 5, 10, 20, 50, 100, and 200 days. The CMA portfoliois an equal-weighted com-

7
bination of the six individual moving average returns.
 International Review of Finance

Table 2 Factor regressions results

Panel A: size sorted portfolios
2
Portfolio α βm βs βh βu R
Low 16.111*** 0.483*** 0.497*** 0.157*** 0.012** 0.653
2 13.737*** 0.625*** 0.579*** 0.190*** 0.010* 0.705
3 11.706*** 0.615*** 0.531*** 0.130*** 0.010* 0.713
4 10.980*** 0.613*** 0.483*** 0.096*** 0.004 0.714
5 9.843*** 0.608*** 0.425*** 0.068*** 0.018*** 0.708
6 9.603*** 0.576*** 0.306*** 0.044*** 0.005 0.691
7 9.346*** 0.586*** 0.246*** 0.069*** 0.009* 0.687
8 8.111*** 0.597*** 0.173*** 0.072*** 0.013** 0.697
9 6.447*** 0.589*** 0.062*** 0.079*** 0.004 0.699
High 2.392*** 0.579*** 0.137*** 0.036*** 0.040*** 0.718
Panel B: book-to-market sorted portfolios.
2
Portfolio α βm βs βh βu R
Low 4.633*** 0.595*** 0.036*** 0.307*** 0.048*** 0.703
2 4.569*** 0.576*** 0.017** 0.062*** 0.009* 0.681
3 5.371*** 0.549*** 0.019*** 0.012 0.023*** 0.663
4 6.551*** 0.589*** 0.011 0.169*** 0.018*** 0.674
5 5.486*** 0.577*** 0.019*** 0.215*** 0.021*** 0.645
6 5.730*** 0.555*** 0.071*** 0.195*** 0.026*** 0.652
7 6.201*** 0.545*** 0.030*** 0.278*** 0.002 0.647
8 6.464*** 0.590*** 0.066*** 0.441*** 0.047*** 0.651
9 7.182*** 0.591*** 0.082*** 0.405*** 0.011** 0.651
High 9.581*** 0.647*** 0.181*** 0.455*** 0.013** 0.642
Panel C: momentum sorted portfolios
2
Portfolio α βm βs βh βu R
Low 14.460*** 0.845*** 0.310*** 0.043*** 0.596*** 0.682
2 7.962*** 0.691*** 0.108*** 0.050*** 0.410*** 0.689
3 5.155*** 0.600*** 0.037*** 0.086*** 0.295*** 0.684
4 5.621*** 0.591*** 0.004 0.118*** 0.162*** 0.672
5 4.603*** 0.566*** 0.001 0.079*** 0.089*** 0.650
6 5.043*** 0.553*** 0.024*** 0.112*** 0.006 0.643
7 5.836*** 0.557*** 0.013* 0.092*** 0.046*** 0.654
8 4.597*** 0.544*** 0.018*** 0.124*** 0.120*** 0.645
9 6.708*** 0.602*** 0.059*** 0.082*** 0.174*** 0.669
High 8.596*** 0.696*** 0.192*** 0.057*** 0.301*** 0.645
Panel D: standard deviation sorted portfolios
2
Portfolio α βm βs βh βu R
High 13.784*** 0.569*** 0.491*** 0.133*** 0.026*** 0.496
2 15.651*** 0.684*** 0.474*** 0.128*** 0.057*** 0.625
3 14.676*** 0.680*** 0.429*** 0.179*** 0.044*** 0.652
4 12.321*** 0.635*** 0.370*** 0.215*** 0.045*** 0.651
5 11.445*** 0.584*** 0.300*** 0.229*** 0.036*** 0.644

8 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 2 (continued)
PanelD:
Panel A:standard
size sorted portfolios
deviation sorted portfolios
2
Portfolio α βm βs βh βu R
6 10.511*** 0.545*** 0.236*** 0.224*** 0.030*** 0.642
7 9.459*** 0.475*** 0.162*** 0.185*** 0.008 0.612
8 8.531*** 0.407*** 0.111*** 0.152*** 0.008* 0.574
9 7.756*** 0.320*** 0.071*** 0.129*** 0.013*** 0.493
Low 6.823*** 0.179*** 0.017*** 0.055*** 0.029*** 0.284
2
This table reports alphas, betas, and adjusted R of the regressions of the CMAP excess returns on
the Carhart four-factors using portfolios sorted by size, book-to-market, momentum, and standard
deviation of return. The alphas are annualized and in percent. The sample period covers January 4,
1960 until December 31, 2013 with value-weighted portfolio returns. The lengths of the moving
average windows are 5, 10, 20, 50, 100, and 200 days. Newey and West (1987) standard errors with
three lags are used in reporting statistical significance of a two-sided null hypothesis at the 1, 5, and
10% level is given by a ***, a **, and a *, respectively.

momentum deciles low and two as well as decile high of the standard deviation
sorted portfolios. Finally, portfolio skewness increases for most of the portfolios
with the exception of book-to-market decile eight and momentum deciles two,
three, and five. Overall, the findings are that the CMA strategy improves all four
of the first moments of all but a handful of portfolios relative to the BH strategy.

A. Abnormal returns
The asset pricing model that I consider in this section is the four-factor Carhart
(1997) model3:

CMAPjt;L ¼ αj þ βj;m rmkt;t þ βj;s rsmb;t þ βj;h rhml;t þ βj;u rumd;t þ εjt ; j ¼ 1; …; N (6)

where rmkt,t is the excess return on the market portfolio at the end of month t, rsmb,t
is the return on the small minus big (SMB) factor at the end of month t, rhml,t is the
return on the high minus low (HML) factor at the end of month t, and rumd,t is the
return of the up minus down (UMD) factor at the end of month t. Note that all of
the risk-adjusted alphas are highly statistically significant (see Table 2). Moreover,
they are all still quite substantial economically ranging between 2.4% and 16.1%
per year. The factor loadings on the market portfolio, SMB, and HML are largely
unchanged across the three sets of decile portfolios while the loadings on the
UMD factor are mostly positive and highly statistically significant (with only a
few exceptions). This suggests that all four factors have a role to play in driving
the performance of the CMAP returns. Nevertheless, the average adjusted R2
values indicate that only between one half and two thirds of the return variation

3 Results for the CAPM and Fama–French three factor models yield very similar and, frequently,
stronger than the results for the Carhart (1997) model. These additional findings are not re-
ported in the paper in the interest of saving space. They are available from the author upon
request.

© 2016 International Review of Finance Ltd. 2016 9

 International Review of Finance

Figure 1 Scatter Plot of Buy-and-Hold returns versus the Combination Moving

Average returns: High ME Decile Portfolio.
Notes: Figure 1 presents a scatter plot of the returns of the high ME decile buy-and-hold
portfolio returns versus the combination moving average strategy returns. The sam-
ple contains 13,592 daily observations and the data covers the January 4, 1960 until
December 31, 2013.

can be explained and accounted for by the market portfolio return, size, value, and
momentum. This leaves a large portion of return variation that cannot be
accounted for. Finally, it is worthwhile noting that all of βm values are negative
and statistically significant, indicating that the market beta of the CMA strategy
significantly exceeds the market beta of the BH strategy. With some exceptions,
this is also the case for βs and βh. The loadings on the momentum factor are both
positive and negative indicating that for some portfolios, the momentum beta
of the CMA strategy is exceeded by the momentum beta of the BH strategy, while
for other portfolios; it is the other way around.

B. Explanation
Before making an attempt at explaining the reasons for the profitability of the
moving average (MA) strategies performance, it is useful to inspect a scatter plot
of the MA strategy returns versus the underlying BH strategy returns for the same
portfolio. For ease of exposition, I provide a plot for a single portfolio only.4
Figure 1 presents the scatter plot for the first decile of the market-capitalization
sorted deciles.
The strategy is clearly triggering false positive signals, where we are told to stay
invested or switch into the underlying asset with a subsequent negative return

4 The scatter plots for the other portfolios sorted on the various characteristics are available from
the author upon request.

10 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Figure 2 Time Series Plot of Cumulative Values of the Buy-and-Hold and the Com-
bination Moving Average returns: High ME Decile Portfolio.
Notes: Figure 2 presents a time series plot of the cumulative values of the buy-and-hold
portfolio returns versus the combination moving average strategy returns for the high
ME decile. The sample contains 13,592 daily observations and the data covers the
January 4, 1960 until December 31, 2013.

(negative quadrant of returns in the figure). Similarly, there are a few instances of
a false negative signal where we switch into the risk-free asset, while the underly-
ing risky asset has a positive excess return in the following period. Nevertheless,
the signal is right about two out of every three times, and in those instances, the
scatter plot resembles the payoff of an at-the-money put option combined with a
long position in the underlying risky asset. This positive convexity is the driving
factor for the relative outperformance of the moving average strategy relative to
the buy-and-hold strategy (see Figure 2). Holding the signal success rate constant,
risky assets with more volatile returns will experience a higher average
outperformance, and this is evidenced in all of the previous tables.

III. ROBUSTNESS CHECKS

In this section, I report my findings for several robustness checks performed on

the performance of the CMA strategy versus the BH strategy for decile portfolios
sorted on market capitalization, book-to-market ratios, momentum, and stan-
dard deviation.

A. Subperiods
In this robustness check, I split the sample in two subperiods of roughly equal
length. The first subperiod runs from January 4, 1960 to December 31, 1987.

© 2016 International Review of Finance Ltd. 2016 11

 International Review of Finance

The second subperiod goes from January 4, 1989 to December 31, 2013. For the
sake of brevity, Table 3 reports the only the annualized abnormal returns and
the goodness-of-fit using the CAPM, the Fama–French and Carhart models in ei-
ther subperiod. The findings show robust and unaccounted for abnormal returns
for portfolios sorted on market capitalization and standard deviation of return in
both subperiods. For the book-to-market and momentum decile portfolios, there
are only a few statistically significant abnormal returns in the second subperiod
compared with statistical significance across the board in the first subperiod.
Nevertheless, the second subperiod abnormal returns of the highest book-to-
market decile portfolio is significant using the CAPM, while book-to-market dec-
iles 5–10 show significant abnormal returns using the Fama–French and Carhart
model. Similarly, the second subperiod abnormal returns of momentum deciles
1–3 are positive and statistically significant using the CAPM. Using the Fama–
French model, it appears that momentum decile portfolios 1–4 have positive
and statistically significant abnormal returns in the second subperiod, while
the Carhart model reveals that only the extreme momentum deciles (1–2 and
10) have positive and statistically significant abnormal returns in the second sub-
period. Furthermore, the abnormal returns in both subperiods are lower for large-
cap portfolios, value portfolios, winner portfolios, and portfolios of stocks with
high standard deviation of return in the past.

B. Buy-and-hold as a benchmark
An alternative way of judging the performance of the CMA strategy is to use the
BH strategy as a benchmark. One simple way of to do this is to regress the CMA
return on a constant and the BH return. A positive and statistically significant in-
tercept indicates a superior performance of the active CMA return relative to the
passive BH benchmark.

f i;t ¼ αi þ βBH;i Ri;t þ εi;t

CR (7)

A modified version of this regression involves testing for any market timing
ability of the active strategy by including the negative BH return in the regression
along the lines of the market timing tests of Henriksson and Merton (1981)
 
f i;t ¼ αi þ βBH;i Ri;t þ γBH;i max Ri;t ; 0 þ εi;t
CR (8)

Table 4 presents the findings of the simple regression of the active CMA return
on the passive BH benchmark as well as the modified regression involving the
negative component of the passive BH return. The most striking finding to
emerge from these regression results is large and statistically as well as economi-
cally significant intercepts. Secondly, the exposure of the CMA return to the BH
return ranges between 0.35 and 0.51 for the first regression specification in (7)
and between 0.37 and 0.63 for the second regression specification in (8). Finally,

12 © 2016 International Review of Finance Ltd. 2016

 Table 3 Factor regressions results in subperiods
Panel A: size sorted portfolios
2 2 2 2 2 2
p α R α R α R α R α R α R
1960/01/04–1987/12/31 1988/01/04–2013/12/31
CAPM Fama–French Carhart CAPM Fama–French Carhart
Low 17.449*** 0.447 19.386*** 0.700 19.394*** 0.700 10.570*** 0.509 11.911*** 0.651 12.040*** 0.652
2 15.896*** 0.511 17.566*** 0.711 17.846*** 0.711 7.078*** 0.577 8.760*** 0.711 8.585*** 0.712
3 14.707*** 0.538 16.224*** 0.717 16.585*** 0.719 5.008*** 0.610 6.304*** 0.718 6.057*** 0.719
4 14.165*** 0.560 15.480*** 0.712 15.963*** 0.715 4.536*** 0.633 5.606*** 0.722 5.363*** 0.723
5 13.352*** 0.578 14.555*** 0.711 15.023*** 0.714 3.906** 0.653 4.731*** 0.714 4.299*** 0.716
6 13.095*** 0.611 14.207*** 0.703 14.778*** 0.707 4.156*** 0.665 4.658*** 0.695 4.295*** 0.696
7 12.579*** 0.629 13.396*** 0.697 13.802*** 0.698 4.264*** 0.676 4.871*** 0.695 4.562*** 0.696

© 2016 International Review of Finance Ltd. 2016

8 11.437*** 0.657 12.237*** 0.693 12.732*** 0.696 3.231** 0.700 3.713*** 0.708 3.305** 0.710
9 10.206*** 0.675 10.654*** 0.685 11.108*** 0.687 1.309 0.713 1.702 0.718 1.448 0.718
High 6.337*** 0.685 5.636*** 0.696 5.958*** 0.697 0.126 0.725 0.362 0.735 1.139 0.741
Panel B: book-to-market sorted portfolios
2 2 2 2 2 2
P α R α R α R α R α R α R
1960/01/04–1987/12/31 1988/01/04–2013/12/31
CAPM Fama–French Carhart CAPM Fama–French Carhart
Combination of Moving Averages

Low 11.742*** 0.684 9.349*** 0.708 9.560*** 0.708 1.374 0.677 0.322 0.706 0.584 0.713
2 9.687*** 0.673 8.892*** 0.681 9.318*** 0.683 0.039 0.690 0.141 0.693 0.506 0.694
3 9.758*** 0.660 9.389*** 0.665 9.907*** 0.667 0.350 0.672 0.292 0.675 0.349 0.675
4 9.943*** 0.657 10.242*** 0.660 10.552*** 0.661 1.309 0.667 2.153 0.694 1.820 0.695
5 8.662*** 0.625 8.909*** 0.628 9.530*** 0.633 0.041 0.628 1.065 0.672 0.581 0.674
6 7.262*** 0.605 8.101*** 0.618 8.454*** 0.620 2.062 0.650 3.015** 0.680 2.559* 0.682
7 7.425*** 0.587 8.715*** 0.599 9.287*** 0.603 1.673 0.625 2.888** 0.681 2.698** 0.681
8 7.060*** 0.608 9.170*** 0.639 9.690*** 0.643 1.476 0.545 3.549** 0.664 2.673* 0.670
9 7.685*** 0.579 10.219*** 0.621 11.053*** 0.630 1.436 0.587 3.155** 0.672 2.914** 0.673
High 9.388*** 0.582 11.959*** 0.632 12.663*** 0.637 4.089** 0.553 6.288*** 0.657 5.808*** 0.659

13
 Table 3 (continued)

14
Panel C: momentum sorted portfolios
2 2 2 2 2 2
P α R α R α R α R α R α R
1960/01/04–1987/12/31 1988/01/04–2013/12/31
CAPM Fama–French Carhart CAPM Fama–French Carhart
Low 20.051*** 0.622 19.904*** 0.649 16.822*** 0.703 17.511*** 0.535 19.209*** 0.559 12.067*** 0.688
2 14.169*** 0.619 14.102*** 0.627 11.818*** 0.670 7.669*** 0.596 8.693*** 0.611 3.905** 0.708
3 9.627*** 0.621 9.532*** 0.628 7.767*** 0.661 4.376** 0.619 5.284*** 0.642 1.929 0.710
4 9.927*** 0.648 9.791*** 0.650 9.114*** 0.655 2.493 0.641 3.362** 0.669 1.393 0.698
5 9.389*** 0.656 9.509*** 0.659 9.439*** 0.659 0.106 0.634 0.609 0.647 0.642 0.660
6 9.554*** 0.640 9.704*** 0.640 10.101*** 0.642 0.978 0.644 0.426 0.661 0.728 0.662
7 9.387*** 0.653 9.644*** 0.655 10.485*** 0.662 0.186 0.655 0.538 0.664 0.773 0.665
8 7.509*** 0.631 7.999*** 0.633 9.534*** 0.657 2.066 0.633 1.718 0.642 0.750 0.652
9 8.989*** 0.631 9.338*** 0.641 11.367*** 0.677 0.031 0.654 0.203 0.655 1.753 0.677
High 9.689*** 0.582 9.781*** 0.605 12.898*** 0.664 1.844 0.567 1.334 0.590 4.221** 0.638
Panel D: standard deviation sorted portfolios
2 2 2 2 2 2
P α R α R α R α R α R α R
1960/01/04–1987/12/31 1988/01/04–2013/12/31
CAPM Fama–French Carhart Fama–French Carhart
High 15.596*** 0.416 17.447*** 0.596 16.860*** 0.598 8.085*** 0.410 9.475*** 0.483 9.168*** 0.484
2 18.273*** 0.523 20.032*** 0.677 19.937*** 0.677 10.333*** 0.574 11.672*** 0.631 10.721*** 0.637
International Review of Finance

3 17.547*** 0.550 19.262*** 0.694 19.398*** 0.694 8.468*** 0.608 9.962*** 0.669 9.150*** 0.674
4 15.135*** 0.552 16.800*** 0.692 16.893*** 0.693 5.976*** 0.602 7.518*** 0.664 6.785*** 0.668
5 14.080*** 0.561 15.650*** 0.690 15.890*** 0.691 5.285*** 0.597 6.774*** 0.662 6.120*** 0.665
6 13.109*** 0.581 14.458*** 0.683 14.681*** 0.684 4.622*** 0.595 5.987*** 0.655 5.461*** 0.658
7 12.037*** 0.585 13.356*** 0.670 13.533*** 0.670 3.881*** 0.575 4.903*** 0.622 4.662*** 0.623
8 10.705*** 0.584 11.955*** 0.653 12.161*** 0.653 3.623*** 0.534 4.405*** 0.574 4.306*** 0.574
9 9.000*** 0.572 10.167*** 0.628 10.241*** 0.628 4.184*** 0.419 4.831*** 0.462 4.814*** 0.462
Low 7.933*** 0.527 9.053*** 0.579 9.022*** 0.579 4.163*** 0.145 4.369*** 0.164 4.515*** 0.165
2
This table reports alphas and adjusted R of the regressions of the CMAP excess returns on the market factor, the Fama–French three-factors, and the
Carhart four-factors in two subperiods using portfolios sorted by size, book-to-market, momentum, and standard deviation of return. The alphas are
annualized and in percent. The sample period covers January 4, 1960 until December 31, 2013 with value-weighted portfolio returns and is split in

© 2016 International Review of Finance Ltd. 2016

two sub-periods of equal length. The lengths of the moving average windows are 5, 10, 20, 50, 100, and 200 days. Newey and West (1987) standard
errors with 24 lags are used in reporting statistical significance of a two-sided null hypothesis at the 1, 5, and 10% level is given by a ***, a **, and a *,
respectively. α is the intercept of the regression, βm is the factor loading on MKT, βs os the factor loading on SMB, βh is the factor loading on HML, βu is
2
the factor loading on UMD, and R is the adjusted R-squared of the regression.
 Combination of Moving Averages

Table 4 Regression of CMA returns on BH returns

Panel A: size sorted portfolios
2 2
Portfolio α βBH R α βBH γBH R
Low 18.781*** 0.406*** 0.551 7.654*** 0.493*** 0.155*** 0.563
2 15.472*** 0.404*** 0.554 12.446*** 0.423*** 0.035*** 0.555
3 14.306*** 0.418*** 0.570 11.405*** 0.435*** 0.033*** 0.570
4 13.383*** 0.411*** 0.563 10.999*** 0.425*** 0.027*** 0.563
5 12.857*** 0.415*** 0.562 10.521*** 0.430*** 0.027*** 0.563
6 12.744*** 0.412*** 0.559 8.964*** 0.436*** 0.046*** 0.560
7 12.465*** 0.411*** 0.558 7.801*** 0.441*** 0.056*** 0.560
8 11.074*** 0.415*** 0.563 7.553*** 0.437*** 0.042*** 0.564
9 9.263*** 0.414*** 0.561 6.319*** 0.433*** 0.036*** 0.561
High 5.538*** 0.417*** 0.564 5.116*** 0.419*** 0.005 0.564
Panel B: book-to-market sorted portfolios
2 2
Portfolio α βBH R α βBH γBH R
Low 8.482*** 0.407*** 0.554 8.005*** 0.409*** 0.005 0.554
2 7.849*** 0.421*** 0.564 4.740*** 0.439*** 0.036*** 0.565
3 8.293*** 0.430*** 0.575 3.717*** 0.458*** 0.055*** 0.576
4 8.669*** 0.423*** 0.569 4.123*** 0.451*** 0.054*** 0.570
5 7.687*** 0.420*** 0.562 6.587*** 0.426*** 0.013 0.562
6 8.381*** 0.430*** 0.579 6.446*** 0.442*** 0.024*** 0.579
7 8.640*** 0.438*** 0.583 5.795*** 0.456*** 0.036*** 0.583
8 8.760*** 0.418*** 0.556 9.216*** 0.415*** 0.006 0.556
9 9.344*** 0.442*** 0.583 7.621*** 0.453*** 0.020** 0.583
High 11.792*** 0.446*** 0.589 6.039*** 0.476*** 0.061*** 0.591
Panel C: momentum sorted portfolios
2 2
Portfolio α βBH R α βBH γBH R
Low 15.343*** 0.353*** 0.509 9.854*** 0.375*** 0.046*** 0.511
2 11.648*** 0.396*** 0.537 10.259*** 0.402*** 0.014* 0.537
3 9.588*** 0.413*** 0.559 7.415*** 0.424*** 0.024*** 0.559
4 8.867*** 0.415*** 0.567 4.933*** 0.437*** 0.046*** 0.568
5 7.442*** 0.419*** 0.564 7.061*** 0.421*** 0.005 0.564
6 7.573*** 0.426*** 0.570 5.782*** 0.437*** 0.022** 0.571
7 7.973*** 0.422*** 0.568 5.750*** 0.436*** 0.027*** 0.568
8 7.091*** 0.460*** 0.607 4.042*** 0.479*** 0.036*** 0.607
9 8.354*** 0.440*** 0.586 6.568*** 0.450*** 0.020** 0.586
High 11.891*** 0.440*** 0.589 12.480*** 0.438*** 0.005 0.589
Panel D: standard deviation sorted portfolios
2 2
Portfolio α βBH R α βBH γBH R
High 28.846*** 0.512*** 0.634 0.407 0.633*** 0.265*** 0.657
2 21.177*** 0.415*** 0.555 6.039*** 0.491*** 0.148*** 0.565
3 18.652*** 0.401*** 0.544 8.735*** 0.455*** 0.103*** 0.549
4 16.815*** 0.401*** 0.542 8.785*** 0.448*** 0.092*** 0.546
5 16.134*** 0.402*** 0.545 8.599*** 0.451*** 0.094*** 0.549

© 2016 International Review of Finance Ltd. 2016 15

 International Review of Finance

Table 4 (continued)
Panel D: standard
A: size sorteddeviation
portfoliossorted portfolios
2 2
Portfolio α βBH R α βBH γBH R
6 15.014*** 0.400*** 0.543 8.145*** 0.449*** 0.093*** 0.547
7 14.155*** 0.404*** 0.542 7.401*** 0.458*** 0.102*** 0.547
8 13.378*** 0.400*** 0.530 4.958*** 0.476*** 0.145*** 0.540
9 12.701*** 0.388*** 0.508 4.759*** 0.476*** 0.168*** 0.522
Low 11.851*** 0.388*** 0.511 4.762*** 0.496*** 0.210*** 0.534
2
This table reports alphas, betas, and adjusted R of the regressions of the CMA returns on the BH
returns as well as the negative component of the BH return using portfolios sorted by size, book-
to-market, momentum, and standard deviation of return. The alphas are annualized and in per-
cent. The sample period covers January 4, 1960 until December 31, 2013 with value-weighted port-
folio returns. The lengths of the moving average windows are 5, 10, 20, 50, 100, and 200 days.
Newey and West (1987) standard errors with three lags are used in reporting statistical significance
of a two-sided null hypothesis at the 1, 5, and 10% level is given by a ***, a **, and a *, respectively.

the market timing coefficient, γBH is almost always positive and statistically sig-
nificant, especially for portfolios sorted on market capitalization and standard de-
viation of return.

C. Statistical significance, trading intensity, and break-even transaction

costs
Table 5 reports the statistical significance in the improvement of the average re-
turn Δμ of the CMA portfolio over the BH portfolio as well as the reduction in the
return standard deviation Δσ. The evidence points towards a substantial improve-
ment in a mean-variance sense for all sets of portfolios under consideration with
the exception of the highest market cap decile and momentum decile eight. The
annualized improvement in the average return ranges from over 1% to just under
14% per annum, while the reduction in the standard deviation is between ap-
proximately 3% to over 12%. The CMA strategy is active more often than not
ranging between 54% and 71% of the sample. Yet, the number of transactions,
number of trades (NT), is never above 6000 and can be as little as under 4000
for decile 10 of standard deviation sorted portfolios. In a sample of 13,592 days
this translates into average holding periods of between 2 and 3 days between
transactions. Next, I report the break-even transaction costs, break-even transac-
tion costs (BETC), calculated as the level of one-way proportional transaction
cost in percent that would eliminate completely the average CMAP portfolio re-
turn. The values of the BETC for the various sets of portfolio range between al-
most 0.00% and as high as 0.15%. Finally, the last two columns report the
fraction of months that the CMA strategy generates a positive return (p1) as well
as a return that is in excess of the risk-free rate (p2). With the exception of three
momentum, all the reported fractions range from 68% to 75% success rate of the
CMA strategy delivering a positive return and 51% to 60% probability of the

16 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 5 Trading frequency and break-even transaction cost

Panel A: size sorted portfolios
Portfolio Δμ Δσ pA NT BETC p1 p2
Low 11.73 6.15 0.62 4247 0.15 0.75 0.58
2 8.43 7.64 0.62 4656 0.10 0.73 0.56
3 6.80 7.40 0.64 4740 0.08 0.72 0.56
4 6.25 7.38 0.64 4882 0.07 0.72 0.56
5 5.48 7.23 0.64 4906 0.06 0.71 0.56
6 5.60 6.87 0.65 4935 0.06 0.71 0.56
7 5.28 6.96 0.65 4950 0.06 0.71 0.55
8 4.11 7.01 0.65 5080 0.04 0.70 0.55
9 2.54 6.90 0.65 5332 0.03 0.70 0.54
High 0.34 7.12 0.64 5655 0.00 0.68 0.53
Panel B: book-to-market sorted portfolios
Portfolio Δμ Δσ pA NT BETC p1 p2
Low 2.95 7.94 0.61 5559 0.03 0.70 0.53
2 1.55 7.06 0.63 5495 0.02 0.69 0.53
3 2.02 6.67 0.64 5513 0.02 0.69 0.53
4 2.43 6.93 0.64 5376 0.02 0.69 0.54
5 1.24 6.87 0.64 5509 0.01 0.68 0.53
6 1.67 6.58 0.65 5256 0.02 0.68 0.54
7 1.65 6.34 0.65 5313 0.02 0.69 0.54
8 1.26 6.94 0.66 5312 0.01 0.68 0.54
9 1.64 6.73 0.66 5209 0.02 0.68 0.54
High 3.53 7.46 0.65 5175 0.04 0.69 0.54
Panel C: momentum sorted portfolios
Portfolio Δμ Δσ pA NT BETC p1 p2
Low 13.72 12.59 0.54 5217 0.14 0.73 0.51
2 6.86 9.36 0.58 5391 0.07 0.71 0.51
3 3.59 7.88 0.62 5458 0.04 0.69 0.52
4 2.92 7.46 0.62 5420 0.03 0.69 0.53
5 1.63 7.03 0.63 5493 0.02 0.69 0.53
6 1.35 6.70 0.64 5478 0.01 0.69 0.53
7 1.76 6.68 0.64 5423 0.02 0.69 0.54
8 0.15 6.33 0.66 5532 0.00 0.69 0.55
9 1.26 7.00 0.65 5445 0.01 0.69 0.54
High 1.98 8.68 0.65 5201 0.02 0.70 0.56
Panel D: standard deviation sorted portfolios
Portfolio Δμ Δσ pA NT BETC p1 p2
High 9.32 7.23 0.67 4316 0.12 0.72 0.58
2 10.82 8.55 0.62 4738 0.12 0.73 0.56
3 9.59 8.32 0.63 4715 0.11 0.73 0.56
4 7.43 7.69 0.65 4711 0.09 0.72 0.57
5 6.81 7.05 0.66 4641 0.08 0.72 0.57

© 2016 International Review of Finance Ltd. 2016 17

 International Review of Finance

Table 5 (continued)
Panel D: standard
A: size sorteddeviation
portfoliossorted portfolios
Portfolio Δμ Δσ pA NT BETC p1 p2
6 6.18 6.57 0.67 4630 0.07 0.71 0.57
7 5.64 5.76 0.68 4603 0.07 0.72 0.58
8 5.21 5.03 0.69 4612 0.06 0.71 0.58
9 5.07 4.22 0.70 4441 0.06 0.72 0.59
Low 5.25 3.14 0.71 3961 0.07 0.73 0.60

This table reports the results for the improvement delivered by the MA switching strategy over the
buy-and-hold strategy, the trading frequency as well as the break-even transaction cost using 10 dec-
ile portfolios sorted by size, book-to-market, momentum, and standard deviation of return. The
sample period covers January 4, 1960 until December 31, 2013 with value-weighted portfolio
returns. Δμ is the annualized improvement in the average in-sample daily return, Δσ isthe annual-
ized improvement in the return standard deviation, pA is the proportion of days during which there
is a hold signal, NT is the number of transactions (buy or sell) over the entire sample period, BETC is
the break-even one-sided transaction cost in percent, p1 is the proportion of days during which a buy
signal was followed by a positive return of the underlying portfolio and p2 is the proportion of days
during which a buy signal was followed by a portfolio return in excess of the risk-free rate. The
lengths of the moving average windows are 5, 10, 20, 50, 100, and 200 days. The moving average
portfolio is an equal-weighted combination of the six individual moving average returns.

CMA strategy having a positive excess return. These values indicate that, more of-
ten than not, the CMA strategy is on the right side of the market. These consid-
erably favorable odds are in line with the evidence reported previously regarding
the superior performance of the CMA switching strategy.

IV. DRIVERS OF ABNORMAL RETURNS

In this section, I investigate the reasons for the superior returns of the CMAP
portfolios. To this end, I control the CMAP performance for economic expan-
sions and contractions as well as other state contingencies like the sign of the
lagged market return. Furthermore, I investigate the conditional performance
of the CMAP returns while controlling for three instrumental variables with doc-
umented predictive power over stock returns.

A. Market timing
The first approach towards testing for market timing ability is the quadratic re-
gression of Treynor and Mazuy (1966)

CMAPjt;L ¼ αj þ βj;m rmkt;t þ βj;m2 r2mkt;t þ εjt ; j ¼ 1; …; N (9)

where statistically significant evidence of a positive βj;m2 can be interpreted as ev-

idence in favor of market timing ability. The second approach is to allow for a

18 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

state-contingent βj,m based on the direction of move of the market return as in

Henriksson and Merton (1981)

CMAPjt;L ¼ αj þ βj;m rmkt;t þ γj;m rmkt;t I frmkt;t >0g þ εjt ; j ¼ 1; …; N (10)

where I frmkt;t >0g is an indicator function of the event of a positive market return. A
statistically significant value of γj,m is usually interpreted as evidence of successful
market timing ability.
Table 6 presents the results of the two market timing regressions for various
sets of value-weighted decile portfolios. Panel TM presents the empirical results
from the Treynor and Mazuy (1966) quadratic regression while Panel HM pre-
sents the results for the state-contingent beta regression of Henriksson and Mer-
ton (1981). In both regressions, both βj;m2 and γj,m are highly statistically
significant, indicating there is strong evidence of market timing ability of the
switching moving average strategy. Nevertheless, a few portfolios have negative
or insignificant values of βma and γm suggesting that the market timing ability
of the CMA strategy does not apply for all of the portfolios under consideration.
This finding suggests that market timing alone is not the sole driver of the abnor-
mal returns generated by the combination moving average strategy.

B. Business cycles and market states

Following Han et al. (2013), I investigate the performance of the CMAP portfolio
returns conditional on the dividend yield of the stock market as well as the short-
term risk-free rate. Table 7 presents the results for the various sets of portfolio dec-
iles. The evidence overwhelmingly indicates that CMAP abnormal returns are
higher small-cap portfolios, high book-to-market portfolio, loser portfolios as
well as both high and low risk portfolios sorted on standard deviation of return.
The first notable finding in Table 7 is the reduced statistical significance of the
abnormal returns. The α s are reduced in value as well, occasionally turning neg-
ative and, for the most part, not statistically significant. This is an indication that
this conditional model is capable of capturing some of the apparent abnormal
performance detected by the unconditional models presented previously in the
paper.
The majority of CMAP portfolios experience an increase in average return
when the market dividend yield increases. Similarly, for most portfolios there is
an increase in average returns when the short-term risk-free rate rises. In term
of the time-varying factor loadings, most market betas of the CMAP portfolios be-
come less negative when the market’s dividend yield increases. The impact of the
risk-free rate on the market betas is the opposite. A less consistent pattern
emerges for the time-varying SMB loading. The SMB beta of the CMAP spread in-
creases with DPm for portfolios sorted on market capitalization is mixed for port-
folios sorted on book-to-market and decreases with DPm for portfolios sorted on
momentum and standard deviation of return. The effect of the risk-free rate on

© 2016 International Review of Finance Ltd. 2016 19

20
Table 6 Market timing Regressions
Panel A: size sorted portfolios
2 2
Portfolio α βm βm2 R α βm γm R
TM HM
Low 11.499*** 0.419*** 0.011*** 0.488 3.257** 0.355*** 0.133*** 0.489
2 10.661*** 0.555*** 0.004*** 0.551 6.084*** 0.522*** 0.068*** 0.552
3 9.370*** 0.558*** 0.003*** 0.581 6.131*** 0.534*** 0.048*** 0.581
4 8.861*** 0.564*** 0.003*** 0.604 6.620*** 0.546*** 0.036*** 0.604
5 8.252*** 0.568*** 0.002** 0.622 6.113*** 0.552*** 0.033*** 0.622
6 7.783*** 0.546*** 0.004*** 0.642 5.484*** 0.527*** 0.041*** 0.642
7 7.375*** 0.557*** 0.005*** 0.656 5.045*** 0.537*** 0.043*** 0.655
8 6.242*** 0.573*** 0.005*** 0.681 4.278*** 0.555*** 0.039*** 0.680
9 4.311*** 0.571*** 0.007*** 0.696 3.706*** 0.560*** 0.027*** 0.695
High 1.881** 0.595*** 0.005*** 0.709 3.550*** 0.600*** 0.004 0.708
Panel B: book-to-market sorted portfolios
2 2
Portfolio α βm βm2 R α βm γm R

TM HM
Low 7.119*** 0.646*** 0.001 0.673 9.258*** 0.661*** 0.030*** 0.673
International Review of Finance

2 4.021*** 0.585*** 0.004*** 0.680 5.087*** 0.587*** 0.001 0.680

3 3.728*** 0.547*** 0.006*** 0.664 3.797*** 0.540*** 0.018* 0.663
4 3.225*** 0.561*** 0.010*** 0.664 2.248* 0.544*** 0.043*** 0.661
5 3.427*** 0.545*** 0.004*** 0.624 4.633*** 0.548*** 0.002 0.623
6 3.506*** 0.522*** 0.005*** 0.631 3.521*** 0.517*** 0.015 0.630
7 2.638*** 0.500*** 0.008*** 0.610 2.004* 0.487*** 0.032*** 0.608
8 2.853*** 0.524*** 0.006*** 0.560 3.087** 0.519*** 0.016 0.559
9 2.803*** 0.523*** 0.008*** 0.581 2.601** 0.513*** 0.026** 0.579
High 4.914*** 0.567*** 0.008*** 0.559 4.149*** 0.554*** 0.033*** 0.557

© 2016 International Review of Finance Ltd. 2016

 Table 6 (continued)
Panel C: momentum sorted portfolios
2 2
Portfolio α βm β m2 R α βm γm R
TM HM
Low 21.010*** 0.874*** 0.009*** 0.546 23.220*** 0.898*** 0.053*** 0.545
2 13.002*** 0.716*** 0.008*** 0.592 16.258*** 0.745*** 0.063*** 0.591
3 8.981*** 0.615*** 0.007*** 0.610 11.261*** 0.637*** 0.049*** 0.609
4 6.017*** 0.588*** 0.002* 0.638 6.966*** 0.592*** 0.007 0.638
5 5.339*** 0.564*** 0.002* 0.638 7.862*** 0.581*** 0.035*** 0.638
6 3.015*** 0.536*** 0.006*** 0.638 4.725*** 0.540*** 0.002 0.637
7 3.502*** 0.538*** 0.006*** 0.651 3.749*** 0.533*** 0.015 0.650
8 0.358 0.513*** 0.010*** 0.631 0.304 0.497*** 0.039*** 0.627
9 2.847*** 0.567*** 0.007*** 0.643 2.071*** 0.554*** 0.031*** 0.641
High 3.543*** 0.659*** 0.010*** 0.572 1.684 0.637*** 0.051*** 0.570

© 2016 International Review of Finance Ltd. 2016

Panel D: standard deviation sorted portfolios
2 2
Portfolio α βm β m2 R α βm γm R
TM HM
High 11.014*** 0.515*** 0.006*** 0.404 5.661*** 0.475*** 0.081*** 0.405
2 12.659*** 0.633*** 0.008*** 0.547 6.690*** 0.587*** 0.096*** 0.547
3 12.054*** 0.625*** 0.005*** 0.578 8.706*** 0.599*** 0.056*** 0.578
4 9.650*** 0.580*** 0.005*** 0.578 7.071*** 0.558*** 0.046*** 0.578
5 8.674*** 0.531*** 0.005*** 0.579 5.550*** 0.506*** 0.054*** 0.579
Combination of Moving Averages

6 7.449*** 0.496*** 0.007*** 0.588 4.258*** 0.468*** 0.059*** 0.587

7 6.669*** 0.435*** 0.006*** 0.576 4.304*** 0.413*** 0.048*** 0.575
8 5.274*** 0.373*** 0.009*** 0.552 2.153** 0.343*** 0.064*** 0.549
9 4.617*** 0.291*** 0.009*** 0.478 0.986 0.259*** 0.071*** 0.474
Low 4.149*** 0.163*** 0.009*** 0.288 0.730 0.133*** 0.067*** 0.282
2
This table reports alphas, betas, and adjusted R of the market timing regressions of the CMAP excess returns on the market factor using portfolios
sorted by size, book-to-market, momentum and standard deviation of return. The TM panel reports the results using the Treynor and Mazuy
(1966) quadratic regression with the squared market factor (βm2 ) while the HM panel reports the results using the Henriksson and Merton (1981) re-
gression with option-like returns on the market (γm). The sample period covers January 4, 1960 until December 31, 2013 with value-weighted port-
folio returns. The lengths of the moving average windows are 5, 10, 20, 50, 100, and 200 days. The moving average portfolio is an equal-weighted
combination of the six individual moving average returns. Newey and West (1987) standard errors with 3 lags are used in reporting statistical signif-

21
icance of a two-sided null hypothesis at the 1, 5, and 10% level is given by a ***, a **, and a *, respectively.
 International Review of Finance

Table 7 Conditional regressions with market dividend yield and treasury bill rate
Panel A: size sorted portfolios

P α DPm rf βm βs βh βu rm × DPm

Low 13.443*** 0.007 1.478*** 0.582*** 0.393*** 0.523*** 0.042*** 0.059***

2 6.037** 0.001 1.289*** 0.853*** 0.592*** 0.638*** 0.051*** 0.101***
3 0.719 0.009*** 0.733* 0.867*** 0.551*** 0.545*** 0.077*** 0.115***
4 0.352 0.009** 0.754** 0.881*** 0.501*** 0.487*** 0.079*** 0.125***
5 2.668 0.013*** 0.345 0.893*** 0.418*** 0.426*** 0.107*** 0.132***
6 1.724 0.012*** 0.428 0.813*** 0.232*** 0.335*** 0.098*** 0.116***
7 0.180 0.009** 0.356 0.826*** 0.168*** 0.417*** 0.061*** 0.110***
8 2.056 0.010** 0.378 0.860*** 0.110*** 0.392*** 0.095*** 0.119***
9 4.844** 0.014*** 0.061 0.833*** 0.042** 0.390*** 0.049*** 0.105***
High 3.076 0.011*** 0.599* 0.796*** 0.154*** 0.132*** 0.138*** 0.092***

Panel B: book-to-market sorted portfolios

P α DPm rf βm βs βh βu rm × DPm

Low 9.240*** 0.021*** 0.401 0.706*** 0.144*** 0.149*** 0.108*** 0.049***

2 6.610** 0.016*** 0.269 0.798*** 0.129*** 0.291*** 0.102*** 0.090***
3 5.211** 0.011*** 0.358 0.757*** 0.140*** 0.367*** 0.037*** 0.093***
4 2.494 0.011*** 0.041 0.809*** 0.085*** 0.567*** 0.034*** 0.090***
5 0.746 0.005 0.317 0.864*** 0.012 0.759*** 0.100*** 0.120***
6 0.849 0.008* 0.042 0.835*** 0.018 0.583*** 0.096*** 0.108***
7 0.288 0.006 0.360 0.770*** 0.032* 0.672*** 0.071*** 0.087***
8 0.393 0.007 0.005 0.756*** 0.006 0.780*** 0.118*** 0.062***
9 1.567 0.007* 0.035 0.830*** 0.037* 0.722*** 0.117*** 0.087***
High 2.134 0.012** 0.392 0.828*** 0.086*** 0.871*** 0.153*** 0.066***

Panel C: momentum sorted portfolios

P α DPm rf βm βs βh βu rm × DPm

Low 12.035*** 0.004 0.838 1.174*** 0.452*** 0.534*** 0.975*** 0.167***

2 0.395 0.008 0.151 1.019*** 0.120*** 0.448*** 0.634*** 0.131***
3 1.453 0.010** 0.363 0.874*** 0.064*** 0.487*** 0.404*** 0.107***
4 2.238 0.009** 0.125 0.809*** 0.131*** 0.507*** 0.242*** 0.082***
5 6.022** 0.012*** 0.242 0.728*** 0.164*** 0.333*** 0.156*** 0.058***
6 7.755*** 0.015*** 0.257 0.709*** 0.166*** 0.418*** 0.052*** 0.054***
7 3.062 0.007 0.744** 0.733*** 0.183*** 0.369*** 0.000 0.073***
8 7.007*** 0.013*** 0.360 0.706*** 0.170*** 0.380*** 0.071*** 0.063***
9 3.984 0.013*** 0.088 0.765*** 0.099*** 0.318*** 0.125*** 0.071***
High 4.622 0.019*** 0.398 0.871*** 0.033 0.063** 0.322*** 0.085***

Panel D: standard deviation sorted portfolios

P α DPm rf βm βs βh βu rm × DPm

High 11.674*** 0.006 0.644 0.554*** 0.182*** 0.508*** 0.012 0.006

2 6.758* 0.010* 0.208 0.768*** 0.218*** 0.636*** 0.171*** 0.054***
3 3.590 0.013** 0.282 0.755*** 0.132*** 0.669*** 0.139*** 0.039***
4 0.744 0.014*** 0.155 0.753*** 0.095*** 0.683*** 0.117*** 0.046***
5 0.480 0.013*** 0.234 0.701*** 0.013 0.681*** 0.113*** 0.044***
6 0.405 0.011*** 0.279 0.687*** 0.009 0.664*** 0.085*** 0.055***
7 0.867 0.007* 0.606* 0.602*** 0.068*** 0.583*** 0.047*** 0.055***
8 1.508 0.006* 0.427 0.494*** 0.095*** 0.479*** 0.022** 0.040***
9 3.678* 0.003 0.354 0.363*** 0.081*** 0.405*** 0.010 0.031***
Low 4.102** 0.003 0.216 0.082*** 0.142*** 0.107*** 0.034*** 0.024***

22 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 7 (continued)
Panel A: size sorted portfolios
2
P rs × DPm rh × DPm ru × DPm rm × rf rs × rf rh × rf ru × rf R
Low 0.024*** 0.152*** 0.017*** 4.879*** 10.772*** 4.080*** 0.329 0.673
2 0.052*** 0.207*** 0.009 3.302*** 7.580*** 6.060*** 1.298** 0.719
3 0.063*** 0.185*** 0.026*** 4.298*** 8.758*** 5.695*** 0.325 0.729
4 0.071*** 0.176*** 0.029*** 5.157*** 10.058*** 6.506*** 0.392 0.732
5 0.081*** 0.145*** 0.039*** 4.937*** 12.411*** 3.947*** 0.076 0.731
6 0.040*** 0.111*** 0.047*** 5.608*** 10.377*** 3.323*** 1.373*** 0.713
7 0.053*** 0.153*** 0.022*** 4.566*** 12.575*** 5.318*** 0.103 0.708
8 0.048*** 0.152*** 0.035*** 4.749*** 11.050*** 6.973*** 0.087 0.718
9 0.014* 0.152*** 0.019*** 3.469*** 8.231*** 6.249*** 0.120 0.718
High 0.010 0.059*** 0.060*** 1.897*** 1.964*** 0.474 2.623*** 0.731

Panel B: book-to-market sorted portfolios

2
P rs × DPm rh × DPm ru × DPm rm × rf rs × rf rh × rf ru × rf R
Low 0.011 0.017 0.056*** 1.274*** 4.049*** 5.340*** 4.373*** 0.711
2 0.016* 0.202*** 0.046*** 2.797*** 4.897*** 10.659*** 1.329*** 0.702
3 0.018** 0.182*** 0.044*** 4.167*** 5.347*** 7.583*** 3.319*** 0.684
4 0.014* 0.210*** 0.010* 1.919*** 3.679*** 7.680*** 2.873*** 0.692
5 0.048*** 0.262*** 0.023*** 2.717*** 7.987*** 7.951*** 1.847*** 0.671
6 0.009 0.213*** 0.029*** 1.627*** 4.547*** 9.329*** 0.287 0.676
7 0.007 0.215*** 0.033*** 1.688*** 5.249*** 9.403*** 0.417 0.669
8 0.031*** 0.160*** 0.004 0.546 7.492*** 2.411*** 4.656*** 0.667
9 0.000 0.192*** 0.052*** 0.353 2.673*** 9.501*** 0.439 0.672
High 0.002 0.212*** 0.048*** 0.129 5.446*** 6.070*** 1.751*** 0.660

Panel C: momentum sorted portfolios

2
P rs × DPm rh × DPm ru × DPm rm × rf rs × rf rh × rf ru × rf R
Low 0.176*** 0.062*** 0.215*** 6.269*** 17.333*** 13.240*** 8.324*** 0.711
2 0.066*** 0.150*** 0.132*** 1.847*** 8.611*** 1.126 5.374*** 0.709
3 0.018* 0.182*** 0.062*** 1.509*** 8.265*** 4.413*** 2.068*** 0.704
4 0.021** 0.198*** 0.036*** 0.797* 4.215*** 6.267*** 0.174 0.690
5 0.019** 0.111*** 0.048*** 0.254 6.230*** 1.826** 2.672*** 0.664
6 0.034*** 0.162*** 0.036*** 0.637 3.551*** 6.418*** 2.636*** 0.660
7 0.040*** 0.150*** 0.023*** 2.500*** 4.107*** 6.996*** 0.549 0.671
8 0.054*** 0.149*** 0.024*** 1.575*** 1.142* 7.424*** 0.564 0.661
9 0.039*** 0.106*** 0.025*** 2.332*** 3.311*** 2.800*** 0.581 0.680
High 0.020* 0.014 0.005 2.520*** 4.598*** 7.884*** 0.478 0.657

Panel D: standard deviation sorted portfolios

2
P rs × DPm rh × DPm ru × DPm rm × rf rs × rf rh × rf ru × rf R
High 0.017 0.145*** 0.006 4.074*** 16.779*** 2.224* 2.597*** 0.523
2 0.016 0.226*** 0.057*** 6.076*** 14.611*** 7.389*** 2.584*** 0.653
3 0.034*** 0.230*** 0.046*** 4.050*** 13.832*** 7.600*** 1.871*** 0.682
4 0.032*** 0.233*** 0.030*** 2.073*** 12.118*** 7.829*** 0.285 0.680
5 0.044*** 0.238*** 0.027*** 1.791*** 10.990*** 8.997*** 0.406 0.679
6 0.035*** 0.223*** 0.021*** 1.821*** 9.622*** 7.633*** 0.147 0.674
7 0.030*** 0.214*** 0.020*** 3.459*** 10.125*** 10.167*** 1.089** 0.651
8 0.024*** 0.181*** 0.019*** 3.502*** 9.827*** 9.726*** 1.836*** 0.615
9 0.006 0.153*** 0.015*** 4.864*** 9.888*** 9.783*** 1.992*** 0.541
Low 0.025*** 0.052*** 0.002 4.414*** 7.472*** 7.701*** 1.492*** 0.358
2
This table reports alphas, betas, and adjusted R of the market timing regressions of the CMAP excess returns on
the C4 factors along with two instrumental variables and interaction terms of the instrumental variables with
the C4 factors using portfolios sorted by various characteristics. Alphas are annualized and in percent. The sam-
ple period covers 1960/01/04 until 2013/12/31. The lengths of the moving average windows are 5, 10, 20, 50,
100, and 200 days. Newey and West (1987) standard errors with 3 lags are used in reporting statistical signifi-
cance of a twosided null hypothesis at the 1, 5, and 10% level is given by a ***, a **, and a **, respectively.

© 2016 International Review of Finance Ltd. 2016 23

 International Review of Finance

the SMB loading of the CMAP portfolios is uniformly negative. A much more
consistent pattern emerges for the HML time-varying loading. All HML betas of
the CMAP spreads increase with DPm and decrease with rf. Furthermore, these ef-
fects are very highly statistically significant. Finally, the UMD loadings of the
CMAP portfolios uniformly increase with DPm but have a mixed reaction to rf.
The risk-free rate has a mixed effect on the UMD loadings of CMAP portfolio
sorted on market capitalization and book-to-market and, largely, a positive
impact for portfolios sorted on momentum and standard deviation of return.

C. Conditional models with macroeconomic variables

Ferson and Schadt (1996) make a strong case for using predetermined variables in
controlling for changes in economic conditions while evaluating investment per-
formance. I augment the four-factor Carhart (1997) model with an intercept that is
a linear function of a set of instruments as well as cross-products of the instrumen-
tal variables with the market return to allow for state-dependent betas with the
market factor. In this conditional model, the state variables Zt consist of a recession
indicator taking on value of one during economic contractions, and a value of zero
during economic expansions as well as a down market dummy variable taking on a
value of one when the portfolio return is negative and a value of zero otherwise

CMAPjt;L ¼ αj þ βj;m rmkt;t þ βj;s rsmb;t þ βj;h rhml;t þ βj;u rumd;t þ

  
γj;Z Z t1 ⊗ 1T ; rmkt;t ; rsmb;t ; rhml;t ; rumd;t þ εjt ; j ¼ 1; …; N:

Table 8 presents the results of the conditional model estimation. The most no-
table result that emerges is that the abnormal returns increase in magnitude and
are uniformly statistically significant. Despite this disconcerting finding, it is still
of interest to note the response of the CMAP returns and factor loadings to the
two state variables. First, it is notable that the abnormal returns of all CMAP
spread portfolios are larger during economic recessions though the coefficients
are statistically significant mostly for portfolios sorted on market capitalization
and standard deviation of return. Secondly, the abnormal returns of the CMAP
portfolios are uniformly reduced during down markets with all of the coefficients
highly statistically significant. Furthermore, the CMAP portfolio loadings on all
four factors are reduced during recessions, while the same effect in down markets
is only apparent for loadings on the SMB factor. Conversely, the loadings on the
momentum factor appear to increase during down markets though this finding is
statistically significant mostly for the CMAP portfolios sorted on market capital-
ization and book-to-market with only a handful of momentum and standard de-
viation portfolios exhibiting statistical significance for this coefficient.
Finally, I consider estimating a conditional model using the market’s dividend
yield and a recession indicator as the two state variables with the four Carhart
(1997) factors as well as interactions between the instrumental variables and
the factor returns. Table 9 presents the empirical findings of this conditional

24 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 8 Conditional regressions with recession indicator and down market indicator
Panel A: size sorted portfolios

P α RI DN βm βs βh βu rm × RI

Low 29.089*** 0.044*** 0.167*** 0.462*** 0.444*** 0.128*** 0.036*** 0.181***

2 40.610*** 0.030*** 0.226*** 0.658*** 0.527*** 0.164*** 0.032*** 0.241***
3 40.359*** 0.030*** 0.231*** 0.663*** 0.477*** 0.102*** 0.028*** 0.215***
4 42.304*** 0.029*** 0.249*** 0.681***0 0.441*** 0.078*** 0.031*** 0.175***
5 41.313*** 0.031*** 0.256*** 0.663*** 0.383*** 0.026** 0.025*** 0.193***
6 39.200*** 0.027*** 0.256*** 0.625*** 0.264*** 0.020* 0.028*** 0.166***
7 38.067*** 0.030*** 0.256*** 0.614*** 0.215*** 0.026** 0.037*** 0.203***
8 38.550*** 0.025*** 0.273*** 0.632*** 0.137*** 0.030*** 0.033*** 0.195***
9 37.909*** 0.023*** 0.280*** 0.625*** 0.020** 0.045*** 0.049*** 0.198***
High 32.747*** 0.012 0.271*** 0.616*** 0.177*** 0.047*** 0.008 0.171***

Panel B: book-to-market sorted portfolios

P α RI DN βm βs βh βu rm × RI

Low 32.447*** 0.027*** 0.237*** 0.631*** 0.070*** 0.274*** 0.022** 0.186***

2 34.914*** 0.019** 0.261*** 0.640*** 0.063*** 0.035*** 0.007 0.150***
3 33.586*** 0.014 0.253*** 0.590*** 0.053*** 0.021* 0.041*** 0.190***
4 34.116*** 0.020** 0.261*** 0.599*** 0.043*** 0.112*** 0.040*** 0.211***
5 36.038*** 0.009 0.253*** 0.607*** 0.013 0.154*** 0.018** 0.227***
6 35.112*** 0.026*** 0.251*** 0.607*** 0.032*** 0.157*** 0.008 0.157***
7 31.625*** 0.017* 0.227*** 0.573*** 0.035*** 0.253*** 0.031*** 0.165***
8 33.432*** 0.007 0.228*** 0.577*** 0.013 0.309*** 0.043*** 0.254***
9 36.562*** 0.009 0.246*** 0.648*** 0.034*** 0.376*** 0.019** 0.148***
High 35.316*** 0.022* 0.215*** 0.673*** 0.117*** 0.398*** 0.025** 0.213***

Panel C: momentum sorted portfolios

P α RI DN βm βs βh βu rm × RI

Low 43.575*** 0.032* 0.249*** 0.836*** 0.273*** 0.056*** 0.410*** 0.224***

2 37.441*** 0.043*** 0.247*** 0.705*** 0.045*** 0.002 0.325*** 0.267***
3 30.378*** 0.041*** 0.211*** 0.604*** 0.022** 0.018 0.224*** 0.251***
4 32.462*** 0.022** 0.239*** 0.592*** 0.057*** 0.089*** 0.080*** 0.260***
5 35.722*** 0.019* 0.252*** 0.605*** 0.037*** 0.058*** 0.049*** 0.226***
6 33.913*** 0.027*** 0.250*** 0.588*** 0.076*** 0.114*** 0.018** 0.202***
7 35.198*** 0.011 0.255*** 0.603*** 0.058*** 0.098*** 0.069*** 0.166***
8 31.968*** 0.015 0.248*** 0.584*** 0.059*** 0.131*** 0.149*** 0.136***
9 37.478*** 0.021** 0.259*** 0.647*** 0.023** 0.031** 0.217*** 0.161***
High 45.521*** 0.020 0.303*** 0.762*** 0.161*** 0.147*** 0.340*** 0.140***

Panel D: standard deviation sorted portfolios

P α RI DN βm βs βh βu rm × RI

High 38.434*** 0.024 0.226*** 0.596*** 0.440*** 0.129*** 0.002 0.172***

2 42.732*** 0.033** 0.261*** 0.703*** 0.411*** 0.139*** 0.012 0.231***
3 44.202*** 0.040*** 0.255*** 0.709*** 0.371*** 0.176*** 0.018* 0.267***
4 43.082*** 0.036*** 0.264*** 0.656*** 0.317*** 0.172*** 0.009 0.300***
5 38.868*** 0.037*** 0.248*** 0.592*** 0.244*** 0.184*** 0.009 0.296***
6 36.464*** 0.038*** 0.245*** 0.541*** 0.176*** 0.166*** 0.021** 0.296***
7 32.819*** 0.035*** 0.220*** 0.482*** 0.116*** 0.153*** 0.020** 0.246***
8 27.828*** 0.030*** 0.200*** 0.408*** 0.091*** 0.128*** 0.023*** 0.194***
9 20.571*** 0.026*** 0.153*** 0.295*** 0.053*** 0.094*** 0.027*** 0.186***
Low 9.808*** 0.020*** 0.072*** 0.141*** 0.001 0.030*** 0.031*** 0.109***

© 2016 International Review of Finance Ltd. 2016 25

 International Review of Finance

Table 8 (continued)
Panel A: size sorted portfolios
2
P rs × RI rh × RI ru × RI rm × DN rs × DN rh × DN ru × DN R
Low 0.171*** 0.069*** 0.063*** 0.052*** 0.026* 0.027* 0.008 0.683
2 0.225*** 0.099*** 0.123*** 0.030*** 0.019 0.047*** 0.034*** 0.737
3 0.216*** 0.092*** 0.124*** 0.040*** 0.033** 0.033** 0.039*** 0.744
4 0.175*** 0.066*** 0.105*** 0.045*** 0.025* 0.032** 0.036*** 0.743
5 0.126*** 0.095*** 0.099*** 0.034*** 0.044*** 0.009 0.043*** 0.741
6 0.097*** 0.057*** 0.086*** 0.009 0.054*** 0.020 0.028*** 0.724
7 0.028* 0.063*** 0.073*** 0.003 0.058*** 0.004 0.029*** 0.725
8 0.045*** 0.077*** 0.078*** 0.007 0.061*** 0.007 0.036*** 0.735
9 0.061*** 0.032** 0.068*** 0.009 0.070*** 0.003 0.039*** 0.742
High 0.016 0.002 0.060*** 0.015* 0.103*** 0.014 0.045*** 0.755

Panel B: book-to-market sorted portfolios

2
P rs × RI rh × RI ru × RI rm × DN rs × DN rh × DN ru × DN R
Low 0.002 0.069*** 0.019 0.019* 0.081*** 0.062*** 0.046*** 0.732
2 0.079*** 0.085*** 0.126*** 0.003 0.087*** 0.013 0.042*** 0.714
3 0.066*** 0.113*** 0.123*** 0.012 0.064*** 0.014 0.022** 0.704
4 0.078*** 0.078*** 0.054*** 0.032*** 0.079*** 0.011 0.036*** 0.715
5 0.012 0.131*** 0.106*** 0.015 0.074*** 0.004 0.027** 0.686
6 0.092*** 0.053*** 0.091*** 0.008 0.052*** 0.015 0.042*** 0.685
7 0.076*** 0.018 0.081*** 0.015 0.115*** 0.009 0.026*** 0.681
8 0.094*** 0.336*** 0.007 0.010 0.103*** 0.030* 0.021** 0.702
9 0.110*** 0.065*** 0.126*** 0.009 0.068*** 0.004 0.019* 0.679
High 0.201*** 0.122*** 0.100*** 0.017 0.054*** 0.002 0.023* 0.669

Panel C: momentum sorted portfolios

2
P rs × RI rh × RI ru × RI rm × DN rs × DN rh × DN ru × DN R
Low 0.095*** 0.319*** 0.172*** 0.032* 0.079*** 0.119*** 0.103*** 0.708
2 0.011 0.149*** 0.041*** 0.031** 0.131*** 0.052*** 0.048*** 0.719
3 0.004 0.133*** 0.055*** 0.021* 0.132*** 0.002 0.046*** 0.718
4 0.032* 0.057*** 0.047*** 0.002 0.103*** 0.034** 0.056*** 0.715
5 0.022 0.069*** 0.082*** 0.032*** 0.077*** 0.044*** 0.008 0.690
6 0.075*** 0.049*** 0.073*** 0.004 0.087*** 0.007 0.015 0.684
7 0.058*** 0.013 0.087*** 0.001 0.080*** 0.032** 0.011 0.690
8 0.051*** 0.055*** 0.037*** 0.016* 0.067*** 0.014 0.000 0.678
9 0.066*** 0.056*** 0.041*** 0.008 0.049*** 0.026* 0.015 0.700
High 0.019 0.131*** 0.048*** 0.013 0.078*** 0.055*** 0.022 0.670

Panel D: standard deviation sorted portfolios

2
P rs × RI rh × RI ru × RI rm × DN rs × DN rh × DN ru × DN R
High 0.122*** 0.054** 0.070*** 0.016 0.061*** 0.013 0.009 0.518
2 0.177*** 0.099*** 0.100*** 0.030** 0.068*** 0.006 0.037** 0.656
3 0.214*** 0.049** 0.099*** 0.015 0.043** 0.002 0.045*** 0.688
4 0.201*** 0.058*** 0.125*** 0.016 0.033** 0.003 0.031** 0.696
5 0.199*** 0.079*** 0.131*** 0.001 0.041*** 0.007 0.020* 0.694
6 0.181*** 0.109*** 0.115*** 0.013 0.049*** 0.005 0.022** 0.700
7 0.131*** 0.039*** 0.123*** 0.013 0.049*** 0.002 0.010 0.667
8 0.060*** 0.000 0.102*** 0.038*** 0.023* 0.016 0.002 0.626
9 0.039*** 0.001 0.080*** 0.056*** 0.021* 0.038*** 0.009 0.550
Low 0.015 0.023* 0.039*** 0.064*** 0.023** 0.045*** 0.018** 0.322
2
This table reports alphas, betas, and adjusted R of the market timing regressions of the CMAP excess returns on
the C4 factors along with two instrumental variables and interaction terms of the instrumental variables with
the C4 factors using portfolios sorted by various characteristics. Alphas are annualized and in percent. The sam-
ple period covers 1960/01/04 until 2013/12/31. The lengths of the moving average windows are 5, 10, 20, 50,
100, and 200 days. Newey and West (1987) standard errors with 3 lags are used in reporting statistical signifi-
cance of a twosided null hypothesis at the 1, 5, and 10% level is given by a ***, a **, and a *, respectively.

26 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 9 Conditional regressions with market dividend yield and recession indicator
Panel A: size sorted portfolios

P α DPm RI βm βs βh βu rm × DPm

Low 10.646*** 0.004 0.042*** 0.484*** 0.338*** 0.322*** 0.077*** 0.021***

2 3.159 0.012*** 0.017* 0.762*** 0.546*** 0.476*** 0.012 0.082***
3 1.767 0.016*** 0.013 0.772*** 0.500*** 0.364*** 0.045*** 0.087***
4 2.847 0.017*** 0.010 0.783*** 0.443*** 0.300*** 0.048*** 0.088***
5 5.114** 0.018*** 0.014 0.789*** 0.366*** 0.208*** 0.074*** 0.096***
6 4.121 0.017*** 0.011 0.713*** 0.180*** 0.122*** 0.069*** 0.076***
7 2.468 0.014*** 0.018** 0.714*** 0.115*** 0.205*** 0.023* 0.077***
8 4.701* 0.015*** 0.014 0.745*** 0.051*** 0.189*** 0.058*** 0.085***
9 6.935*** 0.016*** 0.012 0.736*** 0.087*** 0.232*** 0.009 0.084***
High 4.215* 0.008*** 0.008 0.738*** 0.166*** 0.026 0.111*** 0.088***

Panel B: book-to-market sorted portfolios

P α DPm RI βm βs βh βu rm × DPm

Low 10.162*** 0.019*** 0.015 0.660*** 0.138*** 0.269*** 0.074*** 0.052***

2 8.016*** 0.016*** 0.007 0.727*** 0.171*** 0.218*** 0.093*** 0.074***
3 7.385*** 0.016*** 0.001 0.676*** 0.182*** 0.247*** 0.023* 0.069***
4 4.361* 0.013*** 0.014 0.724*** 0.124*** 0.456*** 0.015 0.081***
5 3.470 0.011*** 0.006 0.761*** 0.035* 0.597*** 0.065*** 0.102***
6 2.439 0.009*** 0.016* 0.759*** 0.022 0.497*** 0.073*** 0.100***
7 2.108 0.010*** 0.007 0.691*** 0.074*** 0.584*** 0.046*** 0.077***
8 1.484 0.010*** 0.006 0.654*** 0.028 0.608*** 0.040*** 0.067***
9 0.203 0.009*** 0.005 0.770*** 0.007 0.675*** 0.104*** 0.088***
High 1.049 0.010*** 0.008 0.757*** 0.059*** 0.776*** 0.116*** 0.070***

Panel C: momentum sorted portfolios

P α DPm RI βm βs βh βu rm × DPm

Low 9.110** 0.005 0.037** 1.004*** 0.394*** 0.088** 0.873*** 0.118***

2 1.898 0.011** 0.029** 0.902*** 0.085*** 0.236*** 0.582*** 0.124***
3 3.370 0.009** 0.029*** 0.767*** 0.100*** 0.314*** 0.359*** 0.100***
4 4.269* 0.011*** 0.017* 0.717*** 0.162*** 0.391*** 0.189*** 0.084***
5 8.053*** 0.016*** 0.006 0.653*** 0.179*** 0.207*** 0.126*** 0.063***
6 9.353*** 0.018*** 0.001 0.640*** 0.191*** 0.345*** 0.023* 0.058***
7 5.313** 0.015*** 0.006 0.660*** 0.219*** 0.271*** 0.023* 0.059***
8 8.445*** 0.017*** 0.003 0.646*** 0.202*** 0.327*** 0.101*** 0.057***
9 5.566** 0.016*** 0.003 0.694*** 0.129*** 0.196*** 0.160*** 0.059***
High 6.139* 0.019*** 0.001 0.826*** 0.037 0.109*** 0.349*** 0.073***

Panel D: standard deviation sorted portfolios

P α DPm RI βm βs βh βu rm × DPm

High 10.497*** 0.003 0.017 0.461*** 0.138*** 0.301*** 0.009 0.026***

2 4.508 0.013*** 0.030** 0.648*** 0.150*** 0.415*** 0.137*** 0.010*
3 1.378 0.016*** 0.030** 0.637*** 0.073*** 0.471*** 0.097*** 0.012**
4 1.597 0.017*** 0.021* 0.640*** 0.047** 0.505*** 0.073*** 0.035***
5 1.907 0.016*** 0.022** 0.589*** 0.036* 0.517*** 0.071*** 0.034***
6 2.018 0.014*** 0.026*** 0.576*** 0.055*** 0.493*** 0.039*** 0.046***
7 1.820 0.013*** 0.022*** 0.488*** 0.127*** 0.415*** 0.018 0.031***
8 0.824 0.011*** 0.023*** 0.390*** 0.152*** 0.328*** 0.003 0.013***
9 1.364 0.007*** 0.024*** 0.251*** 0.148*** 0.231*** 0.013 0.007**
Low 2.571 0.005** 0.023*** 0.002 0.198*** 0.025 0.048*** 0.060***

© 2016 International Review of Finance Ltd. 2016 27

 International Review of Finance

Table 9 (continued)
Panel A: size sorted portfolios
2
P rs × DPm rh × DPm ru × DPm rm × RI rs × RI rh × RI ru × RI R
Low 0.050*** 0.089*** 0.014*** 0.198*** 0.146*** 0.109*** 0.064*** 0.678
2 0.007 0.138*** 0.016*** 0.282*** 0.217*** 0.122*** 0.126*** 0.735
3 0.008 0.113*** 0.028*** 0.264*** 0.208*** 0.104*** 0.127*** 0.742
4 0.002 0.096*** 0.031*** 0.229*** 0.164*** 0.073*** 0.107*** 0.739
5 0.006 0.075*** 0.040*** 0.256*** 0.111*** 0.092*** 0.102*** 0.739
6 0.036*** 0.046*** 0.042*** 0.226*** 0.069*** 0.056*** 0.085*** 0.719
7 0.044*** 0.074*** 0.026*** 0.259*** 0.002 0.073*** 0.075*** 0.720
8 0.038*** 0.067*** 0.038*** 0.257*** 0.016 0.082*** 0.080*** 0.730
9 0.048*** 0.077*** 0.024*** 0.261*** 0.030* 0.042*** 0.073*** 0.734
High 0.005 0.029*** 0.050*** 0.243*** 0.026* 0.013 0.063*** 0.749

Panel B: book-to-market sorted portfolios

2
P rs × DPm rh × DPm ru × DPm rm × RI rs × RI rh × RI ru × RI R
Low 0.036*** 0.015* 0.036*** 0.239*** 0.027 0.070*** 0.017 0.726
2 0.055*** 0.110*** 0.040*** 0.198*** 0.037** 0.062*** 0.123*** 0.709
3 0.059*** 0.099*** 0.028*** 0.237*** 0.026 0.090*** 0.121*** 0.699
4 0.044*** 0.142*** 0.010** 0.260*** 0.048*** 0.112*** 0.060*** 0.709
5 0.019*** 0.185*** 0.036*** 0.272*** 0.016 0.162*** 0.105*** 0.689
6 0.025*** 0.137*** 0.033*** 0.210*** 0.065*** 0.071*** 0.094*** 0.685
7 0.034*** 0.136*** 0.033*** 0.207*** 0.046*** 0.045*** 0.080*** 0.678
8 0.023*** 0.132*** 0.035*** 0.291*** 0.068*** 0.366*** 0.011 0.697
9 0.019*** 0.125*** 0.054*** 0.197*** 0.083*** 0.083*** 0.120*** 0.678
High 0.033*** 0.160*** 0.058*** 0.243*** 0.161*** 0.158*** 0.088*** 0.672

Panel C: momentum sorted portfolios

2
P rs × DPm rh × DPm ru × DPm rm × RI rs × RI rh × RI ru × RI R
Low 0.042*** 0.078*** 0.185*** 0.287*** 0.117*** 0.291*** 0.203*** 0.719
2 0.001 0.105*** 0.109*** 0.331*** 0.017 0.132*** 0.028* 0.727
3 0.048*** 0.123*** 0.056*** 0.300*** 0.039** 0.138*** 0.051*** 0.724
4 0.056*** 0.134*** 0.042*** 0.304*** 0.012 0.079*** 0.043*** 0.715
5 0.065*** 0.074*** 0.036*** 0.273*** 0.023 0.084*** 0.075*** 0.684
6 0.059*** 0.101*** 0.025*** 0.243*** 0.035** 0.020 0.069*** 0.676
7 0.075*** 0.082*** 0.022*** 0.211*** 0.012 0.009 0.083*** 0.682
8 0.066*** 0.087*** 0.025*** 0.179*** 0.009 0.029* 0.033*** 0.671
9 0.064*** 0.066*** 0.026*** 0.211*** 0.025 0.075*** 0.040*** 0.693
High 0.052*** 0.005 0.001 0.218*** 0.042* 0.130*** 0.062*** 0.663

Panel D: standard deviation sorted portfolios

2
P rs × DPm rh × DPm ru × DPm rm × RI rs × RI rh × RI ru × RI R
High 0.137*** 0.081*** 0.006 0.180*** 0.048* 0.018 0.059*** 0.519
2 0.121*** 0.124*** 0.046*** 0.250*** 0.098*** 0.032 0.081*** 0.656
3 0.130*** 0.135*** 0.039*** 0.282*** 0.128*** 0.019 0.080*** 0.691
4 0.115*** 0.147*** 0.030*** 0.324*** 0.124*** 0.118*** 0.112*** 0.697
5 0.121*** 0.149*** 0.032*** 0.316*** 0.119*** 0.139*** 0.116*** 0.699
6 0.102*** 0.144*** 0.024*** 0.323*** 0.114*** 0.160*** 0.105*** 0.701
7 0.107*** 0.116*** 0.016*** 0.268*** 0.066*** 0.087*** 0.114*** 0.668
8 0.103*** 0.087*** 0.012*** 0.212*** 0.000 0.048*** 0.093*** 0.625
9 0.088*** 0.056*** 0.008** 0.196*** 0.009 0.044*** 0.072*** 0.548
Low 0.090*** 0.023*** 0.007* 0.101*** 0.026** 0.018 0.031*** 0.346
2
This table reports alphas, betas, and adjusted R of the market timing regressions of the CMAP excess returns on
the C4 factors along with two instrumental variables and interaction terms of the instrumental variables with
the C4 factors using portfolios sorted by various characteristics. Alphas are annualized and in percent. The sam-
ple period covers 1960/01/04 until 2013/12/31. The lengths of the moving average windows are 5, 10, 20, 50,
100, and 200 days. Newey and West (1987) standard errors with 3 lags are used in reporting statistical signifi-
cance of a twosided null hypothesis at the 1, 5, and 10% level is given by a ***, a **, and a *, respectively.

28 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

model specification. The first notable finding is that once again, the abnormal
returns are reduced just as in Table 7, previously. Most of the α s turn negative
with the exception of market-cap, momentum and standard deviation decile
one which have all positive and statistically as well as economically significant
values. Nevertheless, it is reassuring that for all other CMAP portfolios the abnor-
mal returns are negative and, mostly, insignificant with a few deciles having neg-
ative and highly significant α s.
The dividend yield on the market appears the affect positively and signifi-
cantly the abnormal returns for all decile portfolios under investigation. At the
same time, the recession indicator has a positive and statistically significant im-
pact on the abnormal returns of the standard deviation deciles only. Positive
and significant coefficients on the RI dummy variable obtain only for market-
cap deciles one and seven as well as momentum deciles one through four. The
rest of the CMAP portfolios’ α s appear to be unaffected by general economic con-
ditions. Next, I investigate the impact of the state variables on the loadings of the
four factors. The market’s dividend yield has a strong and positive effect on both
the market and HML beta for all CMAP portfolios. Conversely, the SMB and UMD
factor loadings are reduced whenever the market’s dividend yield increases. The
recession indicator has mostly a negative impact on all factor loadings with a
few exceptions.

D. Comparison with simple moving average strategies

In this subsection, I compare the performance of the CMA strategy against the
simple moving average strategy (SMA). Specifically, I use two popular window
lengths of 10 days (SMA(10)) and 20 days (SMA(20)). Table 10 reports the trading
frequency (NT) and the BETC for all three strategies across the four sets of portfo-
lios investigated in this article. The NT and BETC for the CMA require a bit more
discussion regarding the fair comparison with the SMA strategy. The values for
NT and BETC reported previously in Table 5 have been adjusted by a factor of
six in order to make the comparison with the SMA more equitable. Specifically,
the NT value for CMA in Table 5 has been divided by a factor of six and reported
in Table 10 to reflect the fact that the CMA is an equal-weighted portfolio of six
SMA strategies. Similarly, the BETC value for CMA in Table 5 has been multiplied
by a factor of six and reported in Table 10.
One thing that emerges from Table 10 is that the trading intensity of the CMA
strategy is lower than the trading intensity of the two SMA strategies, while the
break-even transaction cost of the CMA strategy exceeds the BETC of SMA(10)
and SMA(20). The NT and BETC values are almost uniformly monotonic from
low to high decile portfolios sorted by market capitalization of equity, book-
to-market ratios, and price momentum. The trading intensity and break-even
transaction costs are also monotonic for decile portfolios sorted by standard devi-
ation with the exception of the reverse sorting where the first decile is high and

© 2016 International Review of Finance Ltd. 2016 29

 International Review of Finance

Table 10 CMA versus Simple MA: trading Frequency and BreakEven Transaction
Costs
Panel A: size sorted portfolios
Portfolio CMA SMA(10) SMA(20)
NT BETC NT BETC NT BETC
Low 708 0.90 1484 0.58 932 0.81
2 776 0.59 1714 0.38 1072 0.51
3 790 0.47 1726 0.32 1090 0.43
4 814 0.42 1778 0.29 1148 0.38
5 818 0.36 1808 0.24 1140 0.31
6 823 0.37 1742 0.29 1166 0.30
7 825 0.35 1810 0.22 1146 0.29
8 847 0.26 1832 0.20 1194 0.21
9 889 0.16 1938 0.11 314 0.09
High 943 0.02 2116 0.01 1502 0.04
Panel B: book-to-market sorted portfolios
Portfolio CMA SMA(10) SMA(20)
NT BETC NT BETC NT BETC
Low 927 0.17 2106 0.08 1410 0.12
2 916 0.09 2032 0.08 1340 0.08
3 919 0.12 2070 0.07 1346 0.09
4 896 0.15 1956 0.11 1316 0.11
5 918 0.07 2070 0.03 1408 0.01
6 876 0.10 1948 0.10 1260 0.07
7 886 0.10 2016 0.03 1302 0.06
8 885 0.08 2012 0.05 1316 0.05
9 868 0.10 1950 0.08 1310 0.07
High 863 0.22 1896 0.18 1282 0.15
Panel C: momentum sorted portfolios
Portfolio CMA SMA(10) SMA(20)
NT BETC NT BETC NT BETC
Low 870 0.86 1832 0.53 1198 0.74
2 899 0.42 1932 0.27 1296 0.36
3 910 0.21 2060 0.14 1316 0.23
4 903 0.18 2064 0.10 1360 0.15
5 916 0.10 2060 0.07 1362 0.08
6 913 0.08 2098 0.06 1402 0.05
7 904 0.11 1996 0.08 1356 0.04
8 922 0.01 2086 0.00 1390 0.01
9 908 0.08 2016 0.04 1436 0.03
High 867 0.12 1938 0.07 1308 0.09
Panel D: standard deviation sorted portfolios
Portfolio CMA SMA(10) SMA(20)
NT BETC NT BETC NT BETC
High 719 0.70 1558 0.52 936 0.75
2 790 0.74 1650 0.53 1098 0.68
3 786 0.66 1634 0.48 1042 0.63
4 785 0.51 1674 0.36 1054 0.48

30 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 10 (continued)
(continued)
Panel A:
Panel D:size sorted portfolios
standard deviation sorted portfolios
Portfolio CMA SMA(10) SMA(20)
NT BETC NT BETC NT BETC
5 774 0.48 1624 0.34 1026 0.44
6 772 0.44 1646 0.32 1048 0.35
7 767 0.40 1664 0.27 1092 0.32
8 769 0.37 1640 0.25 990 0.36
9 740 0.37 1546 0.27 928 0.38
Low 660 0.43 1350 0.32 832 0.40

This table reports findings from a comparison of the performance of the combination MA (CMA)
strategy to a simple MA strategy with 10 and 20 days window. Below I report the equivalent trading
frequency, NT, for the combination MA strategy as well as the equivalent break-even transaction
cost for the combination MA strategy, BETC, using 10 decile portfolios sorted by size, book-to-mar-
ket, momentum, and standard deviation of return. The sample period covers January 4, 1960 until
December 31, 2013 with value-weighted portfolio returns. SMA(q) refers to the simple MA strategy
with q days in the window. The lengths of the moving average windows in the combination MA
strategy are 5, 10, 20, 50, 100, and 200 days. The combination moving average portfolio is an
equal-weighted combination of the six individual simple moving average returns.

contains 10% of the stocks with the highest historical volatility, while the tenth
decile is low and contains stocks with the lowest historical volatility.
Next, I turn to the comparison of the mean and variance improvement of the
CMA strategy relative to SMA(10) and SMA(20). Table 11 reports the mean im-
provement, Δμ, and the risk reduction, Δσ, for all three strategies. A very consis-
tent and curious pattern of findings emerges from all four panels. The CMA
strategy uniformly produces a lower improvement in average return relative to
buy-and-hold when compared with both the SMA(10) and SMA(20). However,
it is also always the case that the risk reduction attained by the CMA strategy rel-
ative to buy-and-hold uniformly exceeds the risk reduction achieved by both
SMA(10) and SMA(20). Both Δμ and Δσ are almost exhibit interesting patterns
cross-sectionally, in particular, return improvement is almost monotonic across
the 10 decile portfolios, while the risk reduction is almost the same across deciles
for size and book-to-market portfolios while monotonic for momentum and
volatility-sorted portfolios.
Finally, I turn to a simple predictive regression comparison where I use a con-
ditional version of the Carhart 4-factor model with a moving average indicator as
a predetermined state variable

CMAPjt;L ¼ αj þ βj;m rmkt;t þ βj;s rsmb;t þ βj;h rhml;t þ βj;u rumd;t þ

  
ϕ j;Z Z t1 ⊗ 1T ; rmkt;t ; rsmb;t ; rhml;t ; rumd;t þ εjt ; j ¼ 1; …; N :

where Zt  1 is represented by a moving average indicator variable equal to one of

the MA signal indicates a buy and zero otherwise for the SMA strategy. For the
CMA strategy, the moving indicator is the equal-weighted average of the six
SMA indicator variables. In the interest of brevity, I report only the adjusted

© 2016 International Review of Finance Ltd. 2016 31

 International Review of Finance

Table 11 CMA versus simple MA: improvement in mean and ariance

Panel A: size sorted portfolios
Portfolio CMA SMA(10) SMA(20)
Δμ Δσ Δμ Δσ Δμ Δσ
Low 11.73 6.15 15.90 5.14 13.86 5.24
2 8.43 7.64 11.93 6.12 10.11 6.19
3 6.80 7.40 10.19 5.95 8.63 6.13
4 6.25 7.38 9.37 5.93 7.98 6.01
5 5.48 7.23 8.08 5.83 6.42 5.92
6 5.60 6.87 9.21 5.60 6.35 5.60
7 5.28 6.96 7.49 5.65 6.03 5.66
8 4.11 7.01 6.82 5.67 4.57 5.71
9 2.54 6.90 3.80 5.60 2.21 5.72
High 0.34 7.12 0.44 5.71 1.09 5.86
Panel B: book-to-market sorted portfolios
Portfolio CMA SMA(10) SMA(20)
Δμ Δσ Δμ Δσ Δμ Δσ
Low 2.95 7.94 3.20 6.40 3.08 6.60
2 1.55 7.06 2.87 5.77 2.02 5.83
3 2.02 6.67 2.70 5.42 2.33 5.49
4 2.43 6.93 4.00 5.65 2.68 5.59
5 1.24 6.87 1.16 5.37 0.29 5.52
6 1.67 6.58 3.49 5.21 1.56 5.33
7 1.65 6.34 1.15 5.05 1.45 5.14
8 1.26 6.94 1.95 5.36 1.17 5.55
9 1.64 6.73 2.90 5.52 1.66 5.51
High 3.53 7.46 6.12 6.02 3.65 6.05
Panel C: momentum sorted portfolios
Portfolio CMA SMA(10) SMA(20)
Δμ Δσ Δμ Δσ Δμ Δσ
Low 13.72 12.59 17.74 8.44 16.21 9.07
2 6.86 9.36 9.56 6.75 8.63 7.08
3 3.59 7.88 5.15 5.67 5.63 5.91
4 2.92 7.46 3.86 5.44 3.67 5.76
5 1.63 7.03 2.66 5.43 1.88 5.42
6 1.35 6.70 2.35 5.41 1.41 5.33
7 1.76 6.68 3.07 5.29 1.10 5.41
8 0.15 6.33 0.07 5.43 0.18 5.37
9 1.26 7.00 1.44 5.95 0.73 6.06
High 1.98 8.68 2.61 7.51 2.07 7.73

32 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages
Panel A: size sorted portfolios
Portfolio CMA SMA(10) SMA(20)
Table 11 (continued)
Δμ Δσ Δμ Δσ Δμ Δσ
Panel D: standard deviation sorted portfolios
Portfolio CMA SMA(10) SMA(20)
Δμ Δσ Δμ Δσ Δμ Δσ
High 9.32 7.23 14.85 6.08 12.83 6.21
2 10.82 8.55 16.23 6.73 13.64 7.06
3 9.59 8.32 14.47 6.56 12.10 6.76
4 7.43 7.69 10.94 6.07 9.22 6.18
5 6.81 7.05 10.21 5.61 8.31 5.66
6 6.18 6.57 9.76 5.23 6.69 5.23
7 5.64 5.76 8.30 4.62 6.42 4.71
8 5.21 5.03 7.46 4.12 6.55 4.24
9 5.07 4.22 7.71 3.59 6.40 3.65
Low 5.25 3.14 7.98 2.74 6.06 2.75

This table reports findings from a comparison of the performance of the combination MA (CMA)
strategy to a simple MA strategy with 10 and 20 days window. I report the improvement in the
mean return, Δμ, and the improvement in the standard deviation, Δσ, for the combination MA
strategy relative to the BH strategy using 10 decile portfolios sorted by size, book-to-market, mo-
mentum, and standard deviation of return. The sample period covers January 4, 1960 until Decem-
ber 31, 2013 with value-weighted portfolio returns. SMA(q) refers to the simple MA strategy with q
days in the window. The lengths of the moving average windows in the combination MA strategy
are 5, 10, 20, 50, 100, and 200 days. The combination moving average portfolio is an equal-
weighted combination of the six individual simple moving average returns.

goodness-of-fit for the CMA, SMA(10), and SMA(20) in Table 12.5 The reported
findings indicate that the adjusted R2 from these conditional predictive regres-
sions for the CMA strategy exceeds that goodness-of-fit of the same regression
for the SMA(20) strategy which in turn exceeds to goodness-of-fit of the SMA
(10) strategy. This finding is robust across all four sets of decile portfolios and is
probably due to the fact that the CMA strategy aggregates six signals from six dif-
ferent SMA strategies with various window lengths. To the extent that there may
be any incremental value-added across SMA strategies of varying lengths then an
aggregative signal like the CMA equal-weighted averaging over several SMA sig-
nals may improve the chances of the signal being on the right side of the market.

E. Individual stocks
In this subsection, I report results on the performance of moving average strate-
gies with individual stocks in the Center for Research in Security Prices (CRSP)
database starting in January 4, 1988 until December 30, 2011 that have continu-
ously non-missing daily return observation during this entire sample period. This

5 The full results of these conditional regressions are available from the author upon request.

© 2016 International Review of Finance Ltd. 2016 33

 International Review of Finance

Table 12 CMA versus simple MA: predictive regressions goodness-of-fit

Panel A: size sorted portfolios
2 2 2
Portfolio RCM A RM Að10Þ RMAð20Þ
Low 0.687 0.488 0.557
2 0.746 0.506 0.586
3 0.772 0.515 0.621
4 0.769 0.508 0.610
5 0.767 0.509 0.605
6 0.736 0.494 0.589
7 0.735 0.484 0.582
8 0.739 0.494 0.582
9 0.754 0.483 0.580
High 0.776 0.492 0.580
Panel B: book-to-market sorted portfolios
2 2 2
Portfolio RCM A RM Að10Þ RMAð20Þ
Low 0.740 0.457 0.558
2 0.703 0.451 0.546
3 0.690 0.457 0.540
4 0.666 0.432 0.532
5 0.649 0.437 0.506
6 0.671 0.459 0.534
7 0.666 0.448 0.517
8 0.676 0.477 0.558
9 0.686 0.460 0.550
High 0.643 0.464 0.527
Panel C: momentum sorted portfolios
2 2 2
Portfolio RCM A RM Að10Þ RMAð20Þ
Low 0.672 0.487 0.593
2 0.715 0.517 0.619
3 0.710 0.514 0.608
4 0.680 0.481 0.577
5 0.664 0.450 0.535
6 0.659 0.431 0.514
7 0.684 0.451 0.518
8 0.700 0.437 0.519
9 0.700 0.446 0.515
High 0.722 0.448 0.518
Panel D: standard deviation sorted portfolios
2 2 2
Portfolio RCM A RM Að10Þ RMAð20Þ
High 0.466 0.363 0.403
2 0.689 0.507 0.581
3 0.713 0.521 0.590
4 0.719 0.508 0.597

34 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Table 12 (continued)
Panel D: standard
A: size sorteddeviation
portfoliossorted portfolios
2 2 2
Portfolio RCMA RMAð10Þ RMAð20Þ
5 0.728 0.523 0.607
6 0.724 0.525 0.589
7 0.715 0.516 0.584
8 0.693 0.491 0.579
9 0.619 0.447 0.528
Low 0.362 0.280 0.310

This table reports goodness-of-fit results in predictive regressions of the conditional Carhart model
with an MA indicator as a state variable for the combination MA (CMA) strategy with the simple
2
MA strategy with 10 and 20 days window. I report the adjusted R for the combination MA strategy
relative to the BH strategy using 10 decile portfolios sorted by size, book-to-market, momentum
and standard deviation of return. The sample period covers January 4, 1960 until December 31,
2013 with value-weighted portfolio returns. SMA(q) refers to the simple MA strategy with q days
in the window. The lengths of the moving average windows in the combination MA strategy are
5, 10, 20, 50, 100, and 200 days. The combination moving average portfolio is an equal-weighted
combination of the six individual simple moving average returns.

Figure 3 Histograms of the Relative Performance of CMA versus BH using Individual

US Stocks.
Notes: Figure 3 presents histogram plots of Δμ, Δσ, pA, NT, BETC, and p1 of the relative
performance of the combination moving average strategy versus the buy and hold
strategy using 1040 individual US stocks. The sample contains 6052 daily observations,
and the data covers the January 4, 1988 until December 30, 2011.

© 2016 International Review of Finance Ltd. 2016 35

 International Review of Finance

results in 1040 individual stocks. Instead of reporting the results in tabular form,
I report the key attributes in Figure 3 as histograms.
The performance of the CMA strategy with individual stocks is largely consis-
tent with the performance of the CMA strategy with portfolios. The risk of the
CMA strategy is uniformly always smaller than the risk of the underlying stock.
The difference in average returns between the CMA and BH strategies is negative
for 1011 or more than 97% of all individual stocks I investigate. The findings for
the BETC are identical as the latter is calculated as a function of the former. The
inferior performance of the CMA strategy relative the BH strategy for individual
stocks comes as a surprise compared with the findings for the portfolios pre-
sented in the previous section. Nevertheless, there is a universal reduction in risk
indicated by the uniformly positive values for Δσ. The majority of negative values
for Δμ could be due to the shorter time period under investigation relative to the
findings for the stock portfolios.
Figure 4 presents histograms of the distribution of bβ m2 and bγ m across 1040 indi-
vidual stocks. The findings of both market timing specifications are both qualita-
tively and quantitatively similar. Only 30% (TM) to 34% (HM) of the market
timing coefficients are positive with less than a half of those statistically signifi-
cant. At the same time, 66% (HM) to 70% (TM) of the market timing coefficients

Figure 4 Histograms of the Market Timing Regression Coefficients of CMA Strategy

using Individual US Stocks.
Notes: Figure 4 presents histogram plots of the market timing regression coefficients βm2
and γm as well as the associated t-statistics of the combination moving average strategy
using 1040 individual US stocks. The sample contains 6052 daily observations and the
data covers the January 4, 1988 until December 30, 2011.

36 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

are negative with the majority being statistically significant. These findings indi-
cate that the CMA strategy performed on individual stocks fails to time the mar-
ket in the right direction.

F. Discussion
The large values of the risk-adjusted abnormal returns presented in the previous
subsection demonstrate the profitability of the CMA switching strategy. This
raises the question as to what ultimately drives of the performance of the CMA
strategy. So far, the evidence points towards a strategy that is contrarian, with a
focus on large-cap growth stocks and short the market. However, the goodness-
of-fit statistics indicate that this is at most only half the story. A more fundamen-
tal question that arises is how can this strategy survive in competitive financial
markets. A few potential reasons seem plausible.
First, there is ample evidence that stock returns are predictable at various fre-
quencies at least to a certain degree. This level of predictability is not perfect
but is sufficient to improve forecasts of future stock returns when stock return
predictability is ignored. Some of the early evidence presented in Fama and
Schwert (1977) and Campbell (1987) as well as more recent work by Cochrane
(2008) clearly demonstrates that stock return predictability is an important fea-
ture that investors should ignore at their own peril.
Evidence regarding the performance of the moving average technical indicator
is present in Brock et al. (1992) in the context of predicting future moments of
the Dow Jones Industrial Average. Lo et al. (2000) provide further evidence using
a wide range of technical indicators with wide popularity among traders showing
that this adds value even at the individual stock level over and above the perfor-
mance of a stock index. More recently, Neely et al. (2010) provide evidence in fa-
vor of the usefulness of technical analysis in forecasting the stock market risk
premium.
Second, early work on the performance of filter rules by Fama and Blume
(1966); Jensen and Benington (1970) concluded that such rules were dominated
by buy and hold strategies especially after transaction costs. Malkiel (1996) makes
a forceful and memorable point against technical indicators: “Obviously, I’m bi-
ased against the chartist. This is not only a personal predilection but a professional
one as well. Technical analysis is anathema to the academic world. We love to pick on
it. Our bullying tactics are prompted by two considerations: (1) after paying transaction
costs, the method does not do better than a buy-and-hold strategy for investors, and (2)
it’s easy to pick on. And while it may seem a bit unfair to pick on such a sorry target, just
remember: It’s your money we are trying to save.” In a follow up on Brock et al.
(1992); Bessembinder and Chan (1998) attribute the forecasting power of techni-
cal analysis to measurement errors arising from non-synchronous trading. Ready
(2002) goes even further and claims the results in Brock et al. (1992) are spurious
and due to data snooping. Formal tests using White’s Reality Check are con-
ducted in Sullivan et al. (1999) confirm that Brock et al. (1992) results are robust

© 2016 International Review of Finance Ltd. 2016 37

 International Review of Finance

to data snooping and perform even better out of sample though there is evidence
of time variation in performance across subperiods. A more recent study using
White’s Reality Check and Hansen’s Superior Predictive Ability (SPA) test is Hsu
and Kuan (2005) who find evidence of profitability of technical analysis using
relatively “young” markets like the National Association of Securities Dealers
and Quotes (NASDAQ) Composite index and the Russell 2000 both in-sample
and out-of-sample.
Furthermore, Treynor and Ferguson (1985) make a strong case in favor of in-
vestor’s learning and Bayesian updating conditional on new information re-
ceived rationally combining past prices can result in abnormal profitability.
Sweeney (1988) revisits Fama and Blume (1966) and finds that filter rules can
be profitable to floor traders in the 1970–1982 time period. Neftci (1991) presents
a formal analysis of Wiener–Kolmogorov prediction theory which provides opti-
mal linear forecasts. He concludes that if the underlying price processes are non-
linear in nature then technical analysis rules might capture some useful informa-
tion that is ignored by the linear prediction rules. More involved and inherently
non-linear rules are investigated in the context of foreign currency exchange
rates by Neely et al. (1997) using a genetic programming approach. Gencay
(1998) goes even further in using non-linear predictors based on simple moving
average rules on the Dow Jones Industrial Average over a long time period be-
tween 1897 and 1988. In a similar vein, Allen and Karjalainen (1999) use genetic
algorithms to search for functions of past prices find that can outperform a sim-
ple buy-and-hold strategy and report negative excess returns for most of the strat-
egies they consider.
Thirdly, it is entirely possible that market prices of financial assets can persis-
tently deviate from fundamental values. Those fundamental values themselves
are subject to incomplete information and, perhaps, imperfect understanding
of valuation tools as well as dispersion of beliefs and objective and behavioral
biases across the pool of traders and investors who regularly interact in finan-
cial markets. When investors’ information is incomplete, and they learn con-
tinuously over time, the true fundamental value, Zhu and Zhou (2009) as
well as Han et al. (2016) show theoretically that the moving average price is
a useful state variable that aids in investors’ learning and improves their well-
being and utility.
Behavioral and cognitive biases have been proposed in Daniel et al. (1998) and
Hong and Stein (1999), among others, as a potential driver of both price under-
reaction and over-reaction in conjunction with the observed price continuation
of stock prices. An alternative explanation for price continuation was proposed
in Zhang (2006). He argues that investors sub-optimally underweight newly ar-
riving public information leading to a persistent deviation of the market price
from the fundamental intrinsic value.
Note also that despite the apparent similarity of the CMA switching strategy to
the momentum strategy, the four-factor alphas reported previously are statisti-
cally significant and of large magnitudes. This is perhaps not surprising given
that the payoff of the CMA strategy resembles an at-the-money protective put

38 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

strategy. The non-linearity this induces makes the asset pricing task much more
difficult when linear models are used.

V. CONCLUSION

In this paper, I report results for the performance of a combination moving aver-
age strategy applied to decile portfolios sorted by size, book-to-market, momen-
tum, and standard deviation of return. Further unreported findings for
portfolios of stocks sorted by various measures of yield, past returns, and industry
classification support the reported findings. There is overwhelming evidence that
the combination moving average strategy dominates in a mean-variance sense
buying and holding any of the decile portfolios. The excess returns of the CMA
returns over BH returns of the underlying portfolios are relatively insensitive to
the four Carhart (1997) factors and generate high statistically and economically
significant abnormal returns. Furthermore, the abnormal returns for most deciles
decline substantially after controlling for the market’s dividend yield, the short-
term risk-free rate, recessions, and up/down markets. This CMA strategy does not
involve overly excessive trading when implemented with daily returns and has
positive break-even transaction costs, suggesting that it will be actionable even
for large institutional investors. These findings are robust with respect to portfo-
lio construction, various lag lengths of the moving average, alternative sets of
portfolios, and individual stocks. The risk-adjusted performance is reduced sub-
stantially only in the context of a conditional asset pricing model with the mar-
ket’s dividend yield and a recession indicator as predetermined state variables.
Hence, it appears that the success of the CMA strategy does not represent an
anomaly and is consistent with rational asset pricing. In addition, any abnormal
returns surviving the previously mentioned tests may not be actionable in prac-
tice because of limits to arbitrage and price impact of trading on illiquid risky as-
sets with low trading volumes.
Further work would be necessary to investigate the potential link between the
returns of the CMA switching strategy and the payoffs of protective put options
on the underlying asset. One potential alternative is to combine all first four mo-
ments using a utility function over them and convert the gains into certainty
equivalent utility gains. Comparing the certainty equivalent utility gains to the
break-even transaction costs will provide further evidence into whether the
CMA switching strategy is desirable for investors who care about the first four
moments of asset returns. In addition, more theoretical studies along the lines
of Zhu and Zhou (2009) as well as Han et al. (2016)would provide a further justi-
fication of the practical application of technical analysis and the continued in-
vestigation of technical analysis role in empirical asset pricing.
Considering the vast literature on technical analysis and the numerous techni-
cal indicators following by some traders in practice, this study is just a first step
towards investigating the performance and implementation of one common
technical indicator. Future work will determine which other technical indicators

© 2016 International Review of Finance Ltd. 2016 39

 International Review of Finance

perform well, and whether they produce significant abnormal returns over and
above the relevant transaction costs.

Paskalis Glabadanidis
Accounting and Finance, Business School
University of Adelaide
Adelaide
SA 5005
Australia
[email protected]

REFERENCES

Allen, F., and R. Karjalainen (1999), ‘Using Genetic Algorithms to Find Technical
Trading Rules’, Journal of Financial Economics, 51, 245–72.
Balduzzi, P., and A. W. Lynch (1999), ‘Transaction Costs and Predictability: Some
Utility Cost Calculations’, Journal of Financial Economics, 52, 47–78.
Bessembinder, H., and K. Chan (1998), ‘Market Efficiency and the Returns to Technical
Analysis’, Financial Management, 27, 5–17.
Blume, L., D. Easley, and M. O’Hara (1994), ‘Market Statistics and Technical Analysis:
The Role of Volume’, Journal of Finance, 49, 153–81.
Brock, W., J. Lakonishok, and B. LeBaron (1992), ‘Simple Technical Trading Rules and
the Stochastic Properties of Stock Returns’, Journal of Finance, 47, 1731–64.
Brown, D. P., and R. H. Jennings (1989), ‘On Technical Analysis’, The Review of Financial
Studies, 2, 527–51.
Campbell, J. Y. (1987), ‘Stock Returns and the Term Structure’, Journal of Financial
Economics, 18, 373–99.
Carhart, M. M. (1997), ‘On Persistence in Mutual Fund Performance’, Journal of Finance,
52, 57–82.
Cochrane, J. H. (2008), ‘The Dog that Did Not Bark: A Defense of Return Predictability’,
Review of Financial Studies, 21, 1533–75.
Daniel, K., D. Hirshleifer, and A. Subrahmanyam (1998), ‘Investor Psychology and
Security Market Under- and Overreactions’, Journal of Finance, 53, 1839–85.
Faber, M. T. (2007), ‘A Quantitative Approach to Tactical Asset Allocation’, Jouranl of
Wealth Management, 9, 69–79.
Fama, E. F., and M. E. Blume (1966), ‘Filter Rules and Stock-Market Trading’, Journal of
Business, 39, 226–41.
Fama, E. F., and W. Schwert (1977), ‘Asset Returns and Inflation’, Journal of Financial
Economics, 5, 115–46.
Ferson, W. E., and R. W. Schadt (1996), ‘Measuring Fund Strategy and Performance in
Changing Economic Conditions’, Journal of Finance, 51, 425–61.
Fung, W., and D. A. Hsieh (2001), ‘The Risk in Hedge Fund Strategies: Theory and
Evidence from Trend Followers’, Review of Financial Studies, 14, 313–41.
Gencay, R. (1998), ‘The Predictability of Security Returns with Simple Technical Trading
Rules’, Journal of Empirical Finance, 5, 347–59.
Glabadanidis, P. (2014), ‘The Market Timing Power of Moving Averages: Evidence from
US REITs and REIT Indexes’, International Review of Finance, 14(2), 161–202.

40 © 2016 International Review of Finance Ltd. 2016

 Combination of Moving Averages

Glabadanidis, P. (2015a), Market Timing and Moving Averages: An Empirical Analysis of

Performance in Asset Allocation. New York: Palgrave Macmillan.
Glabadanidis, P. (2015b), ‘Market Timing with Moving Averages’, International Review of
Finance, 15(3), 387–425.
Goh, J., F. Jiang, J. Tu, and G. Zhou (2012), ‘Forecasting Government Bond Risk Premia
Using Technical Indicators’, Working paper.
Han, Y. (2006), ‘Asset Allocation with a High Dimensional Latent Factor Stochastic
Volatility Model’, Review of Financial Studies, 19, 237–71.
Han, Y., K. Yang, and G. Zhou (2013), ‘A New Anomaly: The Cross-Sectional Profitabil-
ity of Technical Analysis’, Journal of Financial and Quantitative Analysis, 48(5),
1433–61.
Han, Y., G. Zhou, and Y. Zhu (2016), ‘Trend Factor: Any Economic Gains from Using
Information over Investment Horizons?’, Journal of Financial Economics, forthcom-
ing. http://dx.doi.org/10.1016/j.jfineco.2016.01.029.
Henriksson, R. D., and R. C. Merton (1981), ‘On Market Timing and Investment Perfor-
mance. II. Statistical Procedures for Evaluating Forecasting Skills’, Journal of Business,
54, 513–33.
Hong, H., and J. C. Stein (1999), ‘A Unified Theory of Underreaction, Momentum
Trading, and Overreation in Asset Markets’, Journal of Finance, 54, 2143–84.
Hsu, P.-H., and C.-M. Kuan (2005), ‘Reexamining the Profitability of Technical Analysis
with Data Snooping Checks’, Journal of Financial Econometrics, 3, 606–28.
Huang, D., and G. Zhou (2013), ‘Economic and Market Conditions: Two State Variables
that Predict the Stock Market’, Working paper.
Jensen, M. C., and G. A. Benington (1970), ‘Random Walks and Technical Theories:
Some Additional Evidence’, Journal of Finance, 25, 469–82.
Jiang, F. (2013), ‘Trend-Based Conditional Asset Pricing: Explaining the Cross-Section
of Technical Analysis Profitability’, Working paper.
Kilgallen, T. (2012), ‘Testing the Simple Moving Average across Commodities, Global
Stock Indices, and Currencies’, Journal of Wealth Management, 15, 82–100.
Lo, A., H. Mamaysky, and J. Wang (2000), ‘Foundations of Technical Analysis:
Computational Algorithms, Statistical Inference, and Empirical Implementation’,
Journal of Finance, 55, 1705–65.
Lynch, A. W., and P. Balduzzi (2000), ‘Predictability and Transaction Costs: The Impact
on Rebalancing Rules and Behavior’, Journal of Finance, 66, 2285–309.
Malkiel, B. G. (1996), A Random Walk Down Wall Street. New York: W. W. Norton &
Company, Inc..
Neely, C. J., D. E. Rapach, J. Tu, and G. Zhou (2010), ‘Out-of-Sample Equity Premium
Prediction: Fundamental vs. Technical Analysis’, Unpublished working paper,
Washington University in St. Louis.
Neely, C. J., D. E. Rapach, J. Tu, G. Zhou (2011), ‘Forecasting the Equity Risk Premium:
The Role of Technical Indicators’, Unpublished working paper, Federal Reserve Bank
of St. Louis.
Neely, C., P. Weber, and R. Dittmar (1997), ‘Is Technical Analysis in the Foreign
Exchange Market Profitable? A Genetic Programming Approach’, Journal of Financial
and Quantitative Analysis, 32, 405–26.
Neftci, S. N. (1991), ‘Naive Trading Rules in Financial Markets and Wiener-Kolmogorov
Prediction Theory: A Study of “Technical Analysis”’, Journal of Business, 64, 549–71.
Newey, W. K., and K. D. West (1987), ‘A Simple, Positive Semi-Definite,
Heteroscedasticity and Autocorrelation Consistent Covariance Matrix’, Econometrica,
55, 703–8.
Ready, M. J. (2002), ‘Profits from Technical Trading Rules’, Financial Management, 31,
43–61.
Sullivan, R., A. Timmerman, and H. White (1999), ‘Data-Snooping, Technical Trading
Rule Performance, and the Bootstrap’, Journal of Finance, 54, 1647–92.

© 2016 International Review of Finance Ltd. 2016 41

 International Review of Finance

Sweeney, R. J. (1988), ‘Some New Filter Rule Tests: Methods and Results’, Journal of
Financial and Quantitative Analysis, 23, 285–300.
Taleb, N. N. (2012), Antifragile Things that Gain from Disorder. New York: Random House,
Inc..
Treynor, J. L., and R. Ferguson (1985), ‘In Defense of Technical Analysis’, Journal of
Finance, 40, 757–73.
Treynor, J. L., and K. Mazuy (1966), ‘Can Mutual Funds Outguess the Market?’, Harvard
Business Review, 44, 131–6.
Wang, J. (1993), ‘A Model of Intertemporal Asset Prices Under Asymmetric
Information’, Review of Economic Studies, 60, 249–82.
Zhang, F. X. (2006), ‘Information Uncertainty and Stock Returns’, Journal of Finance, 61,
105–36.
Zhu, Y., and G. Zhou (2009), ‘Technical Analysis: An Asset Allocation Perspective on
the Use of Moving Averages’, Journal of Financial Economics, 92, 519–44.

42 © 2016 International Review of Finance Ltd. 2016

View publication stats