APPLIED FINANCE LETTERS

VOLUME 05, ISSUE 02, 2016

MOVING AVERAGE TRADING RULES FOR NASDAQ

COMPOSITE INDEX

MASSOUD METGHALCHI1, CHIEN-PING CHEN2*, MASSOMEH HAJILEE3

1. Professor of Finance, School of Business Administration, University of Houston-Victoria, USA
2. Associate Professor of Economics, School of Business Administration, University of Houston-
Victoria, USA
3. Associate Professor of Economics, School of Business Administration, University of Houston-
Victoria, USA

* Corresponding Author: Chien-Ping Chen, R 3-66 Building 2, 2002 W. Grand Parkway N, Katy,
TX 77449, USA.  +1(281)7967979  [email protected]

Abstract: This paper tests a few moving average technical trading rules for the
NASDAQ Composite index from 1972 to 2015. Our results indicate that
moving average (MA) rules do exhibit strong predictive power for
NADSAQ composite index. Can a trader use this predictive to beat the
Buy and hold strategy? We show that MA-100 days can, most of the time,
make an abnormal profit in the case of NASDAQ composite index by
considering both transaction costs and risk. In addition RSI and MACD
trading rules have also strong predictive power.

Keywords: moving average, trading rules, abnormal profit, market efficiency,

transaction costs

1. Introduction
The Efficient Market Hypothesis (EMH) has provided the foundation of many academic
textbooks in finance and investments. The EMH suggests that investors cannot expect
to outperform the market consistently. This is because securities prices fully reflect all
available information; any new information will be quickly and instantaneously
reflected in prices (Fama, 1970). Since securities prices incorporate all known
information and new information comes randomly, day-to-day price changes follow a
random walk over time, although with a positive drift. A random walk implies any price
pattern is accidental and if securities prices follow a random walk, trading rules and
other Technical Analysis (TA) methods of predicting securities prices will be useless.

Contrary to the above suggestion, there are many traders using TA principles to predict
future prices. TA can be dated back hundreds of years ago. According to historical
records, a great Japanese rice trader by the name of Homma Munehisa (1724-1803)
who fathered candlestick charting at today’s value, would have made over $100 billion
in profits and been considered as the greatest trader in history of financial markets. In
the U.S., technical analysis was initiated by Charles Dow (1851-1902), the founder of Wall
Street Journal and Dow Chemical. Today many traders still follow his Dow Theory for buy
and sell signals and the Dow Jones Industrial Index still serves as one of the most
important reflections for the U.S. stock market.

45
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

The development of technical analysis is based upon three assumptions: 1) the market
discounts everything. In other words, all of financial positions of a company are
reflected in its stock price; 2) price moves in trends; that is, a trend line will be of
tremendous help to predict the future prices. Early detection of a trend is essential to
the success of TA. One of the most important trend determining techniques is the use
of moving average (MA) employed in this paper; and 3) history tends to repeat itself;
this implies traders and investors will react in same way to the same conditions which
will create opportunity for profitable trading. As Meyers (2002) states: “Technicians
record, usually in chart form, historical price and volume activity and deduce from that
pictured history the probable future trend of prices.”

The purpose of this paper is twofold. First, we examine whether moving average (MA)
trading rules have predictive power in the NASDAQ composite Index (NASDAQ); and
secondly, if MA trading rules do exhibit predictive power, could a trader design a
strategy to beat the profitability of the buy and hold (B&H) strategy, considering
transaction costs and risk? Our results indicate that MA trading rules do exhibit strong
predictive power for NADSAQ composite index. A trader employing MA-100 trading rule
could most of the time make an abnormal profit even considering both transaction
costs and risk. We also develop two strategies associated with MA-100 for different risk-
tolerance traders to beat handsomely the buy-and-hold NASDAQ strategy. In addition
we show that the RSI and MACD trading rules have also strong predictive power.

The remainder of this article is organized as the following. Section 2 details some of the
relevant literature. Section 3 describes the data and methodology. The empirical results
of various moving average trading rules are exhibited in Section 4. Section 5 compares
different strategies to locate the most profitable strategies to beat the buy-and-hold
strategy considering transaction costs and market risk. The final section provides
concluding remarks.

2. Literature Review
It is possible to trace the history of TA back to 17th century, an Amsterdam trader, poet,
and philosopher, Joseph Penso de la Vega. However as we have mentioned above,
the fathers of modern TA are the Japanese rice trader, Homma Munehisa (candlestick
charting) and Charles Dow the founder of Wall Street Journal and Dow Theory. In the
1920s and 1930s, Richard W. Schabacker published several books which continued the
work of Charles Dow and William Peter Hamilton in their books Stock Market Theory and
Practice and Technical Market Analysis. In 1948 Robert D. Edwards and John Magee
published Technical Analysis of Stock Trends, which is widely considered, even today,
to be one of the seminal works of the TA. Other pioneers contributed to TA including
Ralph Elliot (the Elliot wave principles) and William Gann (the Gann angles and arcs).
Most technicians were Wall Street traders and most finance professors were believer of
EMH. In early 1970s and 1980s, the Random Walk Hypothesis and its close relative EMH
had become icons of modern financial economics that continue to have many
followers in academic circles as well as professional fund manager in today’s world. As
Lo and MacKinley (1990) point out, even after three decades of research and
thousands of journal articles, finance professors and economists have not yet reached
a conclusion about whether financial markets are efficient or not. Early well-known
empirical studies supported weak form market efficiency, implying that a trader cannot
use past prices to forecast future prices. See for example Larson (1960), Cowles (1960),
Granger and Morgenstern (1963), Mandelbrot (1963), Alexander (1964), Fama (1965),
Fama and Blume (1966), Van Horn and Parker (1967), and Jensen and Benington (1970).

46
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

However since the mid-1980s, technical trading has enjoyed a renaissance both on Wall
Street and in academic circles. Several papers have questioned the validity of EMH by
demonstrating that simple technical trading strategies possess significant power to
predict future security prices. The cornerstones of this renaissance in technical analysis
were articles by Sweeney (1986), Lukac et al. (1988), and Brock, Lakonishok and LeBaron
(1992, BLL hereafter). Sweeney (1986) applies some filter rules for ten currencies and find
that various filters were profitable in more than 80 percent of the cases. Lukac et al.
(1988) find that moving average rules statistically beat the buy and hold strategy. In
1992 a seminal paper by BBL analyzes moving averages and trading rules on the Dow
Jones Industrial Index for a period of 89 years from 1897 to 1985. They use various short
and long moving averages of prices to generate buy and sell signals. They point out
that ‘all buy-sell differences are positive and the t-tests for these differences are highly
significant…’ and they go on to conclude that their ‘results are consistent with technical
rules having predictive power’.

Park and Irwin (2007) provide an excellent survey of technical trading rules by
differentiating the early studies from the modern studies; they conclude that early
studies do not support the predictive power of TA for equity market, and 56 out of a
total 95 modern studies support profitable trading rules. The bulk of modern studies
suggest that trading rules, especially the moving average rules, exhibit predictive
power. However, whether applying those trading rules to obtain abnormal profits when
including transaction costs and risk is not clear for most indexes

3. Data and Methodology

In this paper, we employ a simple technical analysis approach to test the predictive
power for Nasdaq Composite index. The exchange traded fund (ETF) that mimics
NASDAQ composite index has been listed on the NASDAQ National Market and has
been traded since October 1, 2003 (ticker symbol: oneq).

We use Datastream’s daily closing price of the NASDAQ composite index over the
period of 1/3/1972 to 10/14/2015 and define the daily returns as changes in logarithms
of the each index price. Although estimating the return this way does not consider
dividend yield, Mills and Coutts (1995) review the literatures regarding dividends
exclusion and conclude that any bias in the results due to dividend exclusion will be
minimal. This conclusion is also supported by Draper and Paudyal (1997). The proxy
interest data (i.e. money market rate) used in this study is from Datastream’s daily
Federal Fund rate. In order to get the daily interest return, we follow Lucke (2002) by
dividing the annual rates by 260.

The art of technical analysis – in fact it is an art to identify trend changes at an early
stage and to maintain an investment position until the weight of evidence indicates
that the trend has been changed (Pring, 1991). As Kirkpatrick and Dahlquist (2015) point
out, one of the most successful techniques of identifying and profiting trends is the use
of moving averages. According to MA rules, buy (sell) signals are generated when a
short-term moving average exceeds (is less than) the long-term moving average by a
specified percentage. In this study we use the long-term moving averages of 20, 50,
100, 150 and 200 days. As for the short-term moving average, like the BLL study, we use
1 day (the raw price) moving average. Thus, a buy signal is emitted when the index
price level breaks the long MA from below and a sell signal is emitted when the index
level breaks the long MA from above.

47
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

We define Pt as the short-term moving average or the raw index level at time t, and
define the long-term moving average of M-day at time t as:

M −1
1
MAt(M) =
M
∑P
i =0
t −i . (1)

An investor/trader who follows MA trading rules could presumably estimate the index
price that would trigger the buy and sell signal just before the day’s close and initiate a
conditional limit order at the close of the market to perform various trading rules. For
example a trader that has been out of the market following a trading rule based on
MA50 (if P > MA50, in the market and if P ≤ MA50 then out of the market) could have
the following conditional limit order at the close of the day: If the price index is above
previous close of MA50, then buy market on close (MOC). Therefore the trader will be
in the market the next day by buying the index at the closing limit price (i.e. next day
will be a buy day). The next day’s return will be the change in logarithms of the index
level. The same reasoning holds if the trader has been in the market, sell the index MOC
if price is below previous close of MA50 (i.e. next day will be a sell day). The next day’s
return will be the change in logarithms of the index level. If the conditional limit prices
are not filled, then the trader will not switch position. The trader is either in the market
“buy” days or out of the market “sell” days, again the sell days mean the trader is out
of the market and not shorting the market.

We define mean buy and mean sell day returns as follows:

1
µb=
nb
∑R b (2)

1
µs=
nS
∑R s (3)

where, Rb and Rs are daily returns of buy and sell days; nb and ns are the total number
of buy and sell days respectively. We will test whether the returns of any moving
average trading rules are greater than a B&H strategy and whether the mean buy is
different from the mean sell. The null and alternative hypotheses are expressed in Table
1:

Table 1: Test Hypotheses

Test One Test Two Test Three

Null: H0 µb – µh = 0 µs – µh =0 µb – µs = 0
Alternative: Ha µb – µh ≠ 0 µs – µh ≠ 0 µb – µs ≠ 0

where µh is the mean of the B&H strategy. Following Kwon and Kish (2002), the test statistic
for the Test is:

t= µb − µ s , (4)
(σ / n b ) − (σ / n s )
2
b
2
s

48
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

where µ represents the mean returns, σ2 is the estimated variance, and n is the number
of observations in each situation. Statistically significant differences in buy-sell day index
returns implies the effectiveness of the MA rules to forecast equity returns, this is the same
procedure used by BBL. The above formula is also used for Test One and Test Two by
replacing the appropriate variables in the t-statistic formula.

4. Empirical Results

In Table 2, we exhibit the summary statistics of daily returns for NASDAQ index in the
entire 43.875 years with four sub-periods for comparison. For the entire period
(01/72~10/15), the average daily return is 0.033% with a standard deviation of 1.23 over
11423 observations or days. The t-statistics for the mean returns of B&H in NASDAQ is
2.84. At the 5 percent confidence level for large numbers of observations, compared
with the critical t-value 1.96 for two-tailed test, the unconditional means of NASDAQ for
the entire period are significantly different from zero. The skewness implies that return
distributions are almost symmetric for NASDAQ index. The Kurtosis is higher than 3,
implying that the return distributions may not be normal, and the Jarque-Bera test
rejects normality of returns in the entire period and all four sub-periods. All of the first and
second order autocorrelations are low except for the first and second sub-periods.

Table 2: Summary Statistics of Returns: NASDAQ Composite Index

NASDAQ Composite Index

Period Mean % SD % Skewness Kurtosis ρ1 ρ2 JB n
01/72 –
0.033% 1.23% -0.29 10.08 0.05 -0.01 24046 11423
10/15
01/72 –
0.023% 0.76% -0.77 3.91 0.32 0.08 453 3391
12/84
01/85 –
0.043% 0.86% -2.28 30.79 0.26 0.05 86191 2609
12/94
01/95 –
0.041% 1.80% 0.01 4.05 0.00 -0.01 120 2610
12/04
01/05 –
0.028% 1.33% -0.25 7.85 -0.07 -0.01 2788 2813
10/15

Note: SD = Standard Deviation; JB = Jacque Bra; critical value = 6; ρ1 & ρ2 = first and second order return
correlations; n = total number of days in the period.

Table 3 summarizes the results of various moving average (MA) trading rules for the
NASDAQ index. For each MA rule, we report mean buy returns, mean sell returns, the
mean buy minus sell returns, standard deviations of returns on buy and sell days, total
number of buy and sell days, and the number of signals generated. The numbers in the
parentheses are the t-statistics defined in Equation (4) to test the difference of the mean
buy and mean sell from the unconditional mean, and buy-sell from zero. For example,
the first row of NASDAQ Composite Index shows the results of MA-20 trading rule. The
trader will be in the market (buy days) if the index level is greater than MA20 and out of
the market (sell days) if the index level is less than or equal to MA-20. Similarly, the other
rows report the results of other MA-days trading rules.

49
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

Table 3: Statistical Results for Moving Average Rules

NASDAQ Composite Index

No. of
Rules Buy Sell Buy-Sell SDb SDs nb ns Signals
0.00091 -0.00054 0.00145
MA-20 0.00966 0.01539 6847 4576 1090
(3.60)* (-3.46)* (5.76)*
0.00081 -0.00047 0.00128
MA-50 0.00949 0.01583 7101 4322 605
(3.01)* (-2.98)* (4.82)*
0.00068 -0.00030 0.00098
MA-100 0.00929 0.01630 7300 4123 381
(2.22)* (-2.23)* (3.53)*
0.00055 -0.00013 0.00068
MA-150 0.00927 0.01681 7639 3784 347
(1.45) (-1.54) (2.33)*
0.00054 -0.00013 0.00067
MA-200 0.00948 0.01688 7835 3588 273
(-0.20) (0.23) (-0.31)

Moving average trading results for daily data from 1972-2015. nb and ns are the number of buy
and sell days, SDb and SDs are standard deviation of buy and sell days respectively. The numbers
in the parentheses are the t-statistics testing the difference of the mean buy and mean sell from
the unconditional 1-day mean, and buy-sell from zero. Numbers marked with asterisks are
significant at the 5% level for a two-tailed test, tcrit., 0.05 =1.96.

The testing results of significance in Table 3 are very strong for NASDAQ. The mean buy
and sell returns are shown in Columns 2 and 3. For MA20, MA50, and MA100 in NASDAQ,
the mean buy returns are all positive with significant t-statistic and the mean sell returns
are all negative with significant t-statistic. All the buy minus sell differences (Column 4)
are positive and the t-test statistics are highly significant to reject the null hypothesis of
equality with zero, except for MA-200. Therefore, the four out of five MA trading rules,
MA20, MA50, MA100 and MA150 have predictive power in the NASDAQ Composite
Index.

It is interesting to note that the standard deviations for buy days are always smaller than
those for sell days in Columns 5 and 6. This implies that the down markets are more
volatile than the up markets. Columns 7 and 8 report the number of buys and sells for
various rules. For example when applying MA20 trading rule for NASDAQ, 60% of the
days we are in the market (buy days) and 40% of the days out of the market (sell days).
Finally the last column reports the number of signals for in and out of the market, as the
MA days increases the number of in and out of the market decreases. It is also
noteworthy to point out the negative returns for sell days is problematic for the
proponents of EMH. As BLL indicates, these returns cannot be explained by seasonality
since they are based approximately on 40% of all trading days. This predictability of
returns can reflect either (1) changes in expected returns generated from an
equilibrium model, or (2) market inefficiency. Although changes in expected returns are
possible, it is hard to imagine an equilibrium model that predicts negative returns over
such a large fraction of trading days.

The results in Table 3 indicate that moving average rules do indeed have predictive
power for NASDAQ and can discern recurring-price patterns for profitable trading.
Given the predictive power of MA rules, the next section discuss how can we design
various trading strategies to beat the B&H strategy considering both transaction costs
and market risk.

50
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

5. Trading Strategies

Now that we have confirmed the predictive power of MA rules for NASDAQ, we
investigate whether it is possible to design various trading strategies for MA rules to beat
the B&H strategy considering both transaction costs and risk. For each MA trading rule,
the profitability will be varied with the position a trader takes when the rule emits sell
signals. For example, if a trader does not invest in any alternative on the sell days (out of
market), then his or her return on the sell days will be zero. Then the trader’s mean return
can be counted as simple as (nb /n)*µb + (ns /n)*0. If a trader chooses to invest in the
money market on the sell days, then the trader’s mean return will include the interest
earnings at money market rate on those sell days.

In this study, following Metghalchi et al. (2015), we consider a total of four strategies as
the following:

Strategy 1 ─ The trader will be in the stock market when MA rules emit buy signals and
be in the money market when a MA rules emit sell signals (long/money).

Strategy 2 ─ The trader will be in the stock market when MA rules emit buy signals and
short the market when the rules emit sell signals (long/short).

Strategy 3 ─ The trader will borrow at the money market rate and double stock
investment when trading rules emit buy signals and be in the money market
when trading rules emit sell signals (leverage/money).

Strategy 4 ─ The trader will borrow at the money market rate and double stock
investment when trading rules emit buy signals; short the market when the
trading rules emit sell signals (leverage/short). Note that the total return on
buy days for the leverage strategy is TRt = 2Rt – Mt, where Rt is the index return
on day t and Mt is the daily money market rate.

For each strategy, we estimate the daily return then subtract it from the daily return of
B&H strategy to get the daily difference return. To test whether the average daily
difference (ddif) is greater than zero or not, we express the null and alternative
hypotheses as:

H0 : ddif ≤0
Ha : ddif >0
The t-statistic for the above test is:

µ ( ddif ) ,
t= (5)
σ 2 ( ddif ) / n

where µ (ddif) is the average daily difference of returns of each strategy over the B&H
strategy and σ2(ddif) is the variance of daily difference returns, and n is the total number
of days. Table 4 reports the results of the above six strategies for MA rules.

51
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

Table 4 shows the strong results with positive daily difference returns and significant t-
statistics. At first glance, MA20 and MA50 rules with Strategies 3, 4 are the most profitable
rules and strategies. If market risk and transaction costs are not considered, then the best
strategy would be to apply MA20 rule using Strategy 4, an extra return of 0.085% per day
over the B&H strategy.

Table 4: Trading Strategies of MA Rules in NASDAQ

Strategy 1 Strategy 2 Strategy 3 Strategy 4

µ(ddif) µ(ddif) µ(ddif) µ(ddif)
0.00030 0.00043 0.00072 0.00085
MA-20 (3.33)* (2.36)* (6.31)* (4.27)*
0.00026 0.00035 0.00064 0.00073
MA-50 (2.89)* (1.94)* (5.59)* (3.65)*
0.00019 0.00021 0.0005 0.00052
MA-100 (2.09)* (1.16) (4.33)* (2.59)*
0.00012 0.00009 0.00036 0.00032
MA-150 (1.35) (0.48) (3.13)* (1.60)
0.00012 0.00009 0.00036 0.00032
MA-200 (1.35) (0.48) (3.13)* (1.60)

µ(ddif) is the average of daily difference between the return of each strategy and the buy-and-
hold strategy. The numbers in the parentheses are the t-statistics testing whether the average
daily difference is greater than zero. Asterisks imply significant at the 5 percent level or less for
one-tail test, tcrit., 0.05 =1.645.

However, we must consider risk and transaction costs of each strategy in order to
choose the best rule/strategy. Table 5 reports the one way “break-even” transaction
costs and the risk of various MA rules for the above four strategies. The one-way break-
even transaction cost (BEC) eliminates the extra return from MA trading rules. Following
Bessembinder and Chan (1995), we estimate the one way BEC by adding the daily
excess returns (Beyond B&H) produced by each trading rule and strategy over the
11423 days and then divide it by the number of trades over the entire period. Since
Strategies 2 and 4 imply shorting the NASDAQ index, we divide the sum of the daily
excess return by 2 times the number of trades. We also assume that investing in a money
market does not incur any transaction cost. The estimation of risk is the standard
deviation of daily returns of each strategy which should be compared with the daily
standard deviation of B&H strategy of 1.23 % in Table 2 for the entire period.

Table 5 provides BEC and risk of each trading rule and various strategies; the first number
in each cell is the BEC and the second number is risk, both in percent. Strategy 1 has an
average risk of 0.76 % much lower than the B&H risk of 1.23 % of Table 2. The risk of
Strategy 2 is similar to the risk of the B&H strategy. Finally, the average risk of the
Strategies 3 and 4 are 1.51% and 1.80% respectively, both higher than the risk of B&H
strategy. In comparison, Strategy 1 is superior to Strategies 2 due to its higher BEC and
lower risk. The risk and return trade-off implies that if a trader prefers a lower-than-market
risk, then Strategy 1 in combination with MA-100 or MA-150 or MA-200 would be the best
trading rules with BEC of 0.22, 0.24, and 0.29 percent. On the other hand, if a trader has
a little higher risk tolerance, then Strategy 3 is superior to Strategies 4, since strategy 3
implies lower risk and higher BEC than Strategies 4. Strategy 3 with either MA rules of 100,
150, or 200 days will provide profitable trading if transaction cost of trading NASDAQ
composite ETF is less than 0.29 %. In conclusion, MA-100, MA-150, and MA-200
associated with Strategies 1 and 3 serve as the most profitable choices for traders.

52
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

Table 5: Break-Even Costs and Risk of Various Strategies

Strategy 1 Strategy 2 Strategy 3 Strategy 4

MA-20 0.10/0.75 0.09/1.23 0.15/1.49 0.11/1.78

MA-50 0.15/0.75 0.12/1.23 0.20/1.50 0.15/1.79

MA-100 0.22/0.74 0.17/1.23 0.29/1.49 0.21/1.78

MA-150 0.24/0.76 0.19/1.23 0.33/1.52 0.23/1.80

MA-200 0.29/0.79 0.22/1.23 0.38/1.57 0.26/1.83

The break-even cost (BEC) estimated by dividing total daily excess returns into total number of
trades over the entire period from 1972-2015. Risk is the standard deviation of daily returns. In each
cell the first number is the BEC in percent and the second number is risk in percent. Each cell
shows (BEC/Risk).

To test the robustness of results, we divide the entire sample into four sub-periods and
provide the estimated BECs for Strategies 1 and 3 for MA-100, MA-150, and MA-200 in
Table 6. Table 6 presents the risk and BECs of our best three rules for four sub-periods.
The risks (standard deviation of returns) of the B&H strategy are 0.761 %, 0.856 %, 1.795%
and 1.332 % for four sub-periods respectively. The BECs are estimated the same way, by
dividing total excess return over the B&H strategy into the total number of trades in each
sub-period.

Noted in Table 6, the BEC are relatively high in the first three sub-periods for both
Strategies 1 and 3. Compared with the risk of B&H in each period, Strategy 1 again has
a lower risk in each sub-period with high BEC in the first three sub-periods. Strategy 3 has
a bit higher risk in each sub-period than B&H but has very high BEC implying strong
possibility of profitable trading. A trader with a bit more risk tolerance than B&H would
adopt MA-100 with combination of Strategy 3 to gain higher BEC in each sub-period,
including the fourth sub-period. For risk-averse choice, a trader could apply MA-100 with
Strategy 1 and gain very well in the first 3 sub-periods but would lose not much (since
BEC is a small negative number) in the fourth sub-period. Our findings also partially echo
Feng et al (2013) to indicate that MA trading rules have not been very successful
recently, since the publication of BLL.

In order to see whether other well-known trading rules can do as well as moving
average rules, we apply two addition popular indicators to NASDAQ composite index.
The Relative Strength Index (RSI) indicator, which measures the velocity of directional
movement by providing the internal strength of a single security or index, was created
by Wells Wilder (1978). Wilder suggests using 14 days for estimating the RSI‘s value which
ranges from 0 to 100. In this study we use two variants of RSI trading as follow:

I. In the market if RSI >50;

Out of the market if RSI ≤ 50.

II. Many traders believe if RSI is above 85 it implies that the market is overbought and
if it is below 15, then the market is oversold. Thus RSI model 2’s rules are as follow:

In the market if: 50 ≤ RSI ≤ 85 or if: RSI ≤ 15;

Out of the market if: 15 < RSI <50 or if RSI > 85.

53
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

Table 6: Break-Even Costs and Risk of Various Strategies

Strategy 1
MA-100 MA-150 MA-200
Sub-Period BEC % RISK % BEC % RISK % BEC % RISK %
1/72 – 12/84 1.63 0.48 2.46 0.50 1.72 0.51
1/85 – 12/94 0.89 0.56 0.67 0.56 0.39 0.57
1/95 – 12/04 0.34 1.06 0.20 1.10 0.51 1.16
1/05 – 10/15 -0.09 0.78 -0.35 0.79 -0.32 0.81
Strategy 3
Sub-Period BEC % RISK % BEC % RISK % BEC % RISK %
1/72 -12/84 2.74 0.99 4.1 1.00 2.75 1.02
1/85 – 12/94 2.54 1.11 2.26 1.13 1.74 1.14
1/95 – 12/04 1.27 2.13 1.04 2.19 2.01 2.31
1/05 – 10/15 0.34 1.56 -0.24 1.59 0.04 1.61

Results for four sub-periods. BEC is the break-even cost, estimated by dividing total daily excess
returns for each sub-period into total number of trades in that sub-period. Risk is the standard
deviation of daily returns in each sub-period.

The second popular indicator allied to NASDAQ Composite is the Histogram based on
Gerald Appel’s (1980) Moving Average Convergence Divergence (MACD), Stochastic.
MACD is the difference between two exponential moving averages (EMA). We follow
the Appel’s recommendation and use 26 and 12 day EMAs. A 9-period EMA of the
MACD (the signal line) is then plotted on top of the MACD. The trading rule is as follow:
in the market if MACD is above the signal line and out of the market if MACD is below
signal line. In Table 7 we present the results for the above three models, two based on
RSI and on one MACD rules.

Table 7: Statistical Results for RSI and MACD Rules

NASDAQ Composite Index

No. of
Rules Buy Sell Buy-Sell SDb SDs nb ns Signals
RSI-1 0.00091 -0.00219 0.00310
0.00938 0.01473 7029 4394 983
(3.64)* (-10.04)* (12.45)*
RSI-2 0.00088 -0.00050 0.00138
0.00964 0.01540 6842 4581 1113
(3.37)* (-3.23)* (5.38)*
MACD 0.00065 -0.00130 0.00195
0.01065 0.01202 5789 6534 859
(1.76) (-8.27)* (9.17)*

Results are for daily data from 1972-2015. nb and ns are the number of buy and sell days, SDb and
SDs are standard deviation of buy and sell days respectively. The numbers in the parentheses are
the t-statistics testing the difference of the mean buy and mean sell from the unconditional 1-day
mean, and buy-sell from zero. Numbers marked with asterisks are significant at the 5% level for a
two-tailed test, tcrit., 0.05 =1.96.

The results of Table 7 strongly support the predictive power of RSI and MACD trading
rules; all buy minus sell t-statistics are highly significant rejecting the hypothesis that the

54
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

mean buy days is equal to the mean sell days. In Table 8 we report the average buy
minus sell days for each sub-period with their corresponding t-statistics.

Table 8: Mean buy minus mean sell for each sub-period

Sub-Period 1 Sub-Period 2 Sub-Period 3 Sub-Period 4

RSI 1 (12.73)* (7.72)* (5.74)* (3.02)*

RSI 2 (5.38)* (4.28)* (2.48)* (-0.97)

MACD (9.47)* (6.32)* (3.34)* (2.80)*

Mean buy minus mean sells are the difference of average buy days minus average of sell days
for each sub-period. The numbers in the parentheses are the t-statistics testing the difference of
buy-sell from zero. Numbers marked with asterisks are significant at the 5% level for a two-tailed
test, tcrit., 0.05 =1.96.

Again, Table 8 concludes a very strong predictive power of technical trading. All except
one buy minus sell averages are highly significant rejecting the equality of mean buy
days with mean sell days. This conclusion does not support the efficiency of NASDAQ
composite index.

4. Conclusion

In this paper, we investigate a few moving average trading rules for the NASDAQ
Composite index over the period of 1/3/1972 to 10/14/2015. Overall our results strongly
support the predictive power of MA trading rules for NASDAQ. Almost all the buy-sell
differences are significantly positive to reject the null hypothesis of equality of buy days
returns with sell days returns for NASDAQ. For NASDAQ, the t-statistics for most buy and
most sell are significant, rejecting the null hypothesis that the mean buy and sell returns
equal to the mean of B&H returns. Therefore, we can conclude that MA rules have
predictive power for NADAQ composite index.

To investigate the most profitable strategies for MA rules for NASDAQ when considering
both transaction costs and risk, we design a total of 4 strategies to test the significance
and robustness in profitability. There are two driftnet risk tolerance strategies found to
be very profitable when using MA-100 in NASDAQ. For risk-averse investors, Strategy 1,
in which a trader will be in NASDAQ when MA-100 emits buy signals and be in the money
market when MA-100 emits sell signals, is the choice to beat the B&H strategy for entire
period (BEC of 0.22 %, and a risk lower than B&H) and 3 out of 4 sub-periods with very
high BECs. For a more risk-taker trader, applying MA-100 with Strategy 3 (i.e. a bit more
risk than B&H) leveraging at money market rate for buy days and being in the money
market for sell days, can beat handsomely the B&H for the entire period and each sub-
period. Finally we apply two very popular indicators (RSI & MACD) to NASDAQ
composite index and find that both also have very strong predictive power for entire
period and each sub-period. We would note that both RSI and MACD trading rules
imply more in and out of the market than MA 100, therefore traders with higher
transaction costs should be careful to apply these trading rules.

55
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

References
Alexander, S., 1964. Price movements in speculative markets: trends or random walks.
Industrial Management Review 5(2), 25-46.

Appel, Gerald., Hitschler F., 1980. Stock Market Trading System: A guide to Investment
Strategy for the 1980s. Homewood, Il: Dow Jones-Irwin.

Bessembinder, H., Chan, K., 1995. The profitability of technical trading rules in the Asian
stock markets. Pacific-Basin Finance Journal 3, 257-284.

Brock, W., Lakonishok, J., LeBaron, B., 1992. Simple technical trading rules and the
stochastic properties of stock returns. Journal of Finance 47(5), 1731-1764.

Cowles, A., 1960. A revision of previous conclusions regarding stock market behavior.
Econometrica 28, 909-915.

Draper, P., Paudyal, K., 1997. Microstructure and seasonality in the UK equity market.
Journal of Business Finance and Accounting 24(7-8), 1177-1204.

Fama, E., 1965. The behavior of stock market prices. Journal of Business 38(1), 34-105.

Fama, E., 1970. Efficient capital markets: a review of theory and empirical work. Journal
of Finance 25(2), 383-417.

Fama, E., Blume, M., 1966. Filter rules and stock market trading profits. Journal of Business
39, 226-341.

Granger, C., Morgenstern, O., 1963. Spectral analysis of New York stock market prices.
Kyklos 16, 1-27.

Jensen, M., Benington, G., 1970. Random walks and technical theories: some additional
evidence. Journal of Finance 25(2), 469-482.

Kirkpatrick II, C., Dahlquist, J., 2015. Technical Analysis: The Complete Resource for
Financial Market Technicians, 3rd edition, FT Press.

Kwon, K., Kish, R., 2002. Technical trading strategies and return predictability: NYSE.
Applied Financial Economics 12(9), 639-653.

Larson, A., 1960. Measurement of random process in futures prices. Food Research
Institute 1(3), 313-324.

Lo, A., MacKinlay, A., 1990. Data-snooping biases in tests of financial asset pricing models.
Review of Financial Studies 3(3), 431–468.

Lucke, B., 2002. Are technical trading rules profitable? Evidence for head-and shoulder
rules. Applied Economics 35, 33-40.

Lukac, L., Brorsen, B., Irwin, S., 1988. A test of futures market disequilibrium using twelve
different technical trading systems. Applied Economics 20, 623-639.

Mandelbrot, B., 1963. The variation of certain speculative prices. Journal of Business 36(4),
394-419.

56
 MOVING AVERAGE TRADING RULES FOR NASDAQ COMPOSITE INDEX

Metghalchi, M., Chen, C., Hayes, L., 2015. History of share prices and market efficiency of
the Madrid general stock index. International Review of Financial Analysis 40, 178-184.

Meyers, T., 2002. The Technical Analysis Course, McGraw-Hill Co.

Mills, T., Coutts, J., 1995. Calendar effects in the London Stock Exchange FT-SE indices.
European Journal of Finance 1(1), 79-93.

Park, C., Irwin, S., 2007. What do we know about the profitability of technical analysis?
Journal of Economic Surveys 21(4), 786-826.

Pring, M., 1991. Technical Analysis: Explained, McGraw-Hill Co.

Sweeney, R., 1986. Beating the foreign exchange market. Journal of Finance 41(1), 163-
182.

Van Horn, J., Parker, G., 1967. The random walk theory: an empirical test. Financial Analyst
Journal 23, 87-92.

Wilder, W. (1978) New Concepts in Technical Analysis, Trend Research, McLeansville, NC

57