Trend salience, investor behaviors and momentum profitability

Paul Docherty* and Gareth Hurst

Newcastle Business School, The University of Newcastle, Australia.

Abstract

Trend extrapolation in financial markets has been well documented, however it is contentious

as to which trends will be extrapolated or mean reverted. We examine whether investors are

more likely to extrapolate trends that they perceive to be salient by examining an investment

strategy that considers both the magnitude and the strength of the trend. Consistent with

behavioral models of momentum, our investment strategy based on trend salience

significantly outperforms traditional momentum strategies and is not explained by the Carhart

[1997] four-factor model. The relative performance of the trend salience signal is robust

across different investment horizons and size-sorted portfolios, although is time-varying; the

strategy does not outperform momentum in “down” markets where volatility is high and

salient trends are more difficult to identify.

JEL Classification: G02, G11, G12

Key Words: Momentum; trend salience; extrapolation; market states.

Acknowledgements: We thank seminar participants at The University of Newcastle and

Peter Brooke. Funding provided by Platypus Asset Management is gratefully acknowledged.

*Corresponding Author: Newcastle Business School, The University of Newcastle.

University Drive Callaghan, NSW, Australia 2308. Email: [email protected]

Electronic copy available at: http://ssrn.com/abstract=2395718

 I. Introduction

There is now such voluminous evidence in support of the profitability of momentum

strategies that momentum is recognized as the “premier anomaly” (Fama and French [2008]).

Despite the weight of evidence supporting momentum, there is no consensus view on the

rationale for the existence of momentum returns. In light of the inability of risk-based models

to explain momentum (Fama and French [1996], Grundy and Martin [2001]), behavioral

models have been developed to explain this phenomenon. Daniel, Hirschlerifer and

Subramanyam [1998] propose a feedback model that incorporates investor overconfidence

due to biased self-attribution. Hong and Stein [1999] suggest an alternate model and argue a

market is made up of two heterogeneous investors; newswatchers and momentum traders that

demonstrate bounded-rationality. A model that incorporates investor forecasts is proposed by

Barberis, Schleifer and Vishny [1998], who suggest investors, initially underreact to news

due to conservatism, resulting in positive autocorrelation, and ultimately overreact over long

periods due to the representative heuristic (Tversky and Kahneman [1974]). Within each of

these models, momentum profits and subsequent momentum reversals may be explained by

market inefficiency due to either individual investor behavior (Barberis et al. [1998], Daniel

et al. [1998]) or market imperfections (Hong and Stein [1999]).

Taken together, these behavioral models predict a short-run continuation in prices followed

by long-run momentum reversals. Part of this process involves investors extrapolating recent

price changes. Evidence of trend extrapolation exists in both psychology and finance settings

(see for example Frankel and Froot [1988], Smith, Suchanek and Williams [1988], Delong,

Shleifer, Summers and Waldman [1990], Haruvy, Lahav and Noussair [2007], Choi, Gabaix

and Madrian [2010], Fuster, Laibson and Mendek [2010], He and Shen [2010]). Extrapolation

may occur as individuals react to the strength of information rather than its statistical weight

Electronic copy available at: http://ssrn.com/abstract=2395718

(Griffin and Tversky [1992]). This process involves investors using the ‘law of small

numbers’ to overestimate the probability of a particular stock belonging to a particular

distribution. A stock that has experienced particularly strong recent performance may induce

investors to extrapolate that recent performance, due to the salience of that signal. A formal

model of investor forecasts, proposed by Andreassen and Kraus [1990], demonstrates how a

trend becomes salient. This model accounts for trends in individual forecasts by using the

Holt-Winters linear trend exponential smoothing model, and incorporates the following

equations:

( )( ) (1)

( ) ( ) (2)

( ) (3)

where is the realised value at time t, is the forecast at time t, is the previous period

forecast, and are the trends at t and t-1 respectively and α and β are measures of

salience given to each variable. If β = 0, the estimates of the trend become 0 and the model

reduces to a simple smoothing model given by:

( ) (4)

This implies that if α is less than 1, if today’s realized value is higher than the observed value

at t-1, the future forecast is less than today’s forecast. In essence, if investors do not consider

a trend salient, they are less likely to extrapolate previous price movements. A β > 0 indicates

investors are incorporating this trend into their forecasts. If investors consider the rate of

change in a stock’s price history to be an important factor for predicting future returns, then it

Electronic copy available at: http://ssrn.com/abstract=2395718

follows that a positive trend that is improving should be more salient than a positive trend

that is deteriorating.

Traditional momentum strategies only consider the magnitude of past performance to

determine whether a stock is a “winner” or a “loser”. We apply behavioral theories of trend

extrapolation by also considering the strength of the trend as a measure of the trend salience,

and examine whether the momentum premium can be improved by including (excluding)

stocks with a salient (non-salient) trend. Specifically, we argue that investors will be more

likely to extrapolate a recent trend if they consider that trend to be salient. As traditional

momentum strategies do not account for the salience of the trend within the formation period,

a strategy that identifies winner stocks and loser stocks with strengthening recent trends

should outperform the traditional momentum strategy. Consistent with Andreassen and

Kraus’s [1990] model, we show that stocks with a strengthening trend exhibit stronger price

continuations than stocks with a deteriorating trend. We develop an investment strategy based

on the salience of a momentum trend whereby the investor purchases salient winners and

sells salient losers. This investment strategy significantly outperforms a traditional

momentum strategy and is robust to size of the stocks in the sample and a risk based

explanation. However the premium is sensitive to the market state, which suggests the

strength of a price signal, is altered by the condition of the market.

II. Data

The data used in this study is sourced from the Australian Graduate School of Management

(AGSM) database and comprises of stocks listed on the Australian Stock Exchange for the

4
period January 1992 to December 20111. The use of Australian data provides an interesting

laboratory; on average the size of the individual firms in the Australian market are much

smaller than the United States and the overall capitalization of the Australian market is

dominated by a few large firms (Gaunt and Gray [2003]). Further, the Australian Stock

Market is concentrated by institutional and foreign ownership contributing to 80% of direct

share ownership (Black and Kirkwood [2010]). This is significant, as numerous behavioral

studies find that experienced traders are less likely to exhibit the behavioral biases described

in the models (for example see Edwards [1968], Smith et al. [1988], DeBondt [1993],

Caginalp, Porter and Smith [2000], Greenwood and Nagel [2008]). Therefore, the use of

Australian data may provide interesting insights into the existing behavioral theories.

For a stock to be included in the sample it must have been listed for a full 12 months prior to

the holding period. After filtering the data, all stocks were ranked based on their market

capitalizations as at 31 December of each year. The 500 largest stocks ranked on market

capitalization were included in the sample and all other listed stocks were excluded. As a

result of this process, the constituents within the sample may differ each year.

The use of the 500 largest stocks is motivated by prior research into Australian equities. The

Australian equity market has shown a significant negative relationship between size and

returns (Beedles [1992]), independent of the stock’s price (Beedles, Dodd and Officer [1988],

Gaunt, Gray and McIvor [2000]). Beedles [1992] finds a positive relationship between size

and liquidity. Whilst these findings are consistent with the United States market, the

concentration of a few large firms ensures the Australian market has notable characteristics

that differ from its United States counterpart (Gaunt and Gray [2003]). The number of small

1
The first year in the sample is 1990, as the 12 month/12 month momentum strategy requires 24 lagged months

returns data; the first full calendar year that can be examined is 1992.

5
firms in the Australian equity market has implications for momentum studies. Brailsford and

O’Brien [2008] find an interaction between firm size and momentum in Australian equities.

Demir, Muthuswamy and Walter [2004] stress that momentum strategies rely heavily on short

selling loser stocks. The short selling of very small, illiquid stocks can incur high transaction

costs such as large bid-ask spreads and higher borrowing costs. The restricted sample of the

top 500 stocks by market capitalization is used to limit the bias caused by these effects and to

focus on an implementable strategy.

The length of sample period can also impact the results of momentum studies. Brailsford and

O’Brien [2008] note that many studies of momentum in Australia use short sample periods.

The use of 20 years of data makes this one of the longest studies of momentum in Australia.

The use of a long sample period is not trivial. Prior research demonstrates the relationship

between market up and downswings and momentum profits (Cooper, Gutierrez and Hameed

[2004], Phua, Chan, Faff and Hudson [2010]). By using a long sample period, trend salience

can be examined over both bull and bear markets, including the impacts of the recent global

financial crisis.

III. Methodology

The construction of the salient strategies involves a two-step process, first momentum

portfolios are formed using a method consistent with Jegadeesh and Titman [1993, 2001]. For

each year, stocks in the sample are ranked based on the cumulative returns over the past J

months and allocated into quintile portfolios. Stocks that have the highest cumulative returns

are allocated to the top quintile and are categorized as winners (W). Stocks with the lowest

cumulative returns are allocated to the bottom quintile of stocks and are categorized as losers

(L). The second step involves categorizing winner (losers) stocks as either salient or non-

6
salient winners (losers) by measuring the slope of their recent return performance as the ratio

of their past M month to 12 month geometric average rate of return given by:

(5)
∏ ( )
∏ ( )

where is the geometric average rate of return (GARR) over the past M months for

stock i as at period t.

The median value of the ratio of M month GARR and 12 month GARR is calculated. If the

return of a “winner” stock has a ratio that is greater than the median, the stock is allocated to

the salient winner (SW) portfolio, otherwise the stock is allocated to the non-salient winner

(NW) portfolio. If the returns of a loser stock has a ratio that is less than the median, the stock

is allocated to the salient loser (SL) portfolio, otherwise the stock is allocated to the non-

salient loser (NL) portfolio. For easy comparison with traditional momentum strategies, M

months are defined as 3, 6 and 9 months. To avoid short-term reversals (Jegadeesh [1990],

Lehmann [1990]) and the bid ask bounce (Lehmann [1990]), the Jegadeesh and Titman

[2001] methodology is adopted and one month is skipped between the formation and holding

periods.

Yu and Chen [2011] provide a formation strategy that differentiates winner and loser stocks

according to whether the trend in the formation period is strengthening or deteriorating. Our

construction of the salient trend portfolios is consistent with Yu and Chen [2011] with a

notable difference; Yu and Chen [2011] benchmark SW (SL) and NW (NL) with the ratio of

M month GARR and 12 month GARR of 1. The difference in benchmark has two effects;

firstly the use of the median in this study eliminates the bias of market states in determining

whether a stock is an SW or NW. This choice of benchmark is important in the study of trend

7
salience. If the aggregate market is negative in the most recent M months of the formation

period, the rate of change of winners would, on average, be decreasing; however the negative

rate of change of the loser stocks would be increasing. This would lead to a larger number of

NW and a larger number of SL. If the benchmark of 1 is adopted the results of this study

would not be independent of market conditions. Secondly, the effect of changes in market

states results in a different number of stocks in each portfolio each month therefore the results

can be influenced by portfolio concentration. The use of the median eliminates the effect of

market states and ensures the same number of stocks in each portfolio each month.

To analyze the effects of trend salience in portfolio returns, two trading strategies that utilize

this information in the formation period are constructed. The first strategy buys the SW

portfolio and sells the SL portfolio. As the first strategy involves buying (selling) winner

(loser) stocks that display a strengthening trend, the strategy is denoted as the ‘salient

momentum’ strategy. The second strategy buys the NW portfolio and sells the NL portfolio.

As the second strategy involves buying (selling) winner (loser) stocks that display a

deteriorating trend, this strategy is denoted as the ‘non-salient momentum’ strategy.

By decomposing the formation period in this way, it can be seen whether trend salience

influences investors to extrapolate recent performance. If the salient momentum strategy

outperforms the non-salient momentum strategy it implies that investors are sensitive to the

strength of the price signal in the formation period. Stocks with a strengthening trend are

likely to be extrapolated, which is consistent with the behavioral models that are predicated

on short-run correlation of prices.

Grundy and Martin [2001] show that momentum cannot be explained by standard risk

models. To similarly ensure that the returns generated by our salient momentum strategy can

be attributed to behavioral factors and not the systematic selection of risky stocks, we regress

8
the returns generated by the salient momentum strategy against the Carhart [1997] four

factors. To perform these tests, the following equation is estimated:

[ ] (6)

where is the return on investment strategy p at time t, and , , ,

are the excess return on the market and the size, value and momentum factors

respectively.

IV. Results

Table 1 shows the raw2 monthly returns for 6 portfolios with equal-weighted returns; the W

and L portfolios with a 12-month formation period; the SW (NW) portfolios, which consists

of stocks from the W portfolio where the ratio of the past M-month GARR to 12-month

GARR is greater (equal to or less) than the median; the NL (SL) portfolios, which consists of

stocks from the L portfolio where the ratio of the past M-month GARR to 12-month GARR is

equal to or greater (less) than the median. For brevity we only report the results of the M to

12-month strategies3; however we have tested M to 9-month and M to 6-month strategies and

found the results to be qualitatively similar.

[TABLE 1 ABOUT HERE]

The results of the portfolio analysis shows that the winner portfolios returns decrease as the

length of the holding period increases, whilst the loser portfolios have increasing returns as

2
We use raw returns throughout this paper, as investors are influenced by a change in price, rather than a

change in risk-adjusted prices.

3
For papers that examine momentum using a 12-month formation period see Fama and French [1996], Carhart

[1997], Moskowitz and Grinblatt [1999], Grundy and Martin [2001], Grinblatt and Moskowitz [2004]

9
the holding period increase. This decay is consistent with previous momentum studies

(Jegadeesh and Titman [1993, 2001]) and is consistent with the long-run mean reversal in the

behavioral models. Consistent with trend salience, SW in each strategy outperform the NW.

Further, SL experience lower returns across all holding periods than the NL. For example the

6 to 12-month SW portfolio held for 6 months yields a raw monthly return of 1.47% that is

statistically significant at the 1% level; this compares with a 0.71% return for the NW

portfolio. This feature is not consigned to the particular strategy or holding period. All SW

returns are significant at the 1% level, whilst all NW returns other than 6 to 12 and 9 to 12-

month strategy held for 3 months are insignificant from zero. The loser portfolios offer the

same pattern of returns, whilst all returns are insignificant from zero, across all strategies and

holding periods SL yield lower returns than the NL portfolios.

A possible explanation for the outperformance (underperformance) of the SW (SL) portfolios

is that these portfolios systematically hold winners (losers) with extreme performance. To test

whether SW and SL hold stocks with extreme performance, the winner and loser stocks in the

momentum strategy using decile performance was decomposed and found that the average

number of SW and NW in the winner portfolio was 24 and 26 respectively and the average

number of NL and SL in the loser portfolio was 24 and 26 respectively. In unreported results,

a chi-squared test found no statistical difference between the proportion of SW and NW and

NL and SL in the extreme momentum strategy. Further, the difference in average monthly

formation period returns between SW and NW is -31 basis points, which indicates that, on

average, NW have higher formation period returns than SW. However, the difference in

average monthly formation period returns between NL and SL is 45 basis points indicating

that SL has lower returns in the formation period. Taken together, it is clear that the SW and

10
SL portfolios do not systematically include stocks with extreme performance in the formation

period4.

Table 2 presents the raw monthly returns from the traditional momentum, salient momentum

and non-salient momentum strategies. The traditional momentum strategy uses the Jegadeesh

and Titman [1993] methodology with a 1-month lag between formation and holding periods.

The salient momentum strategy is a zero investment strategy that is long in SW and short in

SL, the non-salient momentum strategy is a zero investment strategy that is long in NW and

short in NL. Returns from these three strategies are reported, in Panel A the 3 to 12-month,

Panel B the 6 to 12-month and Panel C the 9 to 12-month. The results show a strong

momentum effect for 3, 6 and 9-month holding periods, with the 12-month holding period

positive but insignificant from zero. Consistent with trend salience, all salient momentum

strategies yield positive monthly returns, significant at the 1% level; whilst none of the non-

salient momentum strategies earn significant returns, except for the 6 to 12-month and 9 to

12-month held for 3 months which are significant at the 5% level. Further, the Sharpe ratios

for the salient momentum strategies are all positive whilst the non-salient strategies are

negative for holding periods of longer than 3 months.

[TABLE 2 ABOUT HERE]

The statistically significant predictive power of stocks with increasing rates of return in the

formation period suggests investors are conditioned to the salience of a trend. Whilst this is

not a direct test of autocorrelation these results are supported by Gaunt and Gray [2003] that

find positive autocorrelations in Australian equity markets for large stocks. To examine a

risk-based explanation for the salience premia, salient momentum and non-salient momentum

4
To robust test these results, the salient strategy is compared to the 10% momentum strategy and outperforms at

all holding periods, statistically significant at the 1% level, at the 9 and 12-month holding periods.

11
returns of the 6 to 12 strategies for all holding periods were regressed on the Carhart [1997]

factors. If systematic risk explains the salient momentum returns, the alpha values should be

zero. Results from Table 3 shows that at each holding period, the alpha for the salient

momentum strategy is positive and significant at the 1% level. Salient momentum returns are

negatively related to the market and size factors and positively related to the momentum

factor. In contrast, the alphas for non-salient momentum strategies are insignificant for the 3

and 6-month holding periods and become negative and significant at the 5% level at 9 and

12-month holding periods. The result of the Carhart [1997] regression show three important

facts: the first is that the salient momentum strategies generates a positive return even after

controlling for momentum, second, the significance of the salient momentum strategy alpha

increases over longer holding periods and third, the negative loading of the salient

momentum strategies on the market, size and value factors suggest trend salience is not a

proxy for systematic risk. Overall the results of the Carhart [1997] regressions are difficult to

reconcile with a risk-based explanation for the salient momentum premium.

[TABLE 3 ABOUT HERE]

Momentum marginal returns are characterized by short-run continuations of about 12-

months, followed by long-run reversals after 12-months. The coefficients of the Carhart

[1997] regression indicate interesting results regarding the dynamics of trend salience. The

theory of trend salience predicts that the marginal salient momentum strategy returns should

be statistically positive for longer than non-salient momentum strategies. Table 4 shows that

the 6 to 12 salient momentum strategy’s marginal returns are positive for 11 months post-

formation, with the first 9 months significant. In contrast, only the first 2 months post-

formation are significant for the non-salient momentum strategy and returns are negative

beyond 6 months. A possible explanation for this result is that stocks in the non-salient

12
momentum strategy enter the overreaction phase of momentum earlier than the salient

momentum stocks and therefore enter the reversal phase earlier. It is interesting to note that

beyond 12 months post-formation, the marginal returns of the salient momentum and non-

salient momentum strategies are statistically identical. Therefore an alternate interpretation of

this result is that during the formation period the salience of the salient momentum strategy’s

trend is statistically larger than the non-salient momentum trend. As both strategies start to

exhibit mean reversion in the holding period, the salience of both strategies reduces and the

statistical difference between them also reduces.

[TABLE 4 ABOUT HERE]

V. Robustness Testing

Trend salience and firm size

Hong, Lim and Stein [2000] show that information diffusion is slow in small firms;

suggesting the momentum premium is driven by small stocks. It is therefore possible that the

salient momentum strategy systematically select small stocks with high expected returns,

which will make this strategy difficult to implement with higher trading costs. To test

whether this effect is implementable, the sample is split at the median based on market

capitalization and the performance of the 6 to 12 strategy is tested across these two

subsamples. If salience is stronger amongst small stocks or if institutional traders are less

likely to extrapolate salient trends, the effect should be stronger in the bottom 250 sample.

Table 5 reports the results of the size splits of the traditional momentum, salient momentum

and non-salient momentum strategies. Panel A reports the results using the top 250 stocks in

the sample and Panel B reports the results of the bottom 250 stocks in the sample. Whilst the

magnitude of the salient momentum premium is larger in the bottom 250 stocks, this is

13
largely due to a stronger momentum effect within that segment of the market. In the sample

of the 250 largest stocks, salient momentum outperforms traditional momentum at the 1%

confidence level across all holding periods. These results demonstrate that the effect of trend

salience is not confined to small stocks.

[TABLE 5 ABOUT HERE]

Hold-Out Periods

To examine the pervasiveness of the outperformance of salient momentum strategies and

ensure that our results are not driven by a particular period of the business cycle, we also

undertake sub-period analysis to examine the performance of each investment strategy across

five-year and ten-year sub-periods. The results of this sub-period analysis are reported in

Table 6. The salient momentum strategy yields returns that are significantly greater than the

traditional momentum returns for all five year holding periods except 2007 to 2011.

Similarly, salient momentum outperformed traditional momentum in the decade from 1992 to

2001, although the difference between returns on these two strategies was not statistically

significant in the second ten year sub-period.

[TABLE 6 ABOUT HERE]

Market State Analysis

Whilst the sub-period results show salient momentum strategies consistently outperform

traditional momentum, it could be inferred that the performance of the overall market may

have an impact on salient momentum and traditional momentum returns, as no salient

momentum strategy yielded a significant return in the 2007-2011 sub-period. Cooper et al.

[2004] find that whilst momentum is not sensitive to the business cycle, the momentum

premium is sensitive to periods of market up and downswings. Phua et al. [2010] confirm the

14
results of Cooper et al. [2004] in Australian equities. To explore the effect of market states on

trend salience, the Cooper et al. [2004] method is utilized whereby two market states are

defined: up markets and down markets. Up markets are defined as periods where the average

equity risk premium over the past three years is positive and down markets are defined as

periods where the average equity risk premium over the past three years is negative.

The relationship between market states and momentum returns has been linked with

behavioral theories of momentum. Daniel et al. [1998] argues that investors have a self-

biased attribution in which external events that confirm an investors original beliefs lead to an

overconfidence of the skill on the part of the individual investor, whilst external events that

contradict an investors original beliefs are marked down to chance. Overconfidence leads to

individuals to act on their beliefs (Griffin and Tversky [1992]). If the market experiences a

general rise, then investors will mark this down to skill therefore aggregate overconfidence

rises leading to greater momentum profits.

The Hong and Stein [1999] model predicts that a decreasing of risk aversion of the

momentum traders leads to greater momentum profits. Barberis, Huang and Santos [2001]

use Prospect Theory to describe how individuals become less loss averse after periods of

increasing wealth and more loss averse after periods of wealth declines. In this model, after

periods of market gains, the risk aversion of the Hong and Stein [1999] momentum traders

may decrease and momentum profits increase.

The salience of a trend may also be affected by market states. Volatility in markets is higher

in market downswings than market upswings (Schewert [1989], French and Sichel [1993],

Hamilton and Lin [1996]). If this is the case then salience of a trend may also decrease in

market downswings as it may be difficult to identify a trend. Andreassen and Kraus [1990]

propose that extrapolation of trends is less likely when the variance of past price changes is

15
large. If the increased volatility in down markets causes confusion in the price signal,

investors may be unlikely to identify and extrapolate a trend.

Table 7 presents the results of the market state analysis for the 6 to 12 strategy. Consistent

with prior evidence, traditional momentum returns are shown to be insignificant in down

markets across all holding periods (Cooper et al. [2004], Phua et al. [2010]). Of particular

significance is that salient momentum only outperforms traditional momentum in up markets

across all holding periods. This suggests that trend salience is stronger after periods of market

gains, when volatility is typically lower.

[TABLE 7 ABOUT HERE]

Seasonality

Jegadeesh and Titman [1993] show that in the United States, January momentum returns are

strongly negative whilst February to December, momentum returns are positive. Jegadeesh

and Titman [1993] also find that months November and December are particularly strong

which is consistent with tax-loss selling. The theory of tax-loss selling is a follows; in the

months prior to the end of the financial year investors sell stocks that have performed

particularly poorly in the months preceding the end-of-financial year. The increased selling

pressure pushes down loser stocks making momentum profits larger in that period. In the first

month of the new financial year, these losers stocks are bought pushing up prices and causing

momentum profits to become negative. Reinganum [1983]) finds empirical evidence that past

loser stocks exhibit a larger January effect than past winner stocks; this effect is largest

amongst small firms.

Table 8 reports the performance of our traditional momentum and 6 to 12 month salient

momentum investment strategies in each calendar month. Panel A reports results for all

16
stocks in the sample, Panel B and C report results for the top 250 and bottom 250

respectively. If tax-loss selling is factor with trend salience5, the traditional momentum and

salient momentum premium should be the largest in the months leading to the end-of-

financial year, May and June and negative in July. Comparable with evidence from the

United States, May and June are the strongest months for the traditional momentum and

salient momentum strategies with July the weakest month. The significant negative returns

occur in the bottom 250 stocks, which is consistent with the results in Jegadeesh and Titman

[1993] and shows that tax-loss selling is stronger in smaller firms. Whilst the returns for each

portfolio are negative in January, the returns are not statistically significant. Window dressing

is also a feature of momentum premiums in the United States, as institutional investors sell

their losers in the month preceding quarterly reporting to avoid disclosing they held poorly

performing stocks in that quarter (Sias [2007]). Table 8 reports weak evidence of window

dressing, with positive end-of-quarter returns followed by negative start-of-quarter returns.

[TABLE 8 ABOUT HERE]

An alternate method of testing seasonal effects of salience is to examine the proportion of

positive returns in a month. Table 9 shows the proportion of positive monthly returns for the

6 to 12 strategy, 6 and 12-month holding periods. The p-values calculated represent the

probability of achieving the hit rate or higher purely due to randomness. Overall, salient

momentum returns are positive 63.75% (p-value 0.00) and 67.92% (p-value 0.00) at 12 and

6-month holding periods respectively. This compares to non-salient momentum with positive

monthly returns 51.25% (p-value 0.35) and 56.25% (p-value 0.03) at 12 and 6-month holding

periods respectively. The hit rate results are consistent with the results in Table 8, which

5
The end-of-financial year in Australia is 30 June.

17
show that end-of-quarter positives are stronger than the start-of-quarter months, with June the

strongest result.

[TABLE 9 ABOUT HERE]

VI. Summary

Andreassen and Kraus [1990] have identified the conditions in which past prices are likely or

unlikely to be extrapolated. They show that if subjects are exposed to a price history that has

a clearly identifiable trend, then subjects are likely to extrapolate that trend. If the trend is

unclear, investors are unlikely to extrapolate the trend. The extent that a trend will be

extrapolated is determined by how salient that trend appears to the investor. A stock with a

strengthening trend in the formation period is likely to have a more salient trend than a stock

that has a deteriorating trend in the formation period. Momentum strategies that take

advantage of trend salience are likely to outperform strategies that do not account for

salience.

We find zero investment strategies that are long (short) in winner (loser) stocks with a

strengthening trend in the formation period earn returns that are significantly higher than the

traditional momentum strategies. This result is not explained by the Carhart [1997] four

factor model, nor does it seem to be explained by the size of stocks in the portfolio. The

salient momentum strategies are not systematically exposed to extreme performance in the

formation period. Whilst the salient momentum returns is strong throughout the sample

period, the results from the global financial crisis suggest that trend salience is sensitive to

changes in the market states. This sensitivity may be due to trend salience reducing during

periods of higher volatility, which makes trends harder to identify.

18
Marginal returns for 60 months after the formation period was also examined with salient

momentum returns significantly positive for 9 months post-formation, compared to only 2

months for the non-salient momentum strategy. The implication of this result is that non-

salient stocks may have entered early into the overreaction phase, leading to early reversals or

it could demonstrate that trend salience is driving momentum performance, as salience

decreases, the level of extrapolation decreases to a point whereby reversals are likely to

occur.

The evidence of trend salience reported in our paper supports the behavioral models of

investor extrapolation. Further, trend salience also has implications for the disposition effect

(Shefrin and Statman [1985]). Dacey and Zielonka [2008] show that investors are likely to

avoid the disposition effect when the future growth prospects of winners (losers) are high

(low). Trend salience may be a factor that investors use to determine the future growth of

stocks, thereby an explanation for the outperformance of the salience strategy. Strengthening

trends in the formation periods has been identified as one determinant of salience; further

research could explore other factors that could identify salience in the minds of investors.

19
 References

Andreassen, P. and S. Kraus. “ Judgmental extrapolation and the salience of change.” Journal of
Forecasting, 9, (1990), pp. 347-372.
Barberis, N., M. Huang and T. Santos. “Prospect theory and asset prices.” Quarterly Journal of
Economics, 116, (2001), pp.1-53.
Barberis, N., A. Shleifer and R. Vishny. “A model of investor sentiment.” Journal of Financial
Economics, 49, (1998), pp. 307-343.
Beedles, W. L. “Small firm equity cost: evidence from Australia.” Journal of Small Business
Management, 30, (1992), pp.57-65.
Beedles, W. L., P. Dodd and R. R. Officer. “Regularities in Australian share returns.”, Australian
Journal of Management, 13, (1988), pp. 3-29.
Black, S. and J. Kirkwood. “Ownership of Australian equities and corporate bonds.” RBA Bulletin
(September), (2010), pp. 25-33.
Brailsford, T. and M. A. O'Brien. “Disentangling size from momentum in Australian stock returns.”
Australian Journal of Management, 32, (2008), pp. 463-484.
Caginalp, G., D. Porter and V. Smith. “Overreactions, momentum, liquidity, and price bubbles in
laboratory and field asset markets.” Journal of Psychology and Financial Markets, 1, (2000), pp. 24-
48.
Carhart, M. “On persistence in mutual fund performance.” Journal of Finance, 52, (1997), pp. 57-82.
Choi, J., D. Laibson and B. Madrian. “Why does the law of one price fail? An experiment on index
mutual funds.” Review of Financial Studies, 23, (2010), pp. 1405-1432.
Chordia, T., and L. Shivakumar. “Momentum, business cycle and time-varying expected returns.”
Journal of Finance, 57, (2002), pp. 985-1019.
Cooper, M. J., R. C. Gutuerrez and A. Hameed. “Market states and momentum.” Journal of Finance,
59, (2004), pp. 1345-1365.
Daniel, K., D. Hirshleifer and A. Subrahmanyam. “Investor psychology and security market under-
and overreactions.” Journal of Finance, 53, (1998), pp. 1839-1885.
Dacey, R., and P. Zielonka. “A Detailed Prospect Theory Explanation of the Disposition Effect.”
Journal of Behavioral Finance, 9, (2008), pp. 43–50.
De Bondt, W. “Betting on trends: intuitive forecasts of financial risk and return.” International
Journal of Forecasting, 9, (1993), pp. 355-371.
DeLong, B. J., A. Shleifer, L. H. Summers and R. J. Waldmann. “Positive feedback investment
strategies and destabilizing rational speculation.” Journal of Finance, 45, (1990), pp. 379-395.
Demir, I., J. Muthuswamy and T. Walter. “Momentum returns in Australian equities: the influences of
size, risk, liquidity and return computation.” Pacific-Basin Finance Journal, 12, (2004), pp. 143-158.
Edwards, W. “Conservatism in human information processing”, In Formal Representation of Human
Judgement, B. Kleinmutz (eds.), (1968), New York: John Wiley and Sons.
Fama, E. and K. French. “Multifactor explanations of asset pricing anomalies.” Journal of Finance,
51, (1996), pp. 55-84.
Fama, E. and K. French. “Dissecting anomalies.” Journal of Finance, 63, (2008), pp. 1653-1678.
Frankel, J. and K. Froot. “Explaining the demand for dollars: international rates of return and the
expectations of chartists and fundamentalists'.” In Macroecenomics, Agriculture, and the Exchange
Rate, R. Chambers, and P. Paarlberg (eds.), (1988), Boulder, CO: Westfield Press.

20
French, M. and D. Sichel. “Cyclical patterns in the variance of economic activity”, Journal of
Business & Economic Statistics, 11, (1993), pp.113-119.
Fuster, A., D. Laibson and B. Mendel. “Natural expectations and macroeconomic fluctuations.”
Journal of Economic Perspectives, 24, (2010), pp. 67-84.
Gaunt, C. and P. Gray. “Short-term autocorrelation in australian equities.” Australian Journal of
Management, 28, (2003), pp. 97-117.
Gaunt, C., P. Gray and J. McIvor. “The impact of share price on seasonality and size anomalies in
Australian equity returns.” Accounting and Finance, 40, (2000), pp. 33-50.
Greenwood, R. and S. Nagel. “Inexperienced investors and bubbles.” Journal of Financial
Economics, 93, (2008), pp. 239-258.
Griffin, D. and A. Tversky. “The weighting of evidence and the determinants of confidence.”
Cognitive Psychology, 24, (1992), pp. 411-435.
Griffin, J., X. Ji, and J. S. Martin. “Momentum investing and business cycle risk: evidence from pole
to pole.” Journal of Finance, 58, (2003), pp. 2515-2547.
Grinblatt, M. and T. Moskowitz. “Predicting stock price movements from past returns: the role of
consistency and tax-loss selling.” Journal of Financial Economics, 71, (2004), pp. 541-79.
Grundy, B. D. and J. S. Martin. “Understanding the nature of the risks and the source of the rewards
to momentum investing.” Review of Financial Studies, 14, (2001), pp. 29-78.
Hamilton, J. and G. Lin. “Stock market volatility and the business cycle.” Journal of Applied
Econometrics, 11, (1996), pp. 573-593.
Haruvy, E., Y. Lahav and C. Noussair. “Traders' expectations in asset markets: experimental
evidence.” American Economic Review, 97, (2007), pp. 1901-1920.
He, W. and J. Shen. “Investor extrapolation and expected returns.” Journal of Behavioral Finance, 11,
(2010), pp. 150-160.
Hong, H., T. Lim and J. Stein. “ Bad news travels slowly: size, analyst coverage, and the profitability
of momentum strategies.” Journal of Finance, 55, (2000), pp. 265-295.
Hong, H. and J. C. Stein. “A unified theory of underreaction, momentum trading, and overreaction in
asset markets.” Journal of Finance, 54, (1999), pp. 2143-2184.
Jegadeesh, N. “Evidence of predictable behaviour of security returns.” Journal of Finance, 45, (1990),
pp. 881-898.
Jegadeesh, N. and S. Titman. “Returns to buying winners and selling losers: implications for stock
market efficiency.” Journal of Finance, 48, (1993), pp. 65-91.
Jegadeesh, N. and S. Titman. “Profitability of momentum strategies: an evolution of alternative
explanations.” Journal of Finance, 56, (2001), pp. 699-718.
Lehmann, B. “Fads, martingales, and market efficiency.” Quarterly Journal of Economics, 105,
(1990), pp. 1-28.
Moskowitz, T. and M. Grinblatt. “Do industries explain momentum?” Journal of Finance, 54, (1999),
pp. 1249-90.
Phua, V., H. Chan, R. Faff and R. Hudson. “The influence of time, seasonality and market state on
momentum; insights from the Australian stock market.” Applied Financial Economics, 20, (2010), pp.
1547-1563.
Reinganum, M. “The anomalous stock market behavior of small firms in january.” Journal of
Financial Economics, 12, (1983), pp. 89-104.
Schwert, W. “Why does stock market volatility change over time?” Journal of Finance, 44, (1989),
pp. 1115-1153.

21
Shefrin, H. and M. Statman. “The disposition to sell winners too early and ride losers too long: theory
and evidence.” Journal of Finance, 40, (1985), pp. 777-90.
Sias, R. “Causes and seasonality of momentum profits.” Financial Analysts Journal, 63, (2007), pp.
48-54.
Smith, V., G. Suchanek and A. Williams. “Bubbles, crashes and endogenous expectations in
experimental spot asset markets.” Econometrica, 56, (1988), pp. 1119-1151.
Tversky, A. and D. Kahneman. “Judgement under uncertainty: heuristics and biases.” In Judgement
Under Uncertainty: Heuristics and Biases, D. Kahneman, P. Slovic and A. Tversky (eds.), (1974),
Cambridge: Cambridge University Press.
Yu, H. and L. Chen. “Momentum-reversal strategy.” SSRN Paper Series. Available at SSRN:
http://ssrn.com/abstract=1663266 (2011).

22
 Table 1 – Returns of portfolios formed on momentum and trend salience

This table reports the mean monthly returns for portfolios formed on momentum and trend salience
across the period 1992-2011. The winner (W) and loser (L) portfolios are the top and bottom
performing quintiles based on the past 12-month returns. The salient winner (SW) portfolios
include stocks where the ratio of the past M month GARR and the past 12-month GARR is greater
than the median. The non-salient winner (NW) portfolios include stocks where the ratio of the past
M month GARR and the past 12-month GARR is equal to or less than the median. The non-salient
loser (NL) portfolios include stocks where the ratio of the past M month GARR and the past 12-
month GARR is equal to or greater than the median. The salient loser (SL) portfolios include
stocks where the ratio of the past M month GARR and the past 12-month GARR is less than the
median. All portfolios are held for K months, defined as 3, 6, 9 and 12. HAC t-statistics are
reported in parentheses under their associated means.

Portfolios
W SW NW L NL SL
Panel A: 3 to 12 Strategy
3 0.0130 0.0176 0.0083 0.0010 0.0032 -0.0011
(3.10)** (4.22)** (1.88) (0.18) (0.58) (-0.18)
6 0.0109 0.0146 0.0072 0.0020 0.0038 0.0003
(2.63)** (3.56)** (1.64) (0.36) (0.72) (0.04)
9 0.0097 0.0129 0.0065 0.0029 0.0046 0.0012
(2.30)* (3.17)** (1.44) (0.54) (0.93) (0.20)
12 0.0088 0.0114 0.0062 0.0042 0.0060 0.0024
(2.10)* (2.84)** (1.39) (0.82) (1.26) (0.44)
Panel B: 6 to 12 Strategy
3 0.0130 0.0172 0.0088 0.0010 0.0020 0.0000
(3.10)** (4.02)** (2.04)* (0.18) (0.37) (0.01)
6 0.0109 0.0147 0.0071 0.0020 0.0033 0.0007
(2.63)** (3.54)** (1.65) (0.36) (0.63) (0.12)
9 0.0097 0.0125 0.0069 0.0029 0.0050 0.0007
(2.30)* (3.04)** (1.56) (0.54) (1.02) (0.12)
12 0.0088 0.0110 0.0067 0.0042 0.0062 0.0022
(2.10)* (2.71)** (1.50) (0.82) (1.30) (0.39)
Panel C: 9 to 12 Strategy
3 0.0130 0.0171 0.0089 0.0010 0.0026 -0.0006
(3.10)** (4.03)** (2.07)* (0.18) (0.49) (-0.09)
6 0.0109 0.0138 0.0080 0.0020 0.0042 -0.0002
(2.63)** (3.34)** (1.87) (0.36) (0.80) (-0.02)
9 0.0097 0.0118 0.0076 0.0029 0.0052 0.0005
(2.30)* (2.88)** (1.72) (0.54) (1.05) (0.09)
12 0.0088 0.0105 0.0071 0.0042 0.0061 0.0024
(2.10)* (2.62)** (1.60) (0.82) (1.26) (0.43)
* denotes significant at the 5% level
** denotes significant at the 1% level.

23
 Table 2 – Investment strategy returns

This table reports the monthly returns of three investment strategies formed on the basis of trend salience across the period 1992-2011. The first strategy (MOM) is a momentum strategy
that takes a long position in the quintile of “winners” and a short position in the quintile of “losers”. The second strategy (SAL) involves buying the SW portfolio and selling the SL
portfolio. The third strategy (NON) involves buying the NW portfolio and selling the NL portfolio. SAL-MOM is the difference between the salient momentum and traditional
momentum strategies. The HAC t-statistics and Sharpe ratios for each strategy are reported underneath the mean returns.

3 to 12 6 to 12 9 to 12
MOM SAL NON SAL-MOM MOM SAL NON SAL-MOM MOM SAL NON SAL-MOM

3 0.0120 0.0188 0.0052 0.0068 0.0120 0.0172 0.0068 0.0052 0.0120 0.0176 0.0063 0.0057
(3.40)** (4.14)** (1.54) (3.60)** (3.40)** (3.61)** (2.08)* (2.55)* (3.40)** (3.82)** (2.09)* (3.38)**
Sharpe 0.17 0.24 0.01 0.17 0.22 0.05 0.17 0.23 0.04
6 0.0089 0.0143 0.0034 0.0055 0.0089 0.0139 0.0038 0.0051 0.0089 0.0139 0.0038 0.0051
(2.82)** (3.56)** (1.14) (3.32)** (2.82)** (3.33)** (1.36) (3.02)** (2.82)** (3.55)** (1.40) (4.08)**
Sharpe 0.11 0.19 -0.03 0.11 0.19 -0.02 0.11 0.20 -0.02
9 0.0068 0.0117 0.0019 0.0049 0.0068 0.0118 0.0018 0.0050 0.0068 0.0113 0.0023 0.0045
(2.48)* (3.34)** (0.74) (3.52)** (2.48)* (3.35)** (0.74) (3.67)** (2.48)* (3.43)** (0.93) (4.43)**
Sharpe 0.06 0.16 -0.07 0.06 0.16 -0.08 0.06 0.16 -0.06
12 0.0046 0.0090 0.0002 0.0044 0.0046 0.0088 0.0004 0.0042 0.0046 0.0082 0.0010 0.0036
(1.90) (3.05)** (0.09) (3.82)** (1.90) (3.01)** (0.17) (3.81)** (1.90) (2.93)** (0.44) (4.18)**
Sharpe 0.00 0.11 -0.13 0.00 0.11 -0.13 0.00 0.09 -0.11
* denotes significant at the 5% level
** denotes significant at the 1% level.

24
Table 3 – Four-factor model regression analysis

This table reports the coefficients of the Carhart [1997] four factor model over the period from January 1992 to
December 2011. The model that is estimated is given as follows:

[ ]

where is the return on investment strategy p at time t, and , , , are the excess
return on the market and the size, value and momentum factors respectively. Results are reported for two
different investment strategies as the dependent variable, which are created using a double sort on momentum
and trend salience. Salient momentum strategies are formed by taking a long position in “winners” from the past
twelve months that have a ratio of past 6-month to past 12-month geometric average rate of return (GARR) that
is above the median (salient winners) and taking a short position in “losers” from the past twelve months that
have a ratio of past 6-month to past 12-month GARR that is below the median (salient losers). Non-salient
momentum strategies are formed by taking a long position in “winners” from the past twelve months that have a
ratio of past 6-month to past 12-month GARR that is below the median (non-salient winners) and taking a short
position in “losers” from the past twelve months that have a ratio of past 6-month to past 12-month GARR that
is above the median (non-salient losers). HAC adjusted t-statistics are reported in parenthesis below their
associated coefficients.

α β1 β2 β3 β4 Adj. R2
Panel A: 3-Month Holding
Non-salient 0.0008 -0.0146 -0.0527 -0.0138 0.4984 0.50
(0.37) (-0.29) (-1.31) (-0.21) (14.27)**
Salient 0.0064 -0.0677 -0.0478 0.0472 0.8133 0.75
(3.19)** (-1.44) (-1.27) (0.76) (24.95)**

Panel B: 6-Month Holding

Non-salient -0.0015 0.0112 -0.0124 0.0730 0.5400 0.62
(-0.93) (0.30) (-0.42) (1.50) (18.65)**
Salient 0.0070 -0.0748 -0.0815 -0.0373 0.7959 0.78
(4.31)** (-1.95) (-2.62)** (-0.74) (26.70)**

Panel C: 9-Month Holding

Non-salient -0.0027 0.0349 0.0151 0.0720 0.5888 0.69
(-2.13)* (1.13) (0.60) (1.76) (22.04)**
Salient 0.0077 -0.0825 -0.0960 -0.0674 0.7824 0.81
(5.84)** (-2.60)** (-3.71)** (-1.60) (28.42)**

Panel D: 12-Month Holding

Non-salient -0.0023 0.0388 0.0207 0.0507 0.6270 0.75
(-2.13)* (1.48) (0.97) (1.46) (25.27)**
Salient 0.0069 -0.0537 -0.0968 -0.0540 0.7753 0.84
(6.42)** (-2.04)* (-4.48)** (-1.54) (31.02)**

* denotes significant at the 5% level

** denotes significant at the 1% level.

25
Table 4 – Marginal returns to trend salience strategies after portfolio formation

This table presents the marginal and cumulative returns for the salient momentum strategy (Panel A) and the non-
salient momentum strategy (Panel B) in each month following the formation period. Salient momentum strategies
are formed by taking a long position in the salient winners portfolio and a short position in the salient losers
portfolio from the previous year. Non-salient momentum strategies are formed by taking a long position in the
non-salient winners portfolio and a short position in the non-salient losers portfolio from the previous year. HAC
t-statistics are in parenthesis underneath their associated mean values.

Panel A: Salient momentum marginal returns

Marginal Cumulative Marginal Cumulative Marginal Cumulative

1 0.0246 0.0246 21 -0.0022 0.1016 41 0.0010 0.0992
(3.99)** (-0.60) (0.39)
2 0.0194 0.0445 22 -0.0005 0.1010 42 0.0015 0.1009
(3.24)** (-0.14) (0.53)
3 0.0133 0.0584 23 -0.0005 0.1005 43 0.0022 0.1033
(2.32)* (-0.13) (0.83)
4 0.0129 0.0720 24 -0.0020 0.0983 44 -0.0026 0.1004
(2.30)* (-0.65) (-0.98)
5 0.0131 0.0861 25 -0.0028 0.0952 45 -0.0023 0.0979
(2.61)** (-0.80) (-0.87)
6 0.0132 0.1004 26 -0.0033 0.0916 46 -0.0052 0.0922
(3.02)** (-0.84) (-1.74)
7 0.0149 0.1168 27 -0.0023 0.0891 47 -0.0072 0.0843
(3.54)** (-0.63) (-2.66)**
8 0.0113 0.1294 28 -0.0008 0.0882 48 -0.0079 0.0758
(3.07)** (-0.22) (-3.01)**
9 0.0087 0.1392 29 -0.0010 0.0871 49 -0.0077 0.0675
(2.53)* (-0.27) (-2.99)**
10 0.0056 0.1456 30 0.0010 0.0882 50 -0.0101 0.0567
(1.82) (0.27) (-3.57)**
11 0.0014 0.1472 31 0.0043 0.0929 51 -0.0072 0.0491
(0.50) (1.30) (-2.50)*
12 -0.0035 0.1432 32 0.0046 0.0979 52 -0.0062 0.0426
(-1.14) (1.63) (-2.20)*
13 -0.0026 0.1402 33 0.0015 0.0996 53 -0.0056 0.0368
(-0.86) (0.44) (-2.24)*
14 -0.0048 0.1348 34 0.0011 0.1008 54 -0.0058 0.0307
(-1.48) (0.39) (-1.98)*
15 -0.0044 0.1298 35 0.0019 0.1029 55 -0.0024 0.0283
(-1.37) (0.78) (-0.89)
16 -0.0062 0.1228 36 -0.0007 0.1021 56 -0.0015 0.0267
(-1.97)* (-0.22) (-0.52)
17 -0.0051 0.1170 37 0.0001 0.1022 57 -0.0004 0.0263
(-1.59) (0.03) (-0.14)
18 -0.0036 0.1130 38 -0.0013 0.1008 58 0.0025 0.0289
(-1.22) (-0.47) (1.07)
19 -0.0036 0.1090 39 -0.0017 0.0989 59 0.0030 0.0320
(-1.09) (-0.75) (1.26)
20 -0.0045 0.1040 40 -0.0007 0.0981 60 0.0013 0.0333
(-1.27) (-0.27) (0.44)
* denotes significant at the 5% level
** denotes significant at the 1% level.

26
Panel B: Non-salient momentum marginal returns

Marginal Cumulative Marginal Cumulative Marginal Cumulative

1 0.0136 0.0136 21 0.0000 -0.0144 41 -0.0034 -0.0026
(3.18)** (-0.01) (-1.13)
2 0.0084 0.0221 22 -0.0018 -0.0162 42 -0.0043 -0.0068
(2.08)* (-0.52) (-1.34)
3 0.0051 0.0273 23 -0.0025 -0.0186 43 -0.0045 -0.0113
(1.38) (-0.73) (-1.32)
4 0.0031 0.0305 24 0.0002 -0.0185 44 -0.0071 -0.0183
(0.84) (0.05) (-2.22)*
5 0.0020 0.0326 25 0.0029 -0.0156 45 -0.0077 -0.0258
(0.64) (0.95) (-2.92)**
6 0.0015 0.0341 26 0.0027 -0.0129 46 -0.0070 -0.0326
(0.43) (0.93) (-3.04)**
7 -0.0007 0.0334 27 0.0019 -0.0110 47 -0.0048 -0.0373
(-0.21) (0.64) (-2.04)*
8 -0.0017 0.0317 28 0.0016 -0.0094 48 -0.0047 -0.0418
(-0.51) (0.55) (-1.71)
9 -0.0025 0.0291 29 0.0018 -0.0077 49 -0.0050 -0.0466
(-0.73) (0.69) (-1.67)
10 -0.0013 0.0277 30 0.0011 -0.0066 50 -0.0050 -0.0513
(-0.39) (0.47) (-1.94)
11 -0.0023 0.0254 31 0.0010 -0.0057 51 -0.0060 -0.0570
(-0.67) (0.39) (-2.35)*
12 -0.0060 0.0192 32 0.0005 -0.0052 52 -0.0011 -0.0580
(-1.93) (0.17) (-0.45)
13 -0.0043 0.0149 33 0.0015 -0.0036 53 -0.0010 -0.0590
(-1.51) (0.60) (-0.45)
14 -0.0073 0.0074 34 0.0013 -0.0023 54 -0.0004 -0.0594
(-2.34)* (0.55) (-0.17)
15 -0.0082 -0.0008 35 -0.0004 -0.0027 55 -0.0007 -0.0600
(-2.25)* (-0.14) (-0.28)
16 -0.0053 -0.0061 36 0.0008 -0.0020 56 -0.0038 -0.0636
(-1.57) (0.33) (-1.66)
17 -0.0055 -0.0116 37 0.0015 -0.0005 57 -0.0019 -0.0654
(-1.51) (0.58) (-0.89)
18 -0.0017 -0.0133 38 -0.0006 -0.0010 58 -0.0022 -0.0674
(-0.47) (-0.19) (-0.92)
19 0.0004 -0.0128 39 0.0025 0.0015 59 -0.0029 -0.0701
(0.12) (0.79) (-1.21)
20 -0.0015 -0.0143 40 -0.0007 0.0008 60 -0.0010 -0.0710
(-0.37) (-0.20) (-0.45)
* denotes significant at the 5% level
** denotes significant at the 1% level.

27
Table 5 – Salient momentum returns across sub-samples formed on market capitalization

This table shows raw monthly returns to the traditional momentum, salient momentum and non-salient momentum
strategies for the period 1992-2011 with HAC t-statistics in parenthesis. Sharpe ratios (Sharpe) use the 90-day bank
accepted bill rate as the risk-free rate. Salient momentum strategies are formed by taking a long position in
“winners” from the past twelve months that have a ratio of past 6-month to past 12-month geometric average rate
of return (GARR) that is above the median (salient winners) and taking a short position in “losers” from the past
twelve months that have a ratio of past 6-month to past 12-month GARR that is below the median (salient losers).
Non-salient momentum strategies are formed by taking a long position in “winners” from the past twelve months
that have a ratio of past 6-month to past 12-month GARR that is below the median (non-salient winners) and taking
a short position in “losers” from the past twelve months that have a ratio of past 6-month to past 12-month GARR
that is above the median (non-salient losers). The momentum strategies are formed using the Jegadeesh and Titman
[1993] equal-weighted portfolios using a 12 month/K month strategy where K months are the holding periods,
defined as 3, 6, 9 and 12. Panel A reports the raw monthly returns for a sub-sample consisting of the 250 largest
stocks in the sample and Panel B reports the raw monthly returns for a sub-sample of the 250 smallest stocks in the
sample.

Holding Period MOM SAL NON SAL-MOM

Panel A: Top 250
3 0.0115 0.0171 0.0059 0.0056
(3.35)** (3.69)** (1.95) (2.97)**
Sharpe 0.15 0.22 0.03
6 0.0084 0.0143 0.0024 0.0060
(2.67)** (3.29)** (0.93) (3.45)**
Sharpe 0.09 0.19 -0.05
9 0.0058 0.0114 0.0002 0.0056
(2.11)* (3.17)** (0.07) (4.22)**
Sharpe 0.03 0.15 -0.12
12 0.0036 0.0084 -0.0012 0.0048
(1.46) (2.74)** (-0.55) (4.15)**
Sharpe -0.03 0.09 -0.17

Panel B: Bottom 250

3 0.0127 0.0188 0.0065 0.0062
(3.22)** (3.55)** (1.56) (2.29)*
Sharpe 0.16 0.21 0.04
6 0.0095 0.0147 0.0042 0.0052
(2.68)** (3.17)** (1.23) (2.54)*
Sharpe 0.11 0.17 -0.01
9 0.0074 0.0121 0.0028 0.0047
(2.38)* (2.99)** (0.92) (2.71)**
Sharpe 0.07 0.14 -0.04
12 0.0055 0.0092 0.0019 0.0036
(2.00)* (2.67)** (0.68) (2.56)*
Sharpe 0.02 0.10 -0.07

* denotes significant at the 5% level

** denotes significant at the 1% level.

28
Table 6 – Sub-sample analysis of trend salience strategies

This table shows raw monthly returns to the traditional momentum, salient momentum and non-salient
momentum strategies within four, 5-year and two, 10-year sub-periods. The sample period is January 1992 to
December 2011. Salient momentum strategies are formed by taking a long position in the salient winners
portfolio and a short position in the salient losers portfolio over the past 6 to 12-month ratio of geometric
average rate of return (GARR). Non-salient momentum strategies are formed by taking a long position in the
non-salient winners portfolio and a short position in the non-salient losers portfolio over the past 6 to 12-month
ratio of GARR. The traditional momentum strategies are formed using the Jegadeesh and Titman [1993] equal-
weighted portfolios using a 12-month/K month strategy where K months are the holding periods, defined as 3,
6, 9 and 12. HAC t-statistics are reported in parenthesis beside their associated mean values.

MOM SAL NON SAL-MOM

1992-1996
3 0.0050 (1.92) 0.0118 (3.43)** -0.0017 (-0.47) 0.0068 (2.81)**
6 0.0020 (0.77) 0.0070 (2.56)* -0.0029 (-0.86) 0.0050 (2.99)**
9 -0.0002 (-0.05) 0.0024 (0.82) -0.0027 (-0.80) 0.0026 (1.72)
12 -0.0012 (-0.42) 0.0005 (0.17) -0.0030 (-0.85) 0.0011 (1.39)

1997-2001
3 0.0192 (3.00)** 0.0296 (3.34)** 0.0087 (1.32) 0.0105 (2.33)**
6 0.0133 (2.52)* 0.0242 (3.41)** 0.0024 (0.46) 0.0109 (3.27)**
9 0.0090 (1.94) 0.0197 (3.18)** -0.0015 (-0.32) 0.0106 (3.85)**
12 0.0057 (1.38) 0.0146 (2.74)** -0.0032 (-0.77) 0.0089 (3.67)**

2002-2006
3 0.0184 (3.33)** 0.0214 (3.53)** 0.0153 (2.73)** 0.0030 (1.60)
6 0.0160 (3.35)** 0.0210 (3.81)** 0.0110 (2..42)* 0.0050 (3.06)**
9 0.0136 (3.27)** 0.0195 (3.72)** 0.0078 (2.24)* 0.0059 (3.85)**
12 0.0104 (2.73)** 0.0151 (3.12)** 0.0057 (1.79) 0.0047 (3.12)**

2007-2011
3 0.0053 (0.54) 0.0059 (0.43) 0.0047 (0.60) 0.0006 (0.11)
6 0.0042 (0.46) 0.0036 (0.29) 0.0048 (0.64) -0.0006 (-0.14)
9 0.0046 (0.60) 0.0054 (0.56) 0.0039 (0.54) 0.0008 (0.22)
12 0.0035 (0.52) 0.0050 (0.64) 0.0021 (0.32) 0.0014 (0.56)

1992-2001
3 0.0121 (3.23)** 0.0207 (4.03)** 0.0035 (0.88) 0.0086 (3.25)**
6 0.0077 (2.42)* 0.0156 (3.69)** -0.0002 (-0.08) 0.0079 (3.95)**
9 0.0045 (1.54) 0.0111 (2.85)** -0.0021 (-0.73) 0.0066 (3.64)**
12 0.0022 (0.84) 0.0075 (2.25)* -0.0031 (-1.11) 0.0053 (3.38)**

2002-2011
3 0.0118 (1.96) 0.0137 (1.70) 0.0100 (1.97) 0.0018 (0.61)
6 0.0101 (1.84) 0.0123 (1.68) 0.0079 (1.77) 0.0022 (0.84)
9 0.0091 (1.97) 0.0125 (2.09)* 0.0058 (1.47) 0.0033 (1.66)
12 0.0070 (1.73) 0.0100 (2.06)* 0.0039 (1.06) 0.0031 (1.96)
* denotes significant at the 5% level
** denotes significant at the 1% level.

29
Table 7 – Trend salience strategies in up and down markets

This table shows raw monthly returns to the traditional momentum, salient momentum and non-salient
momentum strategies within up and down markets. Up markets are defined as periods where the average equity
risk premium over the past three years is positive and down markets are defined as periods where the average
equity risk premium over the past three years is negative. Salient momentum strategies are formed by taking a
long position in the salient winners portfolio and a short position in the salient losers portfolio from the previous
year. Non-salient momentum strategies are formed by taking a long position in the non-salient winners portfolio
and a short position in the non-salient losers portfolio from the previous year. The traditional momentum
strategies are formed using the Jegadeesh and Titman [1993] equal-weighted portfolios using a 12 month/K
month strategy where K months are the holding periods. Mean monthly returns are reported for investment
strategies with a 3, 6, 9 and 12-month holding period. The HAC t-statistics (T-stat) and number of observations
(Obs) are reported underneath their associated mean values.

MOM SAL NON SAL-MOM

3-month
Down Return -0.0016 -0.0011 -0.0021 0.0005
T-stat (-0.25) (-0.12) (-0.34) (0.13)
Obs 62 62 62 62
Up Return 0.0167 0.0235 0.0099 0.0068
T-stat (5.28)** (6.02)** (3.15)** (4.28)**
Obs 178 178 178 178

6-month
Down Return -0.0019 -0.0036 -0.0003 -0.0017
T-stat (-0.32) (-0.47) (-0.04) (-0.52)
Obs 62 62 62 62
Up Return 0.0126 0.0200 0.0053 0.0074
T-stat (4.53)** (6.00)** (1.93) (5.97)**
Obs 178 178 178 178

9-month
Down Return -0.0004 -0.0014 0.0006 -0.0010
T-stat (-0.07) (-0.22) (0.11) (-0.43)
Obs 62 62 62 62
Up Return 0.0093 0.0164 0.0023 0.0070
T-stat (3.63)** (5.33)** (0.92) (6.49)**
Obs 178 178 178 178

12-month
Down Return -0.0005 -0.0007 -0.0003 -0.0002
T-stat (-0.10) (-0.13) (-0.05) (-0.14)
Obs 62 62 62 62
Up Return 0.0064 0.0121 0.0006 0.0057
T-stat (2.69)** (4.27)** (0.29) (5.87)**
Obs 178 178 178 178
* denotes significant at the 5% level
** denotes significant at the 1% level.

30
Table 8 – Monthly seasonality effects in the trend salience momentum strategy

This table presents the monthly returns to the traditional momentum, salient momentum and non-salient momentum strategies. Salient momentum and non-salient momentum
strategies are formed on double sorts based on past returns and the ratio of 6 to 12-month geometric mean returns. All portfolios are rebalanced annually. Traditional
momentum strategies are formed using a 12-month formation period and a 12-month investment period. The average returns for the size based sub-samples are also reported.
HAC t-statistics are reported in parenthesis beneath their associated mean values. The sample period is January 1992 to December 2011.

All Stocks Top 250 Bottom 250

Mom Sal Non Mom Sal Non Mom Sal Non
Jan -0.0082 -0.0063 -0.0101 -0.0037 -0.0010 -0.0064 -0.0086 -0.0072 -0.0101
(-1.64) (-0.98) (-1.63) (-0.73) (-0.17) (-1.11) (-1.31) (-0.93) (-1.14)
Feb 0.0200 0.0261 0.0138 0.0239 0.0310 0.0168 0.0188 0.0252 0.0125
(2.01) (2.50) (1.34) (1.89) (2.26)* (1.34) (1.89) (2.40)* (1.18)
Mar 0.0085 0.0146 0.0025 0.0007 0.0025 -0.0011 0.0129 0.0195 0.0063
(1.21) (1.49) (0.47) (0.08) (0.22) (-0.18) (1.88) (1.70) (1.15)
Apr -0.0160 -0.0177 -0.0144 -0.0178 -0.0169 -0.0187 -0.0161 -0.0205 -0.0117
(-1.74) (-1.73) (-1.54) (-1.93) (-1.61) (-2.22)* (-1.53) (-1.68) (-1.00)
May 0.0198 0.0278 0.0118 0.0163 0.0208 0.0119 0.0232 0.0317 0.0147
(3.90)** (4.06)** (2.36)* (3.45)** (3.10)** (2.25)* (3.67)** (3.87)** (2.35)*
Jun 0.0305 0.0407 0.0202 0.0173 0.0249 0.0097 0.0357 0.0463 0.0252
(4.56)** (4.67)** (3.63)** (3.28)** (3.09)** (2.83)* (4.58)** (4.83)** (3.30)**
Jul -0.0188 -0.0228 -0.0147 -0.0052 -0.0025 -0.0080 -0.0273 -0.0337 -0.0208
(-2.85)* (-3.68)** (-1.98) (-0.62) (-0.27) (-0.99) (-4.28)** (-5.15)** (-3.04)**
Aug 0.0030 0.0076 -0.0017 0.0062 0.0103 0.0021 0.0038 0.0116 -0.0040
(0.43) (0.89) (-0.26) (0.73) (1.04) (0.28) (0.72) (1.59) (-0.59)
Sep 0.0087 0.0150 0.0025 0.0017 0.0071 -0.0037 0.0142 0.0224 0.0060
(1.63) (2.65)* (0.40) (0.33) (1.30) (-0.59) (2.08) (3.17)** (0.80)
Oct -0.0029 -0.0007 -0.0050 -0.0067 -0.0013 -0.0120 0.0009 0.0029 -0.0010
(-0.72) (-0.11) (-1.31) (-2.02) (-0.27) (-2.66)* (0.16) (0.30) (-0.21)
Nov 0.0019 0.0056 -0.0019 0.0027 0.0109 -0.0055 0.0008 -0.0011 0.0026
(0.28) (0.72) (-0.28) (0.38) (1.32) (-0.85) (0.10) (-0.15) (0.27)
Dec 0.0086 0.0154 0.0018 0.0073 0.0146 -0.0001 0.0079 0.0128 0.0030
(3.36)** (4.76)** (0.62) (1.66) (2.61)* (-0.02) (1.47) (1.69) (0.67)
* denotes significant at the 5% level
** denotes significant at the 1% level.

31
Table 9 – Hit rates for trend salience strategies

This table reports hit rates, which are the proportion of months in which the abnormal return is positive for three investment strategies: the traditional momentum (MOM),
salient momentum (SAL) and non-salient momentum (NON) strategies. Salient momentum and non-salient momentum strategies are formed on double sorts based on past
returns and the ratio of 6 to 12-month geometric average rate of return (GARR). Traditional momentum strategies are formed using a 12-month formation period. Panel A
(Panel B) reports the results for portfolios that are rebalanced semi-annually (annually). The right tail p-value is reported in brackets under the associated hit rates. These p-
values represent the probability of randomly achieving the measured hit rate or higher.

Panel A: Hit rates for trend salience strategies with a 6-month holding periods

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total
MOM 0.40 0.75 0.75 0.40 0.75 0.85 0.45 0.50 0.60 0.60 0.55 0.80 0.62
[0.81] [0.01] [0.01] [0.81] [0.01] [0.00] [0.67] [0.50] [0.19] [0.19] [0.33] [0.00] [0.00]
SAL 0.55 0.90 0.75 0.35 0.80 0.90 0.40 0.65 0.70 0.70 0.65 0.80 0.70
[0.33] [0.00] [0.01] [0.91] [0.00] [0.00] [0.81] [0.09] [0.04] [0.04] [0.09] [0.00] [0.00]
NON 0.40 0.65 0.60 0.45 0.75 0.75 0.45 0.50 0.60 0.50 0.50 0.60 0.56
[0.81] [0.09] [0.19] [0.67] [0.01] [0.01] [0.67] [0.50] [0.19] [0.50] [0.50] [0.19] [0.03]

Panel B: Hit rates for trend salience strategies with a 12-month holding periods

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Total
MOM 0.30 0.65 0.70 0.45 0.80 0.95 0.45 0.55 0.55 0.40 0.50 0.60 0.62
[0.96] [0.09] [0.04] [0.67] [0.00] [0.00] [0.67] [0.33] [0.33] [0.81] [0.50] [0.19] [0.00]
SAL 0.40 0.80 0.70 0.35 0.90 0.95 0.30 0.70 0.65 0.55 0.60 0.75 0.68
[0.81] [0.00] [0.04] [0.91] [0.00] [0.00] [0.96] [0.04] [0.09] [0.33] [0.19] [0.01] [0.00]
NON 0.30 0.65 0.60 0.40 0.65 0.80 0.45 0.45 0.55 0.40 0.50 0.40 0.56
[0.96] [0.09] [0.19] [0.81] [0.09] [0.00] [0.67] [0.67] [0.33] [0.81] [0.50] [0.81] [0.35]

32