Introduction
Traditional asset pricing theory assumes that investors rationally utilize all available information to assess risky assets (Fama, 1970) This foundation underpins expected utility theory and standard choice theory, which social scientists frequently employ to predict stock returns However, extensive research indicates that risk attitudes often deviate from the independence axiom of expected utility (Kahneman & Tversky, 1979; Barberis & Huang, 2008) In response, Bordalo et al (2012) introduced "salience theory," which posits that decision-makers focus on unusual or salient attributes, leading them to overemphasize these features while neglecting non-salient payoffs Empirical evidence supports salience theory's predictions regarding stock returns in both the U.S market (Cosemans & Frehen, 2021) and in developed and emerging markets (Cakici).
This study investigates the presence of the salience effect in the Vietnam stock market, highlighting its distinctive characteristics compared to the U.S and other developed markets Key differences include unique information disclosure practices (Craig & Diga, 2002), varying levels of market efficiency (Rizvi & Arshad, 2014), and specific institutional arrangements (Nguyen et al.).
This study aims to explore whether salience models, which address various decision-making theory challenges like the Allais paradox, can shed light on decision makers' behaviors in the increasingly accessible Vietnam stock market Given the rising interest from both academics and investors in examining the salience effect on return predictability within this market, understanding these dynamics is becoming increasingly important.
Empirical findings strongly validate the predictive capabilities of salience theory in the context of Vietnam's stock market Time-series analysis reveals that stocks with prominent upside potential yield significantly higher returns in the following month compared to those with notable downside risks The disparity between high and low portfolio performance is both economically and statistically significant throughout the sample period from 2010 to 2021.
Adjusting for risk reveals significant differences in returns, with six-factor alphas reaching 1.64% Utilizing Fama-MacBeth regression confirms a robust cross-sectional relationship between short-term (ST) factors and expected returns The ST effect remains both economically and statistically significant, even when accounting for additional predictive variables This aligns with the findings of Cosemans & Frehen (2021), which identify a negative correlation between the ST effect and future returns.
This study investigates the impact of salience components, specifically salience-weight (SW) and equal-weight (EW) past returns, on the capital market It finds that salience distortion and salient trading volume have significant economic and statistical implications Utilizing Fama-MacBeth analysis, the results indicate a negative correlation between both salience distortion and salient trading volume with future expected returns This aligns with the hypothesis that stocks exhibiting high levels of salience distortion tend to have lower expected returns compared to those with low levels of salience distortion.
To examine the effectiveness of salience theory, I adjusted the degree of probability and salience My findings indicate that the salience theory measure remains both economically and statistically significant, even with variations in salience measurement However, it has minimal impact on the predictability of stock returns Specifically, when the degree of salience approaches 1 (δ → 1), it suggests that heightened rationality among investors diminishes the predictability of the salience effect on asset pricing theory.
This study is structured into several key sections: Section 2 discusses decision-making puzzles and the implications of salience theory on asset pricing Section 3 presents a descriptive summary and details the dataset utilized In Section 4, time-series analysis and firm-level cross-sectional regressions are reported Section 5 delves deeper into the mechanisms underlying the effects of salience theory Finally, Section 6 concludes the findings of the research.
Literature Review
Salience theory
In Bordalo et al.'s (2012) model, known as BGS, local thinkers are presented with a decision-making scenario involving two lotteries, L1 and L2 These lotteries are structured over a state space S, which comprises N distinct states, with each state s in S occurring with a specific probability.
In the context of lotteries, the local thinker assesses payoffs \( x_i \) in state \( s \) using a linear value function \( v(x) = x \), with evaluations centered around a reference point of zero This approach indicates that, in the absence of salience distortions, local thinkers perceive the lottery \( L_i \) based solely on its straightforward payoffs.
The decision-maker diverges from the initial equation by assigning less weight to the least prominent outcome in the lottery's set of states This salience distortion occurs in two phases: first, the lottery payoff generates a ranking of salience among the states, and second, this ranking leads to the substitution of the original probabilities with transformed, lottery-specific decision weights Bordalo et al (2012) define the salience of each lottery state as a continuous and bounded function, characterized by three specific conditions, with x_s_min and x_max_s representing the smallest and largest payoffs within the set.
1 Ordering: If for states s, sS, and x s min ,x s max is a subset of x s min ,x s max , then
2 Diminishing sensitivity: If x s j 0 for any j =1, 2, then for any 0,
3 Reflection: For any two states s , sS such that x x s j , s j 0 for j =1, 2, we have
To measure the salience of the payoff x i s of lottery i in state s , Bordalo et al
= − + + (2) where 0 and N s i / x = i x N , with N denoting the number of lotteries
The salience effect in (2) must satisfy the above three conditions: (i) ordering, (ii) diminishing sensitivity, and (iii) reflection If x min s ,x s max is a subset of min max s , s x x
In the context of decision-making, ordering indicates that the state transition effect from state s to state s' is greater than the initial state s Diminishing sensitivity suggests that as the absolute payout of a lottery decreases, its perceived significance increases Additionally, reflection emphasizes the importance of the magnitude of gains over their sign, implying that reflecting gains in losses does not alter the perceived importance of outcomes, as perceptions remain sensitive to differences in absolute values.
Given the emphasis function in (2), the salient thinkers rank states and skew their decision weights as follows:
Given states s s, S, lottery L i state s is more salient than s if ( ,x x s i s i ) ( ,x x s i s i )
In the salience ranking system, let \( k_{s_i} \) represent the salience of state \( s \) for individual \( L_i \), where a lower \( k_{s_i} \) signifies greater salience States sharing the same level of salience receive identical rankings, allowing the salient thinker to effectively adjust the odds based on these rankings.
of s relative to s into the odds i s i s
= − where (0,1] The decision weight attached by salient thinkers to a generic state s in the evaluation of L i is: i i s s s
= where s i denotes the salience-weighted subjective state probability, s is the objective probability, and s i is the salience weight defined as:
5 where each state s occurs with probability s such that s i 1 s
, the decision weights are normalized to sum to 1 (the expected distortion is zero ( [ s i ] 1= )
The parameter δ in Equation (3) quantifies the extent of local thinking by assessing cognitive ability and how salience influences decision weights When δ equals 1, the decision-maker aligns with standard economic theory, where decision weights reflect objective probabilities (ωsi = 1 for all s ∈ S) Conversely, if δ is less than 1, the decision-maker is characterized as a great thinker who tends to overweight salient states (ωsi > 1) while underestimating non-salient ones (ωsi < 1) As δ approaches 0, these thinkers concentrate solely on the primary payout of a lottery, disregarding all other potential payouts.
Given the condition above, the local thinker computes the value of lottery L i as:
Thus, L i ’s evaluation always lies between the value of its highest and lowest payoffs
In choice between two lotteries, equation (4) implies that – due to the symmetry of salience function, the local thinker prefers L 1 to L 2 if and only if:
When δ equals 1, local thinkers align their decision weights with objective probabilities However, when δ is less than 1, local thinking favors option L1 over L2 when L1 offers greater rewards in more salient states, which are perceived as less discounted.
The implications of salience theory and asset pricing
The salience theory asset pricing model proposed by Bordalo et al (2013) highlights the influence of salience effects on investment decisions and stock prices Their consumption-based model features two periods and assumes identical investors with linear utility regarding current (t = 0) and future (t = 1) consumption values At time t = 0, each investor is endowed with a consumption good worth w0, along with one unit of each asset.
6 of the F 1 available assets f =1, ,F At t = 1, there are S states of nature
1, , s= S, each occuring with probability s with
S s = s Asset j pays dividend x js in state sS at t =1
The salience of a generic asset's payoff, denoted as x fs, is influenced by its comparison to the average market payoff, represented as x s, which is calculated by summing the payoffs of all available assets in the same state This relationship defines the salience effect, highlighting how the perceived value of an asset's payoff is relative to the broader market context.
"narrow framing" of objective asset payoffs, and unaffected by investor-specific characteristics such as portfolio size or assets In particular, the salience function (x js ,x s )
satisfies two properties: (i) ordering: if interval x y , is a subset of interval
x y ', ' , then ( , ) x y ( ', ') x y ; (ii) diminishing sensitivity: if all x y, 0 and any
0, additional of payoffs x and y, cause to state s is more salient, ( , )x y (x ,y )
If an investor trades an amount f of each asset f , his expected utility at 0 t = is:
F f f f c = − p , 1, F ( f 1) f s fs c = + x , where f +1 is the endowment of asset f plus any additional amount bought or sold by the investor
The first-order condition for a solution to this problem is:
The investor's payoff valuation typically follows standard practices, but it is influenced by distorted state probabilities When the potential gains (or losses) of stock f are prominent, an investor who focuses on these salient aspects tends to purchase more (or less) of stock f compared to an expected utility maximizer, who relies on undistorted probabilities for stock evaluation.
By integrating the optimal trading decisions of all investors with the market clearing condition, we can determine the price impact resulting from an increase in stock demand In equilibrium, every investor possesses the market portfolio, establishing the stock price accordingly.
[ fs fs ] [ fs ] cov[ fs , fs ] p f = x = x + x
In the absence of salient distortions, the first term on the right-hand side of Eq
The price of a stock reflects the future value of its expected payouts, determined by objective probabilities A stock is considered overvalued when its maximum payout captures investors' attention, indicating a positive correlation between expected returns and perceived potential Conversely, when the minimum payout becomes more prominent, investors concentrate on the associated risks, leading them to demand a lower valuation than its fair price to justify holding the stock.
Following Cosemans & Frehen (2021), Dividing both sides of Eq (6) by p j yields the implications of salience theory for expected returns:
The equation (7) illustrates the core hypothesis of salience theory in asset pricing, suggesting that future returns are diminished for stocks with notable positive salience (salient upsides) compared to those with significant negative salience (salient downsides) When investors exhibit rational behavior (δ = 1), biases are absent, resulting in equal salience across all states Consequently, the covariance between the salience weight and future returns is zero, leading to an expected return of zero, as risk-neutral investors do not discount future outcomes.
Salience theory posits that securities with noticeable positive attributes attract investor attention, leading to overpricing and high returns during periods of heightened interest, followed by a decline in returns as the overvaluation normalizes Conversely, stocks with prominent negative features deter investors, resulting in low returns during selling phases, with subsequent periods reflecting justified undervaluation This article aims to empirically analyze the predictions of salience theory regarding returns during market correction phases.
2.2.2 Construction of salience theory measure
To test the hypothesis that a stock's salience theory value inversely correlated to its future returns, I need to specify the state space specifications that can occur and
The experimental design presents participants with a choice between two lotteries, highlighting the relationship between payoffs and their probabilities In empirical applications, the concept of state space can become ambiguous According to Barberis et al (2016), and further explored by Cosemans & Frehen (2021), investors tend to represent each stock by analyzing the distribution of its historical returns They utilize past returns to infer potential future return states when making stock selections Consequently, the state is defined by daily returns from the preceding month, with each return assigned an objective probability based on the number of trading days within that month.
To determine the value of salience theory, I first calculate the salience of each stock's daily payoffs using a specific measurement method According to Bordalo et al (2013), investors consider all available stocks in the market when making decisions, which means they evaluate a company relative to its competitors Consequently, the salience effect value of a stock's return on a given day is influenced by its deviation from the average market payoff on that day This relationship establishes the salience of daily payoffs.
In line with Cosemans & Frehen (2021), I maintain a constant θ = 0.1 Additionally, the average return (r_s) is computed using an equal-weighted approach for all stock prices in the Vietnam market This method preserves the ordering, diminishes sensitivity, and reflects the properties of the salience function.
In the second step, individual stock payoffs (r_i_s) are assessed based on their salience, as defined by Equation (8) Each stock is then ranked (k_i_s) from 1, indicating the most salient, to S, representing the least salient Here, S signifies a set of potential states that reflect the estimated number of transaction days within the month.
Subsequently, the salience-weighted probabilities of return i s , is computed, which substitute the objective probability:
9 where the salience weight of security ( i s , ) on day s is defined as follows:
The parameter in equation (9) as proxies for the decision-marker's cognitive ability, which obtained by Bordalo et al (2012) from experimental data, namely,
=0.7 In addition, the decision weights are normalized to sum to 1, so that
Lastly, the salience theory value is computed as the covariance between salience weights and daily returns as:
S S i t is t is t s t is t is t s t is t s s r r r
Equation (10) illustrates the key prediction of salience-based asset pricing, indicating that stocks anticipated to have low future returns are associated with high salience theory values Conversely, stocks expected to yield high returns tend to exhibit low salience theory values.
This article examines cross-sectional predictors of stock returns recognized in asset pricing theory, focusing on key factors such as size, book-to-market ratio, momentum effect, and short-term reversal anomaly Size (ME) is represented by the natural logarithm of a company's market capitalization, calculated by multiplying the current share price by the total number of outstanding shares The book-to-market ratio (BM) is derived from the natural logarithm of the ratio of the six-month lagged book value of equity, as outlined by Fama & French (1992) The momentum effect (MOM), following Carhart (1997) and Fama & French (1996), is determined by the monthly cumulative return of a stock from month t−12 to t−2 Additionally, the short-term reversal anomaly (REV) is assessed through a stock's return at one-month lagged, as discussed by Lehmann (1990) and Jegadeesh.
1990) The ILLIQ effect of Amihud (2002) is captured by the stock market liquidity ratio, as follow:
The return of stock i on day s, denoted as r, is influenced by the daily trading volume, represented as DVol, which is quantified in trillions of VND Additionally, NoTD i t indicates the number of trading days for stock i within month t.
A large body of research shows that a group of “skewness” variable help to explain the cross-section of stock return's predictability (Bali et al., 2011, Barberis,
To analyze the relationship between stock returns and related skewness variables, I incorporate several metrics into our tests Max represents a stock's highest daily return over the past month, while Min indicates the negative value of the stock's lowest daily return during the same period Skewness (SKEW) refers to the asymmetry of daily stock returns, and idiosyncratic skewness (ISKEW) is derived from regression residuals, estimated using the Fama & French (1993) model, as demonstrated by Boyer et al (2009) This approach follows the methodologies outlined by Harvey & Siddique.
2002, co-skewness (COSKEW) is derived by running the following regression using the past year of daily data:
Empirical research
Recent research by Cosemans & Frehen (2021) reveals a significant inverse relationship between salience theory effects and expected stock returns in the U.S market Their cross-sectional regression analysis supports the salience model, demonstrating that stocks with prominent upsides yield lower returns over the following month compared to those with notable downsides A univariate portfolio analysis highlights a statistically significant and economically substantial return difference between the highest and lowest salience deciles, with zero-cost strategies showing average returns of -1.28% per month for equal-weighted portfolios and -0.60% for value-weighted portfolios These discrepancies are not accounted for by traditional market factors, as five-factor alphas range from -1.44% (EW) to -0.80% (VW) per month Furthermore, the salience effect persists even when considering the investment and profitability factors outlined in the Fama and French (2018) six-factor model, with six-factor alphas of -1.32% (EW) and -0.64% (VW).
Cosemans & Frehen (2021) reveal a significant correlation between stock trading (ST) and future returns, particularly in stocks with high retail ownership and limited arbitrage opportunities Their research indicates that the predictive ability of ST is enhanced during periods of high market sentiment, when less experienced investors are more active Additionally, the salience effect diminishes when ST is calculated using open-to-open returns instead of close-to-close returns, suggesting that retail investors base their trading decisions on observed close-to-close returns This finding challenges risk-based explanations for the salience effect, supporting a behavioral interpretation of the connection between ST and future returns.
Cosemans & Frehen (2021) investigate the salience effect to confirm it is not merely a rebranding of existing return anomalies by creating double-sorted portfolios and conducting firm-level Fama-MacBeth regressions Their findings indicate that the salience theory effect continues to demonstrate substantial explanatory power for returns, even after accounting for a comprehensive set of variables.
Twelve firm characteristics have been identified that explain the cross-sectional variation in stock returns, including various indicators of lottery demand Additional tests demonstrate that the relationship between stock salience and future returns remains consistent, even when using different methods to measure return salience and varying the market index for comparison.
Cosemans & Frehen (2021) investigate two alternative explanations for the negative relationship between short-term (ST) returns and future returns They suggest that ST may reflect short-term return reversals, which can be attributed to investors over-extrapolating past returns when predicting future performance This aligns with behavioral theories, such as those proposed by Subrahmanyam (2005) and Greenwood and Shleifer (2014) However, salience theory posits that investors' reactions to information are context-dependent; they tend to overweight a stock's past returns only if they are notable compared to the market return, while downplaying non-salient past returns Consequently, salience influences which returns are extrapolated, leading to distorted expectations not from overreaction but from biases in how returns are perceived ST, defined as the difference between salience-weighted and equal-weighted returns, effectively captures these distortions in return expectations.
Cosemans & Frehen (2021) conduct several tests to distinguish the salience effect from the one-month return reversal Their findings indicate that the salience effect continues to be significant even when an additional month is omitted between the construction of the returns.
The study examines the measurement of subsequent returns using short-term (ST) metrics over periods longer than one month, where stock returns do not show reversal patterns It enhances the five-factor model by incorporating a short-term reversal factor, revealing that the alphas of the high-low ST portfolio remain substantial and statistically significant, with monthly returns of 102 basis points (t-stat = -10.35) for the equal-weighted (EW) portfolio and 32 basis points (t-stat = -2.24) for the value-weighted (VW) portfolio Additionally, by integrating a stock's past one-month return in bivariate portfolio sorts and Fama-MacBeth regressions, the research indicates that while accounting for reversal diminishes the salience effect's magnitude, it continues to be considerable and significant.
In a study of 13 bivariate sorts, the average return spread between high- and low-short-term (ST) deciles is observed to be 48 basis points (bps) per month with equal weighting and 22 bps with value weighting The five-factor alpha spreads range from 60 bps (equal weighting) to 30 bps (value weighting) per month, translating to annual figures of 7.2% and 3.6%, respectively Fama-MacBeth regression analysis reveals that the coefficient on ST is statistically significant at the 1% level (t-stat = -6.80), even after accounting for a stock’s previous one-month return and 13 additional firm characteristics Furthermore, a two-standard-deviation increase in ST is associated with a predicted decrease of 26 bps in the following month’s stock return, highlighting both statistical and economic significance.
In a robustness check, the authors reanalyze stock returns using quote midpoints, revealing that the salience effect is not influenced by microstructure issues like bid-ask bounce, while short-term reversal diminishes by 40% with midpoint returns They observe diverging trends in the salience and reversal effects, confirming that return reversal has significantly weakened in recent decades due to enhanced market liquidity, whereas the salience effect remains robust and statistically significant Additionally, they find that returns linked to the salience effect occur entirely intraday, in contrast to short-term reversal returns, which are realized overnight This distinction in intraday and overnight returns supports the notion that salience and reversal effects are separate phenomena with different underlying mechanisms.
According to Cosemans & Frehen (2021) and Cakici & Zaremba (2021), the salience theory (ST) effect is evident in international stock returns, with stocks exhibiting the highest ST values underperforming those with the lowest by an average of 0.34% per month In a global analysis, a long-short strategy that involves buying the top decile and selling the bottom decile of ST stocks results in a six-factor alpha of -0.66% While the salience anomaly appears to be significant on an international scale, it is important to note three key caveats related to cross-sectional, time-series, and the interplay with existing anomalies.
The salience effect is most pronounced during extreme market conditions, where abnormal returns are heightened by factors such as market-wide illiquidity, high-bill yields, and economic uncertainty This anomaly is particularly influenced by past market returns and volatility, leading to significant mispricing following major down markets The salience strategy yields substantial profits during these turbulent times, while its impact diminishes during stable market periods Additionally, the salience effect shares a close relationship with short-term return reversal, exhibiting similarities in time-series dynamics and portfolio composition Notably, the anomaly is strongest over a one-month estimation period but weakens or vanishes with longer time frames The salience effect also relies heavily on daily return reversals, as its significance decreases when the most recent day’s return is excluded from analysis Overall, while the short-term reversal does not fully encompass the salience effect, their interconnectedness is clear.
In consequence, the short-term reversal effect accounts for a substantial share of the
Cakici & Zaremba (2021) highlight key aspects of the salience anomaly, noting that its evidence is significantly influenced by the study period Their analysis of a long-run US sample indicates that the salience effect was strong and consistent in the past, but has notably diminished in recent decades As a result, over the last 30 years, the abnormal returns linked to salience have primarily persisted in a limited capacity, specifically among the smallest firms in the equity market.
The salience anomaly varies significantly across countries, particularly influenced by the level of idiosyncratic risk within each market Countries characterized by high average idiosyncratic volatility experience a more pronounced salience effect compared to their "safer" counterparts Notably, high-risk markets yield salience strategy profits that surpass those of low-risk markets by 0.60% to 0.79%.
15 risk countries The idiosyncratic risk has proven to be the key driver of international variation within salience-driven mispricing
Cakici & Zaremba (2021) validate their findings through additional tests and robustness checks, including extensive portfolio sorting across individual countries and pooled samples of global, developed, and emerging market stocks They perform cross-sectional regressions with firm size restrictions and weighted least squares regressions, as demonstrated by Hou et al (2020) By relaxing various assumptions related to their baseline calculations—such as different reference benchmarks, alternative estimation periods, modified filtering rules for the equity universe, and different portfolio formation methods—they consistently arrive at the same conclusion The ST effect, while observable globally, is significantly influenced by the estimation period and firm size Mispricing primarily arises in small firms, risky countries, and extreme market conditions, indicating that the ST anomaly is exacerbated by severe arbitrage constraints, making practical attempts to capitalize on this phenomenon challenging, if not impossible.
Data
My data set comes from standard sources - State Security Commission of
Vietnam (SSC) and includes all Hochiminh Stock Exchange (HSX) and Hanoi Stock
This analysis focuses on non-financial firms listed on the Hanoi Stock Exchange (HNX) from January 2010 to December 2021 Following the methodology established by Cosemans & Frehen (2021), the study includes stocks with available daily prices for a minimum of 15 days, allowing for the calculation of the ST effect.
Table 1 summarizes the statistical characteristics of stocks, with Panel A displaying the means and standard deviations, while Panel B presents the pairwise correlations Notably, the results labeled as ST in Panel B highlight significant correlations among the stock characteristics.
The study reveals that ST exhibits a positive correlation with risk measures such as Beta and DBeta, as well as with gambling variables like Max and Min Conversely, ST shows a negative correlation with volatility measures (Ivol) and an inverse relationship with skewness metrics, including iskew, coskew, and skew.
The findings support the salience-based asset pricing theory, indicating that stocks with prominent upside potential tend to yield lower subsequent returns compared to those with notable downside risks Additionally, stocks exhibiting a strong short-term effect (ST-effect) are often associated with higher market capitalizations, likely due to their classification as growth stocks.
Panel A Mean and standard deviation
ST ME BM Illiq Mom SVR Max Min Iskew Skew Coskew Dbeta Beta Ivol
ST ME BM Illiq Mom SVR Max Min Iskew Skew Coskew Dbeta Beta Ivol
The article provides a comprehensive overview of the research analysis variables, detailing their mean and standard deviation in Panel A, while Panel B illustrates the pairwise correlation among these variables for the entire sample Key metrics include the salience-theory effect (ST), the natural logarithm of firm market value (ME), and the book-to-market ratio (BM) as defined by Fama & French (1992) Additionally, momentum (Mom) is measured as cumulative returns from month t-12 to t-2, and short-term reversal (REV) reflects the lagged monthly return The illiquidity measure (Illiq), based on Amihud’s (2002) methodology, is scaled by 10^9 VND over a one-month window The analysis also includes maximum and minimum daily returns (Max and Min) for month t-1, skewness of daily stock returns (Skew), expected idiosyncratic skewness (Iskew) derived from the Fama & French (1993) model, and co-skewness (Coskew) calculated as per Harvey & Siddique (2002) over a one-year period Furthermore, idiosyncratic volatility (Ivol) is obtained from regressing daily excess returns on market excess returns, with market beta (Beta) and downside beta (Dbeta) also calculated from this regression, focusing on days with below-average market returns The analysis spans from January 2010 to December 2021, highlighting coefficients significant at the 1% level in bold.
Cross-sectional relation between salience and stock returns
Time-series test
The primary hypothesis posits that the salience theory's impact on a stock's historical return distribution can effectively forecast its future returns in cross-sectional analysis Specifically, the equation (10) embodies the key prediction of the salience-based asset pricing model: during favorable market conditions, stocks exhibiting a high salience threshold (salient upside) are likely to experience lower future returns compared to those with a low salience threshold (salient downside).
Decile Raw return FF4 alpha FF5 alpha FF6 alpha Raw return FF4 alpha FF5 alpha FF6 alpha
The table presents the average monthly raw excess returns and alphas of decile portfolios categorized by the salience theory effect (ST), as outlined in Section 2.2 At the end of each month, stocks are organized into decile portfolios based on their ST values For each portfolio, we provide equal-weight (EW) and value-weight (VW) raw excess returns, along with four-factor alpha per Carhart (1997), five-factor alpha that includes a liquidity factor, and six-factor alpha based on Fama & French’s (2015) model incorporating momentum Stocks with the highest and lowest ST values are designated as High and Low, respectively The table also highlights the average differences in returns and alphas (H-L) between the High and Low deciles This study covers the period from January 2010 to December 2021, with Newey & West’s (1987) t-statistics adjusted and noted in parentheses.
In my empirical analysis, I employ univariate portfolio sorts, categorizing all stocks into decile portfolios based on their ST value at the end of each month For each decile portfolio, I provide equal-weight (EW) and value-weight (VW) raw excess returns, along with four-factor alpha as outlined by Carhart (1997) Additionally, I calculate five-factor alpha, which incorporates a liquidity factor into the Carhart four-factor model, and six-factor alpha as derived from Fama's framework.
& French’s (2015) model with momentum factor The stocks with highest and lowest
ST value are labeled High and Low The table also report the average differences in returns and alphas (H-L) on the High (Low) decile The study spans the years January
2010 to December 2021 Newey & West’s (1987) t-statistics are adjusted in parenthesis
Table 2 demonstrates a significant impact of salience theory, particularly evident in equal-weighted portfolio returns The average raw return difference between the highest (decile 10) and lowest (decile 1) salience theory portfolios is 1.48% per month, supported by a Newey & West t-statistic of 58.05 Additionally, the Fama-French alphas reveal a notable difference, with decile 1 showing an alpha of -0.73% and decile 10 at 0.91% per month, resulting in a 1.64% difference and a t-statistic of 30.69 Furthermore, Table 3 indicates a U-shaped relationship between salience theory and raw excess returns.
During challenging economic periods, investors tend to favor value stocks over growth stocks due to the prominent risks associated with the latter, resulting in the undervaluation of these assets and consequently lower expected returns.
To get better understanding of what is it actually make stock with low ST’s effect that appeal to investor, I look more closely on characteristics of low and high
During challenging market conditions, investors often exhibit risk aversion, gravitating towards value stocks instead of growth stocks Consequently, I anticipate that stocks with low short-term (ST) metrics will also reflect low historical returns and elevated idiosyncratic volatility.
Table 3 Characteristics of ST-sorted portfolio
Decile ST Price Size BM Max Min Illiq Ivol Mom Iskew Skew Coskew Dbeta SVR BETA
The table presents the characteristics of the ST-sort portfolio, which is organized into deciles based on ST value each month It calculates the average return for each characteristic portfolio, where ST represents the salience-theory effect expressed in percentages Stock prices are reported in thousands of VND, while market equity (ME) is the firm's market value at the end of the previous month, converted to millions of VND The book-to-market ratio (BM) is calculated using the logarithmic method as outlined by Fama & French (1992) Momentum (Mom) reflects cumulative returns from the beginning of month t-12 to the end of month t-2, also expressed in percentages Short-term reversal (REV) is based on the lagged monthly return, while illiquidity (Illiq) is measured using Amihud’s (2002) formula, scaled by millions of VND and averaged over one month Additionally, the maximum and minimum daily returns for month t-1 are provided in percentage terms.
The article discusses key financial metrics related to stock returns, focusing on skewness, idiosyncratic skewness (Iskew), and co-skewness (Coskew) Skewness measures the asymmetry of daily stock returns over the past year, while Iskew is derived from the Fama & French model, as referenced by Boyer et al Coskew is calculated based on the methodology of Harvey & Siddique, also over a one-year period The article highlights that idiosyncratic volatility (Ivol) and market beta (Beta) are obtained through regression analysis of daily excess returns against market excess returns Additionally, downside beta (Dbeta) is calculated from this regression, specifically considering days when the market return falls below the average daily market return over the same timeframe The analysis covers a sample period from January 2010 to December 2021.
This analysis evaluates the relationship between stock characteristics and returns by calculating the average values for each characteristic across all stocks in each decile The findings, presented in Table 2, indicate that stocks with low salience value (ST) exhibit higher past returns, as measured by market equity (ME), book-to-market (BM), and price Additionally, the mean market capitalization of these stocks increases from low to high ST values, while the book-to-market ratio decreases This supports the hypothesis that Vietnamese investors tend to focus on the prominent downsides of value stocks, resulting in overvaluation and lower future returns Furthermore, stocks with lower ST values also demonstrate higher illiquidity and idiosyncratic volatility.
Time-series analysis demonstrates a robust predictability of salience theory's impact on future stock returns Consistent with our hypothesis, Table 2 indicates preliminary evidence of a negative relationship between the salience theory effect and expected stock returns The disparity in returns between high- and low-salience theory portfolios is both economically and statistically significant, though it is only evident in the equal-weighted portfolio.
Robustness checks of time-series
Prior to performing the Fama-MacBeth regression analysis, I conduct a series of robustness time-series checks to ensure the validity of my conclusions presented in Table 4 This table features four panels, each representing a distinct robustness check The two right columns summarize the five-factor alpha for the high-ST portfolio minus the low-ST portfolio, utilizing both value-weighted and equal-weighted returns.
First, I consider whether my result not only is statistically significant in full sample, but also in alternative period: Jan 2010 – Jun 2015, and July 2015 – Dec
2021 The result in Table 4 support my hypothesis that long-short ST portfolio, include value-weighted portfolio, economically and statistically significant in two subperiod
Table 4 Robustness check of time-series
The table presents the outcomes of three robustness tests conducted for time series analysis, highlighting equal-weighted (EW) and value-weighted (VW) five-factor alphas for a long-short portfolio that invests in stocks from the highest and lowest deciles of short-term (ST) performance each month The first panel details results from two distinct sub-periods, while the second panel utilizes daily returns from the past 3, 6, and 12 months to derive performance metrics for ST In the third panel, the benchmark return is adjusted to Eq (8), incorporating raw returns and returns exceeding the risk-free rate The analysis covers a sampling period from January 2010 to December 2021, with t-statistics adjusted per Newey & West (1987).
Return in excess of the risk-free rate -3.76 -2.89
In constructing the past return distribution for a stock, I analyze daily returns over the previous month To validate the findings presented in Tables 2 and 3, I extend my analysis to daily returns over three, six, and twelve-month periods to assess the effect of ST The results in the second panel of Table 4 indicate that the returns for both High- and Low-ST portfolios align closely with those in Table 2 Additionally, I modify the benchmark return in Equation (8) by incorporating raw returns and returns exceeding the risk-free rate, yet the outcomes remain consistent Overall, the robustness checks across different time frames reinforce the conclusion that ST has a significant predictive power for future stock returns.
Fama-MacBeth Firm-level regressions
To evaluate the predictive power of salience theory while controlling for established return predictors, I employ cross-sectional regressions based on Fama & MacBeth’s (1973) methodology The analysis reveals a significant and substantial relationship between average monthly returns and the salience theory (ST) variable, even when accounting for key asset pricing predictors.
In month t+1, cross-sectional regression analysis are estimated on a firm's ST variable and a vector of control variables Y i t , measured at the end of month t :
The return of firm i in month t+1, denoted as r_i_t+1, is influenced by the salience theory value of stock i in month t (ST_i_t) and a set of standard control variables, including size (ME), book-to-market ratio (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), an illiquidity measure (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (Coskew).
Table 5 presents the time-series means of the slope coefficient (γ) for the independent variables, accompanied by t-statistics calculated using Newey & West (1987) standard errors The ten columns in the table represent nine distinct regression specifications, each varying in the number of control variables included.
The findings in the table support my hypothesis, revealing that the ST variable negatively predicts one-month stock returns, maintaining both economic and statistical significance The salience anomaly demonstrates predictive power not only in univariate portfolios but also retains a significant economic effect at the 1% level, even when additional variables are included The average slope γ1,t from the monthly regression of predicted returns on ST alone is -0.007, with a t-statistic of -3.18 Despite the inclusion of related-skewness variables in columns 7-10, the predictive power of ST remains statistically significant at the 1% level Notably, the coefficient of ST continues to show strong significance after accounting for related-skewness variables Furthermore, the t-hurdle for the ST variable in Table 5 exceeds 3.0, surpassing the stringent threshold of 3.0 proposed by Harvey et al (2016).
Table 5 Firm-level Fama-MacBeth regression
The table presents the average slope coefficients derived from cross-sectional regression employing the Fama-MacBeth methodology Each month, stocks are categorized into decile portfolios based on their ST values The findings are detailed in the most comprehensive regression model outlined in column (10).
The study analyzes various financial metrics, including size (ME), book-to-market (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), an illiquidity measure (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (coskew) The data spans from January 2010 to December 2021, with t-statistics adjusted according to Newey & West (1987).
Impact of limit to arbitrage
In the salience-based asset pricing model proposed by Bordalo et al (2013), it is assumed that all investors act rationally However, market participants exhibit varying cognitive biases that influence the salience effect differently Rational investors can address mispricing when there are no constraints on arbitrage Consequently, I anticipate that the salience theory will have a more pronounced effect on securities.
In Table 6, I present the findings of Fama-Macbeth regression by interacting
ST with three proxies for limits to arbitrage: market capitalization, illiquidity, and idiosyncractic volatility My regression result is based on the full sample from January 2010 to December 2021
Table 6 Fama-MacBeth regression: limits to arbitrage
This table presents the findings from the Fama-MacBeth regression, analyzing stock returns in relation to the effects of short-term trading (ST) and its interaction with three variables that represent limits to arbitrage: market capitalization (ME), illiquidity (Illiq), and idiosyncratic volatility (Ivol) The monthly cross-sectional regression was conducted to assess these relationships.
In this study, Z i t, j serves as a firm-level indicator of limits to arbitrage, while Y i t, j encompasses a comprehensive vector of firm characteristics detailed in section 2.2.3 The analysis covers a sample period from January 2010 to December 2021, with t-statistics adjusted according to the Newey & West (1987) methodology.
The interaction terms in the three regressions presented in Table 6 confirm my hypothesis that the predictive strength of TK is particularly robust for stocks characterized by low market capitalizations, low liquidity, and high idiosyncratic volatility.
Mechanism
This section explores the predictive power of two components of salience theory: salience-weighted (SW) and equal-weighted (EW) effects Salience-weighted values indicate that stocks with higher SW values are often overpriced due to salient thinking, resulting in lower future returns Conversely, the equal-weighted component, which reflects salient trading volume, suggests that investors concentrating on stocks with high EW values are likely to experience lower returns compared to those focusing on stocks with lower EW values.
I employ the Fama-MacBeth methodology to explore the salience components effect in the cross-section of predictability returns Monthly cross-section regression is run as follow:
The return of firm i in month t+1, denoted as r_it+1, is influenced by the salience component value vector Z_it and a set of standard control variables outlined in Section 2.2.3 These control variables include Size (ME), book-to-market ratio (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), an illiquidity measure (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (Coskew).
In column (1) in Table 7, the result provide support for my predictions that the
SW variable strong negatively correlate to future stock return with coefficient is -
The coefficient of SW remains economically and statistically significant, with a t-statistic of -4.36, even after controlling for firm size, market beta, book-to-market ratio, and momentum variables Columns (2) to (6) demonstrate that the inclusion of the acknowledge power control variable does not diminish the significant impact of SW on expected returns Furthermore, when accounting for skewness-related variables (Skew, Iskew, and Coskew), the SW coefficient continues to exhibit a large economic significance and a t-statistic of -5.66, surpassing the stringent hurdle of 3.0 proposed by Harvey et al (2016).
Table 7 Firm-level Fama-MacBeth regression
The table presents the mean slope coefficients derived from cross-sectional regression using the Fama-MacBeth methodology Stocks are categorized into decile portfolios based on their SW value at the end of each month In the comprehensive regression model outlined in column (10), the dependent variable Y_it incorporates factors such as size (ME), book-to-market (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), and an illiquidity measure.
The study analyzes various financial metrics, including illiquidity (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (coskew) The data spans from January 2010 to December 2021, with t-statistics adjusted according to the methodology of Newey & West (1987).
Table 8 Firm-level Fama-MacBeth regression
The table presents the mean slope coefficients derived from cross-sectional regression using the Fama-MacBeth methodology Each month, stocks are organized into decile portfolios based on equal-weighted (EW) values In the comprehensive regression model outlined in column (10), the dependent variable \(Y_{it}\) incorporates factors such as size (ME), book-to-market ratio (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), an illiquidity measure (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (coskew) The analysis covers the period from January 2010 to December 2021, with t-statistics adjusted according to Newey & West (1987).
The findings in Table 8 support the hypothesis that stocks with high salient trading volume attract concentrated investor interest, resulting in lower expected returns Specifically, the results indicate a negative correlation between salient trading volume and expected returns, with the equal-weighted (EW) model predicting a decrease of 0.97 basis points in one-month future returns at a 99% confidence level Furthermore, the EW coefficient remains economically significant and statistically robust at the 1% level, evidenced by a t-statistic of -7.46, even after controlling for various asset pricing variables such as Beta, Market Equity (ME), Book-to-Market (BM), Momentum (MOM), Short-term Volatility (SVR), Illiquidity (Illiq), and Idiosyncratic Volatility (Ivol) Additional analysis in columns (7) to (9, which incorporates skewness as a control variable, reveals minimal impact on the predictability of stock returns based on salient trading volume Ultimately, the final column demonstrates that a two standard deviation increase in salient trading volume correlates with a decrease of 0.96 basis points in one-month future stock returns when all 14 control variables are included.
Before conclude the predictive ability of salience theory effects and its components, I will conduct a robustness checks on different value of the parameter
In the previous section, the parameters used to measure the ST value were based on the experimental findings of Bordalo et al (2012), specifically with values of θ = 0.1 and δ = 0.7 I first adjusted the δ parameter to 0.6 and 0.8 while maintaining θ at 0.1 In the third and fourth columns of Table 9, I kept δ fixed at 0.7 and varied θ to 0.05 and 0.15, respectively The last four columns of the table present these variations clearly.
Table 9, Panel A demonstrates that the predictive power of the ST value remains both economically and statistically significant despite variations in the parameters and However, the findings indicate that altering both parameters leads to a notable decrease in the salience theory's impact on future return predictability Specifically, the coefficients for ST in columns 7 and 8 are -0.007 and -0.004, accompanied by t-statistics of -1.69 and -1.04, respectively.
Table 9A Fama-MacBeth regressions that vary the degree of probability weighting
The table examines the effects of salience theory by adjusting various parameters using the Fama-MacBeth methodology, presenting the average slope coefficients from cross-sectional regression At the end of each month, stocks are categorized into decile portfolios based on their ST values In the most comprehensive regression model (column 10), the dependent variable \(Y_{it}\) incorporates factors such as size (ME), book-to-market ratio (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), an illiquidity measure (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (coskew) The analysis covers a sample period from January 2010 to December 2021, with t-statistics adjusted according to Newey & West (1987).
A Fama-MacBeth regressions without skewness control
Table 10B Fama-MacBeth regressions that vary the degree of probability weighting
The table examines the effects of salience theory by varying parameter degrees, utilizing the Fama-MacBeth methodology to derive average slope coefficients from cross-sectional regression At the end of each month, stocks are categorized into decile portfolios based on their ST values The most comprehensive regression specification, presented in column (10), incorporates variables such as size (ME), book-to-market (BM), market beta (Beta), momentum (MOM), short-term reversal (SVR), an illiquidity measure (Illiq), idiosyncratic volatility (Ivol), downside beta (DBeta), maximum and minimum daily returns (Max/Min), skewness (Skew), idiosyncratic skewness (Iskew), and coskewness (coskew) The analysis covers a sample period from January 2010 to December 2021, with t-statistics adjusted according to Newey & West (1987).
B Fama-MacBeth regressions with skewness control
Panel B of Table 9 indicates that the most statistically significant variables in salience theory correspond to a value of δ below 0.7, suggesting an overweighting of salient states Notably, the t-statistics decreased significantly from -4.06 to -0.99 when δ was set at 0.8, alongside a model that included additional skewness-related variables.
In conclusion, the predictive power of two elements of salience theory is both economically substantial and statistically significant, even when accounting for a subset of power predictors in asset pricing Additionally, while varying the degree of probability weighting does not alter the influence of the salience effect on future returns, it does reveal that as the value increases, the predictability of the salience effect diminishes This suggests that as investors become more rational, the salience effect on stock return predictability in asset pricing decreases.
Conclusion
This article explores the predictive power of salience anomaly on stock returns in the Vietnam stock market, revealing a statistically and economically significant relationship in cross-sectional stock returns Univariate portfolio analyses indicate that stocks with the lowest past returns, which are salient, tend to earn lower returns in the following month compared to those whose lowest past returns are more pronounced This phenomenon may be attributed to investors being drawn to value stocks' downturns, resulting in growth stocks being undervalued and offering high future returns during challenging market conditions The disparity between high and low past returns is both economically and statistically significant, primarily focusing on equal-weighted portfolios.
A firm-level Fama-MacBeth regression indicates that the short-term (ST) effect on stocks is negatively correlated with expected stock returns, remaining largely unaffected even after controlling for various asset pricing predictors such as beta, size, book-to-market, and momentum Notably, the predictive power of ST effects is minimally impacted by these controls, and the magnitude of the ST coefficient remains unchanged even when accounting for variables related to lottery demand, including implied volatility, skewness, and coskewness This analysis aligns with a two-step framework for understanding the psychology of tail events, where individuals first assess the probability of such events and then make decisions based on that probability judgment.
33 preferences Therefore, the ST effect (related-psychology variables) remain unaffected and statistically significant after including skewness variables which provide the evidence that ST effects is new asset pricing factor
The findings of this study support a behavioral interpretation, revealing that the predictive power of salience theory components—salience distortion and trading salient volume—negatively correlates with future expected returns Notably, the salience theory effect retains its strength and significance even when adjusting the weighted probability However, under the assumption of fully rational investors, the salience effect diminishes as the parameter δ approaches 1, resulting in reduced predictability of future returns.
While my findings support the salience theory effect, they are limited to the securities market Future research should explore other asset classes, including options, bonds, and gold, to provide additional evidence for the implications of salience theory on asset pricing.
In his 1953 work, Allais critiques the assumptions and axioms of the American school regarding rational behavior in the face of risk, highlighting the complexities of decision-making under uncertainty Meanwhile, Amihud's 2002 study explores the relationship between illiquidity and stock returns, examining both cross-sectional and time-series effects, which underscores the importance of liquidity in financial markets.
Journal of Financial Markets, 5(1), pp 31-56
Ang, A., Chen, J & Xing, Y., 2006 Downside Risk The Review of Financial Studies, 19(4), p 1191–1239
In their 2011 study published in the Journal of Financial Economics, Bali, Cakici, and Whitelaw explore the relationship between stocks and lottery-like behavior, examining how this affects the cross-section of expected returns Barberis (2013) contributes to the discourse by addressing the psychological aspects of tail events, highlighting both advancements and ongoing challenges in understanding investor behavior.
Barberis and Huang (2008) explore the concept of stocks functioning as lotteries, emphasizing how probability weighting affects security prices Their findings suggest that investors perceive stock returns through a lens similar to gambling, which can lead to mispricing in the market In a subsequent study, Barberis, Mukherjee, and Wang (2016) empirically test the implications of Prospect Theory on stock returns, revealing that investor behavior and risk perception significantly influence market outcomes Together, these studies highlight the psychological factors driving investment decisions and their impact on financial markets.
Bordalo, P., Gennaioli, N & Shleifer, A., 2012 Salience Theory of Choice Under Risk
The Quarterly Journal of Economics, 127(3), p 1243–1285
Bordalo, P., Gennaioli, N & Shleifer, A., 2013 Salience and Asset Prices American Economic Review, 103(3), pp 623-628
Boyer, B., Mitton, T & Vorkink, K., 2009 Expected Idiosyncratic Skewness The Review of Financial Studies, 23(1), p 169–202
Cakici, N & Zaremba, A., 2021 Salience theory and the cross-section of stock returns: International and further evidence Journal of Financial Economics
Carhart, M M., 1997 On Persistence in Mutual Fund Performance The Journal of Finance, 52(1), pp 57-82
Cosemans, M & Frehen, R., 2021 Salience theory and stock prices: Empirical evidence
Journal of Financial Economics, 140(2), pp 460-483
Cox, J C & Epstein, S., 1989 Preference Reversals Without the Independence Axiom
Craig, R & Diga, J., 2002 Corporate Accounting Disclosure in ASEAN Journal of International Financial Management & Accounting, 9(3), pp 246-274
Fama, E F., 1970 Efficient Capital Markets: A Review of Theory and Empirical Work
The Journal of Finance, 25(2), pp 383-417
Fama, E F & French, K R., 1992 The Cross-Section of Expected Stock Returns The
Fama, E F & French, K R., 1993 Common risk factors in the returns on stocks and bonds Journal of Financial Economics, 33(1), pp 3-56
Fama, E F & French, K R., 1996 Multifactor Explanations of Asset Pricing Anomalies The Journal of Finance, 51(1), pp 55-84
Fama, E F & French, K R., 2015 A five-factor asset pricing model Journal of Financial Economics, 116(1), pp 1-22
Fama, E F & MacBeth, J D., 1973 Risk, Return, and Equilibrium: Empirical Tests
Journal of Political Economy, 81(3), pp 607-636
Greenwood, R., 2014 Expectations of Returns and Expected Returns The Review of Financial Studies, 27(3), p 714–746
Harvey, C R., Liu, Y & Zhu, H., 2016 … and the Cross-Section of Expected Returns
The Review of Financial Studies, 29(1), pp 5-68
Harvey, C R & Siddique, A., 2002 Conditional Skewness in Asset Pricing Tests The
Jegadeesh, N., 1990 Evidence of Predictable Behavior of Security Returns The Journal of Finance, 45(3), pp 881-898
Kahneman, D & Tversky, A., 1979 Prospect Theory: An Analysis of Decision under Risk Econometrica, 47(2), pp 263-292
Lehmann, B N., 1990 Fads, Martingales, and Market Efficiency The Quarterly Journal of Economics, 105(1), pp 1-28
Newey, W K & West, K D., 1987 A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix Econometrica,
Nguyen, T G., Nguyen, L & Nguyen, T., 2021 Capital Structure and Its Determinants: Evidence from Vietnam The Journal of Asian Finance, Economics and Business, 8(10), pp 1-10
Rizvi, S A R & Arshad, S., 2014 Investigating the efficiency of East Asian stock markets through booms and busts Pacific Science Review, 16(4), pp 275-279
Subrahmanyam, A., 2005 Distinguishing Between Rationales for Short-Horizon Predictability of Stock Returns The Financial Review, 40(1), pp 11-35
Appendix A Fama-French Factor Model
To control for the multidimensional structure of equity returns, I estimate the portfolio using Fama-French factor model, as follow: t MKT t SMB t HML t WML t RMW t CMA t t
R = + MKT + SMB + HML + WML + RMW + CMA + where R t indicates the excess return of portfolio in month t, t denotes the error term,
The slope coefficients in financial analysis include several key factors: MKT t represents the market risk factor, while SMB t indicates the return difference between small and large stock portfolios HML t reflects the return disparity between high and low book-to-market stocks, and WML t measures the return difference between winning and losing stocks Additionally, RMW t captures the return variation between stocks with strong and weak profitability, and CMA t highlights the return difference between low and high investment firms.
Sort Breakpoints Factors and their components
Size and Mom, or Size and Inv
Size: HSX median SMB B M / = ( SH + SN + SL ) / 3 ( − BH + BN + BL ) / 3
SMB OP = SR + SN + SW − BR + BN + BW
SMB Inv = SC + SN + SA − BC + BN + BA
B/M: 30 th and 70 th HSX percentiles
HML = SH + BH − SL + BL
OP: 30 th and 70 th HSX percentiles
Mom: 30 th and 70 th HSX percentiles
W ML = SH + BH − SL + BL
Inv: 30 th and 70 th HSX percentiles
CMA = SC + BC − SA + BA
ST Salience theory value of a stock’s historical return distributions It is calculated by Eqs (8)-(10)
ME The natural logarithm of the firm’s market capitalization on
December of the previous year
BM The natural logarithm of the firm’s book-to-market ratio, where the book value and market value are of June of the previous year
The illiquidity measure is calculated as the average ratio of the absolute daily return to the volume over the month of t-1, by following Amihud (2002)
MOM A control for momentum It is calculated as the cumulative return from month t-12 to t-2
SVR Short-term reversal measure, which is the lagged return Datastream
Max Stock’s maximum one-day return in month t-1 Datastream
Min Stock’s minimum one-day return in month t-1 Datastream
Iskew Idiosyncratic skewness (ISKEW) is computed following Fama &
French (1993), as in Boyer et al (2009)
Skew The skewness of stock return over one – year window Datastream
Coskew Co-skewness is calculated by following Harvey & Siddique’s
A stock’s downside beta computed following Ang et al (2006) using daily excess return over a the past 12-month returns on market return that below average market return over one – year window
Beta The beta of stock return over one – year window Datastream
Ivol Idiosyncratic volatility, which is calculated as the volatility of the stock’s daily idiosyncratic returns over month t-1
NHẬN XÉT CỦA GIẢNG VIÊN HƯỚNG DẪN
(Ký và ghi rõ họ tên)