Banking Academy of Vietnam Faculty of Finance ----------------------- GRADUATION THESIS TOPIC: Salience Theory and The Cross-section of Stock Returns in Emerging Market: Empirical evidence in Vietnam Student name: Gia Tan NGUYEN Class: K21CLCC Student ID: 21A4010507 Instructor: Dr. Ngoc Mai TRAN Hanoi, May 2022 17014128986491000000 Banking Academy of Vietnam Faculty of Finance ----------------------- GRADUATION THESIS TOPIC: Salience Theory and The Cross-section of Stock Returns in Emerging Market: Empirical evidence in Vietnam Student name: Gia Tan NGUYEN Class: K21CLCC Student ID: 21A4010507 Instructor: Dr. Ngoc Mai TRAN Hanoi, May 2022 Declaration I certify that the thesis I have presented for examination for the Bachelor degree of the Banking Academy of Vietnam is solely my own work other than where I have clearly indicated that it is the work of others (in which case the extent of any work carried out jointly by me and any others is clearly identified in it). The copyright of this thesis rests with the author.
Quotation from it is permitted provided that full acknowledgement is made. This thesis may not be reproduced without my prior written consent. I warrant that this authorization does not, to the best of my belief, infringe the rights of any third party. I declare that my thesis consists of 8684 words.
Signature (full name) Nguyen Gia Tan i Acknowledgements Through the writing of this thesis I have received a great deal of support and assistance. I am deeply indebted to my advisor, Dr Tran Ngoc Mai, for her continuous guidance and encouragement. Secondly, I am very thankful for my cat – his name is “Meo”. He is like my little brother who helps me to reduce stress, anxiety, and boredom during this period.
I am very happy to include his contribution here. Last but not least, I am extremely grateful to my mother, my father. I would not be submitting my thesis without their limitless support and reassurance, especially during multiple lockdowns. ii Abstract This article aims to test the prediction of salience theory effect that stocks with high salience effect should have low future return in “good” times.
Consistent with recent empirical evidence, I find that salience theory effect is negatively correlated to future returns. Specifically, the difference of six-factor alpha between lowest ST value and highest ST value is 1. In addition, the coefficient on ST effect in Fama-MacBeth cross-sectional regressions are negative and significant, even after controlling a list of acknowledgement predictor in asset pricing. (JEL D03, G11, G12) iii Table of Contents Declaration.
iii List of Tables. The implications of salience theory and asset pricing. Salience-based asset pricing. Construction of salience theory measure.
Cross-sectional relation between salience and stock returns. Time-series test. Robustness checks of time-series. Fama-MacBeth Firm-level regressions.
Impact of limit to arbitrage. Fama-French Factor Model .39 iv List of Tables Table 1. Decile portfolio analysis. Characteristics of ST-sorted portfolio.
Robustness check of time-series. Firm-level Fama-MacBeth regression. Fama-MacBeth regression: limits to arbitrage. Firm-level Fama-MacBeth regression.
Firm-level Fama-MacBeth regression. Fama-MacBeth regressions that vary the degree of probability weighting. Fama-MacBeth regressions that vary the degree of probability weighting. Introduction A crucial assumption in any model of traditional asset pricing theory is that investors always rationally use all available information to evaluate risky assets (Fama, 1970).
Expected utility theory and the standard theory of choice under risk, based on this assumption, have been most used by social scientists to reliably predict the cross-section of stock returns. However, a large body of research shows that the decision attitudes to risk can depart from the independence axiom of expected utility (Kahneman & Tversky, 1979; Barberis & Huang, 2008). In an effort to develop non- expected utility models, Bordalo et al. (2012) argue these cognitive biases, then propose new novel theory –“ salience theory”, in which decision marker's attitudes are drawn by the unusual or “salient” attributes of different situations.
Consequently, decision-makers local thinkers overemphasize the salient attributes, and neglect non- salient payoffs. There is substantial empirical evidence showing that the salience theory has directly derived the predictions, for the cross-section of stock return in the U. In this study, we examine whether the salience effect also exists in the Vietnam stock market. Compared to the U.S market, or developed market, the Vietnam market has its salient features in information disclosure (Craig & Diga, 2002), market efficiency (Rizvi & Arshad, 2014), and institutional arrangement (Nguyen et al.
Our goal is to see if salience models, which account for a variety of decision- making theory puzzles, such as Allais paradox, can also help us understand decision makers' attitudes in the Vietnam market. Moreover, the Vietnam stock market has been becoming more accessible to foreign investors. Examining the salience effect in the cross-section of predictability returns in this market is therefore of growing interests to both academic and investor professionals. The empirical results provide strong support for the predictive power of salience theory effect in the cross-sectional of Vietnam stocks.
First, time-series analyses show that stock with salient upside earn higher return over the next month than salient downside. A difference between high portfolio and low portfolio is economically and statistically significant over the sample period 2010-2021. After 1 adjusting for risk, the difference in return are still pronounced, with six-factor alphas of 1. By employing Fama-MacBeth regression, the result confirms stronger cross-sectional relation between ST and expected return.
The effect of ST is even economically and statistically significant, after controlling additional power predictor. This finding is consistent with Cosemans & Frehen (2021) who finds a negative relationship between ST effect and future returns. In addition, I further study the effect of salience components on the cross- section of capital market, namely salience-weight (SW) and equal-weight (EW) past returns, I explore that salience distortation and salient trading volume is larger economically and statistically significant. Through Fama-MacBeth analsys, the result provide the evidence that both salience distortation and salient trading volume negatively correlate to future expected return.
It is consistent with a hypothesis that stock with high level of salience distortation are lower expected return than stock with low level of salience distortation. To ensure the effect of salience theory, I also alter the degree of probability degree of salience. My salience theory measure remains larger economically and statistically significant after varying a degree of salience measurement, but it still has little impact on predict ability of salience theory on stock return. Specifically, the degree to which salience distorts decision weights, could reduce the effect of salience theory if this degree is near 1 ( → 1 ).
It implies that the more we assume the investors are rational, the less predictability of salience effect on asset pricing theory. The study is organized as follow. Section 2 provides the decision-making puzzles and the implications salience theory on asset pricing. Section 3 report the descriptive summary and dataset.
Section 4 reports time-series analysis and firm-level cross-sectional regressions. Section 5 focuses more mechanism of salience theory effect. Salience theory In Bordalo et al.’s (2012) model (henceforth BGS), the local thinkers faced with a choice set that consists of two lotteries L1 and L2. The lotteries are defined over a state space S that contain N states.
Each state s S occurs with probability 2 s and lottery Li delivers payoffs xsi in state s. I assume the local thinker uses a linear value function v ( x ) = x , where lottery payoffs are evaluated to a reference point of zero. Without any salience distortions, the local thinkers evaluate Li as: V ( Li ) = s v ( xsi ). (1) sS The decision-maker departs from equation (1) by underweighting the lottery's least salient state in S.
There are two steps in salience distortion. First, the lottery payoff Li generate a salience ranking among the states in S. Second, based on this salience ranking, the probability in equation (1) is replaced with a transformed, lottery-specific decision weight s. Let xs , xs , respectively denote the smallest i min max and largest payoffs in xs , then Bordalo et al.
(2012) define the salience of state for each lottery is a continuous and bounded function that satisfies three conditions: 1. Ordering: If for states s , s S , and xsmin , xsmax is a subset of xsmin , xsmax , then ( xsi , xs−i ) ( xsi , xs−i ). Diminishing sensitivity: If xs 0 for any j = 1, 2 , then for any 0 , j ( xsi + , xs−i + ) ( xsi , xs−i ). Reflection: For any two states s , s S such that xs , xs 0 for j = 1, 2 , we j j have ( xsi , xs−i ) ( xsi , xs−i ) if and only if ( − xsi , − xs−i ) ( − xsi , − xs−i ).
To measure the salience of the payoff xsi of lottery i in state s , Bordalo et al. (2012) propose the function xsi − xs ( x , xs ) = i (2) xsi + xs + s where 0 and x = i xsi / N , with N denoting the number of lotteries. N 3 The salience effect in (2) must satisfy the above three conditions: (i) ordering, (ii) diminishing sensitivity, and (iii) reflection. If xsmin , xsmax is a subset of xsmin , xsmax , then ordering means that the ST effect of state s ' is larger than state s.
Diminishing sensitivity means that the significance increases as the absolute payout level of the lottery continues to decrease. Reflection means that the focus is on the magnitude of the gains rather than their sign. Reflecting gains in losses does not change the importance of countries because perceptions are 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 Li state s is more salient than s if ( xsi , xs− i ) ( xsi , xs−i ) .,| S |} be the salience ranking of state s for Li , i with lower k s indicating higher salience.
All states with the same salience obtain the s same ranking. Then the salient thinker transforms the odds of s relative to s into s si the odds , given by: si si = k −k s i i s s s i s where (0,1]. The decision weight attached by salient thinkers to a generic state s in the evaluation of Li is: si = s si where s denotes the salience-weighted subjective state probability, s is the i objective probability, and s is the salience weight defined as: i k i s = i , (0,1]. i The parameter in Eq.
(3) measures the degree of local thinking by capturing the degree of cognitive ability in which salience distorts decision weights and proxies for the decision-cognitive maker’s ability. If = 1 , the decision-cognitive maker is a standard economic decision-cognitive maker, and its decision weight is equal to the objective probability ( s = 1 for all s S ). If 1 , the decision-cognitive maker is i a great thinker who will overweight the most salient states ( s 1 ) and underestimate i the non-salient states ( s 1 ).