10 An Extension of Arnott’s and Asness’s Research
Tải bản đầy đủ
17
The main variables of interest were the monthly values of dividend yield, earnings
yield, a retail price index, a stock market index level and the payout ratio. The sample
period differs by country because the availability of data differs by index. Descriptive
statistics show that the mean payout ratio in the U.K. is the highest one with a
percentage of 53% (payout ratio is 0.53), during the period from 1973 till 2004. The
Netherlands have a payout ratio of 0.48 for the same period of time. They have
concluded that there exists a positive and mostly significant relation between future
earnings growth and payout ratios.
In other words, Gwilym et al. (2006) have concluded that “substantial reinvestment of
retained earnings does not lead to faster future earnings growth, although it does lead to
faster real dividend growth (Gwilym et al., 2006)”. For the U.K., the U.S., France and
Japan the adjusted R2 was reasonably high. Unfortunately for the other countries,
including the Netherlands, this R2 was low. In the end, Gwilym et al. (2006) have not
proved a significant positive relationship between the payout ratio and the returns.
2.11 Hypotheses Development
In summary, some researchers have concluded that a negative relationship exists
between dividend payouts and future earnings (growth). Other researchers found a
positive relationship, for example Nissim & Ziv (2001) and Arnott & Asness (2003).
There are any numbers of possible explanations for these opposite relationships. For
example, differences in sample, sample size, firm cultures, branches and sectors could
result in different relationships. Further research is needed to find and explain some
possible explanations for these opposite relationships. Now, this research focuses on
two specific hypotheses.
18
2.11.1 Hypothesis 1: The Payout Ratio is positively correlated to the Expected
Future Earnings Growth listed firms on HOSE and HNX
Concluding from the literature review, this research will focus on the influence of
payout ratios on expected future earnings growth. Obviously, there are much more
determinants for the level of future earnings growth. However, during this research I
have focused on the most important variables related to future earnings growth. This
research has used the basis model developed by Arnott and Asness (2003), as
represented in equation (2.4) in this research. Arnott and Asness (2003) developed first
this basic model. Afterwards, they optimize the model by doing some robustness tests.
This research follows the method used in the research of Arnott and Asness (2003). I
am curious whether the positive relation between dividend distribution and future
earnings growth also exists in the Vietnamese market stock. Therefore, the main
hypothesis during this research is The payout ratio is positively correlated to the
expected future earnings growth of listed firms on HOSE and HNX
2.11.2 Hypothesis 2: The Dividend yield is positively correlated to expected future
earnings growth of listed firms on the HOSE and HNX
In addition to the first hypothesis, this research studies the following hypothesis The
dividend yield is positively correlated to the expected future earnings growth of listed
firms in HOSE and HNX. The dividend yield expresses the dividend per share as a
percentage of the share price. It is expected that dividend yield is a substitute variable
instead of the payout ratio to measure dividend distribution. These two variables have
the same numerator but a different denominator. Additionally, the payout ratio is
scaled with earnings and the dividend yield is scaled with the share price. In this
research, I expect a positive relation between payout ratio and expected earnings
19
growth. For this reason I also expect a positive relation between dividend yield and
expected earnings growth.
20
Chapter 3: Data & Methodology
This chapter discusses the data and methodology for this research.
3.1 Introduction
Before discussing in details, we present here some analysis methods.

Time series data analysis: a time series is a sequence of data points,
measured typically at successive time instants spaced at uniform time
intervals. Examples of time series are the daily closing value of the Dow
Jones index or the annual flow volume of the Nile River at Aswan.

Panel data analysis: Panel (data) analysis is a statistical method. The data are
usually collected over time and over the same individuals. A common panel
data regression model looks like
, where y is the
dependent variable, x is the independent variable, a and b are coefficients, i
and t are indices for individuals and time. The error
is very important in
this analysis.

Cross sectional data analysis: is a type of regression model in which the
explained and explanatory variables are associated with one period or point
in time. This is in contrast to a timeseries regression or longitudinal
regression in which the variables are considered to be associated with a
sequence of points in time.
3.2 Panel data analyses
To analyze hypothesis 1, 2 The payout ratio or dividend yield is positively related to
the expected future earnings growth of listed firms on HOSE and HNX; this research
makes use of Panel data analyses.
21
3.2.1 Sample construction
This research focuses on firms of HOSE and HNX which have enough data from 2006
to 2011. From these sources, HOSE includes 39 firms. HNX includes 39 firms, too.
3.2.2 Variable description and model building
Data
for
research
comes
from
the
website
http://www.cafef.vn
and
http://www.vietstock.com . The dividend yield mean represents an average number of
the individual yields of the firms on HOSE and HNX. And, the payout ratio mean is an
earningsweighted average of all individual payout ratios. See Appendix I for detail.
In the first set of regressions, the PR is the independent variable and the expected
earnings growth (EEG) is the dependent variable. This study focuses on different EEG
levels, successively the 1, 2, and 4 years annualized growth (EEG1YR, EEG2YR and
EEG4YR). Arnott and Asness (2003) use the same independent and dependent variable
in their first models. The first simple model is presented in Equation (3.1).
EEGi,t years = α + β(PRi,t)+ ui,t (3.1)
Where, t = 1, 2 or 4 years
The different levels of ‘annualized’ expected earnings growth are calculated by the
change in EPS for t years, dividing by t years. For example, EEG4YRt= (EPSt+4/EPSt 1)/4. The same method is used to calculate EEG1YR and EEG2YR
3.2.3 Methodology for panel data analyses
This research makes use of simple linear regression analyses. This method makes it
possible to estimate the relationship between a dependent variable (EEG) and an
independent variable (PR or DY). This research develops a panel data model. This
22
panel data model explains the relationship between the yearly aggregate payout ratio
and the yearly aggregate future earnings growth of listed firms on HOSE and HNX. In
summary, in these models one focuses not on the firm specific information but focuses
on the average values of the firms. A general form of a simple linear model can be
expressed as follows:
Yi,t+k = α + βXi,t+ ui,t+k(3.2)
Where, t = 1, 2, ….., N and k = 1, 2, or 4 years
The ordinary least squares method (OLS) is the used point estimator. The method of
least squares is a standard approach to the approximate solution of over determined
systems, i.e., sets of equations in which there are more equations than unknowns.
"Least squares" means that the overall solution minimizes the sum of the squares of the
errors made in the results of every single equation.
There are a few assumptions of the OLS regression. Firstly, error terms are statistically
independent of other independent variables. In addition, the expected value of the
errors is always zero. The independent variables are not too strongly collinear.
Secondly, all error terms together have an expected value of zero. Furthermore, an
assumption related to the error term is its constant variance. The error terms are
normally distributed. The model is linear in parameters.
The least squares criterion can be described as:
Minimize ∑ ȗi2 = ∑ (Yi – Ŷi)2
23
3.2.4 Advantages of Panel Data
Some of the advantages of panel data analyses are used for my average model over
time. Brüderl (2005) mentions a few advantages of using panel data. Below, the
advantages that applied to my average models are presented. Firstly, panel data is more
informative because the data is more variable and there is less collinearity. Secondly,
with panel data one can study individual dynamics of the different firms in more detail.
Furthermore, with panel data one can control for individual unobserved heterogeneity
24
Chapter 4: Results
The results are presented in this chapter. The first part of this chapter presents the
regression results of the payout ratio and expected earnings growth analyses. The
second part of the chapter describes the dividend yield and expected earnings growth.
In the end, describes the dividend yield and payout ratio with expected earnings
growth.
4.1 The regression results of the payout ratio and expected earnings growth
analyses.
4.1.1 Descriptive statistics
Table 4.1 presents the descriptive statistics. This table includes the mean, median,
standard deviation, minimum and maximum for the independent and dependent
variables of the model. The firms in HOSE presents higher average payout ratio than
the firms in HNX; an average payout ratio of 42.27% with a corresponding average 4
years annualized earnings growth of 108.40%. The firms index of HNX have average
payout ratio of 28.54% with a corresponding average 4 years annualized earnings
growth of 166%. One possible explanation is that the HNX includes smaller firms,
which are growing faster compared to the firms of the HOSE. For this reason, firms in
the HNX have a higher retention and payout less dividends. These firms need their
money to finance future growth. Notice, this reasoning supports the life cycle theory as
mentioned in chapter 2 of this research. If the firm ends up in a more mature stage of
their cycle, the firm pays more dividends. Indeed, the HOSE firms are in more mature
stage of their life cycle compared to the HNX firms. The highest payout ratio paid by
indices of this research is the payout ratio of 42.27%. Focusing on the dependent
variable, expected earnings growth, one sees that the EEG increases over time for all
25
indices. In other words, for all indices the EEG4YR is a larger percentage compared to
the EEG1YRL. Therefore, the HOSE and HNX firms increase their future profitability,
which is expressed in expected earnings growth.
Table 4.1 Descriptive Statistics
PR
EEG1YR
EEG2YR
EEG4YR
Mean
43.45%
14.81%
71.80%
108.40%
Median
43.07%
8.508%
69.22%
110.66%
St. Deviation
25.33%
70.14%
78.21%
77.72%
Minimum
0.00%
291.48%
260.58%
4.16%
Maximum
143.57%
292.91%
280.48%
274.14%
N
229
182
140
68
Mean
28.54%
23%
76.84%
166%
Median
43.07%
13.64%
76.35%
133.95%
St. Deviation
25.92%
59.48%
68.00%
85.65%
Minimum
0.00%
96.72%
32.74%
11.29%
Maximum
148.22%
325.03%
336.00%
385.25%
N
233
191
150
69
Panel A: HOSE
Panel B: HNX
In this table the two indices are presented. The testing periods of per index is from 2006 to 2012. The
independent variable: The payout ratio which is calculated by dividing DPSi,t by EPSi,t. The different levels of
“annualized” expected earnings growth are calculated by the change in EPS for t years, dividing by t years. For
example, EEG4YRi,t = (EPSi,t+4/EPSi,t 1)/4. The same method is used to calculate EEG1YRi,t and EEG2YRi,t.
For all indices the average annualized expected earnings growth increases with the
number of prior years that are taken into account. So, the expected earnings growth for
one year (EEG1YR) on the HOSE and HNX is smaller than the five years annualized
expected earnings growth (EEG5YR). In this research, the sample size differs with the
26
level of annualized earnings growth, the sample size decreases if the level of earnings
growth increases.
4.1.2 Panel data regressions
This research run some simple linear regression models, based on equation (3.2)
EEGi,t+k = α + β (PR)i,t + ui,t+k
(4.1)
The three different levels of annualized earnings growth are used in the models. For
this reason, table 5.2 presents three horizon (k=1, 2, or 4) estimations per index,
displayed in Panel A, B and C. This table shows that in all models the payout ratio is
significant and positively related to the expected annualized earnings growth. Arnott
and Asness (2003) found the same positive relationship. Notice, Arnott and Asness
(2003) investigate an older and much longer time period, namely from 1871 till 2001.
In these models, the independent variables are all significant at a 0.001 and 0.05 level.
There 06 figures plotted the relation between the average payout ratio of all two indices
and the annualized 1, 2 and 4 years expected earnings growth of the two indices. In
addition, the lines in figures are regression lines of these observations. The values are
measured in the time period from 2006 to 2011. This is an increasing line which means
there is a positive relationship between PR and EEG1YR, EEG2YR or EEG4YR.
From table 4.2, a notable result is the difference between the HOSE listedfirms and
HNX listedfirms. In HNX, the payout ratio influences the expected annualized
earnings growth less. In other words, the coefficients for the HNX listedfirms models
are lower.
27
Table 4.2 panel data analyses between EEG and PR
Panel A: HOSE
Mode I – EEG1YR
Model II – EEG2YR
Model III – EEG4YR
Intercept (α)
35.865
25.941
84.928
**
**
1.974*
β(PR)
1.318
Ttest
4.350
3.851
2.084
R2
0.097
0.097
0.059
N
178
140
71
Panel B: HNX
Mode IV – EEG1YR
Model V – EEG2YR
Model VI– EEG4YR
Intercept (α)
3.075
65.326
115.615
β(PR)
0.386*
0.481*
0.737*
Ttest
2.255
2.092
2.029
R2
0.027
0.029
0.059
N
187
147
68
1.546
** *
, Significant at a 0.001; 0.05 level
Figure 4.1: Scatter plot of the average PR (Xaxis) versus EEG1YR (Yaxis) for firms on HOSE