TFIN605 Data Analytics in Finance Spring 2020
1
Final Project Guidelines
Analysis of Corporate Dividends in Australia and Canada
Project Due Date: Saturday, 07 November, 2020, by 6:00 pm.
Report due via Turnitin.
Jupyter Notebook should be uploaded to the appropriate folder on Moodle.
For the final project you group will conduct analysis of the dividends of companies from the
following two countries: Australia and Canada.
Relevant financial data is available on Moodle in one excel file: dividend_au_ca.xlsx. This file is in the
‘Final_Project_Documents’ folder on Moodle. You will have to download the data, upload to your
Jupyter notebook and then conduct your analysis using Pandas and other Python Li
aries. You will
have to submit the Jupyter notebook you used to do the analysis for this project.
You will have to write up your analysis in a report of up to 2,000 words. Your report should also
include tables and graphs from your analysis. These tables and graphs have to be produced using
Python and you will have to submit all the relevant codes in a Jupyter notebook.
The objectives of your analysis are as follows:
ï‚· Document and discuss the distribution and trends in dividend payout ratio (dividend/net
income) and the number and percentage of dividend payers (positive dividend) over time in
each country:
o Use dividend to net income ratio as the measure of dividend payout ratio.
o Dividend payout ratio attempts to measure what percentage a firm’s earnings is paid out
in dividends.
o Dividend payout ratio is not a meaningful measure in the following two cases, so you
need to deal with these cases in the data pre-processing step:
1. When a firm has negative net income, the dividend payout ratio is not a meaningful
measure.
 So exclude observations (rows) with negative net income from your sample
2. When a firm has a dividend payout ratio higher than 1 (dividend is higher than net
income), the dividend payout ratio is not reliable as a firm cannot pay out more dividend
than net income in the long run.
 So cap the value of dividend payout ratio at XXXXXXXXXXset any value higher than 1.0
to 1.0.
o You will conduct the analysis for Australia and Canada and you will discuss how the
dividend payout ratios of the two countries compare with each other and if they show
similar or different trends over time.
TFIN605 Data Analytics in Finance Spring 2020
2
o You should perform similar analysis of dividend payers in the two countries. In two
separate graphs, you should show the number and percentage of dividend paying firms
in the two counties and how these have changed over time.
o You will document the distribution of dividend payout ratio in each country in 2007 and
2017 to see if the distribution has changed over time. You can use histograms, kernel
density plots and percentile plots to show the distributions.
ï‚· Analyse the determinants of dividend payout ratio in each country. So you will have two sets of
esults.
o Initially, explore the relations between various firm characteristics (such as firm size,
profitability, growth opportunity etc.) and dividend payout ratio using scatter plot.
o You will then conduct co
elation analysis to determine if there are significant
co
elations between these characteristics and dividend payout ratio.
o Then use simple linear regressions to quantify the relation between leverage and these
characteristics one at a time. Here you will use regressions with one independent
variable (see lecture 7).
o Finally you will use multiple linear regression analysis to consider the effects of all the
different firm characteristics on dividend payout ratio.
o You will compare and contrast the results you get from the above analysis for the two
countries in your sample: Australia and Canada.
ï‚· Finally, you should estimate two Machine Learning models and evaluate the predictive
performance of these models.
o The first model will try to predict the dividend payout ratio of a firm. You can use the
Boston House Price example as a template for this analysis and do similar analysis on
dividend payout ratio (instead of house price).
 As X (or independent) variables, use the four firm characteristics we used in the
group project: Firm size (Logsale), Profitability, Tangibility and Market to book
atio.
 The y variable or dependent variable in your model would be the dividend
payout ratio.
 You should to the train-test split and evaluate the model’s performance on the
test dataset and interpret the results.
o The second model will try to predict whether a firm pays dividends --- that is, whether
the dividend of a firm is positive.
 Create a variable in your dataframe called PAYER which should be 1 if a firm has
positive dividend (and therefore positive dividend payout ratio) and 0 otherwise.
This variable will be the categorical dependent variable in your supervised
classification model.
 Same as in the first model, as X (or independent) variables, use the four firm
characteristics we used in the group project: Firm size (Logsale), Profitability,
Tangibility and Market to book ratio.
 Use the K Nearest Neighbor model or KNN model for this analysis
 You can use Iris flower example (covered in lecture 8) as a template for this
analysis and do similar analysis on dividend PAYER (instead of Iris flower types).
 You should do the train-test split and evaluate the model’s performance on the
test dataset and interpret the results.
TFIN605 Data Analytics in Finance Spring 2020
3
I have posted two papers on Moodle for you to read for this assignment. These papers analyse
dividend payout ratio, but they do not use the same firm characteristics as independent variables
that your dataset has --- but these papers will help you understand the general research background
and how to interpret the results. As independent variables in your analysis, you should use the same
firm characteristics that we used in the leverage analysis (such as firm size (Logsales), Profitability
etc.).
You should also do additional research via google on the determinants of dividend payout ratio and
use those sources as references in your report.
You will summarise you main finding is a report of 2000 words. The report will:
1. Summarise the relevant literature (research papers) and research question.
2. Report and discuss descriptive data analysis and data visualisation.
3. Report and discuss co
elation and regression analysis
4. Report and discuss Machine Learning (ML) analysis of dividend payout ratio using the linear
egression model (LinearRegression: covered in lecture 11 in the Boston House Price example). Fit an
ML model to predict dividend payout ratio and evaluate the performance of the model.
5. Report and discuss Machine Learning (ML) analysis of dividend payers (if dividend payout ratio > 0
then dividend payer = 1, otherwise dividend payer = 0) using the K Nearest Neighbor model or KNN
model (covered in lecture 8). Fit an ML model to predict whether a firm pays dividend and evaluate
the performance of the model.
6. Draw inference and conclusion and relate the findings to existing research.
TFIN605 Data Analytics in Finance Spring 2020
4
Marks Distribution:
Marks will be distributed as follows:
Report write-up with graphs and tables (2000 words): 20%
80% of the marks will be based on the Juypyter Notebook and the following sections of the
notebook (you should comment the Notebook so it is easy to follow your work):
Data cleaning and pre-processing: 10%
Descriptive analysis and data visualisation: 25%
Co
elation and regression analysis 15%
Machine learning analysis
Linear regression 15%
K Nearest Neighbor 15%
Introduction:
Electronic copy available at: http:
ssrn.com/abstract=2642679
International Review of Business Research Papers
Vol. 10. No. 2. September 2014 Issue. Pp. 62 – 80
Impact of Firm Specific Factors on Cash Dividend Payment
Decisions: Evidence from Bangladesh
Md. Faruk Hossain*, Rashel Sheikh** and S.M. Akterujjaman***
This study aims to explore the impact of firm specific factors on cash dividend
payment decisions for a sample of 41 non financial firms listed in Dhaka Stock
Exchange (DSE) in Bangladesh during XXXXXXXXXXThis study tests a null
hypothesis that none of the firm specific factors namely profitability, size,
liquidity, growth, earnings volatility, and managerial ownership has significant
effects on cash dividend payments using fixed-effect regression model under
the assumption that intercepts vary for each firm and the slope coefficients are
constant across firms. Checking multicollinearity, cross-sectional dependence,
autoco
elation and controlling heteroskedasticity in the regression analysis it
is found that profitability has statistically significant positive effects on cash
dividend payments. This study has discovered a significant negative effect of
earnings volatility and managerial ownership on dividend payments which
were unfolded before this study. On the other hand, size, growth and liquidity
were not found to be significant explanatory variables of dividend payments.
Thus, profitability, earnings volatility and managerial ownership are functioning
as the key determinants of cash dividend payments in Bangladesh.
Keywords: Bangladesh; Dhaka Stock Exchange (DSE); Dividend Payments; Firm Specific
Factor; Listed Companies; Panel Data.
1. Introduction
Should a corporation pay dividends to common stockholders? Perhaps the answer of this
question mostly depends on the effects of dividend payments on share price of the firm that
ultimately yields a concern of dividend payment decisions. It implies payout policy, in which
managers decide the size and pattern of cash distribution to shareholders over time. The
presence of significant effect of dividend payments on share price has been raised by many
theoretical as well as empirical researches done by Lintner (1956), Gordon (1959), Pradhan
(2003), Ho (2003), Myers & Bacon (2004), Pani (2008), and Khan et al XXXXXXXXXXOn the
other hand, Miller and Modigliani XXXXXXXXXXare the first advocates of proving that dividends are
i
elevant and insignificant factor in maximizing firm‟s value under the assumptions of
perfect and efficient markets.
In addition to Miller and Modigliani (1961), insignificant influence of the dividend on equity
share price was also found in the study of Black and Scholes (1974), Uddin & Chowdhury
(2005),