Impact of Price Indexes on Stock Market Prices of Banks in Financial
Crises
NURSEL SELVER RUZGAR
Ted Rogers School of Management,
Toronto Metropolitan University,
350 Victoria Street, Toronto, ON M5B 2K3,
CANADA
Abstract: - During times of crises, stock markets often experience heightened volatility, and the banking sector
is particularly susceptible. This study aims to investigate the impact of index values on the daily closing prices
of five banks during five major financial crises in recent decades, using logistic regression analyses. The results
show that in five crisis periods, different indexes have a significant impact on the daily stock price of banks.
Although there is no pattern found for different crisis periods because each bank has different investment
instruments, the index, ind38- CFMRC (VWI) Over $2, seems to have a highly significant impact on the crisis
periods I-IV and ind37- CFMRC (DEWI) Over $2 plays a significant role in predicting the outcomes. The
findings indicate that banks should give particular focus to their investment instruments, particularly value-
weighted indexes (VWI) over $2 and equal-weighted indexes (DEWI) over $2 when any indications of a crisis
arise. This is crucial because these index values influence the daily closing prices of banks and could
potentially contribute to economic crises. Moreover, larger banks are more sensitive to changes in the index
values than smaller banks, attributed to variations in their investment amounts.
Key-Words: - Financial crises, Daily closing prices of banks, Indexes, Logistic regression, Value weighted
index over $2, Equal weighted index over $2.
Received: April 29, 2023. Revised: October 9, 2023. Accepted: October 21, 2023. Published: November 3, 2023.
1 Introduction
Economic crises drastically affect everything in
socio-economic life. During crises, the
unemployment rate, prices, bankruptcies, and
insolvencies increase, whereas GDP, production,
and purchasing power decrease. The stock market
also experiences significant volatility in crisis
periods, [1]. Index values are considered critical
indicators of the stock market's performance and can
have a significant impact on individual stock prices,
including banks, [2], [3]. Banks are integral players
in the financial system and their performance is
closely tied to the health of the broader economy.
During the crisis periods, the stock market indexes
fluctuate, going up and down, these lead to
significant drops in market capitalization and stock
prices and increases in volatility and risk, often
resulting in large losses for investors. Since the
banks are seen as a fundamental stone of the overall
health of the economy, banks can be heavily
impacted by the crises. Therefore, it is essential to
understand the relationship between the daily
closing price of banks and index values and also
how the crises have an impact on the closing price
of banks. This can help investors and financial
institutions make more informed decisions about
buying or selling bank stocks during crisis periods.
Stock prices are influenced by many economic
factors, such as investors' psychology and
expectations, the macroeconomic conditions of a
country, the movement of other stock markets,
political events, etc., [3]. In the stock market,
investment instruments in portfolios, such as stocks,
bonds, mutual funds, and other derivative
instruments are the main factors that affect the daily
closing price of banks, [1]. In the stock market,
indexes are the main indicators and they have a
fundamental role in the daily stock prices of banks.
Stock prices change moment by moment in response
to any kind of activity, such as economic factors,
industry performance, news, and investor
sentiments, [4]. “The price of a stock is largely
determined by supply and demand.” With high
demand, the stock price tends to up, and with high
supply, the price tends to go down, [4], [5].
The impact of index values on the daily stock
prices of banks is a fundamental area of
investigation for investors and financial institutions.
This study aims to provide valuable insights for
banks and investors regarding their investment
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2436
Volume 20, 2023
strategies during crisis periods and the impact of
various indexes on their investment portfolios. For
this purpose, this study examines the effect of index
values on the daily closing prices of five major
banks in Canada during five different crisis periods,
using logistic regression (LR) analysis. By utilizing
LR, the study provides valuable information on the
fluctuation of stock prices of banks during the
crises.
The remainder of this paper is organized as
follows: Section 2 presents a literature review.
Section 3 presents the aim and methodology.
Section 4 discusses findings and finally, Section 5
provides conclusions and suggestions for further
studies.
2 Literature Review
The stock market is one of the most important
components of a country’s economy because it
contains all economy-related sectors and
automatically reflects their impacts on the economy.
During crises, the performance of the stock market
becomes even more important as it reflects the
impact of the crisis on the economy. Stock market
indices are an important instrument for investors to
understand the overall performance of a particular
sector or the entire stock market, [6], indicating that
“stock investment provides benefits in the form of
dividends as a share of company profits and in the
form of capital gains, namely the difference between
the selling price of the shares and the buying price
of the shares.” Investors want to gain more. To do
this they monitor the stock prices and changes in the
prices of each investment instrument. Stock
investment provides huge benefits in the long term,
high levels of liquidity investing with small capital,
[6].
The banking sector is a significant part of the stock
market, and the performance of banks plays a
critical role in the overall performance of the
market. Most of the financial indexes are composed
of banking sectors. For example, the Royal Bank of
Canada (RBC), Toronto-Dominion Bank (TD),
Bank of Montreal (BMO), and Bank of Nova Scotia
(BNS) are in the top 10 constituents of the
S&P/TSX Composite Index that is the primary
gauge for Canadian-based, Toronto Stock Exchange
listed companies in Canada.
The stock market is a very complex and
substantial financial system, so various economic
and political factors affect the changes in the stock
market at every moment, [1]. The stock prices go up
and down, it can be difficult to predict how much or
when it will go down, [4]. Changes in stock prices
are the most important concern to the stockholder in
the market, [1]. The stock prices of individual banks
are also affected by various market factors,
including stock market indices. The study, [4],
grouped the main factors that affect stock prices,
into four parts, company news and performance,
industry performance, investor sentiment, and
economic factors. Therefore, accurately predicting
the upward and downward trends of stock prices
remains a significant challenge for all investors, [7].
Stock prices are formed by the prices of instruments
and their percentage amounts in their portfolio, it is
an essential issue to learn more about the factors
that can affect stock prices. Since banks play an
important role in financial systems and their
performance has a significant impact on the
economy, the stock prices of banks are easily
affected during crises. Analyzing the factors that
affect the stock prices of banks can help investors
better understand how to respond to fluctuations in
the market and make informed decisions about
investments, risks, and regulations.
In recent years, there has been an increasing
interest in the relationship between the daily closing
price of banks and various index values during
crises. In literature, many scholars studied stock
market prices, including the effects of indexes on
stock market prices or returns as well as forecasting
of stock indexes, [7], [8], [9], [10]. They used
various methods such as LR, [2], [3], [11], penalized
logistic regression, [7], multiple linear regression
(MLR), [1], [10], decision trees (DT), [2],
discriminant analysis (DA), [8], Partial linear
regression (PLS), [8], data mining (DM), [9],
machine learning (ML), [1], [9], and other
techniques. One scholar used the LR model with the
gradient-boosted decision trees (GBDT) and support
vector machine (SVM) aiming to predict and engage
in the trading of stock indexes, [9]. Other scholars
combined technical analysis with group penalized
logistic regressions, and proposed group
SCAD/MCP penalized logistic regressions with
technical indicators to predict up and down trends
for stock prices. This novel prediction method was
implemented using three stocks: BAC, Amazon, and
Citibank. The objective was to forecast whether the
stock prices would rise or fall the following day, [3].
Likewise, LR was employed to investigate the
impact of daily trading volume at the Botswana
Stock Exchange on daily stock market movement.
The findings revealed that only the trading volume
from three days prior influenced the current stock
market index movement. Interestingly, no
significant impact was observed from the trading
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2437
Volume 20, 2023
volumes of the past five days on today's stock
market movement, [12].
Another study constructs a global economic policy
uncertainty index through the principal component
analysis of the economic policy uncertainty indices
for twenty primary economies around the world,
[13]. The PCA-based economic policy uncertainty
index demonstrates a positive correlation with both
volatility and correlation in the global financial
market. This indicates that higher levels of global
economic policy uncertainty lead to increased
volatility and stronger correlations among stocks.
Comparatively, the PCA-based global economic
policy uncertainty index performs slightly better, as
it exhibits a more pronounced and significant
relationship with market volatility and correlation,
[13]. Another research paper integrates the LR
model to establish a correlation analysis model
between stocks and the Purchasing Managers' Index
(PMI), [14]. The study employs PMI data from the
National Bureau of Statistics as a sample and
conducts experiments to evaluate the effectiveness
of the proposed system model. The experimental
analysis reveals that the algorithm developed in this
paper yields significant results, further confirming
the robust correlation between PMI and stocks, [14].
Interestingly, one scholar examines the
predictability of the twelve most liquid
cryptocurrencies by employing machine learning
classification algorithms, such as support vector
machines, LR, artificial neural networks, and
random forests. The analysis is conducted at both
daily and minute-level frequencies. The models
utilize historical price data and technical indicators
as features to make predictions, [15].
In another research endeavor, the aim is to
examine the influence of US financial stress on the
risk-return dynamics within the Indian equity
market. This investigation employs a combination
of Markov regime-switching and binary LR models.
The study incorporates the weekly closing local
values of benchmark equity indices, namely 'CNX
Nifty 50 and S&P 500,' along with the St. Louis Fed
Financial Stress Index (SFSI). The findings of the
LR model reveal a positive association between US
financial stress and the probability of a bear regime
being present, [16]. Similarly, another research
paper introduces a learning architecture called
LR2GBDT for forecasting and trading stock indices.
The proposed architecture is evaluated by
comparing its performance with several other
models, namely LR, GBDT, SVM (support vector
machine), NN (neural network), and TPOT (tree-
based pipeline optimization tool), [2]. The
evaluation is conducted using data from three stock
indices belonging to two different stock markets: an
emerging market (Shanghai Stock Exchange
Composite Index) and a mature stock market
(Nasdaq Composite Index and S&P 500 Composite
Stock Price Index). Under the same test conditions,
the cascaded LR2GBDT model not only
outperforms the other models but also demonstrates
statistically and economically significant
improvements in exploiting simple trading
strategies, even when considering transaction costs,
[2].
Financial ratios play a crucial role in shaping
investors' expectations of stock prices and
consequently influence their investment decision-
making process, [17]. These ratios are utilized by
analysts, investors, and researchers to forecast future
trends in stock prices. Ratio analysis has become a
vital tool for fund managers and investors to assess
the intrinsic value of shares, making financial ratios
extensively employed for stock valuation. In the
existing literature, a specific paper aims to predict
the performance of stocks using Multinomial
Logistic Regression (MLR), [17]. The paper
employs financial ratios as practical selection
criteria to categorize stocks into three groups:
GOOD, AVERAGE, and POOR, based on their
returns and variances compared to the market's
returns and variances. The primary objective of this
paper is to explore the applicability of these
financial ratios for predicting stock market returns
within the Indian market context, [17].
In another research paper, the focus is on
analyzing the determinants of debenture issuing
prices in Brazil between the years 2000 and 2004. A
factor model is employed, wherein exogenous
variables are used to explain the behavior of returns
and prices. The variables examined in this study
encompass rating, index selection, maturity, country
risk, basic interest rate, spread between long-term
and short-term rates, stock market index, and
foreign exchange rate. The findings reveal that the
choice of index, probability of default, and bond
maturity significantly influence pricing. Moreover,
the study highlights a correlation between long-term
bonds and higher-rated issuances, [18]. In addition,
a researcher investigates the structural dynamics of
the Egyptian stock market after the implementation
of the economic reform program in 1991, employing
LR analysis. The findings from the LR reveal
noteworthy alterations in market activity, market
size, market liquidity, and market concentration
data, [19]. In another research paper, the authors
employ an LR approach to examine the relationship
between market conditions in Croatia and the
classification problem of market inefficiency, [20].
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2438
Volume 20, 2023
They analyze the daily returns of CROBEX,
spanning from September 1997 to July 2021, to
assess the market efficiency of the Croatian stock
market. Based on empirical findings, the study
reveals the time-varying nature of the stock market,
with price levels and trading volumes emerging as
significant indicators of market efficiency, [20].
Specifically, periods characterized by lower prices
and higher liquidity were more likely to exhibit
inefficiency, potentially indicating trading
opportunities within the Croatian stock market, [20].
Machine learning is another forecasting method,
especially when variables in the model are
nonstationary and weak relationship among the
variables, [9]. A researcher undertakes a
comparative analysis of three predictive models,
namely Support Vector Machine (SVM), Long
Short-Term Memory Recurrent Neural Network
(RNN), and LR, to forecast the direction of price
movement. The study employs data comprising the
share codes from the VN30 list between January 1,
2015, and January 27, 2022, with daily trading
activities considered. The "rolling window"
approach is utilized, resulting in the RNN model
achieving an average forecast accuracy of 82.19%.
Additionally, LR is found to play a crucial role in
determining the statistical significance of input
variables. Based on the experimental results, the
study recommends the adoption of machine learning
algorithms to enhance prediction accuracy.
Simultaneously, investors are advised to consider
company-specific characteristics when formulating
medium and long-term investment plans, [9].
Likewise, in another research investigation, the
study examines various linear classification models,
including LR classification (LR), linear discriminant
analysis (LDA), partial least-squares discriminant
analysis (PLS-DA), penalized discriminant analysis
(PDA), and nearest shrunken discriminant analysis,
[8]. The objective is to compare their performance
in predicting the stock market prices of the top six
banks in Bangladesh. The results indicate that PDA
performs well in the presence of multicollinearity or
the risk of overfitting. LDA yields better
approximations when the data exhibit multivariate
normality, while the nearest shrunken method is
effective when dealing with high-dimensional data.
Interestingly, despite the data in this study
possessing all the aforementioned characteristics,
LR demonstrates a lower misclassification rate or
apparent error rate. Consequently, the study
suggests that if there are predictors with high
correlation, multicollinearity, multivariate
normality, and high dimensionality, LR should be
preferred among the linear classification models,
[8]. In a similar vein, another study focuses on
predicting the stock price movement on the second
day following the release of companies' annual
reports, utilizing a range of models such as decision
trees, LR, random forest, neural networks, and
prototypical networks. The experiments are
conducted using two sets of financial indicators
sourced from the East Money website, which are
disclosed by the companies. However, the results
indicate that these models do not exhibit strong
predictive capabilities for determining stock price
tendencies. Additionally, after applying filters to
include only stocks with a return on equity (ROE)
greater than 0.15 and a net cash ratio greater than
0.9, it is observed that the predictability of stock
price movements on the second day after disclosure
remains weak. The random forest classifier achieves
the highest accuracy of approximately 59.6% and a
maximum precision of about 0.56 on the test set, but
overall performance remains limited. Furthermore,
the study finds that stock filtering does not enhance
the predictive performance significantly. Overall,
random forests demonstrate the best performance
among the models considered, which aligns with
previous research findings, [21].
A different study suggests a novel approach to
creating volatility networks for global stock
markets. This involves constructing both undirected
and directed networks by analyzing the pair-wise
correlation and system-wide connectedness of
national stock indices using a vector auto-regressive
model, [22]. The researchers investigate the impact
and utility of network indicators by employing them
as inputs for various machine learning techniques,
such as LR, support vector machine, and random
forest, [22]. Within this framework, two strategies
are devised: a global stock market prediction
strategy and a regional allocation strategy targeting
developed and emerging markets, [22]. The results
indicate that network indicators serve as valuable
supplementary tools for predicting the global stock
market and determining the relative directions (up
or down) of regional markets, [22].
Aiming to provide a comprehensive understanding
of the factors that affect the stock prices of banks to
help investors and financial institutions during crisis
periods, this study examines how the index values
affect the closing prices of five major banks in
Canada during five crisis periods using the LR
method.
3 Aim and Methodology
The banking sector is a significant part of the stock
market, and the performance of banks plays a
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2439
Volume 20, 2023
critical role in the overall performance of the
market. The corresponding index values in the stock
market play a significant role in determining the
daily closing prices of banks, especially during
crisis periods, indexes in the stock market fluctuate
and directly affect the stock prices of banks.
This paper aims to investigate the impact of
various price indexes on the daily closing prices of
banks during the crisis periods. Additionally, the
study aims to identify any discernible patterns in the
influence of indexes on daily stock market prices
during times of uncertainty.
To achieve this objective, data on the daily
common stock market prices and indexes of five
major Canadian banks (BMO, BNS, CIBC, RBC,
and TD) were collected. The data covered the period
from January 1, 1975, to December 31, 2020. The
data of five banks were obtained from the source,
[23], [24]. The study considered the daily closing
price (DCP) of stocks as the dependent variable.
The independent variables comprised fifteen out of
the total twenty-nine indexes, specifically the daily
price indexes (DPI) represented by ind1 to ind29.
Among these,
Table 1. Price index definitions
CFMRC: Canadian Financial Markets Research Center;
S&P/ TSX: Standard and Poor (stock
performance)/Toronto Stock Exchange; DPI: daily price
index; DEWI: daily equal-weighted index; VWI: value-
weighted index. Source, [23], [24].
the odd-numbered indexes represented daily price
indexes, while the even-numbered indexes
represented daily return indexes. However, the study
excluded the return indexes due to their high
correlation with the closing price variable. Of the
remaining indexes, three (ind33, ind35, and ind37)
were daily equal-weighted indexes (DEWI), three
(ind34, ind36, and ind38) were value-weighted
indexes (VWI), ind31 represented the call loan
interest rate, and ind32 represented the daily foreign
exchange rate. It is worth noting that certain indexes
had different starting time points. Indexes 3-22
covered the period from January 1, 1975, to
December 30, 1987, while indexes 23-30 spanned
from January 1, 1975, to May 31, 2002, [24]. Any
missing periods within each index were excluded
from the analysis, specifically if they occurred
during the crisis periods under investigation. The
model employed the independent variables as listed
in Table 1, [24].
Fig. 1: Daily closing prices of five banks from
January 1, 1975, to December 31, 2020, [24].
To identify periods of crisis, a graph was
constructed, representing the daily closing prices of
each bank (Figure 1). The graph was then analyzed
by referring to documented Canadian economic
crises and focusing on shared periods of decline.
Five distinct crisis periods were identified as
follows:
CR I (Crisis Period I): January 1, 1992, to
April 30, 1993.
CR II (Crisis Period II): July 1, 1998, to
October 30, 1998.
CR III (Crisis Period III): May 1, 2007, to
March 30, 2009.
CR IV (Crisis Period IV): September 1,
2014, to February 29, 2016.
CR V (Crisis Period V): February 1, 2020,
to March 30, 2020.
These crisis periods were determined based on
specific time intervals during which significant
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2440
Volume 20, 2023
financial disruptions or instability occurred, [24],
and using the declining periods of DCPs in Figure 1.
While daily price indexes typically derive from
DCPs within the Toronto Stock Market, this paper
adopts a unique approach by examining price
indexes in the reverse direction to search for a
significant impact on DCP in different crisis periods
and also the price index differences among the crisis
periods in conjunction with the recession or
depression.
In this paper, to determine the most important
price indexes that affect the daily stock closing price
of the banks in Canada, the five largest banks’
secondary data were analyzed by logistic regression
analysis using SPSS software and Excel.
LR is a statistical method that is used to analyze
relationships between a set of independent variables
and a binary (categorical) dependent (response)
variable. The study, [12], indicated that “a LR
model become a popular model because of its ability
to predict, classify and draw relationships between a
dichotomous independent variable and dependent
variables’. The main difference between LR and
Multiple Linear Regression (MLR) is the type of
dependent (response) variable and the outcome and
the normality conditions. MLR and LR
requirements on the dependent variable are
different. While MLR works with the continuous
dependent variable, LR works with the binary
dependent variable. For MLR, all variables, either
dependent or independent variables, should be
normally distributed and the residues must have the
same variance. However, for LR, there is no
requirement for the dependent and independent
variables to be normally distributed and the residues
to have the same variance. LR is modeling the
probability of outcome occurring based on the
values of the independent variables, [17], [25]. LR
results provide a classification table in the output,
[25]. This 2×2 classification table presents the
observed values on the outcome and predicted
values for the outcome and it will show the accuracy
of the LR model in percent, [25]. The LR model
uses the logistic function to map the linear
combination of the independent variables to the
range of [0, 1]. The logistic function is defined as:
(1)
where represents the probability of the
dependent variable being 1 given the values of the
independent variables X and z is the linear
combination of the independent variables and their
coefficients:
(2)
In the equation above, represents the intercept,
are the coefficients corresponding
to the independent variables ,
respectively, [20].
In LR, to estimate the coefficients the maximum
likelihood estimation (MLE) method is commonly
used, [20]. The MLE aims to find the values of
that maximize the likelihood of
observing the given data. This involves minimizing
the log-likelihood function:
= ∑[*log(P(Y=1|)) + (1-) * log(1-
P(Y=1|))]
(3)
where i ranges from 1 to the total number of
observations, is the observed outcome for the ith
observation, and is the predicted
probability of the dependent variable being 1 given
the values of the independent variables in the ith
observation, [19]. Once the coefficients are
estimated, they can be used to predict the
probability of the dependent variable being 1 for
new observations based on the values of the
independent variables. In summary, LR utilizes the
logistic function to model the relationship between a
binary dependent variable and independent
variables. By estimating the coefficients using the
maximum likelihood estimation method, LR allows
us to predict the probability of the dependent
variable's outcome based on the independent
variables' values. In summary, multiple linear
regression predicts continuous numerical outcomes,
while LR predicts probabilities or classifies binary
outcomes, [20]. LR analysis is often used to
investigate the relationship between discrete
responses and a set of explanatory variables. LR
uses logit as a link function, i.e., it takes the log of
the odds of the success ratio, [19].
To facilitate LR analysis, the DCP (z) was
transformed into binary format, distinguishing
positive differences between the current DCP and
the previous day's DCP as "P" which is encoded as
“1” and negative differences as "N" which is
encoded as “0”. Hence, the dependent variable, z, is
defined in the following manner
(4)
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2441
Volume 20, 2023
(5)
where, are the Impact values
of indexes.
After conducting the analysis in SPSS, the
following statistical measures were utilized to assess
the LR models: the Pearson chi-square test, -2 log-
likelihood, the significance of the Hosmer-
Lemeshow test, and the Cox & Snell R-square and
Nagelkerke R-square. The results will be discussed
in the following section.
For LR models, to assess the goodness of fit, the
Pearson Chi-square test is commonly used. It tests
whether the LR model fits the data well or not under
the following hypotheses:
Ho: The LR model fits the data well
Ha: The LR model does not fit the data well.
The test statistic measures if there is a significant
difference between the observed and expected
frequencies of the binary dependent variable and
evaluates the overall goodness of fit of the LR
model, [17]. If the p-value is less than the
significant level, it is concluded that there is
evidence to believe that the LR model does not fit
the model well. Conversely, if the p-value is greater
than the significance level, it is concluded that there
is insufficient evidence to believe that the LR model
does not fit the data well, [17].
The -2 log-likelihood is another measure for LR
models, which assesses the overall goodness of fit
of the LR model. It measures the difference between
the observed data and the predicted probabilities
from the LR model, [17]. The lower -2 log-
likelihood value indicates that the LR model
predicts the outcomes with higher probability,
The Hosmer-Lemeshow test is another test to test
whether the model fits the data well. Similar to the
Pearson Chi-square test, it tests the model under the
following hypotheses:
Ho: The LR model fits the data well
Ha: The LR model does not fit the data well.
If the p-value is less than the significant level, it is
concluded that there is evidence to believe that the
LR model does not fit the model well. Conversely,
if the p-value is greater than the significance level, it
is concluded that there is insufficient evidence to
believe that the LR model does not fit the data well,
[26], [27].
The Cox & Snell R-square is the other measure
that measures the proportion of the total variation in
the binary dependent variable that is explained by
the LR model. Its value changes from 0 to 1, [28],
[29]. Interpretation of it is similar to the R-square in
MLR, but LR R-squared values are typically lower
than those of linear regression due to the nonlinear
nature of the LR model, [28], [29]. When the value
of Cox & Snell R-square is zero, it is concluded that
the model explains none of the variation. On the
other hand, when it is 1, it is concluded that the
model explains all of the variation, [28], [29].
The last measure that is used in this paper is the
Nagelkerke R-squared which is also known as the
Cox & Snell R-squared. The Nagelkerke R-squared
provides an estimate of the proportion of the
variation in the binary dependent variable that is
explained by the LR model, [29].
4 Finding and Discussion
In this section, to explore which indexes had a
positive or negative impact on the DCP of five
major banks in Canada during the five crisis periods
LR results are discussed in detail after giving the
characteristics of the crisis period.
4.1 CR I: Crisis I (1/1992-4/1993)
Canada experienced economic crises and recession
in the early 1990s, then it turned to economic
recession in CR I, [30], [31]. During this crisis, the
Canadian economy significantly declined. While
the real GDP decreased by more than 1 percent,
the unemployment rate increased by more than 10
percent, [31]. Many industries were affected
dramatically, and many people lost their jobs.
Canada also experienced an exchange rate crisis
during this time, [30], [31]. The value of the
Canadian dollar depreciated significantly, leading
to concerns about its stability and impact on trade
and investment.
Table 2. Logistic regression results of five banks
during Crisis I (1/1992-4/1993)
*p<0.001, **p<0.05, ***p>0.05, a: Estimation
terminated at iteration number 6 because parameter
estimates changed by less than 0.001, b: Estimation
terminated at iteration number 7 because parameter
estimates changed by less than 0.001, c: Estimation
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2442
Volume 20, 2023
terminated at iteration number 5 because parameter
estimates changed by less than 0.001
Table 2 depicts the LR analysis results of five
banks for CR I (1/1992-4/1993). The LR results for
the CR I (1/1992-4/1993) dataset indicate the
associations between the independent variables
(ind9Sector25-Consumer Discretionary Daily Price
Index, ind31-Call Loan Interest Rate,
ind34CFMRC-Daily Value Weighted Index,
ind36CFMRC-Value Weighted Under $2 Index, and
ind38CFMRC-Value Weighted Over $2 Index) and
the dependent variables (DCPs of BMO, BNS,
CIBC, RBC, TD). The analysis includes the
coefficients, exponential values (Exp(B)), and
statistical significance levels. ind1SPTSX, ind3,
ind5, ind7, ind11, ind13, ind15, ind17, ind19, ind21,
ind23, ind25, ind27, ind29, ind32 are not included in
any of the models. The LR results are slightly
different from the MLR results in another study that
used the same data set, [24]. In that study, when
they looked at the same data set using MLR, they
found that after ind15 (Financials), the most
influential factor on the DCP of five banks was
ind21 (utilities). Following that were ind3 (Energy),
ind7 (Industrials), and ind13 (Health Care) during
the CR I. Some of these factors had positive impacts
on the banks' DCPs, while others had negative
impacts. This happened because each bank used
different investment tools with varying percentages
in their portfolio, [24]. The differences in LR and
MLR results could be due to the type of dependent
variable used, continuous for MLR and discrete for
LR. The LR results show the coefficients for each
independent variable, along with the significance
level and the exponentiated coefficient (Exp(B)).
The constant term is also included, which represents
the intercept of the regression equation. The
constant terms in the LR model have coefficients of
0.317, 19.15, -0.021, 0.361, and 2.8 are statistically
significant at the p<0.001 level for BMO, CIBC,
and RBC, and the p<0.05 for BNS and TD. This
indicates that when all independent variables are
zero, there is a significant effect on the log odds of
the dependent variable.
Ind9Sector25-Consumer Discretionary Daily Price
Index hurts only the DCP of BNS and similarly, the
ind31-Call Loan Interest Rate has a negative impact
on only the DCP of TD and the ind36CFMRC-
Value Weighted Under $2 Index has a negative
impact on only DCP of BMO. ind9Sector25-
Consumer Discretionary Daily Price Index and
ind31-Call Loan Interest Rate are significant at the
p<0.01 level. However, the ind34CFMRC-Daily
Value Weighted Index is included in the LR models
of BMO and CIBC. Among the independent
variables, ind34CFMRC-Daily Value Weighted
Index and ind36CFMRC-Value Weighted Under $2
Index have statistically significant coefficients at the
p<0.05 level. ind34CFMRC-Daily Value Weighted
Index has a negative coefficient of -55565.65,
suggesting a negative relationship with the DCP of
BMO, and has a positive coefficient of 408.25,
suggesting a positive relationship with the DCP of
CIBC. Ind36 has a positive coefficient of 81.53,
indicating a positive association with the DCP of
BMO. Both variables are significant at the p<0.001
level. The ind38CFMRC-Value Weighted Over $2
Index is included in all LR models, except for the
LR model of CIBC, and is statistically significant at
p<0.001. The exponential values (Exp(B)) provide
the odds ratios associated with each independent
variable. For example, the ind34CFMRC-Daily
Value Weighted Index has an exponential value of
0.0000, indicating that for a one-unit increase in the
ind34CFMRC-Daily Value Weighted Index, the
odds of the dependent variable decrease
significantly. ind36CFMRC-Value Weighted Under
$2 Index has an exponential value of 2.559E+35,
indicating a substantial increase in the odds of the
dependent variable with a one-unit increase in
ind36CFMRC-Value Weighted Under $2 index.
The -2 loglikelihood values for each dependent
variable provide information about the goodness of
fit of the model, [25]. Lower values indicate a better
fit, [25]. The values range from 125.757 to 142.046,
suggesting that the model fits relatively well for
these dependent variables. The Hosmer and
Lemeshow Test is used to assess the goodness of fit
of the model by comparing the observed and
expected frequencies, and its significance levels
indicate whether the model fits the data well, [26],
[27]. The provided significance levels for BMO,
BNS, CIBC, RBC, and TD are greater than 0.05,
suggesting that the model fits the data reasonably
well. The Cox & Snell R Square and Nagelkerke R
Square values provide measures of the proportion of
explained variance in the dependent variables, [28],
[29]. The values range from 0.133 to 0.409,
indicating that the independent variables
collectively explain a moderate amount of the
variance in the DCP of LR models. Moreover,
Omnibus Tests of model coefficients by Chi-square
test are all <0.001, therefore overall all models are
significant.
In conclusion, the LR analysis suggests that
several independent variables have significant
associations with the dependent variables in the CR
I (1/1992-4/1993) dataset. Ind34, ind36, ind9, ind31,
and ind38 are found to have significant effects on
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2443
Volume 20, 2023
the DCP of five banks, as indicated by their
coefficients, exponential values, and statistical
significance levels.
Table 3 shows the classification results of the LR
models. The analysis of the observed and predicted
values for the different variables (BMO, BNS,
CIBC, RBC, TD) reveals the accuracy of the LR
model. The "Percentage Correct" column indicates
the proportion of correctly predicted outcomes.
Table 3 presents the number of observed instances
and the number of predicted instances for each
category, as well as the overall percentage of correct
predictions. Similarly, for BNS, the model has an
overall percentage of 70.6% correct predictions.
Table 3. Classification tablea for Crisis I (1/1992-
4/1993)
a The cut value is .500; N represents price drop; P
represents price up.
For CIBC, the model performs relatively better,
with an overall percentage of 73.8% correct
predictions. For RBC, the overall percentage of
correct prediction is 68%. For TD, the model
achieves an overall percentage of 82.1% correct
prediction, indicating a good predictive
performance. These results suggest that the LR
model can reasonably predict the outcomes for the
dependent variables in this analysis. The models are
applicable to forthcoming crises of a similar nature.
4.2 CR II: Crisis (7/1998-10/1998)
Canada was largely affected by the Asian financial
crisis which started in Thailand in mid-1997, then
spread to other Asian countries, and turned out to be
a global crisis when it spread to the Russian and
Brazilian economies, [32]. The Asian financial crisis
led to a decline in global investor confidence and
increased volatility in international financial
markets. As a result of the Asian financial crisis, the
Canadian dollar (CAD) experienced significant
depreciation against major currencies, including the
US dollar. The depreciation of the Canadian dollar
had adverse effects on various sectors of the
Canadian economy, [32]. The crisis also impacted
the Canadian stock market, leading to a sharp
decline in stock prices. The Toronto Stock
Exchange (TSX) experienced a period of
considerable volatility, causing financial stress for
investors and companies alike. The financial crisis
contributed to an economic contraction in Canada.
The decline in export demand, particularly from
Asia, along with the stock market turmoil,
negatively affected business investment and
consumer confidence. Consequently, Canada's GDP
growth rate slowed down during this period, [32].
The crisis also affected the Canadian banking sector.
Some Canadian banks faced financial strains due to
their exposure to the global financial turmoil and
their ties to the affected regions. However, the
Canadian banking system proved to be resilient, and
major banks were able to weather the storm without
significant failures or government bailouts, [32].
Table 4 shows the results of a LR analysis with
several independent variables (ind7Sector20-
Industrials Daily Price Index, ind17Sector45-
Information Technology Daily Price Index, ind31-
Call Loan Interest Rate, ind32-Daily Foreign
Exchange Rate, ind33CFMRC-Daily Equal
Weighted Index, ind35CFMRC-Equal Weighted
Under $2 Index, and ind38CFMRC-Value Weighted
Over $2 Index) and several dependent variables
(BMO, BNS, CIBC, RBC, and TD).
Table 4. Logistic regression results of five banks for
Crisis II (7/1998-10/1998)
*p<0.001, **p<0.05, ***p>0.05, a: Estimation
terminated at iteration number 6 because parameter
estimates changed by less than 0.001, b: Estimation
terminated at iteration number 7 because parameter
estimates changed by less than 0.001, c: Estimation
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2444
Volume 20, 2023
terminated at iteration number 5 because parameter
estimates changed by less than 0.001
It appears that some of the independent variables,
ind7Sector20-Industrials Daily Price Index,
ind17Sector45-Information Technology Daily Price
Index, ind31-Call Loan Interest Rate, ind32-Daily
Foreign Exchange Rate, ind33CFMRC-Daily Equal
Weighted Index, ind35CFMRC-Equal Weighted
Under $2 Index, and ind38CFMRC-Value Weighted
Over $2 Index, have significant effects on some of
the dependent variables while the other 17 indexes
do not have a significant effect on the DCP of five
banks. For example, the ind38CFMRC-Value
Weighted Over $2 Index has a significant positive
effect on the DCP of BMO, BNS, CIBC, RBC, and
TD, as indicated by the positive coefficient and the
low p-value. The ind31-Call Loan Interest Rate has
a significant positive effect and the ind33CFMRC-
Daily Equal Weighted Index has a significant
negative effect on the DCP of TD. The
ind35CFMRC-Equal Weighted Under $2 Index has
a significant negative effect on BNS, as indicated by
the negative coefficient and the low p-value.
ind7Sector20-Industrials Daily Price Index and has
a significant positive effect on the DCP of BNS,
however, ind17Sector45-Information Technology
Daily Price Index and ind35CFMRC-Equal
Weighted Under $2 Index have a significant
negative effect on the DCP of BNS. The other
variables do not appear to have significant effects on
the dependent variables, as indicated by the non-
significant p-values. Moreover, Omnibus Tests of
Model Coefficients are all <0.001, therefore overall
all models are significant. It is interesting that only
ind38CFMRC-Value Weighted Over $2 Index is the
common index as in crisis I and it has a significant
positive effect on all of the DCPs of the five banks
during crisis II. These significant coefficients
suggest that ind38 has a strong influence on
predicting the DCP of five banks.
The p-values for Hosmer and Lemeshow test4th
row indicate that the model's fit is statistically
significant for all dependent variables except for the
DCP of CIBC, [26], [27]. The p-value for CIBC is
0.008, suggesting that the model's fit for CIBC is
not statistically significant. Regarding the R-squared
measures, [28], [29], Cox & Snell R Square ranges
from 0.177 to 0.535, and Nagelkerke R Square
ranges from 0.241 to 0.714. These measures indicate
the proportion of variance explained by the model,
with higher values indicating better model fit. In this
analysis, BNS and RBC show relatively higher R-
squared values compared to other dependent
variables.
In conclusion, the LR results suggest that some
independent variables have statistically significant
effects on predicting the dependent variables.
However, the model's overall fit varies across the
dependent variables, with some variables showing
statistically significant fits while others do not. The
LR analysis demonstrates the significant impact of
certain independent variables on the likelihood of
specific outcomes represented by the dependent
variables. The results highlight the importance of
considering these variables when predicting the
probabilities of the outcomes of interest.
Table 5 presents the observed and predicted values
for the dependent variables (DCP of BMO, BNS,
CIBC, RBC, TD) in the LR model the number of
observed instances, and the number of predicted
instances for each category, as well as the overall
percentage of correct predictions during Crisis II.
The results are presented for which step of the
classification process ends.
Table 5. Classification tablea for Crisis II (7/1998-
10/1998)
a: The cut value is .500; N represents price drop; P
represents price up.
For BMO, the model achieved an overall
percentage of 75%, indicating that it accurately
predicted 75% of the outcomes. Similar to BMO,
the overall percentage of correct prediction for
CIBC was 73.8%. Although the lowest overall
percentage of correct prediction was 68% for RBC,
it still indicates a satisfactory level of accuracy in
predicting outcomes. When compared with BMO,
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2445
Volume 20, 2023
CIBC, and RBC, the overall percentage of correct
predictions for TD and BNS were very high, 82.1
and 86.9, respectively. This indicates a relatively
high level of accuracy in predicting the outcome for
TD and BNS. In conclusion, the LR model
demonstrated varying levels of accuracy in
predicting the outcomes for different variables. The
LR models show promising predictive performance,
correctly identifying a substantial proportion of the
observed outcomes. These findings suggest that the
model has the potential to be a useful tool for
predicting the outcomes of interest.
4.3 CR III: Crisis III (5/2007-3/2009)
Canada experienced significant challenges during
the economic crisis CR III. This crisis originated in
the United States with the collapse of the subprime
mortgage market and quickly spread to other parts
of the world, [33], [34]. It was the liquidity crisis in
the global financial markets, then turned into a
solvency crisis, [24]. The financial system in the
world is affected dramatically. Mortgage lenders,
insurance companies, and commercial as well as
investment banks were among those that faced
difficulties during this crisis. For example, the big
investment bank, Lehman Brothers, precipitated
during this crisis. Although Canada's banking sector
was generally more conservative and well-regulated
compared to other countries, some Canadian banks
and financial institutions faced challenges,
particularly due to exposure to risky assets and
disruptions in global financial markets, [33], [34].
Canada's major banks were able to remain stable
throughout the crisis due to lending practices,
stricter regulations, and a conservative banking
culture. According to, [35], Canada not only through
international trade, but also by weakening financial
markets, shaking consumer and business
confidence, and postponing capital investments, in
light of the high level of uncertainty, during this
crisis, [24]. The crisis resulted in a significant
economic recession in Canada. The country
experienced a contraction in economic activity,
declining GDP growth, rising unemployment rates,
and reduced business investment. Industries such as
manufacturing, housing, and automotive were
particularly affected, [33], [34].
Table 6 includes coefficients, significance levels,
and exponential values for each independent
variable included in the LR analysis during the
2007-2009 financial crisis. The coefficients
represent the estimated change in the dependent
variable for a one-unit change in the corresponding
independent variable, while the significance levels
indicate the probability that the observed
relationship between the independent and dependent
variables is due to chance.
Table 6. Logistic regression results of five banks
for Crisis III (5/2007-3/2009)
*p<0.001, **p<0.05, ***p>0.05, a: Estimation
terminated at iteration number 6 because parameter
estimates changed by less than 0.001, b: Estimation
terminated at iteration number 7 because parameter
estimates changed by less than 0.001
Based on the results in Table 6, it appears that
some of the independent variables, ind33CFMRC-
Daily Equal Weighted Index, ind34CFMRC-Daily
Value Weighted Index, ind35CFMRC-Equal
Weighted Under $2 Index, ind36CFMRC-Value
Weighted Under $2 Index, ind37CFMRC-Equal
Weighted Over $2 Index, ind38CFMRC-Value
Weighted Over $2 Index, have a statistically
significant relationship with the dependent variable
while the other 18 indexes do not have any
significance effect on the DCP of five banks. For
example, the coefficient for the "ind36CFMRC-
Value Weighted Under $2 Index" variable is
negative and statistically significant for the DCP of
five banks, suggesting that a decrease in this
variable is associated with an increase in the
dependent variable (which is likely related to
financial institution performance during the crisis).
On the other hand, while the ind38CFMRC-Value
Weighted Over $2 Index has a significant positive
effect on the DCP of BMO, CIBC, and TD,
statistically significant, it has a significant negative
effect on the DCP of BNS and RBC. ind34CFMRC-
Daily Value Weighted Index has a significant
positive effect on the DCP of both BNS and RBC.
This shows that the investments of BNS and RBC
are similar during crisis III. ind33CFMRC-Daily
Equal Weighted Index and ind35CFMRC-Equal
Weighted Under $2 Index have a significant
positive effect on the DCP of TD and BMO,
respectively.
During Crisis III, there are two common indexes,
ind36CFMRC-Value Weighted Under $2 Index and
ind38CFMRC-Value Weighted Over $2 Index that
affect the DCP of five banks. Similar to Crisis I, in
the other study, using the same data with the MLR
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2446
Volume 20, 2023
method, [24], different indexes have an impact on
the DCP of banks. In that study, ind15-Financials
that have a positive impact on the DCP of all banks,
ind11-Consumer Staples, ind13-Health Care, ind17
-Information Technology, ind19 -
Telecommunications Services and ind32-Daily
Foreign Exchange Rate are important indexes to
predict the DCP of banks. The differences between
LR and MLR results could be due to the way the
dependent variable is treated, it's continuous for
MLR and discrete for LR.
The -2 loglikelihood values reflect the goodness of
fit of the model, with lower values indicating a
better fit, [25]. In this case, the -2 loglikelihood
values range from 436.321 to 504.306, suggesting
that the model adequately captures the relationships
between the predictor variables and the outcomes.
The Hosmer and Lemeshow test results indicate the
goodness of fit of the model. With p-values ranging
from 0.055 to 0.919, the model does not
significantly deviate from the expected values for
the observed and predicted outcomes, [27]. The Cox
& Snell R Square and Nagelkerke R Square values
provide an estimate of the proportion of variance
explained by the model, [28], [29]. These values
range from 0.282 to 0.501, indicating that the model
explains a moderate amount of the variance in the
outcomes, [28], [29].
In conclusion, the logistic LR reveals significant
relationships between the independent variables and
the dependent variables. The odds ratios provide
information on the direction and magnitude of these
relationships. The model fits the data well, and the
R-squared values suggest that the model explains a
meaningful portion of the variance in the dependent
variables. These findings indicate that the
independent variables considered have an impact on
the likelihood of the observed outcomes. Moreover,
Omnibus Tests of Model Coefficients are all
<0.001, therefore overall all models are significant.
All these models can be utilized for addressing
similar financial crises in the future.
The observed versus predicted values for the LR
model in CR III (5/2007-3/2009) provide insights
into the accuracy of the model's predictions (Table
7). For BMO, the model achieved an overall
percentage of 73.3, indicating that it accurately
predicted 73.3% of the outcomes. Similarly, for
BNS, CIBC, RBC, and TD, the overall percentages
were 74.6, 72.6, 76.5, and 77.1, respectively.
These results suggest that the LR model performed
reasonably well in predicting the outcomes for the
variables under consideration. In conclusion, the LR
model demonstrated a moderate to good level of
accuracy in predicting the outcomes of interest. The
overall percentages indicate a relatively high level
of correctness in the model's predictions.
Table 7. Classification tablea for Crisis III (5/2007-
3/2009)
a The cut value is .500; N represents price drop; P
represents price up.
Overall, the LR model demonstrates reasonably
accurate predictions for the dependent variables
DCP of BMO, BNS, CIBC, RBC, and TD. The
percentages of correct predictions range from 72.6%
to 77.1%. These findings contribute to a better
understanding of the factors influencing the
outcomes and can inform decision-making.
4.4 CR IV: Crisis IV (9/2014-2/2016)
The largest decline in the oil price in the world from
mid-2014 to early 2016, negatively affected the
Canadian economy, [36]. Canada is the world's
third-largest exporter of oil after Saudi Arabia and
Russia. The bulk of the oil reserves are located in
Alberta, [37]. This decline in oil prices had a
negative impact on the energy sector and the regions
heavily reliant on oil production, such as Alberta,
[36], [38]. It led to job losses, reduced investments,
and slower economic growth in those areas. During
CR IV, while the energy sector and mainly Alberta
were negatively affected by CR IV, the other
sectors, such as services, and construction showed
resilience and contributed to economic growth, [36],
[38].
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2447
Volume 20, 2023
Table 8. Logistic regression results of five banks for
Crisis IV (9/2014-2/2016)
*p<0.001, **p<0.05, ***p>0.05, a: Estimation
terminated at iteration number 6 because parameter
estimates changed by less than 0.001, b: Estimation
terminated at iteration number 7 because parameter
estimates changed by less than 0.001
Table 8 presents the LR outcomes for CR IV. The
findings indicate variations in the indexes included
in the LR models, along with differing impacts on
these models. Some indexes positively affect the
DCP of banks, while others have a negative effect.
The negative coefficient for BNS and the positive
coefficient for BMO and RBC for the constant term
suggests that these banks had different trends in
stock price change during the crisis period. Among
the independent variables, the daily equal-weighted
index (ind33CFMRC) had a significant negative
impact on TD's stock price, indicating that TD's
DCP was adversely affected during the crisis period
by changes in this index. The daily value-weighted
index (ind34CFMRC) had a significant negative
impact on the DCP of BMO, BNS, and RBC. In
contrast, the value-weighted over the $2 index
(ind38CFMRC) had a positive impact on the stock
prices of all five banks, indicating that higher values
of this index were associated with higher stock
prices of banks.
For the DCP of BMO, the only significant variable
is the ind34CFMRC-Daily Value Weighted Index
and ind38CFMRC-Value Weighted Over $2 Index.
However, for the DCP of BNS, the significant
variables are the ind7Sector20-Industrials Daily
Price Index, ind15Sector40-Financials Daily Price
Index, ind34CFMRC-Daily Value Weighted Index,
ind35CFMRC-Equal Weighted Under $2 Index and
ind38CFMRC-Value Weighted Over $2 Index. This
shows that the investments of BNS are different
from the others. Overall, the results suggest that
during the crisis period of 2014-2016, the DCPs of
five banks were significantly affected by different
indexes. The impact of these variables varies across
different banks. The common index that has a
significant effect on the DCP of five banks is the
ind38CFMRC-Value Weighted Over $2 Index.
While the ind38CFMRC-Value Weighted Over $2
Index has a positive impact on the DCP of BMO,
BNS, CIBC, and RBC, it has a negative impact on
the DCP of TD.
The -2 log-likelihood values indicate the goodness
of fit of the LR models, [25]. Lower values indicate
better model fit, [25]. In this case, the -2 log-
likelihood values range from 315.425 to 375.327,
suggesting a reasonable model fit for the data. The
Cox & Snell R Square and Nagelkerke R Square
values indicate the proportion of variance explained
by the model, [28], [29]. The values range from
0.318 to 0.558, suggesting moderate to substantial
explanatory power. Overall, the LR results
demonstrate the relationships between the
independent variables and the dependent variables
in CR IV. The significant effects and odds ratios
highlight the importance of certain variables in
predicting the outcomes. Moreover, Omnibus Tests
of Model Coefficients are all <0.001, therefore
overall all models are significant.
The LR analysis for the observed and predicted
values of the dependent variables (BMO, BNS,
CIBC, RBC, TD) in CR IV (9/2014-2/2016) shows
promising results. Table 9 displays the number of
observed and predicted values for each variable,
along with the overall percentage of correct
predictions.
Table 9. Classification tablea for Crisis IV (9/2014-
2/2016)
a The cut value is .500; N represents price drop; P
represents price up.
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2448
Volume 20, 2023
Table 9 shows that the LR model performs
reasonably well in predicting the outcomes. For
BMO, the model correctly predicts 78.1% of the
cases, indicating a good level of accuracy. Similarly,
for BNS, the model achieves an overall percentage
of 80.7% correct predictions, suggesting a reliable
predictive performance. CIBC achieved an overall
percentage of 75.1%. However, it still demonstrates
a reasonable level of accuracy in predicting the
outcomes for this variable. For RBC, the model
performs with an overall percentage of 76.9%
correct predictions. Although slightly lower than
BNS, it still indicates a satisfactory level of
accuracy in predicting the outcomes. Finally, for
TD, the model achieves an overall percentage of
79.6% correct predictions, indicating a good
predictive performance for this variable. These
results suggest that the LR model performed
reasonably well in predicting the outcomes for these
variables.
4.5 Crisis (2/2020-3/2020) (CR V)
The global economy has a severe disruption during
the CR V due to the outbreak of the COVID-19
pandemic, [39], [40]. Antonio Guterres, the
Secretary-General of the United Nations, warned on
April 1, 2020, stating that the global community was
confronting its most dreadful crisis since World War
II, [41], [42]. The novel coronavirus, officially
named SARS-CoV-2, began spreading rapidly
worldwide, including in Canada, leading to a public
health emergency. The rapid spread of the virus led
to health emergencies and the implementation of
strict public health measures, resulting in economic
disruptions, increased unemployment rates, and
significant market volatility. Industries such as
tourism, hospitality, retail, and entertainment were
particularly affected, [39], [43]. The crisis had a
serious impact on financial markets, including the
Toronto Stock Exchange (TSX). Stock prices
experienced extreme volatility and sharp declines as
investors reacted to the uncertainties surrounding
the pandemic and its potential economic impact,
[39], [43].
Table 10 displays the LR coefficients and
significance levels for the five major Canadian
banks (BMO, BNS, CIBC, RBC, and TD) during
the crisis period V. The independent variables
include daily price indices for different sectors and
the equal-weighted and value-weighted indices for
stocks with different market caps. For BMO, the
only significant independent variable is the daily
equal-weighted index (ind33CFMRC), with a
positive coefficient of 141.112 and a significance
level of 0.009. For BNS, the significant independent
variable is the equal-weighted index for stocks over
$2 (ind37CFMRC), with a positive coefficient of
277.279 and a significance level of 0.12. For CIBC,
there are two significant independent variables,
ind36CFMRC-Value Weighted Under $2 Index and
ind37CFMRC-Equal Weighted Over $2 Index.
While ind37CFMRC-Equal Weighted Over $2
Index has a positive effect on the DCP of CIBC,
ind36CFMRC-Value Weighted Under $2 Index has
a significant negative effect on the DCP of CIBC.
Table 10. Logistic regression results of five banks
for Crisis V (2/2020-3/2020)
*p<0.001, **p<0.05, ***p>0.05, b: Estimation
terminated at iteration number 7 because parameter
estimates changed by less than 0.001, d: Estimation
terminated at iteration number 8 because parameter
estimates changed by less than 0.001, e: Estimation
terminated at iteration number 9 because parameter
estimates changed by less than 0.001.
For RBC, the significant independent variable is
the equal-weighted index for stocks over $2
(ind37CFMRC), with a positive coefficient of
351.127 and a significance level of 0.024, and for
TD, ind37CFMRC-Equal Weighted Over $2 Index
has a positive effect with the coefficient of 300.635
on the DCP of TD.
Overall, the results suggest that during Crisis V,
the DCP of BNS, CIBC, RBC, and TD were
positively influenced by ind37CFMRC-Equal
Weighted Over $2 Index, while BMO showed a
significant positive relationship only with
ind33CFMRC-Daily Equal Weighted Index. It
seems that the investments and DCP of banks
change according to the character of the crisis.
While ind38CFMRC-Value Weighted Over $2
Index which has generally a significant effect on the
DCP of banks is the common index for the first four
crises, during the Crisis V, ind37CFMRC-Equal
Weighted Over $2 Index is the common one, except
the DCP of BMO.
When comparing the LR results with another study
that utilized the MLR method on the same dataset,
[24], the indexes that have either positive or
negative impact on the DCP of banks are completely
different. While the DCP of banks is influenced by
certain indices such as ind33CFMRC,
ind36CFMRC-Value Weighted Under $2 Index, and
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2449
Volume 20, 2023
ind37CFMRC-Equal Weighted Over $2 Index in LR
models, none of these indices are encompassed
within the MLR models, [24]. This discrepancy can
be attributed to the nature of the dependent variable;
it is continuous in MLR, whereas it is discrete in
LR.
In terms of model fit, the -2 log-likelihood values
indicate the goodness-of-fit of the LR models, [25].
Lower values indicate a better fit, [25]. The -2 log-
likelihood values range from 18.270 to 36.545,
suggesting that the models provide a reasonably
good fit to the data. The Hosmer and Lemeshow test
evaluates the goodness-of-fit of the model. The
higher p-values (>0.05) for all variables indicate that
the models do not show evidence of a lack of fit,
[26], [27]. The Hosmer and Lemeshow Test results
indicate the goodness of fit of the LR model based
on a significance level. In this analysis, all variables
have p-values greater than 0.05, indicating that the
model fits the data well.
The Cox & Snell R Square and Nagelkerke R
Square values provide information about the
proportion of variance explained by the model, [28],
[29]. The values range from 0.387 to 0.809,
indicating that the LR models account for a
substantial amount of variability in the outcome
variables.
In conclusion, the LR results for CR V suggest
that the variables ind33CFMRC-Daily Equal
Weighted Index and ind37CFMRC-Equal Weighted
Over $2 Index have significant effects on predicting
the outcomes for BMO, BNS, CIBC, RBC, and TD.
The models show a reasonably good fit to the data,
and they explain a considerable amount of variance
in the outcome variables. Moreover, Omnibus Tests
of Model Coefficients are all <0.001, therefore
overall all models are significant.
The analysis of the LR results for CR V (2/2020-
3/2020) suggests that the predicted probabilities
align quite well with the observed outcomes for the
variables BMO, BNS, CIBC, RBC, and TD (Table
11).
For BMO, the LR model achieved an overall
percentage of 85.4, indicating that 85.4% of the
observed outcomes were correctly predicted by the
model. Similarly, for BNS and TD, the models both
achieved an overall percentage of 87.8, suggesting a
good fit between the predicted probabilities and the
observed outcomes. For CIBC, the model's overall
percentage was 75.6, indicating a moderate level of
accuracy in predicting the outcomes. For RBC, the
Table 11. Classification tablea for Crisis V (2/2020-
3/2020)
a The cut value is .500; N represents price drop; P
represents price up.
LR model demonstrated higher accuracy, with an
overall percentage of 90.2. The overall percentages
suggest that LR models utilizing the CR V variables
demonstrate a reasonable predictive capacity for the
observed outcomes of BMO, BNS, CIBC, RBC, and
TD. The observed and predicted percentages
suggest a satisfactory alignment, indicating the
potential usefulness of the LR model in predicting
the outcomes of interest.
5 Conclusion
Financial crises can differ significantly in terms of
what causes them and how severe they are. They
might lead to significant drops in stock prices and
market capitalization, as well as increased volatility
and risk, changes in the overall economic situation,
the value of currencies, or external factors like the
Covid-19 pandemic, [44]. Since financial markets
react quickly to unexpected problems, the impact of
these crises is usually felt right away in the markets,
and these effects can stick around for a long time,
[44], [45], [46].
In the stock market, financial institutions and
banks play an important role. For example,
CAD$4.6 trillion of assets are managed by financial
institutions. The banks manage 70% of these assets
and 90% of the banking assets are controlled by the
top six banks: BOM, BNS, RBC, CIBC, TD, and
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2450
Volume 20, 2023
DG (Desjardins Group which is an investment
bank), [45]. The Five Big Banks whose data used in
this work hold over $100 trillion in assets of
Canada, and they are all based in Toronto, [45].
Therefore, it is important for investors,
policymakers, and financial regulators to understand
how banks are affected by crises. With this in mind,
the purpose is to explore the main indicators that
have either positive or negative impacts on the DCP
of banks and to identify patterns during times of
crisis. In order to offer valuable insights that can
help investors, policymakers, and financial
regulators effectively reduce the effects of
upcoming crises, in this paper, the data of five major
banks in Canada were examined by LR models of
banks during five different crisis periods. LR
analysis was conducted with the SPSS software. To
facilitate the analysis The DCP values converted
into binary form, denoted as "P" for an increase in
DCP compared to the previous day and "N" for a
decrease in DCP compared to the previous day.
All the LR models during the five crisis periods
exhibited higher accuracy percentages in the range
of 68 to 90.2. It is concluded that these models
could be used for other similar crises in the future. It
was imperative to note that these crises resulted
from a combination of domestic and global factors,
encompassing economic downturns, financial
market volatility, and external shocks. Each crisis
exhibited its own distinct characteristics and exerted
unique impacts on the Canadian economy and
society. Therefore, conducting separate analyses
allowed for a comprehensive understanding of the
diverse effects that each crisis had on the variables
under investigation.
CR I: 1/1992-4/1993 Crisis
During CR I, the Canadian economy experienced a
significant decrease. The real DCP decreased by
more than 1 percent and the unemployment rate
increased by over 10 percent, [31]. Canadian
companies were dealing with the consequences of
high inflation and this included a significant
decrease in the values of risky investments, the
challenge of managing substantial debts, and also
the impact of falling global commodity prices, [47].
The dollar experienced significant depreciation, and
its impact on trade and investment, [30], [31].
The LR models employed in this study focused on
five specific indexes: ind9Sector25-Consumer
Discretionary Daily Price Index, ind31-Call Loan
Interest Rate, ind34CFMRC-Daily Value Weighted
Index, ind36CFMRC-Value Weighted Under $2
Index, and ind38CFMRC-Value Weighted Over $2
Index to explore their impact on the DCP of five
banks during crisis periods.
The results revealed that among the 23 indexes
assessed, only one of them, the ind38CFMRC-Value
Weighted Over $2 Index, demonstrated a highly
significant effect on DCPs across all the LR models,
except for the LR model of CIBC. This index held
substantial statistical significance at the p < 0.001
level in the models where it was incorporated.
Furthermore, during crisis periods, it is observed
distinct characteristics shaped by a combination of
domestic and global factors, such as economic
downturns, financial market volatility, and external
shocks. The findings suggest that during financial
crises of type CR I, policymakers, investors, and
financial institutions would benefit from focusing
their attention and investments on value-weighted
indexes exceeding $2 along with other factors like
inflation, political news, etc. and these models can
be used for similar future crises.
CR II: 7/1998-10/1998 Crisis
Starting in Asia and spreading all over the world,
this global crisis dramatically affected economies,
including Canada. During this global crisis, CR II,
export demand, particularly from Asia, declined due
to currency, which affected business investment.
The Canadian stock market also experienced a sharp
decline in stock prices, resulting in considerable
volatility on the Toronto Stock Exchange (TSX) and
financial stress for investors, [32]. Therefore,
Canada's GDP growth rate slowed down, and the
unemployment rate increased during this period,
[32]. Regarding the character of the crisis, the
banking sector in Canada was impacted by the crisis
as well, especially, small banks were faced financial
strains due to their ties to the affected regions.
However, Canada's major banks, BMO, BNS,
CIBC, RBC, and TD, exhibited remarkable
resilience and stability during this crisis and they
weathered the storm without significant failures due
to strong regulations, [33], [48].
Developed LR models for CR II incorporated
either one or a pair of the subsequent indexes out of
23 indexes: ind7Sector20-Industrials Daily Price
Index, ind17Sector45-Information Technology
Daily Price Index, ind31-Call Loan Interest Rate,
ind32-Daily Foreign Exchange Rate, ind33CFMRC-
Daily Equal Weighted Index, ind35CFMRC-Equal
Weighted Under $2 Index, but surprisingly, the
index ind38CFMRC-Value Weighted Over $2 Index
was a consistent component in all LR models.
Similar to the LR outcome of the crisis CR I, the
ind38CFMRC-Value Weighted Over $2 Index is a
common index and shows a strong significant
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2451
Volume 20, 2023
positive effect on the DCPs of all five banks during
Crisis II in predicting the DCPs. The rest of the
indexes showed no significant effect on the DCP of
these banks.
The findings of LR models provide valuable
insights into the influence of specific indexes on the
DCP of five banks during the global financial crisis.
The conclusions indicate that both investors and
banks should exercise caution when selecting
investment instruments and consider directing their
investments towards value-weighted indexes
exceeding $2 in similar crisis scenarios that may
arise in the future. Furthermore, these models can be
employed during comparable crises to make
informed investment decisions.
CR III: 5/2007-3/2009 Crisis
This study revealed that several indexes exhibited a
statistically significant relationship with the DCP of
the five banks, while the remaining 18 indexes did
not show any significant effect on their DCPs.
Particularly, ind33CFMRC-Daily Equal Weighted
Index, ind34CFMRC-Daily Value Weighted Index,
ind35CFMRC-Equal Weighted Under $2 Index,
ind36CFMRC-Value Weighted Under $2 Index,
ind37CFMRC-Equal Weighted Over $2 Index, and
ind38CFMRC-Value Weighted Over $2 Index were
found to be significant factors impacting the DCP.
During Crisis III, we identified two common
indexes, namely ind36CFMRC-Value Weighted
Under $2 Index and ind38CFMRC-Value Weighted
Over $2 Index, as key drivers affecting the DCP of
all five banks which contributes to the findings of
[49]. These indexes displayed a significant impact
on the DCP during this crisis period. The character
of Crisis III, which spanned from May 2007 to
March 2009, was marked by a significant financial
crisis known as the global financial crisis or the
Great Recession. Originating in the United States
with the collapse of the subprime mortgage market,
the crisis quickly spread worldwide, leading to a
severe credit crunch, financial instability, and a deep
economic recession. While Canada's financial
system was relatively more stable than some other
countries, it still faced substantial challenges during
the crisis, [33], [38]. Despite the challenging
economic climate, Canada's major banks exhibited
remarkable resilience and stability throughout the
crisis. This can be attributed to prudent lending
practices, stricter regulations, and a conservative
banking culture, [33], [34]. Canada experienced a
contraction in economic activity, declining GDP
growth, rising unemployment rates, and reduced
business investment. Industries such as
manufacturing, housing, and automotive were
particularly affected by the downturn, [33], [34].
Based on the outcomes of LR models, findings
provide valuable insights into the relationships
between specific indexes and the DCP of banks
during Crisis III. The identification of significant
indexes can aid financial institutions and investors
in understanding and navigating through such
challenging economic conditions. In similar future
crises, it is advisable for investors and banks to
consider investing in value-weighted indexes that
have a value exceeding $2 as found in the outcomes
of LR models for CR I and CR II.
CR IV: 9/2014-2/2016 Crisis
The LR models encompassed the examination of
several indexes, including ind7Sector20-Industrials
Daily Price Index, ind15Sector40-Financials Daily
Price Index, ind34CFMRC-Daily Value Weighted
Index, the daily equal-weighted index
(ind33CFMRC), ind35CFMRC-Equal Weighted
Under $2 Index, ind36CFMRC-Value Weighted
Under $2 Index, and ind38CFMRC-Value Weighted
Over $2 Index. Among the 23 indexes assessed,
only the ind38CFMRC-Value Weighted Over $2
Index demonstrated a significant impact on the DCP
of the five banks. Although one of the indexes listed
above was included in the LR models, the common
index with a noteworthy effect on the DCP of all
five banks was the ind38CFMRC-Value Weighted
Over $2 Index. However, it is important to note that
the impact of this index varied among the banks.
While the ind38CFMRC-Value Weighted Over $2
Index positively influenced the DCP of BMO, BNS,
CIBC, and RBC, it had a negative impact on the
DCP of TD.
During the period from September 2014 to
February 2016, Canada did not experience a specific
crisis; however, it faced multiple challenges due to
global economic uncertainties, declining oil prices,
and regional economic impacts, [36], [38]. Being a
significant oil-producing nation, the sharp decline in
global oil prices during this period directly affected
Canada's energy sector and regions heavily reliant
on oil production, such as Alberta, [36], [38]. The
repercussions included job losses, reduced
investments, and slower economic growth in these
areas. Amidst the complex economic landscape, our
findings shed light on the influence of specific
indexes on the DCP of banks. The significance of
the ind38CFMRC-Value Weighted Over $2 Index
underscores its relevance in predicting DCP
fluctuations during this challenging period.
Moreover, the varying impact on different banks
highlights the importance of considering individual
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2452
Volume 20, 2023
bank characteristics and strategies when analyzing
the effects of market indexes on their performance.
In conclusion, this research contributes valuable
insights into the relationships between specific
indexes and the DCP of banks during a time of
economic challenges in Canada. The findings
highlight the role of the ind38CFMRC-Value
Weighted Over $2 Index as a common influential
factor and emphasize the need for prudent risk
management practices for financial institutions,
especially in the context of global economic
fluctuations. This contributes to the findings in [13],
where indexes like the number of circulating stocks,
value, and price-earnings ratio do not have a
significant effect on the stock price.
CR V: 2/2020-3/2020 Crisis
The LR models focused on three specific indexes,
namely ind33CFMRC-Daily Equal Weighted Index,
ind36CFMRC-Value Weighted Under $2 Index, and
ind37CFMRC-Equal Weighted Over $2 Index, out
of the 23 indexes assessed. The results of our LR
analysis provided valuable insights into the
influence of these indexes on the DCP of five banks
during Crisis V.
During Crisis V, which was primarily triggered by
the outbreak of the COVID-19 pandemic, the LR
analysis yielded insightful results. The DCP of
BNS, CIBC, RBC, and TD were positively
influenced by the ind37CFMRC-Equal Weighted
Over $2 Index, while BMO showed a significant
positive relationship only with the ind33CFMRC-
Daily Equal Weighted Index. These findings
suggest that the investments and DCP of banks were
influenced differently based on the character of the
crisis. In previous crises, the ind38CFMRC-Value
Weighted Over $2 Index was the common index
with a generally significant effect on the DCP of
banks. However, during crisis V, the ind37CFMRC-
Equal Weighted Over $2 Index emerged as the
common influential index, except for the DCP of
BMO. The LR results for Crisis V further
underscore the significance of ind33CFMRC-Daily
Equal Weighted Index and ind37CFMRC-Equal
Weighted Over $2 Index in predicting the outcomes
for BMO, BNS, CIBC, RBC, and TD during this
unprecedented crisis period, which contributes the
finding in [49]. The CR V was characterized by the
rapid spread of the COVID-19 virus, leading to
health emergencies and the implementation of strict
public health measures. These measures resulted in
substantial economic disruptions, increased
unemployment rates, and significant market
volatility, [39], [43]. This research highlights the
importance of equal-weighted indexes exceeding $2
in influencing the DCP of banks during CR V. Since
the different indexes have an impact on the DCP of
banks during CR V when compared with the other
crises CR I -CR IV, the banks need to adapt their
investment strategies and risk management
approaches depending on the economic
environments.
This study has several limitations that need to be
acknowledged. Firstly, the analysis focused only on
decreasing periods, as indicated in Table 1, which
may not fully capture the entire scope of the actual
crisis period, potentially limiting the depth of
insights gained. Secondly, missing data during
certain periods were excluded from the assessments,
potentially introducing some bias in the results.
Another constraint is the lack of access to
information about the specific investment tools used
and their respective amounts by the banks. This
missing data could have provided valuable context
and potentially influenced the findings. In this
study, the DCP of banks was utilized as the main
variable of interest. However, using daily returns, as
commonly employed in existing literature, such as,
[11], [14], [15], [17], [20], might offer more
comprehensive insights into the dynamics of bank
performance during crisis periods.
In conclusion, the study revealed that the DCP of
banks depends on the characteristics of crises and
investment instruments of banks in their portfolios,
[9]. Economic, financial, political, and other factors
form the character of crises, [52]. Therefore, each
crisis has its unique characteristics and has a
different level of impact on the DCP of banks.
For further studies, including different factors such
as bank volatiles, inflation rate, open price, daily
call changes, and bid price in the data can be useful
to get more specific indexes that affect the daily
closing prices. Furthermore, using different
methods, which are commonly used in literature,
such as machine learning, [2], [15], [17], [50], [51],
data mining, [2] or structural models can enhance
the robustness of the findings. Although no specific
pattern was found for crises in this study and in
[50], a comparative study may identify a pattern for
different crises. That will help investors and banks
to manage their investments accordingly for the
upcoming crises. In addition, this study can be
repeated with return indexes which are commonly
used in literature.
References:
[1] Cao, M., Predicting the Link between Stock
Prices and Indices with Machine Learning in
R Programming Language, Journal of
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2453
Volume 20, 2023
Mathematics, Vol. 2021, 2021, Article ID
1275637, 10 pages.
htttps://doi.org/10.1155/2021 /1275637.
[2] Zhou, F., Zhang, Q., Sornetted, D., Jiang, L.,
Cascading logistic regression onto gradient
boosted decision trees for forecasting and
trading stock indices, Applied Soft Computing
Journal, Vol. 84, 2019, Article: 105747.
[3] Yang, Y., Hu, X., Jiang, H.,
Group penalized
logistic regressions predict up and down
trends for stock prices, North American
Journal of Economics and Finance, Vol.
59, 2022, Article:101564
[4] OSC: Ontario Securities Commission,
Factors that can affect stock prices,
[Online], https://www.
getsmarteraboutmoney.ca/invest/investment
-products/stocks/factors-that-can-affect-
stock-prices/. (Accessed Date: August 19,
2023).
[5] Egan, J., How Are Stock Prices
Determined: The Factors that Affect Share
Prices of Listed Companies, Personal
Finance, March 29, 2023, [Online],
https://time.com/personal-
finance/article/how-are-stock-prices-
determined/ (Accessed Date: August 19,
2023).
[6] Nugroho, A. P., Kesdu, P. A.C., Fatonah, S.,
Susanti, N. I., The Analysis of Fundamental
Factors Affecting Company Stock Prices
(Case Studies of Companies, East Asian
Journal of Multidisciplinary Research
(EAJMR), Vol. 1, No. 7, 2022, pp. 1429-1448,
ISSN-E: 2828-1519.
[7] Breheny, P., Huang, J., Group descent
algorithms for nonconvex penalized linear and
logistic regression models with grouped
predictors, Stat Comput, Vol, 25, 2015, pp.
173-187. DOI 10.1007/s11222-013-9424-2.
[8] Karimuzzaman, Md., Islam, N., Afroz, S.,
Hossain, M., Predicting stock market price of
Bangladesh: A comparative study of linear
classification models, Annals of Data Science,
Vol. 8, 2021, pp. 21-38.
[9] Khoa, B. T., Vinh, D. P., Hiep, T. T., Quang,
N. D., Huynh, T. T., Predict stock price
movement using long short-term memory
recurrent neural network and support vector
machine, General Management, Vol. 24, No.
194, 2023, pp. 49-59.
DOI:10.47750/QAS/24.194.06.
[10] Tevdovski, D., Stojkoski, V., What is behind
extreme negative returns co-movement in the
South Eastern European stock markets?,
Scientific Annals of Economics and Business,
Vol. 68, No. 1, 2021, pp. 43-61. DOI:
10.47743/saeb-2021-0003.
[11] Jiang, H., Hu, X., Jia, H., Penalized logistic
regressions with technical indicators predict
up and down trends, Soft Computing, Vol 27,
2022, pp. 13677–13688.
https://doi.org/10.1007/s00500 -022-07404-1.
[12] Kambeu, E., Trading volume as a predictor of
market movement: An application of Logistic
regression in the R environment, International
Journal of Finance & Banking Studies, Vol. 8,
No. 2, 2019, pp. 57-69.
[13] Daia, P., Xiong, X., Zhouc, W., A global
economic policy uncertainty index from
principal component analysis, Finance
research letters, Vol. 40, 2021, Article:
101686.
[14] Yang, F., Yang, H., Yang, M., Discrimination
of China's Stock Price m anipulation based
on primary component analysis, 2014
International Conference on Behavioral,
Economic, and Socio-Cultural Computing
(BESC2014), 30 October 2014 - 01 November
2014, DOI: 10.1109/BESC.2014.7059519,
Shanghai, China.
[15] Akyildirim, E., Goncu, A., Sensoy, A.,
Prediction of cryptocurrency returns using
machine learning,
Annals of Operation
Research, Network and Risk Management,
Vol. 297, 2021, pp. 3-36.
https://doi.org/10.1007/ s10479-020-03575-y.
[16] Singh, A., Kaur, P.,
Does US financial
stress explain risk-return dynamics in
Indian equity market? A logistic
regression approach, Vision: The Journal
of Business Perspective, Vol. 1, 2017, pp.
13-22.
https://doi.org/10.1177/0972262917695.
[17] Upadhyay, A., Bandyopadhyay, G., Dutta, A.,
Forecasting stock performance in Indian
market using multinomial logistic regression,
Journal of Business Studies Quarterly, Vol. 3,
No. 3, 2012, pp. 16-39.
[18]
Paivaa, E. V. Savoia, J. R. F., Pricing
corporate bonds in Brazil: 2000 to 2004,
Journal of Business Research, Vol. 62,
2009, Article: 916-919.
[19] Omran, M., Structural break in the Egyptian
stock market: A logistic regression analysis,
International Journal of Business, Vol. 11,
No. 4, 2006.
[20] Mile, B., Evolving efficiency of stock returns
and market conditions: The case from Croatia,
Montenegrin Journal of Economics, Vol. 19,
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2454
Volume 20, 2023
No. 1, 2023, pp. 107-116. DOI:
10.14254/1800-5845/ 2023.19-1.9.
[21] Fengyu, H., Yue, W., Predicting stock price
movement after disclosure of corporate annual
reports: A case study of 2021 China CSI 300
Stocks, IDEAS Working Paper Series from
RePEc, Papers 2206.12528, 2022, arXiv.org,
revised Jul 2022.
[22] Lee, T. K., Cho, J. H., Kwon, D. S., Sohn, S.
Y., Global stock market investment strategies
based on financial network indicators using
machine learning techniques, Expert Systems
with Applications, Vol. 117, 2019, pp. 228-
242.
[23] CHASS Data Centre - University of Toronto,
[Online],
http://clouddc.chass.utoronto.ca.ezproxy.
lib.ryerson.ca/ds/cfmrc/displayTSX.do?ed=20
18 &t=ts&f=daily&lang=en#v2
(Accessed
Date:
May 5,
2021).
[24] Ruzgar, N. S., Chua-Chow, C., Behavior of
banks’ stock market prices during long-term
crises, International Journal of Financial
Studies, Vol. 11, No. 1, 2023,
https://doi.org/10.3390/ ijfs11010031.
[25] Logistic regression,
Ryerson University, on
15 Jan 2022, subject to the Cambridge
Core terms of use, https:/
/doi.org/10.1017/CBO9780511778643.009.
[26] Fagerland, M. W., A generalized Hosmer–
Lemeshow goodness-of-fit test for
multinomial logistic regression models, The
Stata Journal, Vol. 12, No. 3, 2012, pp. 447-
453.
[27] Hosmer, D.W., Lemeshow, S., A goodness-
of-fit test for the multiple logistic regression
model. Communications in Statistics, 1980,
Article: 10, pp. 1043-1069.
[28] Cox, D.R., Snell, E. J., The analysis of binary
data, 1989, 2nd ed. London: Chapman and
Hall.
[29] Newsom, Psy 525/625 Categorical Data
Analysis, Spring 2021, pp.1-13,
[Online],
https://web.pdx.
edu/~newsomj/cdaclass/ho_logistic.pdf
(Accessed Date:
June 5,
2023).
[30] Wilson, T., Dungan, P., & Murphy, S. The
sources of the recession in Canada: 1989-
1992. University of Toronto, 2019,
S2CID 17318693, archived from the original
on 2 March 2019.
[31] Economics Concepts. Recessions,
[Online],
http://www.canadianeconomy.
gc.ca/english/economy/recession.html
(Accessed Date:
March 7, 2022
).
[32] Murray, J., Zelmer, M., Antia, Z.,
International financial crises and flexible
exchange rates: some policy lessons from
Canada, 2022, pp. 358-384,
[Online],
https://www.bis.org/publ/confer08r.pdf
(Accessed Date:
March 7, 2022
).
[33] Bauer, G., Pasricha, G., Sekkel, R., Terajima,
Y., The global financial cycle, monetary
policies, and macroprudential regulations, in
small, open economies, Canadian Public
Policy, Vol. 44, No. 2, 2018, pp. 81-99.
[34] Gordon, S., Recession of 2008–09 in Canada.
The Canadian Encyclopedia, 2017,
[Online],
https://www.
thecanadianencyclopedia.ca/en/article/recessi
on-of-200809-in-canada
(Accessed Date:
March 8, 2022
).
[35] Bank of Canada. Strengthening international
capital and liquidity standards: A
macroeconomic impact assessment for
Canada, 2010, [Online],
https://www.bankofcanada.ca/wp-
content/uploads/2010/09/report.pdf
(Accessed Date:
October 30, 2023
).
[36] Stocker, M., Baffes, J., Vorisek, D., What
triggered the oil price plunge of 2014-2016
and why it failed to deliver an economic
impetus in eight charts, Let’s talk
development, 2018,
[Online],
https://blogs.worldbank.org/developmenttalk/
what-triggered-oil-price-plunge-2014-2016-
and-why-it-failed-deliver-economic-impetus-
eight-charts
(Accessed Date:
March 8, 2022
).
[37] Twin, A., The World's 10 Biggest Oil
Exporters, July 28, 2023,
[Online],
https://www.investopedia.com/
articles/company-insights/082316/worlds-top-
10-oil-exporters.asp
(Accessed Date:
August
18, 2023
).
[38] Mark, S. B., Recession in Canada, The
Canadian Encyclopedia, 2017,
[Online],
https://www.thecanadian
encyclopedia.ca/en/author/mark-s-bonham
(Accessed Date:
July 7, 2022
)
[39] Assous, H. F., Al-Najjar, D., Consequences of
COVID-19 on banking sector index: Artificial
neural network model. International Journal
of Financial Studies, 2021, Vol. 9, pp. 67.
https://doi.org/10.3390/ijfs9040067.
[40] Rehman, M, U., Kang, S. H., Ahmad, N., Vo,
X. V., North American Journal of Economics
and Finance, Vol. 58, 2021, 101526.
[41] Associated Press, UN Chief Says COVID-19
is worst crisis since World War II, New York
Times, 2020,
[Online],
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2455
Volume 20, 2023
https://www.nytimes.com/aponline/2020
/03/31/world/ap-un-virus-outbreak-united-
nations-lst-ld-writethru.html
(Accessed
Date:
July 7, 2022
)
[42] Poirier, W., Ouellet, C., Rancourt, M.,
Bechard, J., Dufresne, Y., (Un)Covering the
COVID-19 pandemic: Framing analysis of the
crisis in Canada, Canadian Journal of
Political Science, Vol. 53, 2020, pp. 365-371.
DOI: 10.1017 /S0008423920000372.
[43] Demers, E, Hendrikse, J. J., P., Lev, B., ESG
did not immunize stocks during the COVID-
19 crisis, but investments in intangible assets
did, Journal of Business, Finance &
Accounting, Vol. 48, No. 3-4, 2021, pp. 433-
462. https://doi-org.ezproxy
.lib.torontomu.ca/10.1111/jbfa.12523.
[44] Asteriou, D.., Pilbeamn, K., Sarantidis, A.,
The behaviour of banking stocks during the
financial crisis and recessions. Evidence from
changes-in-changes panel data estimations,
Scottish Journal of Political Economy, Vol.
66, No. 1, 2019. DOI:10.1111/sjpe.12191.
[45] Kenton, W., The big Five: Here are Canada’s
largest banks by total assets. Investopedia,
[Online],
https://www.investopedia.com/
terms/b/bigfivebanks.asp (Accessed Date:
November 21, 2022.
[46] Spyrou, S., Are broad market shocks
anticipated by investors? Evidence from
major equity and index options markets,
International Review of Financial Analysis,
Vol. 20, No. 3, 2011, pp. 127-133,
https://doi.org/10.1016/j.irfa.2011.02.002.
[47] Thiessen, G. S., Recession of 2008–09 in
Canada, The Canadian Encyclopedia, 2017,
[Online],
https://www.thecanadianencyclopedia.ca/en/ar
ticle/recession-of-200809-in-canada
(Accessed Date: March 8, 2022).
[48] Zhang, J., Oshawa This Week Tuesday, June
6, 2023, [Online],
https://www.durhamregion.com/opinion/
canadas-banking-system-a-pillar-of-strength-
amid-global-financial-
turbulence/article_810c93ad-7b67-5651-bc47-
4d397afe9d0b.html (Accessed Date: June 9,
2023).
[49] Zhang, X., Ding, Z., Hang, J., He, Q., How do
stock prices indices absorb the COVID-19
pandemic shocks?, North American Journal of
Economics and Finance, Vol. 60, 2022,
Article: 101672.
https://doi.org/10.1016/j.najef.2022.101 672.
[50] Ayala, J., Garcia-Torres, M., Noguera, J. L.
V., Gómez-Vela, F., Divina, F., Technical
analysis strategy optimization using a
machine learning approach in stock market
indices, Knowledge-Based Systems, Vol. 225,
2021, Article: 107119.
[51] Sheta, A., Prediction stock market exchange
prices for the reserve bank of Australia using
auto-regressive with eXogenous input neural
network model, WSEAS Transactions on
Systems, Vol. 17, 2018, pp. 79-88.
[52] Lv, D., Huang, Z., Li, M., Xiang, Y.,
Selection of the optimal trading model for
stock investment in different industries, PLos
One, Vol. 14, No. 2, e0212137.
https://doi.org/10.1371/journal.pone.
0212137.
Contribution of individual authors to the
creation of a scientific article (ghostwriting
policy)
Nursel Selver Ruzgar collected data, made statistical
analysis, wrote paper, reviewed and edited the paper
The author has read and agreed to the published
version of the manuscript.
Sources of funding for research presented in a
scientific article or scientific article itself
This research received no external funding.
Conflict of Interest
The authors have no conflicts of interest to declare.
Creative Commons Attribution License 4.0
(Attribution 4.0 International, CC BY 4.0)
This article is published under the terms of the
Creative Commons Attribution License 4.0
https://creativecommons.org/licenses/by/4.0/deed.en
_US
WSEAS TRANSACTIONS on BUSINESS and ECONOMICS
DOI: 10.37394/23207.2023.20.209
Nursel Selver Ruzgar
E-ISSN: 2224-2899
2456
Volume 20, 2023
