Chapter Three Research Methodology

CHAPTER THREE
RESEARCH METHODOLOGY
3.0Introduction
This chapter presents the research methodology for the study. Primarily, the methodology
explains the data collection and estimation method used in the study. First, there is a discussion
of the research design which underscores the reliability of the findings and conclusions drawn
from this study. Thus, the population, sample size, and sampling techniques, data sources and
collections as well as data management and analysis. Second, the model estimation and
specification is outlined. Basically, these include: data envelopment analysis and pooled OLS
model with panel corrected standard errors (PCSEs).
3.1Research design
Panel data is a special type of pooled data in which the same cross-sectional units is surveyed
over time. Panel data is preferred in this study because it has the advantages of: (1) providing
more accurate inference of model parameters; and (2) controlling the impact of omitted variables
or time heterogeneity (Gujarati & Porter, 2009). Panel data comes in two-folds: a balanced panel
and an unbalanced panel. With the former, the same units appear on each time period whilst with
the later, some units do not appear in each time period often due to attrition (Gujarati & Porter,
2009). The use of panel data method makes it possible to obtain more data points. There are a
variety of methods of estimating panel data which includes: pooled OLS model, fixed effect
model and random effect model. In pooled OLS model, data are elements of both time series and
cross-sectional data which are combined without recourse to the cross-sectional and time series
nature of the data. The pooled OLS model has the benefit of yielding more accurate predictions
of individual outcomes (Gujarati & Porter, 2009). The fixed effect model allows for
heterogeneity among the entities by allowing them to have their intercept value (Gujarati &
Porter, 2009). On the Other hand, the random effect model assumes that the discrepancy across
entities is random and uncorrelated with the explanatory variables (Gujarati & Porter, 2009). The
random effect model also allows for heterogeneity among entities but the entities have a common
expected value of the intercept (Gujarati & Porter, 2009). To achieve the objectives of the study,
the researcher made use of two stage approach (Banker & Natarajan, 2008). In line with this
approach, a decision had to be made concerning the appropriate model to be used. In order to
decide the appropriate model between FE model and RE model, Hausman test is performed. The
p-value for the Hausman Test is insignificant [chi2 (6): 6.22; p-value: 0.3992]; hence the random
effect model is preferred. The researcher therefore needed to decide the appropriate model upon
which the findings of this study will be interpreted. Thus, a comparison test was conducted using
Breusch and Pagan Lagrangian multiplier test for random effects. The test results showed that
pooled OLS was more appropriate than random effects [chibar2 (01): 0.00; p-value: 1.000].
However, given that there are criticisms of OLS estimates as being inefficient in the face of panel
data since the errors are likely to show panel heteroskedasticity and contemporaneous
correlation (Beck, 2001, p. 278), the researcher adopted Beck and Katz (1995) panel corrected
standard errors (PCSEs) to ensure that the sampling variability in OLS estimates of is
measured correctly, similar to studies by Alhassan & Ohene-Asare (2016). Hence, the results of
this study was based on OLS-PCSE.

3.2Population
The study is based on listed and non-listed banks in Ghana. This population is chosen to give a
true representation of banking industry in Ghana.
3.3Data and sample
Data to be used in this study comprises of year-end financial characteristics of banks in Ghana.
The sample consists of selected commercial banks for the period of 2010 to 2015. We collect
financial data between 2010 and 2015 from annual reports of the banks in the sample. The
observation period from 2010 to 2015 is used in order to capture the effect of the universal
banking license introduced in 2003. Only banks with six years financial data are considered in
order to provide a better representation of banks performance in terms of efficiency. The final
sample will consist of all commercial banks that have operated in Ghana from/before 2010. The
study employs an unbalanced panel data estimation of OLS regression modeling based on 119
bank-year observations. Data on corporate governance variables are manually extracted from the
various years annual reports of the banks in the sample based on content analysis.
3.4 Corporate governance measures
Basically, these measures only include board diversity, board independence and trust tenets of the
corporate governance mechanisms. They have been chosen because of their importance to the
banking industry (Basel Committee, 2006; Walker, 2009).
3.4.1 Control variables
As there is the likelihood of omitted variables bias, four other variables were incorporated to
control for the possible effects of bank efficiency. In essence, we control for banks size (Bokpin,
2013; Afrifa and Tauringana, 2015), leverage (Afrifa and Tauringana, 2015), and capital
adequacy ratio (Bokpin, 2013; Ramly et al, 2015)
3.5 Regression model
The following regression models are estimated to examine whether corporate governance is
related to bank performance.
NIMi,t = i + 1BDi,t + 2BIi,t + 3Tru_WOi,t + 4lnSIZEi,t + 5LEVi,t + 6CARi,t + i,t +

i,t ..(eq.1)
ROAi,t = i + 1BDi,t + 2BIi,t + 3Tru_WOi,t + 4lnSIZEi,t + 5LEVi,t + 6CARi,t + i,t +

i,t .(eq. 2)
ROEi,t = i + 1BDi,t + 2BIi,t + 3Tru_WOi,t + 4lnSIZEi,t + 5LEVi,t + 6CARi,t + i,t +

i,t (eq. 3)
Where i is bank 1 to 119; t is 2010 to 2015 financial year; NIMi,t,; ROAi,t; ROE the performance
indicators for bank i in period t; BDi,t the proportion of female directors on the board for bank i in
period t; Tru_WOi,t Trust Without (dummy variable assigned the value of 1 if bank i experiences
increase in deposit from customers in period t and 0 if otherwise); lnSIZEi,t is natural logarithm
for total assets for bank i in period t; LEVi,t the ratio of total liabilities to total assets for bank i in
period t; CARi,t the ratio of equity to total assets for bank i in period t; ROAi,t the ratio of profit
after tax to total assets for bank i in period t; ROEi,t the ratio of profit after tax to equity
investment for bank i in period t; NIMi,t the difference between interest income and interest
expenses as a percentage of total assets for bank i in period t; i the time invariant effect and i,t
the time variant effects.

References
Banker, R. D., & Natarajan, R. (2008). Evaluating contextual variables affecting productivity
using data envelopment analysis. Operations research, 48-58.
Beck, N. (2001). Time-series-cross-section data: What have we learned in the past few years?
Annual review of political science, 4(1), 271-297.
Galagedera, D., & Silvapulle, P. (2003). Experimental evidence on robustness of data
envelopment analysis. Journal of the Operational Research Society, 54(6), 654-660.
Gujarati, D. N., & Porter, D. (2009). Basic Econometrics (International Edition ed.). McGraw-
Hill Irwin.
Raab, R. L., & Lichty, R. W. (2002). Identifying subareas that comprise a greater metropolitan
area: the criterion of county relative efficiency. Journal of Regional Science, 42(3), 579-
594.

Chapter Three Research Methodology

Uploaded by

Copyright:

Available Formats

Chapter Three Research Methodology

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter Three Research Methodology

Uploaded by

Copyright:

Available Formats

CHAPTER THREE

model with panel corrected standard errors (PCSEs).

this study was based on OLS-PCSE.

true representation of banking industry in Ghana.

3.3Data and sample

3.4 Corporate governance measures

banking industry (Basel Committee, 2006; Walker, 2009).

3.4.1 Control variables

adequacy ratio (Bokpin, 2013; Ramly et al, 2015)

3.5 Regression model

related to bank performance.

NIMi,t = i + 1BDi,t + 2BIi,t + 3Tru_WOi,t + 4lnSIZEi,t + 5LEVi,t + 6CARi,t + i,t +

ROAi,t = i + 1BDi,t + 2BIi,t + 3Tru_WOi,t + 4lnSIZEi,t + 5LEVi,t + 6CARi,t + i,t +

ROEi,t = i + 1BDi,t + 2BIi,t + 3Tru_WOi,t + 4lnSIZEi,t + 5LEVi,t + 6CARi,t + i,t +

the time variant effects.

using data envelopment analysis. Operations research, 48-58.

Annual review of political science, 4(1), 271-297.

Galagedera, D., & Silvapulle, P. (2003). Experimental evidence on robustness of data

envelopment analysis. Journal of the Operational Research Society, 54(6), 654-660.

You might also like