Professional Documents
Culture Documents
0948-BH-BAF-19 Muhammad Ali Haider Financial Econometrics Project
0948-BH-BAF-19 Muhammad Ali Haider Financial Econometrics Project
Introduction:
The dataset on which the commands on STATA software have been run is from
an article the name: Determinants of profitability of Islamic banks in Pakistan.
The aim of this study is to identify the factors that influence Islamic banks in
Pakistan. Data from five Islamic banks in Pakistan were gathered by the author
during a ten-year period, from 2006 to 2015. The dependent variable was
Return on Asset, while the independent variables were GDP, Size, Inflation,
and Liquidity. Making use of STATA software and using panel regression, the
researcher discovered that the size of Islamic banks and liquidity are the only
two significant variables at 5% in the fixed model effect. Size has a detrimental
impact on the profitability of Islamic banks, although liquidity has a beneficial
one. The findings will assist the Islamic banking sector formulate appropriate
policies.
Question No#1:
These were the research question on which this study is entirely based on:
Question No#2:
Dependent Variable:
In this study the researcher has selected the profitability as a dependent
variable. However, profitability is a term which can be signified using various
variables but in this study the researcher opted to use ROA as the dependent
variable
Independent Variable:
1. GDP: GDP stands for Gross Domestic Product. Total production and
services over specified fiscal period.
2. Inflation: an increase in an economy's overall price level for products
and services.
3. Size: The total assets of the company can be used to measure the size of
the bank. The most crucial tool for any company to increase its profit is
its assets. It is not a given that if one company has more assets than
another, they will also make more money. Strategies and policies also
have a crucial influence.
4. Liquidity: Any company's liquidity is its capacity to pay all of its
obligations when they become due. Total loans to deposit ratio is a
liquidity variable used by the author.
Question No#3:
There are five Islamic banks in Pakistan, which are working under Shariah
Compliant mode, taken as sample. These are the strategies used to locate
data, gather data from information sources, and evaluate information. The
information for the financial and economic variables (dependent and
independent variables) is gathered from secondary sources, such as reports
from the State Bank of Pakistan, final reports that have been made public or
financial statements that have been posted online by banks, reports from the
Pakistan Statistics Bureau, and reports from the Economic Survey of Pakistan.
Question No 4:
The type of data that the researcher used in this study is panel data
We used the following method to ascertain whether the data used in the study
is panel data:
Panel data analysis has a number of benefits, including the ability to effectively
handle endogeneity problems, capture individual-level variability, and
compensate for time-invariant unobserved components. Several frequently
used methods for panel data analysis include:
First Differences and Fixed Effects: This approach involves differencing the data
to remove time-invariant heterogeneity and estimating the effects using fixed
effects models.
Dynamic Panel Data Models: These models incorporate lagged dependent
variables or other dynamic specifications to account for the temporal
dependencies within the panel data.
By employing panel data analysis techniques, you can effectively analyze the
determinants of profitability in the Islamic banking sector of Pakistan while
accounting for both time and cross-sectional variation within the dataset.
Question No#5:
Question No#6:
In panel data analysis, the fixed effects model is a statistical technique often
used in econometrics to address variances among individual entities. It is
especially useful for analysing data gathered over time from many entities. The
focus of the fixed effects model is on capturing entity-specific effects, which
persist over time yet vary between entities. These consequences could result
from several elements that are connected to each entity. The author
discovered that the size of Islamic banks and liquidity are the two significant
variables at 5% in the fixed model effect. Here, the profitability of Islamic banks
is positively correlated with liquidity but negatively correlated with size.
The author then applies a random effect to the model of Islamic banks. The
author once more discovered that size and liquidity are strongly related to the
profitability of Islamic banks when using a random effect model. Both factors
have a 5% significance level. Size and liquidity have opposing effects on each
other. It is crucial to use the Hausman Test to choose the best model. The
author chose the most suitable model from both of them using the same test.
The P-value came less than 0.05 which suggests that fixed model is more
suitable than random effect model.
Question No#7:
For this work, the Islamic bank model equation was created. All variables,
including the macroeconomic and financial ones, are supported by pertinent
theories and earlier research. The equation of model is as below:
Equation:
INF = Inflation
σ = Error Term
The author has used log variable transformation equation in order to fulfill
assumption of panel regression analysis. After using the log variable
transformation the following equation emerges:
Question Number 8:
1. Linearity: The researcher expects that the dependent variable and the
independent factors have a linear relationship. This presumption states
that a straight line or a linear combination of variables can appropriately
depict the relationship.
2. Independence: It is assumed that each observation in the dataset is
unrelated to the others. This presumption suggests that none of the
observations are affected by or connected to one another.
3. Normality: It is assumed that each observation in the dataset is
unrelated to the others. This presumption suggests that none of the
observations are affected by or connected to one another.
4. Homoscedasticity: All levels of the independent variables are considered
to have the same variance of the residuals. The spread of the residuals,
in other words, ought to be comparable across the range of values for
the independent variables.
5. No Multicollinearity: It is presumed that there is no correlation between
the independent variables employed in the regression model. Coefficient
estimates can become unstable and incorrect due to multicollinearity,
which develops when independent variables are strongly connected.
Question No#9:
Hausman Test: To decide between the fixed effects and random effects
models, the author used the Hausman test. The Hausman test evaluates the
relationship between the independent variables and the effects that are
unique to each individual. The test implies that the individual-specific effects
are unrelated to the independent variables if it reveals that the random effects
model is suitable.
Question No#10:
Why do you think the research has chosen the specific methodology? Do you
think that there are other estimation/analysis techniques which shall align
with the data type and research question.
Based on the information provided in the article, the research has chosen
panel regression analysis as the methodology for several reasons:
Panel Data Nature: The study makes use of panel data, which are observations
made over time on numerous organisations (banks). In order to analyse both
cross-sectional and time-series variations, panel regression analysis was
created specifically to handle this kind of data. The researcher can accurately
capture the dynamics of profitability in Islamic banks over time and take
individual-specific impacts into account by utilizing panel regression.
Random Effects Model: The research article employs the fixed effect model,
but an alternative approach could be the random effects model. The random
effects model assumes that individual-specific effects are uncorrelated with
the independent variables. If the Hausman test had indicated that the random
effects model is appropriate, it would have been another valid choice.
However, the researcher in this study determined that the fixed effect model
was more suitable based on the results of the Hausman test.
Dynamic Panel Data Models: Dynamic panel data models, such the Arellano-
Bond or the system GMM (Generalised Method of Moments) estimator, may
be taken into account if the research topic entails studying the dynamics and
temporal lags in the relationship between the independent variables and
profitability. These models offer more accurate estimates when serial
correlation is present because they take into account both the fixed effects and
any endogeneity resulting from lagged dependent variables.
Question No#11:
Conclusion:
Appendix:
delta: 1 year
------------------------------------------------------------------------------
------------------------------------------------------------------------------
-------------+----------------------------------------------------------------
-------------+----------------------------------------------------------------
sigma_u | 1.5247605
sigma_e | 1.8525438
------------------------------------------------------------------------------
F test that all u_i=0: F(4, 39) = 5.44 Prob > F = 0.0014
------------------------------------------------------------------------------
------------------------------------------------------------------------------
-------------+----------------------------------------------------------------
-------------+----------------------------------------------------------------
sigma_u | 1.1170932
sigma_e | 1.8525438
------------------------------------------------------------------------------
------------------------------------------------------------------------------
------------------------------------------------------------------------------
chi2(4) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 218.30
Prob>chi2 = 0.0000
------------------------------------------------------------------------------
. xttest2
__e0 1.0000
-------------------------------------------------------------------------------
. xttest3
Modified Wald test for group wise heteroskedasticity in fixed effect regression
model
Prob>chi2 = 0.2420
---------------------------------------------------------------------------------
F( 1, 4) = 0.539
End