Learning
from the works of Sharifi-Renani and Mirfatah, Hamida, Kyereboah-Coleman and
Agyire-Tettey, the relationship between economic growth and revenue can be
modeled as follows [44-46]:
LGDPRt=?0+?1LINFt+?2LFDIt+?3LGNSt+?4LPOGRt+?5LDIt+?6LGTRt+?7LSAVt+
?t …....(4)
Table
4, shows the summary statistics of all the variables under study in their raw
form. It shows the mean, maximum, minimum and standard deviations of all
variables. The skewness, kurtosis and Jarquebera statistics of all variables
shown do not fully indicate the true nature of the data series since the
probability value of Jarquebera statistics of all the series are shown to be
less than the acceptable 0.05 for LGDPR, LGTR, LDI, LSAV, LINF, LFDI, LGNS and
LPOGR indicating non-normality of the series (Table 4).
Table
4 reports the results of the unit root test the ADF test, the study concludes
that variables, LFDIY, LTOY, LGCFY, LREER and LGDPY are of mixed level of
integration. Some at I(0) and others at I(1), then we apply the bound test to
test for cointegration among variables. The KPSS results confirm that LFDI and
LDI are integrated at I(2), therefore our dmax is 2 (Table 5).
The
KPSS test has confirmed that the order of integration is 2. We now use the TY
approach for the causality test. The determination of optimal lag length k is
conducted by applying usual lag selection procedure. When choosing the correct
lag order, the VAR is crucial for accurate inference. Several information
criteria are used to select the optimal lag order, including Akaike Information
Criterion (AIC), Hannan-Quinn Criterion (HQIC), and Schwarz Information
Criterion (SIC). These criteria balance model fit with model complexity,
penalizing models with more parameters.
Table 7 indicates the appropriate lag of 1 since the AIC, SC and HQ all
have an italic on 1 as shown. The existence of cointegration is confirmed in table
4.4 among the variables. The study then estimates the Toda-Yamamoto causality
and Table 8 reports the results in The Toda-Yamamoto causality test is based on
the modified Wald statistic known as the MWald statistic (Table 6).
LR:
sequential modified LR test statistic (each test at 5% level)
FPE: Final prediction error
AIC: Akaike information criterion
SC: Schwarz information criterion
HQ:
Hannan-Quinn information criterion
Given
the trace test and the maximum eigenvalue (?-max) statistics derived under the
Johansen cointegration test, the study confirms that there is cointegration
among the variables. The trace test indicates one cointegrating vector, while
the maximum eigenvalue test suggests two cointegrating vectors (Table 7). After
confirming the existence of cointegration among the variables, the study
estimated the Toda-Yamamoto causality and reports the results in Table 4.5
Unlike the Granger causality test that uses the conventional Wald statistic,
the Toda-Yamamoto causality test is based on the modified Wald statistic known
as the MWald statistic. The augmented lag length for the MWald statistic, p, is
set to 3, which is computed from the sum of the VAR lag length (k) plus the
maximum order of integration (d). That is to say: p = (k + d); leads to p = (2
+ 1) = 3.
The
Toda-Yamamoto causality results in Table 8 suggests that there is
unidirectional significant at 5% causality from real GDP growth (LGDPR) to
government revenue (LGTR), indicating that LGDPR cause LGTR in Uganda for the
period understudy to grow.
This
finding can be partly attributed to GDP growth in Uganda boosts government
revenue primarily by expanding the tax base through increased industrial,
agricultural, and service sector activities, driving higher corporate income
tax, VAT, and PAYE collections. In addition, it has been established from this
study that LGTR unidirectional significantly at 1% causality real growth rate
(LGDPR). This can be so because Government revenue boosts real GDP growth in
Uganda by financing critical public investments, such as infrastructure
development, education, and healthcare, which enhance productive capacity.
Increased revenue, including through digital tax systems (e.g., EFRIS) and tax
base expansion, reduces reliance on debt and funds economic stimulation. The LGDPR is found to granger cause
inflation (LINF) significantly at 10%. This is probably in Uganda, real GDP
growth Granger-causes inflation primarily through demand-pull pressures, where
increased economic activity and income levels outpace the supply of goods and
services. Rapid expansion raises consumer demand, leading to higher prices,
especially when coupled with supply-side bottlenecks characteristic of a
developing economy. In this study, LINF
was found not to granger cause LGDP growth. Also, LGDPR was found to
Granger-cause domestic savings (LSAV) significantly at 5%. This is probably
because past economic growth helps predict and increase future savings, acting
as a one-way, short-to-long-term driver. This occurs because rising incomes
from GDP growth boost disposable income, allowing for increased savings
capacity among households and firms. But LSAV is not found to granger cause
real GDP growth.
The
study established that LGDPR does Granger-cause domestic investment (LDI)
significantly at 5%. This is probably because Real GDP growth in Uganda
Granger-causes domestic investment, specifically private sector investment,
through a demand-pull mechanism, where higher economic growth increases
business confidence, higher consumption, and better market performance. This
means past growth levels in Uganda help predict future increases in domestic
capital formation. But domestic
investment does not granger cause LGDPR as established by the study. The study
established that LGTR Granger-cause LINF significantly at 1%. In Uganda, real
government revenue contributes to inflationary pressures primarily when it
fails to keep pace with expenditures, leading to budget deficits that are
financed through inflationary means. Although higher tax revenues are meant to
fund development, structural inefficiencies and deficits mean that government
fiscal operations often drive-up demand and prices, a relationship validated
through Granger causality tests. Also, LINF in this study is found to
significantly at 5% to ganger cause LGTR. In Uganda, research like Ssebulime
indicate a Granger causal relationship where inflation, often driven by money
supply or external shocks, impacts government revenue, specifically tax
revenue, in both the long and short run. Inflation influences nominal tax
bases, with high inflation potentially causing financial market frictions that
reduce investment and, consequently, long-term tax revenue [47]. The study
established that LGTR Granger-causes institutional quality (LGNS) significantly
at 5%. In Uganda, government revenue, particularly from improved tax
administration, Granger-causes better institutional quality by providing the
fiscal capacity to strengthen regulatory frameworks and enhance government
effectiveness. Increased tax revenue, managed through institutions like the
Uganda Revenue Authority, allows for better service delivery, fostering trust
and reducing corruption, which in turn improves overall governance. But LGNS is not found to granger cause LGTR.
Also, LGTR is found to granger cause domestic savings in Uganda significantly
at 1% level. Simply because In Uganda, government revenue, particularly non-tax
revenue and tax, demonstrates a causal relationship with economic activity,
which in turn influences domestic savings. Evidence suggests a unidirectional,
long-run relationship where increased economic growth, often spurred by
effective government revenue management, leads to increased domestic savings
rather than vice-versa.
But
LSAV is not found to granger cause domestic revenue probably because, some
studies on economics, like Samuel and Abebe, empirically say in Uganda,
domestic savings often do not Granger-cause government revenue due to a
combination of a large, untaxed informal sector, the nature of savings being
channeled into non-productive consumption, and structural issues in the tax
system [48]. In Uganda, the causal relationship tends to run in the opposite
direction—economic growth (GDP) and revenue collection drive savings, rather
than savings driving government revenue. The study established that LGTR does
at 5% significantly Granger-cause Population growth rate (LPOGR). This is
probably government revenue, through several mechanisms, primarily centered on
fiscal policy's impact on public health, social services, and economic
infrastructure. Increased government revenue allows for higher public spending
on healthcare, sanitation, and medical infrastructure, which reduces mortality
rates and can lead to a higher population growth rate. Higher revenue enables
governments to invest in education and social services. Improved access to
education, particularly for women, can shift population dynamics. Revenue
generation helps fund infrastructure (water, electricity, transportation) that
supports a larger population, facilitating growth by improving living
conditions. Revenue is used to create a conducive environment for development.
A stable, funded economy can support higher population growth by enhancing
overall economic security. But LPOGR was
found not to granger cause government revenue. In addition, LGTR was found to
Granger-cause domestic investment (LDI). This is probably because the funds
collected through taxes, duties, and other revenues act as the financial
foundation for public capital expenditures like infrastructure, education and
health that directly facilitate and stimulate private sector investment. In
Granger causality terms, this means past levels of government revenue help
predict future levels of domestic investment, indicating that revenue
mobilization is a necessary precursor to public investment-driven economic
growth. It was also established that LDI
significantly granger causes government revenue at 10% level. This is so
because Domestic investment Granger-causes government revenue because increased
investment activity boosts economic growth, which in turn expands the tax base
and raises non-tax revenue for the government. This relationship is often
characterized as a unidirectional causality where private sector capital
formation, such as investments in infrastructure and machinery, increases
overall production and consumption, allowing governments to collect more taxes
on profits, sales, and income.
The
inflation (LINF) was found to significantly Granger-cause LGTR at 1%. This is
probably because Inflation primarily because rising prices increase nominal tax
bases (income, consumption, and asset values) faster than tax brackets or
exemptions are adjusted, a phenomenon known as bracket creep or the
Olivera-Tanzi effect. It is an economic phenomenon where high inflation leads
to a significant decline in the real value of government tax revenue. The
effect occurs primarily due to the collection lag, which is the delay between
when a taxable event happens (like a sale or earning income) and when the
government actually receives the tax payment. As inflation drives up nominal
prices, sales tax revenues and nominal corporate profits rise, leading to
higher tax collections, allowing inflation to serve as a form of implicit
taxation. Also, LINF is found to Granger-cause LPOGR and LDI significantly at
1% respectively. Inflation Granger-causes population growth rate primarily
because rising prices, particularly for food and basic necessities, force
behavioral shifts in household planning, resource allocation, and, in some
contexts, increased mortality or reduced fertility due to economic hardship.
High, persistent inflation acts as a constraint on disposable income,
influencing demographic decisions over time. In addition, Inflation
Granger-causes domestic investment primarily because rising price levels signal
economic instability, increase input costs, and erode purchasing power, which
directly forces investors to adjust, delay, or reduce capital expenditures.
High or volatile inflation creates uncertainty, reducing the profitability of
long-term projects and shifting capital away from productive investment towards
hedging. Inflation generally hurts domestic investment by reducing real
returns, increasing costs, and creating uncertainty, although low, stable
inflation can sometimes encourage it. High inflation erodes purchasing power,
causes interest rates to rise, and discourages long-term projects due to increased
risks. Key impacts include reduced value of fixed-income assets and a
preference for real assets over financial ones (Table 8).
The
study found out that foreign direct investment (LFDI) does Granger-cause LGTR
significantly at 10%, because Foreign Direct Investment (LFDI) primarily by
stimulate economic activity, which directly boosts tax bases through increased
corporate income taxes, employee income taxes, and consumption-based taxes like
VAT. LFDI inflows act as a catalyst for economic growth, generating employment
and boosting productivity, which consequently leads to higher revenue
generation for the host government. In
addition, LFDI was found to Granger-cause domestic savings (LSAV) and domestic
investment (LDI) significantly at 1% level respectively. This is so because
domestic savings boosts economic growth, raising household incomes, and
enhancing financial sector efficiency, which collectively increase the capacity
for local savings. LFDI acts as a catalyst that stimulates productive economic
activity and, in some contexts, directly drives capital accumulation. Foreign Direct Investment (FDI) boosts
domestic investment because, by acting as a catalyst for local industrial
growth through technology spillovers, knowledge transfers, and increased
competition. FDI fills capital, technology, and skill gaps in host countries,
encouraging domestic firms to invest in upgrades and expand capacity to remain
competitive.
Residual
and stability diagnostics tests
The
Breusch-Godfrey (BG) test is a robust method for detecting serial correlation.
The BG test uses residuals from the original regression as the dependent
variable run against initial regressors plus lagged residuals and null
hypothesis is the coefficients of the lagged residuals are zero. From the
results in Table 9 the null hypothesis is accepted and concluded that there is
no serial correlation in the model. This means there's no statistically
significant relationship between successive values of a variable over time. It
indicates that the current value of a variable is not influenced by its past
values. A serial correlation value of zero suggests this independence. The
current observation is not correlated with its previous observations,
indicating no predictive power from past values. Residue stability tests
determine how well a substance or residue maintains its integrity over time
when stored under specific conditions. These tests are crucial for ensuring the
accurate analysis of residues, demonstrating the stability of pesticides in
crops, and verifying the stability of residues in various products (Table 9).
The results as shown in Table 10 show that there is no heteroscedasticity since
the null of no heteroskedasticity is accepted. No heteroskedasticity means the
errors in a model have a constant variance, meaning the spread of the residuals
is consistent across all values of the independent variable. In simpler terms,
it means the variability of the dependent variable (the thing being predicted)
is the same at all levels of the independent variable (Table 10).
The
Ramsey Reset Test, also known as the Ramsey Regression Equation Specification
Error Test, is a diagnostic tool used to test if the functional form of a
regression model is appropriately specified. Specifically, it checks if
non-linear combinations of the independent variables help explain the dependent
variable, indicating potential model misspecification. In essence, it helps
determine if a linear model is the best representation of the relationship
between variables or if a non-linear model would provide a better fit. From the
results in Table 11 the null is accepted and conclude that there is no
misspecification in the model (Table 11). A CUSUM (Cumulative Sum) chart is a
statistical quality control tool used to monitor a process and detect small shifts
in the process mean. It works by plotting the cumulative sum of deviations from
a target value, helping to identify changes that might be missed by traditional
control charts. CUSUM charts is a valuable tool for monitoring processes and
detecting subtle changes that might not be visible with other control chart
methods, enabling timely corrective actions and improving process stability as
shown in (Figure 3). The CUSUM of Squares test is a statistical test used to
assess the stability of regression models, especially in time series analysis.
It's designed to detect systematic changes in the model parameters, including
the variance of the error term, over time. Specifically, it looks for sudden
shifts or changes in the squared values of the residuals, which can indicate
instability in the model's parameters as indicated in (Figure 4).