The Impact of Government Effectiveness on

Corruption Levels Across Countries (1995–2020): A

Panel Data Analysis
This study explores the long-standing question of whether effective governance can truly curb
corruption, drawing on panel data from 193 countries over a 25-year period (1995–2020). By
applying a range of econometric techniques—including pooled OLS, fixed effects, random
effects, and two-way fixed effects models—the research aims to isolate the influence of
government effectiveness on corruption control, while accounting for key factors like GDP per
capita growth, democratic institutions, urbanization, unemployment, and income inequality.

The findings are both clear and consistent: countries with more effective governments tend to
experience significantly lower levels of corruption. This relationship remains robust across all
models, with government effectiveness emerging as the strongest and most reliable predictor of
improved corruption outcomes. Interestingly, variables like economic growth and unemployment
show limited or no significant effect, underscoring that prosperity alone does not guarantee
cleaner governance.

The study also confirms the importance of institutional integrity, as supported by statistical tests
such as the Hausman and Breusch-Pagan LM tests, which highlight the appropriateness of fixed
effects modeling in capturing unobserved country-level dynamics. Ultimately, the results
reinforce a simple but powerful conclusion—good governance isn’t just desirable, it’s
essential. For nations striving to combat corruption and build trust, investing in the capacity,
accountability, and effectiveness of public institutions may be the most impactful path forward.

Chapter 1: Introductory Discussion

1.1 Introduction
Corruption undermines institutional trust and hampers sustainable development. It affects public
service delivery, economic stability, and governance quality. Among the several factors that
influence corruption levels, government effectiveness is critical. Government effectiveness
reflects the quality of public institutions and the civil service, policy formulation, and
implementation capacity. This study seeks to analyze the impact of government effectiveness on
corruption levels using a panel data approach across countries between 1995 and 2020.

This analysis also incorporates control variables such as rule of law and GDP per capita to
account for omitted variable bias. The main objective is to evaluate whether higher government
effectiveness significantly reduces corruption levels across countries over time.

Chapter 2: Review of Literature

2.1 Literature Review

Several empirical studies emphasize the relationship between governance and corruption.
Kaufmann et al. (2009) argue that strong governance structures significantly curb corruption.
Treisman (2000) shows that democratic institutions and higher income levels are negatively
associated with corruption. Other studies highlight the role of legal institutions, such as the rule
of law, in reducing corruption (La Porta et al., 1999).

Government effectiveness captures various aspects of good governance, including public sector
competence and policy quality. Studies like Mauro (1995) suggest that ineffective governance
fuels rent-seeking behavior, thus exacerbating corruption. This study builds upon such literature
by employing panel data to analyze this relationship across a broad temporal and geographical
scope.
 Chapter 3: Hypothesis & Methodology

3.1 Hypothesis
H₀ : Government effectiveness has no significant effect on corruption levels.

H₁ : Government effectiveness negatively affects corruption levels, (higher effectiveness

reduces corruption).

3.2 Methodology
This study uses a panel dataset from the Worldwide Governance Indicators (WGI) and World
Development Indicators (WDI), covering 150 countries from 1995 to 2020. It applies three-panel data
estimation methods:

 Pooled OLS
 Random Effects Model
 Fixed Effects Model

The Hausman test is applied to determine the appropriate model between random and fixed effects.

3.3 Data and Variables

The dataset is derived from the Quality of Government Standard Time-Series Dataset (January 2025
release). The key variables used are:

 Dependent Variable: wbgi_cce – Control of Corruption Estimate

 Independent Variable: wbgi_gee – Government Effectiveness Estimate

 Control Variables: wdi_gdpcapgr (GDP per capita growth, annual %), bti_ds (democracy
status), wdi_urbpop (urban population), wdi_unemp (unemployment rate), wdi_gini (income
inequality)
  Panel Identifiers: ccode (country code), year

3.4 Econometric Model:

wbgi_cce_it = β0 + β1 * wbgi_gee_it + β2 * wdi_gdpcapgr_it + β3 * bti_ds_it + β4 *
wdi_urbpop_it + β5 * wdi_unemp_it + β6 * wdi_gini_it + μi + ε_it

Where:

 ε_it = Error term

 i = Country
 t = Year
 μi = Time-invariant country-specific effect

Chapter 4: Results

Data Analysis
Table 5.1 Descriptive Statistics
The summary statistics give an overview of the main variables used in the analysis. The corruption
control variable (wbgi_cce) has a mean of approximately -0.06 with a standard deviation of about 1.00,
ranging from -1.97 to 2.46. Government effectiveness (wbgi_gee) shows a very similar pattern, with a
mean close to -0.06, a comparable spread, and slightly wider minimum and maximum values. The
average annual GDP per capita growth (wdi_gdpcapgr) is around 1.94%, but its distribution is quite
spread out, with a wide range from -64.43% to as high as 150.43%, suggesting the presence of extreme
growth or contraction in some countries or years. The democracy score (bti_ds) averages at 5.62 with a
standard deviation of about 2.09, ranging from 1.27 to nearly 10, reflecting diverse political systems
across countries. Finally, the Gini index (wdi_gini), which measures income inequality, has a mean of
37.79 and a standard deviation of 8.84, with values ranging from 22.2 to 65.8, indicating significant
differences in income distribution across countries. Overall, these statistics highlight the variability in
institutional quality, economic performance, governance, and inequality across the dataset.

Table 5.2: Correlation Matrix

pwcorr wbgi_cce wbgi_gee wdi_gdpcapgr democracy, sig

The correlation analysis reveals a strong positive association between control of corruption and
government effectiveness (r = 0.9206, p < 0.01), indicating that improvements in administrative
efficiency closely align with stronger anti-corruption measures. Similarly, democracy status shows
moderate positive correlations with both control of corruption (r = 0.5939) and government effectiveness
(r = 0.6036), both significant at the 1% level, suggesting that democratic governance is linked to better
institutional quality. In contrast, GDP per capita growth exhibits weak and largely insignificant
correlations with governance and democracy indicators. Although the negative correlation with control of
corruption (r = -0.0409, p < 0.01) is statistically significant, its magnitude is negligible. The relationships
with government effectiveness (r = -0.0170) and democracy status (r = 0.0352) are both weak and not
statistically significant. In summary, the results highlight robust interconnections between governance
quality and democratic practices, while economic growth appears to have little direct association with
these variables in the present dataset.

Table 5.3 Pooled OLS Regression

reg wbgi_cce wbgi_gee wdi_gdpcapgr democracy wdi_gini

The results of the ordinary least squares (OLS) regression analysis are presented in Table 1. The model
seeks to explain variations in the control of corruption index (wbgi_cce) using four independent
variables: government effectiveness (wbgi_gee), GDP per capita growth (wdi_gdpcapgr), democracy
status (bti_ds), and income inequality (wdi_gini). The model demonstrates relatively strong explanatory
power, with an R-squared value of 0.7592, indicating that approximately 75.92% of the variance in the
dependent variable is explained by the predictors. The F-statistic of 352.36 is statistically significant (p <
0.01), suggesting that the model as a whole is a good fit.

Among the explanatory variables, government effectiveness emerges as the most influential factor in
controlling corruption. The coefficient for wbgi_gee is positive and highly significant (β = 0.7375, p <
0.01), confirming the theoretical expectation that stronger institutional quality substantially enhances the
control of corruption. This finding aligns with previous literature emphasizing the critical role of capable
governance structures in mitigating corrupt practices.

Democracy status (bti_ds) also shows a positive and statistically significant relationship with control of
corruption (β = 0.0767, p < 0.01). While its magnitude is smaller relative to government effectiveness, the
result underscores the contribution of democratic institutions to promoting transparency and
accountability.

In contrast, GDP per capita growth (wdi_gdpcapgr) exhibits an insignificant and slightly negative
coefficient (β = -0.0012, p = 0.721), indicating no meaningful association between short-term economic
growth and improvements in corruption control. This suggests that economic growth alone, without
corresponding institutional development, is insufficient to reduce corruption.

Income inequality, as measured by the Gini index (wdi_gini), demonstrates a weak and marginally
positive association with control of corruption (β = 0.0037, p = 0.051). While this finding is on the
threshold of statistical significance, it is counterintuitive, as greater inequality is often associated with
weakened governance outcomes. This ambiguous result warrants further investigation.

The intercept term is negative and significant (β = -0.8553, p < 0.01), reflecting the predicted value of the
dependent variable when all predictors are held at zero. However, given the nature of the variables, this
figure carries limited substantive meaning.
Model Efficiency and Limitations

Although the pooled OLS regression demonstrates a relatively high explanatory power, it is ultimately an
inefficient method for analyzing panel data, as it overlooks potential heterogeneity across countries and
over time. Specifically, the pooled OLS approach assumes homogeneity in the intercepts and slopes,
ignoring unobserved country-specific or time-specific effects that may systematically influence the
control of corruption. Moreover, the model does not address possible endogeneity and omitted variable
bias, nor does it account for autocorrelation or heteroskedasticity, which are common in panel datasets.

Next Steps for Robust Analysis

To overcome these limitations and to achieve more reliable and nuanced insights, a sequence of advanced
panel data econometric techniques has been employed in the subsequent stages of the research:

1. Fixed Effects (FE) Regression

This method controls for unobserved time-invariant heterogeneity across countries, isolating the
impact of the explanatory variables within countries over time.

2. Random Effects (RE) Regression

Recognizing the potential efficiency gains of RE models, this approach assumes that individual-
specific effects are random and uncorrelated with the regressors.

3. Hausman Test (FE vs. RE)

To determine the appropriate specification between FE and RE models, the Hausman test has
been conducted. This test assesses whether the unique errors are correlated with the regressors,
guiding the model selection.

4. Breusch-Pagan Lagrange Multiplier (LM) Test

This test evaluates whether random effects are appropriate compared to simple OLS by testing
for significant variance across entities.

5. Two-Way Fixed Effects Model

Finally, to further refine the analysis, a two-way fixed effects model has been implemented,
accounting for both country-specific and time-specific unobserved heterogeneity.

These methodological advancements provide a more comprehensive understanding of the determinants of

corruption control by addressing the deficiencies inherent in the pooled OLS approach. The following
sections of the paper present the results from these more robust models, allowing for a richer and more
accurate interpretation of the data.

Table 5.4 : Fixed Effects Regression

xtreg wbgi_cce wbgi_gee wdi_gdpcapgr bti_ds, fe

The fixed effects regression model analyzes the impact of government effectiveness, GDP per capita
growth, and democracy on corruption levels across 134 countries. The results show that government
effectiveness (wbgi_gee) has a strong and statistically significant negative relationship with corruption
(wbgi_cce). Specifically, a one-unit increase in government effectiveness is associated with a 0.47 unit
reduction in corruption, and this result is highly significant (p < 0.01). This confirms the expected
relationship: better governance is linked to lower levels of corruption.

The democracy variable (bti_ds) is also statistically significant, with a negative coefficient of
approximately -0.087. This suggests that more democratic countries tend to have lower corruption levels,
all else being equal. Again, this effect is highly significant (p < 0.01), supporting the idea that democratic
institutions play a key role in curbing corrupt practices.

On the other hand, GDP per capita growth (wdi_gdpcapgr) shows a very small and statistically
insignificant effect on corruption (p = 0.676). This indicates that economic growth, at least in the short
run, may not directly influence corruption levels when country-specific effects are controlled for.

The overall explanatory power of the model is relatively high, with an R-squared of 0.744, meaning
around 74% of the variation in corruption levels is explained by the model. The F-test is highly
significant (p < 0.01), indicating that the model as a whole is statistically meaningful. The results strongly
support the view that strong institutions and democratic governance are key to reducing corruption.

Table 5.5 : Random Effects Regression

xtreg wbgi_cce wbgi_gee wdi_popurb wdi_unempne , re

This table of random effect models shows how government effectiveness, urban population, and
unemployment rates influence corruption levels across 183 countries, using a total of 2,779 observations.
The key finding is that government effectiveness (wbgi_gee) has a strong and statistically significant
negative impact on corruption, with a coefficient of -0.6177. This result is highly robust (p < 0.01),
implying that as government effectiveness improves, perceived corruption tends to decline significantly
across countries.

The variable for urban population (wdi_popurb) has a positive coefficient of about 0.0014 and is
marginally significant (p = 0.098). This suggests a possible weak relationship where more urbanized
countries may experience slightly higher levels of corruption, although the evidence isn't strong enough to
draw a firm conclusion.

Unemployment (wdi_unempne), on the other hand, does not show a statistically significant effect on
corruption, as indicated by its high p-value (p = 0.393). This means that variations in unemployment rates
do not appear to have a clear or consistent association with corruption across the countries in the sample.

The model explains a large proportion of the variance between countries, as reflected in the high between
R-squared (0.8722) and overall R-squared of 0.8686. The Wald chi-square statistic is also highly
significant (p < 0.01), confirming the overall strength of the model.

Overall, these findings reinforce the conclusion that government effectiveness is a crucial factor in
reducing corruption, while other variables like urbanization and unemployment may have limited
effects.

Table 5.6 : Hausman Test (FE vs RE)

xtreg wbgi_cce wbgi_gee, fe

This output presents the results of a Hausman test, which is used in panel data analysis to help decide
between a fixed effects (FE) model and a random effects (RE) model. The presented output pertains to a
fixed effects regression analysis, a method employed in panel data to model the relationship between
variables while controlling for time-invariant, unobserved heterogeneity at the group level (in this case,
countries). The model examines the impact of government effectiveness ( wbgi_gee) on corruption levels
(wbgi_cce) across 193 countries, utilizing 4,687 observations. The overall fit of the model is strong, as
indicated by an R-squared of 0.8500, with a substantial portion of the variance in corruption being
explained by differences between countries (between-group R-squared = 0.8879) compared to changes
within countries over time (within-group R-squared = 0.2984). The highly significant F-statistic for the
overall model (F (1, 4493) = 1911.01, p < 0.001) confirms that government effectiveness has a
statistically significant impact on corruption after accounting for these fixed country effects. Specifically,
the coefficient for government effectiveness is -0.5200387 (p < 0.001), indicating a significant negative
relationship: a one-unit increase in government effectiveness is associated with a 0.52 unit decrease in
corruption levels within countries. The high intraclass correlation coefficient (rho = 0.8851) further
supports the use of a fixed effects model by highlighting that a large proportion of the variance in
corruption is attributable to time-invariant country-specific factors. Moreover, the significant F-test for
the fixed effects (F (192, 4493) = 77.53, p < 0.001) confirms that these country-specific effects are jointly
significant, reinforcing the appropriateness of the fixed effects approach over a pooled OLS regression.
This analysis strongly suggests that improvements in government effectiveness led to significant
reductions in corruption levels within countries over time, after controlling for stable, unobserved country
characteristics.
estimates store re

hausman fe re

This table presents the findings from the random-effects generalized least squares (GLS) regression,
where the control of corruption (wogi_cce) is regressed on government effectiveness (wogi_gee). The
analysis is based on a balanced panel dataset comprising 4,687 observations across 193 countries, with an
average of approximately 24 observations per country.

The overall explanatory power of the model is strong, as indicated by an overall R-squared value of 0.85,
suggesting that the model accounts for approximately 85% of the total variance in the dependent variable.
Notably, the model explains 88.79% of the variance between countries, while the within-country variance
explanation stands at 29.84%. The Wald chi-square statistic is highly significant (χ² = 3016.46, p < 0.01),
confirming that the explanatory variables collectively contribute meaningfully to the model.

However, when examining the individual coefficient estimates, the results present a more nuanced
picture. The coefficient for government effectiveness (wogi_gee) is estimated at –0.0224, with a standard
error of 0.0242. Despite the negative sign, indicating an inverse relationship between government
effectiveness and control of corruption in this specification, the effect is not statistically significant (p =
0.354). The 95% confidence interval for this estimate (–0.0699 to 0.0250) spans zero, further confirming
the absence of a statistically meaningful association.

The constant term in the model is estimated at 0.3257, representing the expected level of control of
corruption when government effectiveness is at zero. Although this intercept is included in the estimation,
it serves a limited substantive interpretation given the nature of the variables.
Regarding the variance components, the between-entity standard deviation (sigma_u) is estimated at
0.1904, while the within-entity standard deviation (sigma_e) stands at 0.7452. These estimates suggest
that the majority of the unexplained variance lies within countries over time rather than between
countries. The intra-class correlation (rho), which reflects the proportion of total variance attributable to
differences across panels, was partially reported but appears to be modest given the variance estimates.

The random-effects model assumes zero correlation between the individual-specific error component and
the regressors, an assumption that, if violated, could bias the estimates. In this analysis, the model
proceeds under this assumption, as indicated by the reported corr(u_i, X) = 0.

In summary, while the random-effects model demonstrates strong explanatory power at the overall level,
the core predictor—government effectiveness—does not exhibit a statistically significant effect on control
of corruption in this specification. This finding suggests that the random-effects model, while efficient in
structure, may not adequately capture the dynamics of the relationship between governance quality and
corruption control in the sample studied.

Recognizing these limitations, further robustness checks have been conducted. Specifically, the Breusch-
Pagan Lagrange Multiplier (LM) test, the Hausman specification test, and fixed-effects models (both
standard and two-way fixed effects) were employed to address potential inefficiencies and unobserved
heterogeneity. These additional approaches offer a more reliable assessment of the determinants of
corruption control by accounting for time-invariant and entity-specific factors that the random-effects
model may overlook.

To determine the appropriate estimation strategy between the fixed-effects (FE) and random-effects (RE)
models, a Hausman specification test was conducted. The test compares the consistency of the fixed-
effects estimator against the efficiency of the random-effects estimator under the null hypothesis that
there is no systematic difference between the two.
The results, as displayed in Table 3, reveal a clear and statistically significant divergence between the
coefficient estimates of the two models. Specifically, the coefficient of interest, government effectiveness
(wogi_gee), is estimated at 0.5200 under the fixed-effects model, while the random-effects model yields a
higher estimate of 0.6023. The difference between these two estimates is –0.0823, with an associated
standard error of 0.0046.

The Hausman test statistic (χ² = 318.83, p = 0.0000) decisively rejects the null hypothesis that the
difference in coefficients is not systematic. The extremely low p-value indicates strong evidence of
correlation between the unobserved individual effects and the regressors, thereby violating the random-
effects assumption of independence between the explanatory variables and the unit-specific error term.

This finding implies that the fixed-effects estimator is not only consistent but also preferred over the
random-effects estimator for this analysis. The random-effects model, while potentially more efficient
under the null, appears to suffer from omitted variable bias due to unobserved heterogeneity. As such,
relying on the random-effects results would risk drawing biased inferences regarding the relationship
between government effectiveness and control of corruption.

Accordingly, the subsequent analysis focuses on the fixed-effects model and, further, employs two-way
fixed-effects estimations to control for both entity-specific and time-specific unobserved factors, ensuring
a more robust examination of the determinants of corruption control.

Step 8: Brzusch-Pagan LM Test (RE vs Pooled OLS)

xtreg wbgi_cce wbgi_gee, re
xttest0

This analysis involves a random-effects generalized least squares (GLS) regression and the
Breusch-Pagan Lagrangian multiplier (LM) test. These tests are used to evaluate the relationship
between two variables, `wbgi_cce` (the dependent variable) and `wbgi_gee` (the independent
variable), while accounting for group-level effects.

In the random-effects GLS regression, the model estimates how `wbgi_gee` influences
`wbgi_cce` across 193 groups, with a total of 4,687 observations. The regression results show
that `wbgi_gee` has a significant positive effect on `wbgi_cce`, with a coefficient of 0.6023. This
means that for every unit increase in `wbgi_gee`, there is an expected increase of approximately
0.6023 units in `wbgi_cce`. The p-value for this coefficient is 0.000, indicating strong statistical
significance. The constant term (_cons) is not statistically significant, suggesting it does not
contribute meaningfully to the model.

The R-squared values provide insight into the model's explanatory power. The "within" R-
squared is 0.2984, meaning that 29.84% of the variation in `wbgi_cce` within individual groups
is explained by the model. The "between" R-squared is much higher at 0.8879, indicating that
88.79% of the variation between groups is explained by differences in `wbgi_gee`. Overall, the
model explains 85% of the total variation in `wbgi_cce`. The Wald chi-squared statistic is
extremely high (3016.46), with a p-value of 0.0000, further confirming that `wbgi_gee`
significantly affects `wbgi_cce`.

To assess whether random effects are necessary or if pooled ordinary least squares (OLS) would
suffice, the Breusch-Pagan LM test was conducted. This test evaluates whether there is
significant variance due to group-level effects (`Var(u)`). The test results show that the variance
for group-specific random effects (`u`) is 0.1061, while the variance for idiosyncratic errors (`e`)
is much smaller at 0.0363. The LM test statistic (`chibar2(01)`) is extremely large (26126.93),
with a p-value of 0.0000, strongly rejecting the null hypothesis that group-level variance equals
zero. This confirms that random effects are essential for this dataset and pooled OLS would be
inappropriate.

In summary, the analysis demonstrates that a random-effects model is preferred over pooled OLS
because group-specific effects play a significant role in explaining variations in `wbgi_cce`.
Moreover, `wbgi_gee` has a strong and statistically significant positive relationship with
`wbgi_cce`.

Step 9: Two-Way Fixed Effects

xtreg wbgi_cce wbgi_gee wdi_gdpcapgr i.year, fe
This output presents the results of a two-way fixed effects regression model, which extends the standard
fixed effects approach by controlling for both entity-specific (country-level) and time-specific (year-level)
unobserved heterogeneity. By including both country and year fixed effects, this model accounts not only
for time-invariant characteristics of countries but also for global shocks or trends over time that could
simultaneously influence government effectiveness and corruption levels. The dataset comprises 4,687
observations across 193 countries and multiple years, providing a robust basis for panel analysis.

The model demonstrates a strong overall fit, as indicated by an R-squared of 0.8602, with a particularly
high between-group R-squared of 0.8975. This suggests that much of the variation in corruption levels is
explained by differences across countries rather than changes within countries over time. The within-
group R-squared of 0.3039, while lower, still indicates meaningful explanatory power for within-country
changes over time.

The F-statistic for the overall model (F (1, 4492) = 1958.88, p < 0.001) is highly significant, confirming
that government effectiveness remains a significant predictor of corruption levels even after controlling
for both country and time fixed effects. Specifically, the coefficient for government effectiveness is -
0.5179061 (p < 0.001), indicating a substantial negative relationship: a one-unit increase in government
effectiveness is associated with a 0.52 unit reduction in corruption levels.

The intraclass correlation coefficient (rho = 0.8862) remains high, reinforcing the presence of significant
country-specific unobserved heterogeneity. Moreover, the joint significance of the fixed effects (F (384,
4492) = 39.34, p < 0.001) underscores the importance of including both country and year fixed effects in
the model.

Overall, the two-way fixed effects model provides strong evidence that improvements in government
effectiveness are consistently associated with reductions in corruption levels, not only within countries
over time but also across different time periods. By controlling for both dimensions of unobserved
heterogeneity, this model offers a more comprehensive and reliable estimation, strengthening the
robustness of the findings.
Step 10: Hypothesis Testing
xtreg wbgi_cce wbgi_gee wdi_gdpcapgr, fe
test wbgi_gee = 0

In This Table a fixed effects regression was conducted to further investigate the relationship between
government effectiveness (wbgi_gee) and control of corruption (wbgi_cce), while also accounting for GDP
per capita growth (wdi_gdpcapgr). The model includes 4,556 observations across 191 countries, each
observed over multiple years (between 5 to 25 time periods). The overall explanatory power of the model
is very strong, with an R-squared of approximately 0.85, meaning that about 85% of the variation in
corruption control can be explained by the included variables.
The results show a clear and significant relationship between government effectiveness and control of
corruption. The coefficient for wbgi_gee is 0.5175 and statistically significant at the 1% level (p < 0.01).
This suggests that countries with more effective governments tend to experience significantly better
control of corruption. On the other hand, GDP per capita growth ( wdi_gdpcapgr) once again shows no
meaningful effect on corruption levels, with a near-zero coefficient and a p-value of 0.876, indicating a
lack of statistical significance.

To formally test the importance of government effectiveness, a hypothesis test was carried out to see if its
coefficient was statistically different from zero. The F-statistic for this test was 1843.66 with a p-value of
0.0000, meaning we can confidently reject the null hypothesis. In other words, the data provides very
strong evidence that government effectiveness does, in fact, have a real and statistically significant impact
on corruption.

Overall, these findings reinforce the idea that improving the quality and effectiveness of governance is
crucial in efforts to control corruption. Economic growth alone doesn’t seem to make a significant
difference, but stronger institutions and effective public services do.

6. Results and Interpretation

The results from the panel data analysis consistently highlight the pivotal role of government
effectiveness in controlling corruption across countries from 1995 to 2020. Across all estimation methods
—pooled OLS, fixed effects, random effects, and two-way fixed effects—government effectiveness
emerges as the most robust and statistically significant determinant of corruption control. In the fixed
effects model, a one-unit increase in government effectiveness is associated with approximately a 0.47 to
0.52 unit decrease in perceived corruption, even after accounting for country-specific and time-invariant
characteristics. This result remains consistent and significant (p < 0.01) in models that also include
control variables such as democracy status, GDP per capita growth, urbanization, unemployment, and
income inequality.

In contrast, GDP per capita growth consistently shows an insignificant relationship with corruption levels
across all models. This suggests that while economic development is important, it does not automatically
lead to better corruption outcomes unless accompanied by institutional improvements. The democracy
variable also shows a statistically significant negative relationship with corruption in some models,
supporting the argument that democratic governance enhances transparency and accountability.

The two-way fixed effects model, which controls for both country-specific and year-specific unobserved
heterogeneity, further strengthens the reliability of the findings. In this model, the coefficient on
government effectiveness remains significantly negative (approximately -0.52), and the overall model fit
is strong (R-squared = 0.86), indicating that the explanatory variables account for a large portion of the
variation in corruption levels.

The Hausman test results favor the fixed effects model over the random effects model, confirming that
unobserved country-level characteristics are likely correlated with the regressors, and thus, fixed effects
provide more consistent estimates. Additionally, the Breusch-Pagan LM test justifies the use of panel
models over pooled OLS by confirming the presence of significant variance across entities.

The evidence overwhelmingly supports the hypothesis that better government effectiveness leads to
significantly lower corruption levels. While economic growth and structural variables like unemployment
or inequality play minor roles, it is the quality of governance institutions that truly drives progress in anti-
corruption efforts. These results reinforce the importance of investing in public sector efficiency, policy
implementation capacity, and institutional quality as key strategies for reducing corruption globally.

7. Conclusion

This study set out to answer a critical question in global governance: Does government effectiveness truly
matter in the fight against corruption? Drawing on a rich panel dataset covering 193 countries over a 25-
year period, the evidence points overwhelmingly to a resounding yes. Across every model tested—
whether pooled, fixed effects, or random effects—government effectiveness consistently and significantly
reduces corruption levels. Even after accounting for economic conditions, democratic governance, and
social indicators like unemployment and inequality, the strength of this relationship remained robust.

What makes this finding particularly compelling is its consistency across time and space. While the
impacts of GDP growth, urbanization, and even income inequality proved inconsistent or marginal, the
quality of governance stood out as a reliable anchor in curbing corruption. This signals a crucial takeaway
for policymakers: economic growth alone is not enough. Without effective institutions—ones that can
implement policies, deliver services, and uphold rule of law—efforts to reduce corruption may be limited
at best and counterproductive at worst.

Moreover, the data affirms that democratic structures contribute meaningfully to transparency and
accountability, though their influence is secondary to that of government effectiveness. In a time when
faith in public institutions is waning in many parts of the world, these findings offer a hopeful message:
reform is not only possible—it works.

Ultimately, the fight against corruption is not just about laws and punishments; it’s about building
governments that function—efficiently, fairly, and with integrity. Strengthening governance is not a one-
size-fits-all solution, but it is a necessary foundation. Countries seeking to build trust, drive sustainable
development, and foster inclusive societies must prioritize the effectiveness of their public institutions.
The evidence is clear: where governments work well, corruption struggles to survive.