OBJECTIVE 2

4.2 The effect gender inequality on longevity (objective 2)

This section presents the results of the effects of income on health outcomes (life expectancy and infant
mortality) in selected African countries. The analysis starts with descriptive statistics, which describe the
characteristics of the variables of interest used in this study. The correlation analysis, unit root test, slope
heterogeneity, cross-sectional dependence, and cointegration tests results were all presented as pre-estimation
procedures to the final results of the linear regression with feasible generalised least squares (FGLS).
4.2.1 Descriptive Analysis of the gender inequality on longevity
The descriptive statistics presented in Table 4.1 offer a comprehensive overview of the central tendencies and
distributional characteristics of key indicators influencing gender inequality in a panel of developing countries. The
variables analyzed include the Gender Inequality Index (GII), education gender gap (EDG), public health
expenditure as a percentage of GDP (PHE_GDP), employer gender gap (EMP_G), employment-to-population
gender gap (EMP_PG), the logarithm of GDP per capita (LGDPC), labour force participation gender gap (LFP_G),
population growth rate (PGR), urban population growth rate (UPR), and wage and salaried workers gender gap
(WSW_G). These indicators reflect multifaceted dimensions of gender disparities across the social, economic, and
demographic structures of developing nations.

The Gender Inequality Index (GII) has a mean of 0.48, indicating a moderate level of gender disparity on average
across the sampled countries. With a standard deviation of 0.16 and a negative skewness of –0.58, the distribution
leans slightly towards lower inequality in a few countries. However, the Jarque-Bera statistic (77.70, p = 0.00)
rejects the null hypothesis of normality, suggesting significant heterogeneity in gender outcomes across the sample.
The education gender gap (EDG) records an average of –3.53, reflecting persistent female disadvantages in
educational attainment. The wide range of values (–84.56 to 81.14) and a high standard deviation (34.34) indicate
large disparities across countries. The distribution is approximately symmetric (skewness = 0.08) and relatively
mesokurtic (kurtosis = 2.89), though the Jarque-Bera test (1.96, p = 0.37) does not signal a statistically significant
departure from normality.

Public health expenditure as a share of GDP (PHE_GDP) has a mean value of 2.28%, signifying limited investment
in health sectors among developing nations. The variable is positively skewed (0.71), suggesting that while most
countries invest minimally, a few outliers allocate relatively higher percentages. A standard deviation of 1.44 and
significant non-normality (Jarque-Bera = 112.88, p = 0.00) further emphasize the uneven commitment to public
health financing. The employer gender gap (EMP_G) has a mean of –2.71, underscoring the underrepresentation of
women among employers. The extremely negative skewness (–2.64) and high kurtosis (10.61) reflect the
concentration of data around a few extreme values, with the Jarque-Bera statistic (4,831.88, p = 0.00) pointing to a
strong deviation from normality. This suggests that women face structural and cultural barriers to entrepreneurship
in many of these countries.
The employment-to-population gender gap (EMP_PG) shows an average of –24.33, indicating a significant
disadvantage for women in employment relative to men. The wide variation, with a standard deviation of 16.98 and
values ranging from –65.91 to 6.73, highlights stark inequalities. The distribution is slightly negatively skewed (–
0.51), with moderate kurtosis (2.11), but again, the Jarque-Bera statistic (103.79, p = 0.00) confirms a significant
departure from normality. Economic welfare, captured by the log of GDP per capita (LGDPC), has a mean of
7,129.88 with an exceptionally high standard deviation (10,766.43), showing substantial disparities in income levels
across developing countries. The positive skewness (2.28) and high kurtosis (7.78) reflect that a few relatively
wealthier economies drive the upper end of the distribution. The Jarque-Bera test (2,457.51, p = 0.00) further
supports the presence of non-normality.

The labour force participation gender gap (LFP_G) averages –24.79, reaffirming the substantial under-participation
of women in the labour market. The variable is moderately left-skewed (–0.63) with a standard deviation of 18.77.
The Jarque-Bera statistic (123.50, p = 0.00) again shows a significant deviation from normality, suggesting that
structural and institutional barriers are common across many countries. The population growth rate (PGR) has a
mean of 2.03%, characteristic of the youthful demographic profile typical of developing nations. However, the
positive skewness (1.30) and high kurtosis (9.39) suggest outliers and asymmetry. The Jarque-Bera test (2,676.96, p
= 0.00) indicates non-normality, highlighting that demographic pressures vary widely, with implications for social
infrastructure, particularly education and health services.

Urban population growth rate (UPR) presents an average of 3.36%, reflecting rapid urbanization trends in
developing countries. The data is highly right-skewed (5.58) with an extremely leptokurtic distribution (kurtosis =
42.97), and the Jarque-Bera test (96,894.66, p = 0.00) points to severe non-normality. This suggests that while some
countries are urbanizing gradually, others are experiencing explosive urban growth that could strain public
infrastructure. The wage and salaried workers gender gap (WSW_G) has a mean of –5.29 and a standard deviation
of 13.02, indicating significant disparities in access to formal employment for women. The distribution is negatively
skewed (–1.21) and leptokurtic (kurtosis = 8.31), with the Jarque-Bera test (1,913.04, p = 0.00) confirming non-
normality. These findings reinforce the persistent inequality in labour market outcomes, especially in formal wage
sectors.

In summary, the descriptive statistics reveal a landscape of deep-rooted gender disparities and structural inequalities
across developing countries. Most variables demonstrate non-normal distributions, high skewness, and elevated
kurtosis, emphasizing the asymmetry and heterogeneity in gender outcomes. These statistical properties suggest that
policy responses must be context-specific and multidimensional, addressing disparities in education, labour market
access, healthcare investment, and demographic pressures. Moreover, given the presence of non-normality in most
variables, it is crucial to employ appropriate data transformations and robust econometric techniques to ensure
reliable and unbiased empirical results.

Table 4.1,
Summary Statistics for gender inequality on longevity
 PHE_
LEB GII EDG EMP_G WSW_G LFP_G GDP UPR
Mean 69.25 0.48 -3.53 -2.71 -5.29 -24.79 2.28 3.36
Median 68.97 0.50 -0.78 -1.90 -4.52 -19.63 2.07 2.73

Maximum 232.53 0.82 81.14 1.97 46.01 5.85 6.82 38.68

Minimum 14.67 0.03 -84.56 -16.73 -78.36 -68.20 0.06 -3.03

Std. Dev. 18.86 0.16 34.34 3.20 13.02 18.77 1.44 3.76

Skewnes
s 4.89 -0.58 0.08 -2.64 -1.21 -0.63 0.71 5.58
Kurtosis 35.11 2.85 2.89 10.61 8.31 2.21 2.89 42.97
Jarque- 4831.8
Bera 63351.56 77.70 1.96 8 1913.04 123.50 112.88 96894.66

P-Value 0.00 0.00 0.37 0.00 0.00 0.00 0.00 0.00

1350.0 1350.0
Obs. 1350.00 1350.00 1349.00 0 1350.00 0 1350.00 1350.00

Source: Computed by the Author based on the World Bank WDI (2025)

Part A: The model of Developing Countries

4.3.2a Correlation Analysis for gender inequality on longevity

The correlation analysis presented in Table 4.13a reveals several noteworthy relationships between life expectancy
at birth (LLEB) and the selected socioeconomic indicators across the examined African countries. Education gender
gap (EDG) exhibits a positive but weak correlation with life expectancy (r = 0.241), suggesting that improved
educational equality is modestly associated with higher life expectancy. Similarly, public health expenditure as a
percentage of GDP (PHE_GDP) shows a stronger positive association with LLEB (r = 0.347), indicating that
increased investment in healthcare contributes to better health outcomes and longevity. Urban population growth
(UPR) also correlates positively with life expectancy (r = 0.437), implying that urbanization may provide improved
access to health services and infrastructure.
Conversely, gender inequality (GII) and labor force participation gender gap (LFP_G) both exhibit negative
relationships with life expectancy, with correlation coefficients of -0.550 and -0.306, respectively. This suggests that
greater gender disparities and limited labor force participation may negatively affect overall population health. Wage
and salaried workers gender gap (WSW_G) also shows a weak negative correlation (r = -0.033), while employer
gender gap (EMP_G) exhibits a slight positive correlation (r = 0.059).
Importantly, all correlation coefficients remain well below the multicollinearity threshold of 0.8 (Gujarati & Porter,
2008; Hair et al., 2009). This implies that the explanatory variables are not strongly collinear, thereby supporting the
reliability of subsequent regression analysis.
Table 4.13a Correlation Matrix for gender inequality on longevity

LLEB GII EDG EMP_G WSW_G LFP_G PHE_GDP UPR

LLEB 1.000
GII -0.5500 1.0000
EDG 0.2408 -0.3025 1.0000
EMP_G 0.0586 -0.0333 -0.2357 1.0000
WSW_G -0.0332 -0.3228 0.3139 -0.1912 1.000
LFP_G -0.3063 -0.0336 -0.2680 0.2013 -0.1262 1.0000
PHE_GDP 0.3466 -0.5548 0.3201 0.0169 0.4960 0.0172 1.0000
UPR 0.4371 0.0077 -0.1373 0.1672 -0.6104 -0.0609 -0.1647 1.0000
Source: Computed by the Author based on the World Bank WDI (2025)

4.2.3a Test for multicollinearity gender inequality on longevity

The multicollinearity diagnostic, based on the Variance Inflation Factor (VIF) and tolerance values, indicates that
there is no significant multicollinearity among the independent variables included in the regression model. VIF
values quantify how much the variance of a regression coefficient is inflated due to collinearity with other
predictors, while tolerance (the inverse of VIF) measures the proportion of variance in a variable that is not
explained by other variables in the model.

Following the commonly accepted thresholds—VIF values below 10 and tolerance values above 0.10, as suggested
by Miles (2014)—all variables in this study fall within acceptable limits. Specifically, the VIF values range from
1.14 for LFP_G to 2.39 for WSW_G, with corresponding tolerance values ranging from 0.4180 to 0.8798. Notably,
WSW_G has the highest VIF, but it remains well below the critical threshold of concern.

The mean VIF value of 1.60 further supports the absence of multicollinearity, confirming that the regression
estimates are stable and reliable. This ensures the robustness of the analytical framework used to examine the
determinants of gender inequality in relation to longevity across the panel of countries.

Table 4.3a,
Variance Inflation Factor (VIF) and Tolerance level gender inequality on longevity
VARIABLE VIF 1/VIF
WSW_G 2.39 0.418017

UPR 1.81 0.552783

PHE_GDP 1.80 0.554754

LGDPC 1.56 0.640248

GII 1.33 0.751007

EDG 1.18 0.844018

LFP_G 1.14 0.879812

Mean VIF 1.60

Variance Inflation Factor (VIF) and Tolerance level for gender inequality on longevity
4.2.4a Slope Heterogeneity Test for gender inequality on longevity

Table 4.4a presents the results of the Pesaran and Yamagata (2008) slope heterogeneity test examining the
relationship between income and life expectancy. At the 1% significance level, both the Delta and adjusted Delta
statistics yield p-values of 0.000, indicating strong statistical significance. Consequently, the null hypothesis of slope
homogeneity is rejected. This confirms the presence of slope heterogeneity across the panel data, implying that the
relationship between income and life expectancy varies across the selected African countries.

Table 4.4a,
Pesaran and Yamagata Slope Heterogeneity Test for gender inequality on longevity

Pesaran and Yamagata Test statistics p-value

Delta 15.301 0.000

Delta adj 7.373 0.000

Source: Computed by the Author based on the World Bank WDI (2025)

4.2.5a Cross-Sectional Dependency Test for gender inequality on longevity

Assessing cross-sectional dependence (CSD) is a crucial step in panel data analysis, particularly when slope
heterogeneity is present. CSD arises when individual units in a panel such as countries are not independent of one
another, often due to shared economic shocks, regional spillovers, or institutional linkages. Ignoring CSD can result
in biased and inconsistent parameter estimates, especially in macro-panel contexts. The Pesaran (2004) Cross-
sectional Dependence (CD) test remains valid across both homogeneous and heterogeneous panels, as well as for
non-stationary data structures.
A p-value below the 5% significance level leads to the rejection of the null hypothesis of cross-sectional
independence. The CD test results show strong evidence of dependence across the panel. All variables tested
including the Gender Inequality Index (GII), public health expenditure (PHE_GDP), employer gender gap
(EMP_G), employment topopulation gender gap (EMP_PG), life expectancy at birth (LLEB), labour force
participation gap (LFP_G), population growth rate (PGR), urban population growth (UPR), and wage and salaried
workers gap (WSW_Gyield statistically significant CD statistics with p values of 0.000.
The strength of dependence varies across variables, with LLEB (CD = 128.20, corr = 0.678), GII (CD = 86.05, corr
= 0.455), and PGR (CD = 44.18, corr = 0.234) showing particularly strong cross country interdependence. These
findings confirm the presence of cross-sectional dependence and highlight the interconnectedness of gender-related
development indicators across developing nations.
In response, the study adopts second-generation panel econometric techniques that explicitly account for cross-
sectional dependence. Specifically, the Cross-sectionally Augmented IPS (CIPS) unit root test by Pesaran (2007) is
employed, which adjusts for unobserved common factors across countries. This ensures that subsequent inferences
regarding long-run relationships among the variables are both valid and robust.

Table 4.5a,
Pesaran CD Cross-Sectional Dependency Test for gender inequality on longevity
Variable CD-test p-value corr Abs(corr)

GII 86.05 0.000 0.455 0.615

EDG . . . .
PHE_GDP 28.61 0.000 0.151 0.449

EMP_G 8.25 0.000 0.044 0.500

EMP_PG 8.45 0.000 0.045 0.549
LLEB 128.20 0.000 0.678 0.751
LFP_G 6.43 0.000 0.034 0.612
PGR 44.18 0.000 0.234 0.407
UPR 29.77 0.000 0.157 0.412
WSW_G 6.26 0.000 0.033 0.529
Source: Computed by the Author based on the World Bank WDI (2025)

4.2.6a Unit Root Test for gender inequality on longevity

The Cross-sectionally Augmented IPS (CIPS) unit root test, developed by Pesaran (2007), was conducted using data
spanning from 1995 to 2023. This method was chosen because the null hypothesis of cross-sectional dependence
could not be rejected. The results show that all the variablessuch as gender inequality, education growth, public
health expenditure, employment growth, GDP per capita, labour force participation growth, population growth,
unemployment rate, and wage and salary workers growth—exhibit unit roots in their level form, as their test
statistics fall short of the 5% critical value threshold.
However, when the variables were transformed into their first-differenced forms, the null hypothesis of a unit root
was rejected across all series. This indicates that each variable is integrated of order one, I(1), confirming their non-
stationary nature in levels but stationary after first differencing. As a result, panel data cointegration techniques were
applied to examine long-run relationships among the variables.
Table 4.6a,
CIPS root tests (gender inequality on longevity)
Cross-sectionally Augmented Im, Pesaran, and Shin (CIPS) test
Level Stat (critical First Difference Stat (critical Remark
Variables level@5% level@5%
GII -2.545 (-2.08) -4.970(-2.08) I(1)
EDG -2.433(-2.08) -5.059 (-2.08) I(1)

PHE_GDP -2.244(-2.08) -5.236 (-2.08) I(1)

EMP_G -1.727(-2.08) -3.993 (-2.08) I(1)

LLEB -2.289(-2.08) -4.013 (-2.08) I(1)

LFP_G -1.602(-2.08) -4.441 (-2.08) I(1)

PGR -1.844(-2.08) -3.094(-2.08) I(1)

UPR -2.002(-2.08) -3.608(-2.08) I(1)

WSW_G -1.585(-2.08) -3.707(-2.08) I(1)

Source: Computed by the Author based on the World Bank WDI (2025)

4.2.7a Panel data Cointegration Test for gender inequality on longevity

The Westerlund Panel Cointegration test, which accounts for cross-sectional dependence, was employed to
examine the existence of a long-run relationship among the variables. In line with Levin, Lin, and Chu (2002), the
test was adapted to reflect cross-sectional dependence. The test evaluates the null hypothesis of no cointegration
against two alternatives: that at least some panels are cointegrated, or that all panels are cointegrated. As shown in
Table 4.7a, the null hypothesis is rejected when testing for cointegration across all panels, indicating a common
long-term relationship among the variables. However, the null is not rejected under the assumption that only some
panels are cointegrated. Overall, the rejection of the null for all panels confirms the presence of cointegration,
suggesting that a stable long-run relationship exists among the variables.
Table 4.7a,
Modified Westerlund Panel Cointegration Test for Drivers of Gender Inequality
Some panels are cointegrated All panels are cointegrated

statistics p-value statistics p-value

Variance ratio -2.6754 0.0037 -1.7544 0. 0397

Source: Computed based on the data from the World Bank WDI (2025)

4.2.8a Heteroskedasticity Test for gender inequality on longevity

In this study, heteroskedasticity was examined using a modified Wald test for group-wise
heteroskedasticity. The null hypothesis that the model does not contain heteroskedasticity is to be rejected when the
P-value is less than 5% or 0.05. Table 4.8a demonstrates that the null hypothesis cannot be rejected because the
probability value is less than 5%. This suggests that the model contains heteroskedasticity.
Table 4.8a,
Modified Wald test for groupwise heteroskedasticity result for gender inequality on longevity
Modified Wald test Statistics and P-value
chi2 (1431) 9842.635
Prob>chi2 0.0000
Source: Computed based on the data from the World Bank WDI (2025)

4.2.9a Testing for Serial Correlation for gender inequality on longevity

Table 4.9a shows the result of the Breusch-Pagan test, which shows that the probability value of the Chi-
statistics is less than 1%. The null hypothesis that there is no serial correlation should be rejected if the p-value is
less than 5% or 0.05. From Table 4.9a, the null hypothesis that there is no serial correlation is rejected because the p-
value (0.000) is less than 0.05. This suggests that the residuals of the panel data model suffer from serial correlation.
Hence, the model is best estimated using the feasible generalised least squares (FGLS).
Table 4.9a,
Breusch-Pagan LM test of independence for gender inequality on longevity
Breusch-Pagan test of independence Statistics and P-value
chi2(54) 4.4e+05
Prob>chi2 0.0000
Source: Computed based on the data from the World Bank WDI (2025)

4.2.10a Testing for the Endogeneity for gender inequality on longevity When a regression model’s error term
and the predictor variables are correlated, endogeneity problems occur. If the endogeneity problem is disregarded,
the results could be skewed, with the true effect of a variable on the intended outcome being over- or under-
estimated depending on how accurate the estimations were. The endogeneity test in this study was done using the
Durbin-Wu-Hausman Test of Endogeneity. The null hypothesis is based on the fact that endogeneity does not exist.
Table 4.10a shows that both the Durbin and Wu-Hausman tests have p-values that are more than 0.05, which
suggests that the null hypothesis is not rejected. This implies that the model is free of endogeneity issues
Table 4.10a,
Durbin-Wu-Hausman Test of gender inequality on longevity
Durbin (score) chi2(1) Wu-Hausman F(1,115)
Statistics 0.95421 0.952056
P-value 0.3286 0.3294

Source: Computed based on the data from the World Bank WDI (2025)

4.2.11a The Effects of gender inequality on longevity

 The regression results based on the feasible generalized least squares (FGLS) estimation offer important
insights into the socioeconomic determinants of Infant Mortality (LIMR) across five selected African countries
Egypt, Nigeria, Cameroon, Kenya, and South Africa. Table 4.11b indicates that the overall model is statistically
significant, as evidenced by the Wald chi-square statistic (Wald chi2(5) = 461.54, p < 0.01), suggesting that the
explanatory variables collectively explain substantial variation in infant mortality across the countries in the study.
Among the key predictors, GDP per capita (LGDPC) exhibits a negative and statistically significant relationship
with infant mortality rate, with a coefficient of -0.1606 (p = 0.000). This implies that a 1% increase in per capita
income is associated with a 0.161% reduction in infant mortality rate. This finding suggests that higher income
levels improve access to healthcare, education, nutrition, and other determinants of well-being that ultimately lead to
reduction in infant death rate.
Health expenditure as a percentage of GDP (CHE_GDP) is also negatively and significantly associated
with infant mortality, with a coefficient of -0.0149 (p < 0.000). This result indicates that increased public investment
in healthcare contributes positively to health outcomes. Specifically, a one-unit increase in CHE_GDP results in a -
0.015% decrease in infant mortality rate. This underscores the critical role of sustained healthcare financing,
especially in countries like Kenya and Cameroon where healthcare infrastructure may still be underdeveloped.
Governments aiming to reduce infant mortality should prioritize health sector budgetary allocations.

Similarly, the real exchange rate (LREX) has a negative and significant coefficient (-0.0777, p = 0.007), indicating
that exchange rate movements may affect health outcomes. Specifically, at 1% level of significance, a percentage
increase in real exchange rate reduces infant mortality rate by 0.078%, meaning that improvement in exchange rate
reduces death among the infant

Contrarily, inflation (LCPI) shows a strong and statistically significant positive effect on infant mortality rate, with a
coefficient of 0.2391 (p < 0.000). This suggests that rising price levels erode purchasing power and make essential
goods and healthcare services less affordable, ultimately increasing infant mortality. A 1% increase in inflation is
associated with a 0.239% increase in infant death. In volatile economic environments such as Nigeria and Egypt,
inflation control remains essential not only for economic stability but also for public health resilience. In the same
vein, population growth rate (PGR) also shows positive and significant relationship with Infant Mortality
(coefficient = 0.0263, p = 0.059). This suggests that high population growth may strain public services and
healthcare delivery while increasing pressure on available health care resource, thus increasing infant death rate. The
effect at 10% level of significance is not robust enough in this model to establish a definitive link. Nonetheless, the
positive sign is consistent witError! Hyperlink reference not valid.h expectations.

Overall, income, real exchange rate and health expenditure have negative and significant effect on infant
mortality, while inflation rate and population growth have positive and significant effect on infant mortality in the
long run. These findings underscore the importance of economic growth and increased public health expenditure in
reducing infant death in African countries. Governments in the region should invest more in the health sector while
ensuring macroeconomic stability, particularly through inflation and population growth control, to foster positive
health outcomes.
Table 4.11a,
Results on the Effect of gender inequality on longevity

LLEB Coef. Std. Err. Z P>|z| [95% Conf. Interval]

GII -0.6033702 0.0000821 -1.5700000 0.1150000 -0.0002902 0.0000316
EDG 0.0002694 0.0019510 -4.3800000 0.0000000 -0.0123000 -0.0047206
EMP_G 0.0018876 0. 0006464 -2.8200000 0.0050000 -0.0045000 -0.0008166
WSW_G -0.0011071 0. 0006579 -30.6600000 0.0000000 -0.0720000 -0.0633660
LFP_G -0. 0033104 0. 0000211 -1.6900000 0.0910000 -0.0046000 0.0003440
PHE_GDP 0. 0250956 0. 0028112 7.5800000 0.0000000 0.0161000 0.0274420
UPR 0. 0223587 0 .0023997 -7.3200000 0.0000000 -0.0109000 -0.0063238
CONS 4.286452 0.0187776 -9.5100000 0.0000000 -0.0034000 -0.0022561
Source: Computed by the Author based on the World Bank WDI (2025)
Note: P-value < 1%, <5%, <10% indicates significance at 1%, 5%, and 10% respectively