In this section, we will explain the empirical results of developing countries data.
Descriptive Statistics
Descriptive statistics shows the essential characteristics of the data. They represent quantitative description in an accurate form and give a manageable sketch of the data and also show trends in different periods. The descriptive statistics for the factors examined in the study are shown in Table 1.
Table 1
Descriptive Analysis Results
Variables | Obs | Mean | Std. Dev. | Min | Max | Skew. | Kurt. |
---|
Dependent variable | |
---|
LEB | 345 | 69.12 | 7.764 | 47 | 80 | -0.67 | 2.44 |
MMR | 345 | 208.13 | 261.08 | 10 | 1130 | 1.679 | 5.05 |
Independent variables |
CHE | 345 | 5.20 | 2.45 | 1.27 | 14.13 | 0.95 | 3.20 |
PSE | 345 | 4.26 | 1.72 | 1.12 | 10.64 | 0.99 | 4.16 |
PHt | 345 | 37.59 | 5.45 | 24.7 | 47 | -0.19 | 2.18 |
Pun | 345 | 10.31 | 9.92 | 1.5 | 55.2 | 2.00 | 7.41 |
SDW | 345 | 38.08 | 39.62 | 0.01 | 100 | 0.397 | 1.44 |
SMS | 345 | 50.36 | 24.82 | 3.82 | 100 | -0.09 | 2.20 |
Note: values are adjusted to two decimal places. |
Correlational Analysis
Correlational analysis is done to check the association between the dependent and independent variables.
Table 2
Variables | (a) | (b) | (c) | (d) | (e) | (f) | (g) |
---|
(a) LEB | 1.000 | | | | | | |
(b) CHE | -0.152* | 1.000 | | | | | |
(c) PSE | 0.170* | 0.391* | 1.000 | | | | |
(d) PHt | -0.040 | 0.344* | 0.277* | 1.000 | | | |
(e) Pun | -0.690* | 0.165* | -0.213* | 0.023 | 1.000 | | |
(f) SDW | 0.653* | -0.026 | 0.008 | 0.196* | -0.363* | 1.000 | |
(g) SMS | 0.606* | -0.092 | 0.032 | 0.339* | -0.515* | 0.439* | 1.000 |
(*** p < 0.01), (** p < 0.05), (* p < 0.1) |
The above Table 2 indicates that there is an indirect association between CHE, PHt, LEB, PUn and LEB at level one percent. The results of PSE, SDW and SMS have a positive association with LEB and it is statistically significant at level one percent.
Table 3
Pairwise Correlations Analysis II
Variables | (a) | (b) | (c) | (d) | (e) |
---|
(a) MMR | 1.000 | | | | |
(b) CHE | 0.126* | 1.000 | | | |
(c) PSE | -0.312* | 0.391* | 1.000 | | |
(d) SDW | -0.459* | -0.026 | 0.008 | 1.000 | |
(e) SMS | -0.711* | -0.092 | 0.032 | 0.439* | 1.000 |
(*** p < 0.01), (** p < 0.05), (* p < 0.1) |
The above Table 3 indicates that there is a positive association between current health expenditure (CHE) and MMR and this association is statistically significant based on the p-value. The PSE, SDW, and SMS have a negative association with MMR and these associations are significant at the level of 1 percent.
Visualizing for Panel Data Variables
This section explained the visualization for variables LEB and MMR that are used in model estimation for public health nutrition and sustainable development goals in developing countries.
Figure 2 and 3 indicates the information on developing countries according to LEB and MMR ratios. In Afghanistan and Algeria, the ratio of MMR is 16.28, 2.39 percent and LEB 3.90, 4.71 percent respectively. In Bahrain, Bangladesh, Benin and Brazil, the ratio of MMR 0.33, 4.39, 8.90 and 1.33 percent and LEB 4.97, 4.41, 3.72 and 4.64 percent respectively. The ratio of MMR in the Central African Republic, Egypt and Ethiopia, are 18.90, 0.84, and 10.32 while the LEB is 3.25, 4.41and 3.92 percent respectively. The ratio of MMR in India, Indonesia, Iran, Jordan and Kuwait, are 4.16, 4.19, 0.40, 1.03, and 0.23 while the LEB is 4.31, 4.35, 4.68, 4.70 and 4.95 percent respectively. The MMR ratio in Malaysia, Mali, Mexico, Namibia and Oman is 0.62, 12.77, 0.81, 4.44 and 0.39 percent while the LEB is 4.71, 3.62, 4.64, 3.72 and 4.81 percent respectively. The ratio of MMR in Pakistan, Saudi Arabia, South Africa and Turkey, are 3.42, 0.37, 2.98, and 0.42 percent while the LEB is 4.11, 4.80, 3.90 and 4.79 percent respectively.
Panel Unit Root test
The stationarity level of the variable was examined using a variety of panel unit root tests in the study. The findings of the IPS (Im-Pesaran-Shin), Fisher-ADF (Fisher-Based on Augmented Dickey-Fuller), and panel unit root tests are shown in Table 4. The study also used the second-generation unit root test, which consists of CADF and CIPS. The null hypothesis states that all panels contain a unit root, and the alternative hypothesis states that some panels are stationary.
Table 4
Variables | IPS | FADF | CIPS | CADF |
---|
I (0) | I (1) | I (0) | I (1) | I (0) | I (1) | I (0) | I (1) |
---|
LEB | -1.28 | -5.07* | 8.55 | 67.75** | -3.25* | … | -2.30* | …. |
MMR | 1.79 | -6.77 * | 72.43* | …. | -1.93 | -3.22* | -2.03*** | … |
CHE | 0.40 | -3.73* | 53.42 | 138.98* | -1.378 | -3.22* | -1.52 | -2.10*** |
PSE | -1.70** | … | 104.46* | … | -1.55 | -3.198* | -1.19 | -2.39* |
PHt | -1.4e+ 04* | … | 123.80* | … | -2.596* | … | -1.69 | -2.21** |
Pun | -0.51 | -1.99** | 121.52* | …. | -1.203 | -2.13** | -1.20 | -1.20* |
SDW | 5.26 | -3.64* | 52.00 | 147.61* | -2.06 | -2.68* | -1.88 | -2.29* |
SMS | 4.33 | -5.32 * | 68.08** | … | -1.781 | -2.30** | -1.78 | -2.30* |
Note: *, ** & *** show 1%, 5% & 10% level of significance respectively |
The above results imply that some variables are stationary at I(0) means level while others remain stationary at first difference I(1). In the IPS test MMR, LEB, CHE, PUn, SDW and SMS are stationary at 1st difference but other variables PHt and PSE are stationary at level. While LEB, CHE, and SDW are stationary at 1st difference other variables MMR, PSE, PHt, PUn and SMS are stationary at a level in the FADF test. To investigate the existence of the unit root in the variables this study also used CIPS and CADW tests. According to these results, in the CIPS test MMR, CHE, PSE, PUn, SDW and SMS are stationary at 1st difference but other variables LEB and PHt are stationary at level. While in the CADW test CHE, PSE, PHt, PUn, SDW and SMS are stationary at 1st difference but other variables MMR and LEB, are stationary at level.
Testing for slope heterogeneity
Testing for slope heterogeneity. The null hypothesis indicates that the slope is homogenous and the alternative value indicates that the slope is heterogeneous.
Table 5
Testing for Slope Heterogeneity
| Statistic | Statistic |
---|
Models | Delta | Delta adj | Delta HAC | (Delta HAC) adj |
---|
LEB | 0.124 (0.902) | 0.189 (0.850) | 0.042 (0.967) | 0.064 (0.949) |
MMR | -1.191 (0.234) | -1.576 (0.115) | 0.960 (0.337) | 1.270 (0.204) |
Variable’s partial led out: Constant |
The above table shows the slope homogeneity results of the model. According to the Swamy test, the null hypothesis that the slopes are homogeneous is accepted and we admit that the slopes are not heterogeneous.
Specification Test
Multicollinearity
The test used for multicollinearity that estimates the relationship of all explanatory variables simultaneously is the variance inflation factor (VIF). When the VIF is less than 10 there are no issues of multicollinearity. Tables 6 and 7 can conclude that the issue of multicollinearity problem is not present in this regression model.
Table 6
Variables | VIF | 1/VIF |
---|
SMS | 1.85 | 0.54 |
Pun | 1.75 | 0.57 |
PHt | 1.51 | 0.66 |
PSE | 1.40 | 0.71 |
CHE | 1.38 | 0.73 |
SDW | 1.31 | 0.77 |
Mean VIF | 1.53 | |
Note: values are adjusted to two decimal places. |
Table 7
Variables | VIF | 1/VIF |
---|
SMS | 1.255 | 0.797 |
SDW | 1.239 | 0.807 |
CHE | 1.196 | 0.836 |
PSE | 1.187 | 0.843 |
Mean VIF | 1.22 | |
Heteroskedasticity and Serial Correlation
To solve the problem of heteroskedasticity groupwise in the residuals model the Breusch-pagan and Wald test was used in the fixed effect regression model. The Wooldridge test for serial correlation is used for testing the autocorrelation. The outcomes are shown in Table 8 which shows the problem of heteroskedasticity and first-order autocorrelation in the models. To control the problem of heteroskedasticity and serial autocorrelation the regression model uses the PCSEs model.
Table 8
Test for Heteroskedasticity and Wooldridge Test for Autocorrelation
| Breusch-Pagan | Wald test | Wooldridge Test |
---|
Model with Dependent Variable | chi2 (1) (Prob > chi2) | chi2 (23) (Prob > chi2) | F Statistics (Prob > F) |
---|
LEB | 13.79 (0.0002) | 3627.25 (0.0000) | 31.087 (0.0000) |
MMR | 34.61 (0.000) | 2.5e + 05 (0.0000) | 353.32 (0.0000) |
Cross-Sectional Dependence Test (CD)
The cross-sectional dependency test is used to check whether the residual dependency relationship exists or not in our analysis. Through the cross-sectional dependence test and the decision, this study checks residual dependency relationship exists or not in our analysis. Whether correlation is present or not in our panel data is based on the values of chi-square statistics and p-value.
Table 9
Cross-Section Dependence Results
| (Model I, Life expectancy at birth) | (Model II, Maternal Mortality Ratio) |
---|
Tests | Statistics (Prob) | Statistics (Prob) |
---|
Pesaran's | 16.23 (0.00) | -2.203 (0.028) |
Friedman's | 89.53 (0.00) | 1.88 (1.00) |
Frees' | 4.32 (0.22) | 12.63 (0.23) |
Note (I): values are adjusted to two decimal places. |
Note (II): Statistical significance is evaluated at a level of five percent significance. |
Results of Table 9 show that our null hypothesis is accepted based on the criteria that in our panel data analysis there is no correlation found between the residual and cross-sectional dependence. It is accepted at a five percent level of significance which explains that there is no cross-section dependency exit in residual. In micro-panel analysis following tests like Pesaran CD, Friedman's test and Frees' test indicate that in our analysis there is no cross-sectional dependency and correlation in our residuals.
Breusch and Pagan Test and Hausman Results
To check whether the random effect model is present or not in our model. For this purpose, we apply the Breusch and pagan Lagrangian Multiplier (LM) test. The Hausman test is used in the panel models to select the fixed and random effect models. The results of both models are shown below in the Table 4.10
Table 10
Breusch and Pagan Test and Hausman Results
| Breusch and Pagan LM Test | Hausman Results |
---|
Model variable | chibar2 (01) (Prob > chibar2) | Chi-square test value (P-value) |
---|
LEB | 1636.00 (0.000) | 7.22 (0.3013) |
MMR | 1152.52 (0.000) | 92.19 (0.000) |
Table 10 results show that based on the values of chi-square and p- p-value in our model null hypothesis is rejected at the confidence level of one percent indicating that in our model random effect of the panel model is selected. The final Hausman test result indicates the first model is a random effect model. The fixed effect model is the second model we use to estimate it because it is more reliable in such situations.
Estimation of Static Panel Regression Models
The various sustainable nutritional factors are crucial in determining goal three of development sustainability in developing countries. The regression analysis results for static panel model I LEB and model II using MMR as a measure of SDGs are given below in Tables 11 and 12 respectively.
Table 11
Regression results for Static Panel Models I
| Pool-OLS | Random-Effects | Fixed-Effects | PCSEs Model |
---|
LEB | Coefficient |
---|
CHE | -0.12 (0.10) | 0.093 (0.096) | 0.105 (0.098) | 0.021 (0.09) |
PSE | 0.82*** (0.15) | 0.609*** (0.12) | 0.605*** (0.122) | 0.153 (0.10) |
PHt | -0.39*** (0.05) | -0.24*** (0.07) | -0.19** (0.08) | -0.389*** (0.08) |
Pun | -0.237*** (0.028) | -0.078*** (0.027) | -0.064** (0.029) | -0.23*** (0.04) |
SDW | 0.087*** (0.006) | 0.11*** (0.017) | 0.127*** (0.026) | 0.10*** (0.01) |
SMS | 0.106*** (0.012) | 0.109*** (0.019) | 0.111*** (0.022) | 0.08*** (0.02) |
Constant | 74.566*** (1.498) | 66.131*** (3.216) | 63.183*** (3.735) | 77.31*** (2.48) |
Mean | 69.116 | 69.116 | 69.116 | 69.116 |
SD | 7.764 | 7.764 | 7.764 | 7.764 |
R-squared | 0.752 | 0.684 | 0.6517 | 0.992 |
R2 within | … | 0.256 | 0.259 | |
R2 between | … | 0.705 | 0.671 | |
Observation | 345 | 345 | 345 | |
(*** p < 0.01), (** p < 0.05), (* p < 0.1) and standard errors contain Parentheses. |
The above finding of Table (4.11) concluded that good health and wellbeing have a positive and significant improvement on LEB. Hausman test result indicates that we used a random effect model so, the panel regression results show that one percentage increase in CHG will bring the 0.09 increase in LEB which is an insignificant level of significance. PSE also has a positive impact on LEB that shows positive and signification improvement in good health and wellbeing in this model. Results show that a one percent increase in PSE will bring a 0.605 decrease in LEB which is significant at a level of one percent and it has a negative impact on LEB. The finding shows that a one percent increase in the PHt will bring a 0.24 increase in the LEB. The increase in PUn one percentage will bring the 0.24 decrease in LEB is significant at the level of 1 percent. SDW has a direct relationship with LEB. One percent increase in SDW will bring 0.11 increases in LEB which is significant at a one percent level of significance. SMS also has a positive and direct impact on LEB. Results show that the increase in SMS by 1 percent of the total population will bring a 0.109 increase in LEB which is significant at a level of one percent. The results conclude that overall models are significant through chi-square p-value at a 1 percent significance level. The R-squared value of the model shows that 68 percent, within 26 percent and between 71 percent variations in response variables are due to included explanatory variables and the remaining variations are due to other variables which are not included in the model.
The regression model uses the PCSEs model to control for heteroskedasticity and serial autocorrelation the regression results show that one percentage increase in CHG and PSE will bring the 0.021 and 0.153 increase in LEB respectively which is the insignificant level of significance. The PCSEs results show that a one percent increase in the PHt will bring a 0.389 increase in the LEB. The increase in PUn one percentage will bring the 0.23 decrease in LEB at the level of one percent is significant. Furthermore, the one percent increase in SDW and SMS will bring 0.10 and 0.08 increases in LEB which is significant at a one percent level of significance.
Table 12
Regression results for Static Panel Models II
| Pool-OLS | Random-Effects | Fixed-Effects | PCSEs Model |
---|
MMR | Coefficient |
---|
CHE | 22.67*** (3.72) | -5.03 (3.82) | -6.90* (3.88) | -8.82* (5.29) |
PSE | -56.95*** (5.29) | -4.275 (4.848) | -1.42 (4.87) | -13.48*** (5.04) |
SDW | -1.24*** (0.23) | -2.623*** (0.663) | -3.748*** (1.046) | -1.207*** (0.25) |
SMS | -6.28*** (0.38) | -3.779*** (0.745) | -2.874*** (0.858) | -6.151*** (0.75) |
Constant | 696.41*** (30.22) | 542.662*** (56.032) | 537.54*** (60.53) | 582.38*** (57.76) |
Mean d. var | 208.133 | 208.133 | 208.133 | 208.133 |
SD d. var | 261.077 | 261.077 | 261.077 | 261.077 |
R-squared | 0.654 | 0.470 | 0.089 | 0.667 |
R2 within | | 0.082 | 0.3793 | |
R2 between | | 0.493 | 0.3962 | |
No. of Observation | 345 | 345 | 345 | 345 |
Standard errors contain Parentheses and (*** p < .01), (** p < .05), (* p < .1). |
The finding of Table 12 concluded that good health and wellbeing have a positive and significant improvement on MMR. Hausman test result indicates that we used a fixed effect model So, the panel regression results show that one percentage increase in CHE (Current health expenditure) will bring the 6.90 decrease in MMR which is significant at a 10 percent level of significance. The prevailing healthcare system is adequate for reducing the rate of maternal mortality. Public spending on education (PSE) also has a negative impact on MMR which shows positive and signification improvement in good health and wellbeing in this model. The outcome of the analysis shows that a one percent rise in PSE will bring a 56.95 decrease in MMR which is insignificant at the level of significance.
People using safely managed drinking water services (SDW) have an indirect relationship with MMR. A one percent increase in SDW will bring a 0.004 decrease in MMR which is significant at a 10 percent level of significance. People using safely managed sanitation services (SMS) also have negative and indirect impacts on MMR. Results show that the increase in SMS by 1 percent of the total population will bring a 1.24 decrease in MMR which is significant at a level of one percent. The value of R-squared in the model shows that 89 percent within 37 percent and between 38 percent variations in response variables are due to included predictor variables and the remaining variations are due to other variables which are not included in the model.
The PCSEs model results show one percentage increase in CHE will bring an 8.82 decrease in MMR which is significant at a level of ten percent. The finding shows that a one percentage increase in PSE will bring a 13.48 decrease in MMR which is significant at a level of one percent. Furthermore, PCSE's one percent increase SDW will bring a 1.207 decrease in MMR which is significant at a level of one percent. Results show that an increase in SMS by 1 percent of the total population will bring the 6.151 decrease in MMR which is significant at a level of one percent.
The final finding of the panel regression model shows that the CHE impact on MMR has a negative and significant effect and the LEB impact on CHE is a positive and significant effect. This ascertain is also expressed in preceding studies (Aziz et al., 2021; Rana et al., 2018; Chen et al., 2021; Oladosu et al., 2022). The panel regression result of the PSE impact on LEB is positive and significant and its impact on MMR is negative and significant effect (Oladosu et al., 2022; Saleem, S. 2023; Raftakis, M. 2023) has reported the same results.
The PHt has a negative and significant impact on LEB. It is an important indicator to improve public health nutrition and global nutritional goals criteria. Edwards et al., (2023), Alipour, et al., (2023) Bugri, et al., (2023) has reported the same. The Pun indicates the lower growth process of the upcoming generation has a negative and significant impact on LEB. These results are in line El-Fatah et al., (2023); Komarulzaman et al., (2023); Tengepare et al., (2023).
Furthermore, the panel regression result of the SDW and SMS impact on MMR is a negative and significant effect and its impact on LEB life is positive and significant. The result is in line with Aziz et al., (2021), Wayland, (2013), Fink et al., (2011) and Pickbourn & Ndikumana, (2019).