Dynamics of Human Fertility, Environmental Pollution, and Socio-Economic Factors in Aral Sea Basin

Abstract

One of the worst natural, economic, and social catastrophes caused by human activity is the Aral Sea crisis in Central Asia. The Aral Sea’s desiccation, which has an impact on the region’s overall sustainable development, human welfare, security, and survival, is what led to the problem. This study assesses the effects of economic expansion, population ageing, life expectancy, internet usage, and greenhouse gas emissions on the fertility rate in the countries that made up the Aral Sea basin between 1990 and 2021. Several econometric techniques were used in this study, including Pooled OLS (Ordinary Least Squares) with the Driscoll–Kraay estimating method, FMOLS (Fully Modified Ordinary Least Square), and DOLS (Dynamic Ordinary Least Square). Additionally, we used the Hurlin and Dumitrescu non-cause tests to verify the causal links between the variables. The empirical findings verify that a decrease in the fertility rate among women in the nations surrounding the Aral Sea occurs when the population of a certain age (women aged 15–64 as a percentage of the total population) grows and life expectancy rises. Greenhouse gas emissions (GHGs) also have an adverse effect on reproductive rates. Conversely, the region’s fertility rate may rise as a result of increased internet usage and economic growth. Furthermore, this study indicates that certain variables—aside from greenhouse gas emissions (GHGs)—have a causal relationship with the fertility rate.

1. Introduction

Health issues related to reproduction are significant. The World Health Organization acknowledged declining fertility as a global public health concern (Barratt et al. 2017). Infertility is classified as “a disease of the reproductive system defined by the failure to achieve a clinical pregnancy after 12 months or more of regular unprotected sexual intercourse” (Zegers-Hochschild et al. 2017). The rate of infertility among women has gone up recently (Talmor and Dunphy 2015). One in seven couples in wealthy nations are thought to be affected by this illness (Healy et al. 1994). Ovulatory problems, endometriosis, chromosomal abnormalities, and male factors account for the majority of cases of female infertility (Rocca et al. 2015; D’Argenio et al. 2017). Additionally, there is evidence that air pollution may contribute to the pathophysiology of female infertility (Mahalingaiah et al. 2016; Conforti et al. 2018). Environmental changes seem to be a cause for concern for human health (Ataniyazova 1995; Zetterström 1999).

According to recent studies, there is a decline in human fertility (Girardi and Bremer 2022). Environmental pollutants are potentially significant causal agents linked to the dropping fertility rate, notwithstanding the many theories explaining this trend (Radwan et al. 2015). This study uses a proxy to evaluate the effects on human fertility, as indicated by births per woman, using the GDP, population age, internet, life expectancy, and CO₂ emissions. This study examines the degree to which specific variables affect fertility rates in the nations that made up the Aral Sea basin between 1990 and 2021.

The purpose of the study is to assess socioeconomic and environmental elements influencing human fertility in the Aral Sea basin, consisting of Afghanistan, Turkmenistan, Iran, Tajikistan, Kyrgyz Republic, Kazakhstan, and Uzbekistan. The research questions how these parameters influence on human fertility in the region. Due to the toxicity of the area, women in the ecologically degraded Aral Sea basin experience high rates of anemia, kidney disease, and liver disease. There is evidence of mutagenic, teratogenic, and embryotoxic effects from the pesticides and heavy metal salts in the drinking water (Ataniyazova 1995). First-time mothers face the same risks of disease and complicated pregnancies as women who have had several pregnancies because of their lifetime exposure to toxicants (Zetterström 1999). The Aral Sea issue serves as an example of the tensions that exist between economic expansion and environmental degradation, as well as between humans and nature (Ataniyazova 1995).

The research hypothesis is that human fertility is being negatively influenced by economic expansion, population ageing, life expectancy, internet usage, and greenhouse gas emissions in Aral Sea basin. To the best of our knowledge, we are the first to show how socioeconomic and environmental factors affect population growth in the Aral Sea region, focusing on economic growth, the fertile population, greenhouse gas emissions, internet usage, and life expectancy.

2. Literature Review

2.1. Economic Growth, Human Capital, and Fertility

The subject “At what age are women most fertile?” has been the focus of many studies, and the majority of the research determines the age range as (20–25 years) (García et al. 2018). In light of the fact that human capital expenditures per child are significantly higher in regions with lower fertility rates, total spending per child decreases with fertility, but the product of the total fertility rate and human capital spending per child is roughly a constant share of labour income worldwide (Lee and Mason 2009).

It is frequently unclear at what age female fertility begins to decline. The majority of research distinguishes between a minor decline in the late 20s and a significant decline in the late 30s (García et al. 2018). Possible confounding variables include the mother’s occupation (for potential chemical exposures and occupational class), age, smoking habits, the educational attainment of both parents, the mother’s most recent method of contraception (including breastfeeding), and pertinent medical conditions like pelvic inflammatory disease. Another potential confounding factor is maternal parity, which should not be automatically accounted for, because it represents fertility before the exposure of interest (Joffe 1997).

2.2. Greenhouse Gas Emissions (GHGs)–Fertility Rate

In the case of the MENA region, total fertility rate (TFR) is changing negatively and is considerably impacted by environmental degradation (measured by CO₂ emissions), while TFR in ASEAN is significantly positively impacted (Nkalu 2021). The results indicate that a number of strong measures are needed to monitor and control environmental pollution in order to reduce the risks to public health, including fertility (Nkalu 2021). Indeed, (Ma et al. 2019) examined the most recent research relating environmental pollutants to issues with human reproductive health, including infertility. Robust data demonstrated that exposure to environmental pollutants may affect the reproductive systems of adult males and females (Di Renzo et al. 2015).

A study on the range of reproductive toxicants was conducted on harmful effects and found that there is only a tenuous link between environmental contaminant exposure and decreased fertility in humans (Foster et al. 2008). There is a substantial correlation between a number of environmental pollutants and lower fertility parameters in both men and women, as well as a number of potential reproductive problems in women (Chiang et al. 2017). Air pollution’s role in the pathophysiology of human reproductive disorders has been well supported by experimental and epidemiologic studies (Giudice 2021). Accordingly, there appears to be a connection between male and female reproductive function and air pollution and endocrine disrupting substances in food and water (Seli and Taylor 2023). Exposure to chemicals derived from fossil fuels may be associated with cancer and reproductive issues in humans and a decrease in fertility rates (Skakkebaek et al. 2019).

2.3. Internet Use–Fertility Rate

A digital revolution has been brought about by the internet’s explosive growth since the 1990s, which has had far-reaching social ramifications (DiMaggio et al. 2001). The internet has emerged as the primary medium for communication, having a significant impact on people’s behavioural choices and way of life, including women’s plans to have children (Wilcox and Stephen 2012; Lam et al. 2009). Relevant empirical research examined how the internet affected adolescent fertility (Guldi and Herbst 2016), marriage rates (Bellou 2014), and married women’s participation in the labour force (Omotoso et al. 2020). For instance, research on the effects of broadband access on American teenagers’ fertility decisions revealed a negative correlation between the number of adolescent births and a rise in internet broadband access (Guldi and Herbst 2016).

By incorporating gender role attitudes into the research framework of women’s fertility intentions and internet usage frequency, the studies (Liu et al. 2021; Billari et al. 2019) adopt a different approach to examining the effects of internet usage frequency and individual fertility intentions from the perspective of family–work conflicts. Instead, (Liu et al. 2021) investigates the factors influencing individual fertility intentions.

2.4. Life Expectancy–Fertility Rate

With increases in GDP per capita, life expectancy at birth, and the expected number of years that women will attend school, the overall fertility rate is currently drastically declining in the majority of regions. In addition, TFR displayed a minor recovery as societal progress and rising factor levels coincided. Therefore, it is recommended that governments, particularly those in developing nations, should take action to promote reproduction and address a number of issues brought on by a dropping fertility rate (Cheng et al. 2022). Salts serve as an electrolyte, which supports the proper operation of the neurological system and plays a significant role in the body (Welt et al. 1952). Users of Aral Sea water have experienced illnesses such as hypertension, hypercalciuria, cardiovascular disorders, high blood pressure, kidney stones, and bone metabolism because of the basin’s extremely high salt content (Abedin et al. 2020).

Along with a high rate of perinatal and infant mortality, it was discovered that the Aral Sea region had a decline in life expectancy among demographic indices. Indeed, the average reproductive age range is 15–49 years old (Turdybekova et al. 2015). In contrast to women in other Soviet regions, the Aral women more frequently experienced later menarche, menstrual abnormalities, and spontaneous miscarriages, according to a survey on women’s reproductive stages conducted in Kazakhstan (Ataniyazova 1995). For 88.2% of the women in labour, the ideal reproductive age was between 20 and 34 years old (Turdybekova et al. 2015).

Most women experience anomalies in their immunological, pituitary–thyroid, and hematopoietic systems prior to the commencement of pregnancy (Anchita et al. 2021). Women were impacted by inadequate nutrition during pregnancy. Meanwhile, prolonged exposure to the chemicals in the Aral Sea region compromises the immune systems of infants who are not breastfed (Johnson and Lichter 2019). These substances build up in fetuses during the postnatal period through nursing and the placenta. Therefore, toxicants and unfavourable environmental conditions have a major impact on women’s reproductive health (Reimov and Fayzieva 2013).

3. Model Specification

Generally, there exists a negative correlation between fertility rates and economic progress. As nations experience economic development, there is a tendency for fertility rates to decrease. This phenomenon is often ascribed to the growing availability of education, especially for women, more involvement of women in the workforce, enhanced healthcare, and improved family planning services. (Miles 2023) already argued that the decrease in labour input is anticipated to decrease aggregate economic growth, but this prediction is not a reliable indicator of how living standards will change. On the other hand, carbon taxes, which decrease income, may increase fertility rates, exacerbating climate change caused by population growth. Therefore, investing in women’s fertility rates involves several factors, such as leveraging economic development, reducing greenhouse gas emissions, increasing life expectancy, managing population ageing, and utilizing information communication technologies. This study primarily utilizes prior research as the empirical foundation for variable selection. Proceeding on with the previous studies (Kim et al. 2020; Gerlagh et al. 2023), the research explores the impact of environmental factors on a woman’s fertility rate. The empirical assessment was conducted using a basic model as follows:

$${FR}_{it}=f\left({GDP}_{it},AGE, {GHGs}_{it}, {INTERNET}_{it}, {LEX}_{it}\right)$$

(1)

In Equation (1), the indicator $“FR”$ represents fertility rate, which is measured by birth per woman. $“GDP”$ describes the overall economic growth of a country. $“AGE”$ refers to the percentage of women aged 15 to 64 in the total female population. $“INTERNET”$ represents the individuals using the Internet. $“GHGs”$ refers to the total greenhouse gas emissions in kt of CO₂ equivalent which calculated short-cycle biomass burning from the total CO₂ emissions, while $“LEX”$ represents life expectancy at birth and serves as a measure of health status.

For the scope of this study, the Aral Sea basin countries considered, at different time intervals from 1990 to 2021, include Uzbekistan, Tajikistan, Turkmenistan, Afghanistan, Iran, Kyrgyzstan, and Kazakhstan (

Table 1. Explanation of variables.

Variables	Description	Year
FR	Woman’s fertility rate (births per woman)	1990–2021
GDP	Economic growth proxied by GDP per capita (current USD)	1990–2021
AGE	Population ages 15–64, female (% of female population)	1990–2021
GHGs	Total greenhouse gas emissions (kt of CO2 equivalent)	1990–2021
INTERNET	Individuals using the internet (% of population)	1990–2021
LEX	Life expectancy at birth, female (years)	1990–2021

Source: Author’s own contribution.

).

To improve accuracy and reduce the issue of heteroscedasticity among variables, we converted the variables into logarithmic form based on Equation (1), which allows for direct elasticity-based comparisons and enables more precise results:

$$ln{FR}_{it}={\mu }_0+{\mu }_{1}ln{GDP}_{it}+{\mu }_2{lnAGE}_{it}+{\mu }_3{lnGHGs}_{it}+{\mu }_{4}ln{INTERNET}_{it}+{\mu }_{5}ln{LEX}_{it}+{\epsilon }_{it}$$

(2)

where $i=$ 1, N denotes the country, and t = 1, …, T denotes the time.

The variables and their long-run elasticities are donated by ${\mu }_1, {\mu }_2, {\mu }_3, {\mu }_{4,}$ and ${\mu }_5$, respectively. The error term is represented by $\epsilon$. The current study utilizes panel data from 1990 to 2021 for empirical estimation. All data are collected from the World Development Indicators (2023).

Table 2. Descriptive statistics.

Variables	Obs	Mean	Std.Dev.	Min	Max
lnFR	224	1.151	0.381	0.525	2.043
lnGDP	210	4.093	0.109	3.865	4.286
lnAGE	224	7.109	1.120	4.921	9.538
lnGHGs	217	11.177	1.532	8.797	13.700
lnINTERNET	165	1.216	2.775	−7.803	4.510
lnLEX	224	4.241	0.086	3.879	4.370

presents the statistical description of the variables used in the model, including fertility rate, GDP per capita, greenhouse gas emissions, internet use, population ages, and life expectancy, which are crucial for assessing socioeconomic dimensions of well-being and livelihood in the Aral Sea regions. The data table reveals that the variable under examination has an arithmetic mean and median values within the reported range. The mean values of $lnFR, lnGDP, lnAGE, lnGHGs, lnINTERNET$, and $lnLEX$ are 1.15, 4.09, 7.10, 11.1, 1.21, and 4.24, respectively. Consequently, the following variables, which were examined in this study, show a significant level of standard deviation: 0.38, 0.10, 1.12, 1.53, 2.77, and 0.08 for $lnFR, lnGDP, lnAGE, lnGHGs, lnINTERNET$, and $lnLEX$, respectively. The standard deviation indicates a minimal amount of data dispersion around the means of the variables.

Table 3. Correlation matrix and VIF test result.

	lnFR	lnGDP	lnAGE	lnGHGs	lnINTERNET	lnLEX
lnFR	1.0000
lnGDP	−0.9018	1.0000
lnAGE	−0.5694	0.7798	1.0000
lnGHGs	−0.6260	0.6161	0.7434	1.0000
lnINTERNET	−0.1675	0.4884	0.4966	0.0935	1.0000
lnLEX	−0.8503	0.8494	0.6997	0.4915	0.5598	1.0000
Variable			VIF		1/VIF
lnPOP			6.61		0.151350
lnLife			5.52		0.181305
lnGDP			4.51		0.221487
lnGHGs			3.09		0.323292
lnInt			1.87		0.535104

displays the results of using the Pearson correlation coefficient for matrix correlation for evaluating the interrelationships of the variables. Based on the statistics, the dependent variable and the independent variables were found to have a negative correlation. Furthermore, VIF test results were also shown to provide evidence for multicollinearity problem among the variables. According to the findings of VIF, there is no multicollinearity among the variables under the study.

4. Empirical Methodology

The empirical strategies involve three main stages in examining the relationship between fertility rate and environmental factors: $\left(i\right)$ examining the cross-sectional dependence and slope homogeneity of the underlying data and ascertaining the level of integration of the selected variables; $\left(ii\right)$ investigating the model-established variables that are crucial for long-term analysis, including long-run panel regressions; $\left(iii\right)$ a novelty of the final stage involves assessing the Granger non-causality test by Dumitrescu and Hurlin (2012) for heterogenous panel data to check the relationship between selected variables.

4.1. Panel FMOLS and DOLS

The empirical estimation approach involves the crucial step of estimating long-run coefficients, as highlighted in Equations (1) and (2). This step occurs after assessing underlying dataset cointegration characteristics. In this stage, we used both the FMOLS approach (fully modified OLS) and the DOLS approach (Dynamic Ordinary Least Square method) in our study. The empirical study often asserts that the Ordinary Least Square (OLS) approaches for panel data sometimes provide misleading results, rendering them wasteful. Endogeneity and serial correlations are potential problems that might occur when using OLS methods. The FMOLS and DOLS are often used in the literature as panel estimation techniques that emphasize heterogeneity, might potentially alleviate these issues (Pedroni 1999; Phillips and Hansen 1990).

Indeed, the FMOLS panel, based on compelling evidence, can validate a long-term link among variables, offering several benefits such as serial correlation, endogeneity, and cross-sectional heterogeneity, thereby enhancing the understanding of the variables under consideration (Erdal and Erdal 2020; Stock and Watson 1993). (Stock and Watson 1993) suggested a more modern and stronger technique, especially for small samples. This method corrects for potential simultaneity bias among the regressors and includes estimating long-run relationships using Dynamic OLS (DOLS). Equations (3) and (4) provide a detailed explanation of the mathematical expressions for these estimators.

$${\beta }_{FMOLS}={\left[N^{-1}{\sum }_{i=1}^N{\sum }_{t=1}^T{({\mu }_{it}-{\underset{\_}{\mu }}_i)}^2\right]}^{-1}\times \left[{\sum }_{i=1}^T\left({\mu }_{it}-{\underset{\_}{\mu }}_i\right)Q_{it}-T{\mathsf{\Delta }}_{\epsilon u}\right]$$

(3)

$${\beta }_{DOLS}=\left[N^{-1}{\sum }_{i=1}^N{\left\{{\sum }_{i=1}^T{K}_{it}{K}_{it}^{\prime }\right\}}^{-1}\times \left\{{\sum }_{i=1}^T{K}_{it}{P}_{it}\right\}\right]$$

(4)

Here, $\mu$ is the explanatory variable; $P$ describes the dependent variable; $K$ is the vector on regressors where $K=\mu -\underset{\_}{\mu }$.

The FMOLS and DOLS approaches are superior in estimating between groups compared to within groups (Saidmamatov et al. 2024). The measures considered address endogeneity issues by considering temporal precedence and using heteroskedastic standard errors. The DOLS methodology’s computational simplicity and ability to reduce biases make it preferable to FMOLS approaches (Kao 1999). The DOLS method, which utilizes leads and lags, is advantageous in addressing issues related to integration order and co-integration.

4.2. Robustness Check Estimations

To support our results, we applied a novel robustness check, which was the Pooled OLS estimation method based on (Driscoll and Kraay 1998). This method is utilized to provide robust standard errors in panel data, especially in dealing with cross-sectional dependence, where error terms across different entities are correlated. The standard errors for coefficient estimates are determined by calculating the square roots of the diagonal components of the asymptotic covariance matrix, as per Driscoll and Kraay’s method.

4.3. Panel Causality Test

In final step, we apply the Dumitrescu and Hurlin test which is a method developed to evaluate Granger causality in panel data contexts, focusing on variability across cross-sectional units in the panel, rather than time series data (Dumitrescu and Hurlin 2012). This allows for coefficient differences across units while providing a consolidated test for Granger causality. The test is conducted using a linear regression model for each individual cross-sectional unit $i$:

$$y_{i,t}={\alpha }_i+\sum _k^K{\beta }_{i,k}{y}_{i, t-k}+\sum _k^K{\gamma }_{i,k}{x}_{i, t-k}+{ϵ}_{i,t}$$

where

$y_{i,t}$ and $x_{i,t}$ are the observed variables for unit $i$ at a time $t$;

${\alpha }_i$ is the intercept for unit $i$;

${\beta }_{i,k}$ and ${\gamma }_{i,k}$ are the coefficients for the lagged values of $y$ and $x$, respectively;

$K$ is the number of lags;

${ϵ}_{i,t}$ is the error term;

Null hypothesis ${(H}_0):x$ does not Granger-cause $y$ for all cross-sectional units.

$$H_0: {\gamma }_{i,k}=0 \forall i,k$$

For each cross-sectional unit $i$, an individual Wald test statistic is computed to test the null hypothesis ${\gamma }_{i,k}=0$.

The average of the individual Wald test statistic is then calculated:

$$\overline{W}={\displaystyle \frac{1}{N}}\sum _{i=1}^N{W}_i$$

The standardized statistic follows a standard normal distribution under the null hypothesis, as the number of period and cross-sectional units tends to infinity.

$$Z={\displaystyle \frac{\sqrt{N}\left(\overline{W-K}\right)}{\sqrt{2K}}}$$

where $K$ is the number of lags.

5. Results and Discussions

The results of the cross-sectional dependency (CD test) and slope heterogeneity tests are displayed in

Table 4. The results of CD test and slope heterogeneity test.

Cross-Sectional Dependence Test H0: Under cross-sectional independence
lnFR lnGDP lnAGE lnGHGs lnINTERNET lnLEX	10.493 *** 17.271 *** 21.815 *** 11.91 *** 19.569 *** 23.174 ***
Slope heterogeneity test H0: Slope coefficients are homogeneous
Delta statistics	8.546 ***
Adj. Delta statistics	10.260 ***

Note: *** p < 0.01.

. The CD test indicates that the null hypothesis is rejected (Pesaran 2014) and demonstrates the presence of cross-sectional dependency among the variables. Furthermore, we applied the slope heterogeneity test proposed by (Hashem Pesaran and Yamagata 2008) and it indicated that the null hypothesis of slope homogeneity is rejected at a significance level of 1%. This showed that there are heterogeneity issues among variables.

The study utilized a second-generation unit root test due to the existence of cross-sectional dependence among variables. We applied a Cross-sectional Augmented (CADF) unit-root test which allows for unbalanced panel datasets. The results indicate stationarity at the first difference in

Table 5. The results of the CADF test.

	CADF
	I (0)	I (1)
lnFR lnGDP lnAGE lnGHGs lnINTERNET lnLEX	−1.697 −1.789 −0.787 −1.897 −4.317 *** −3.143 ***	−2.609 *** −3.868 *** −3.067 *** −3.287 *** −3.320 *** −2.935 ***

Note: *** p < 0.01.

, except for lnINTERNET and lnLEX in CADF. These findings serve as a foundation for further research on cointegrated interactions among variables, highlighting the importance of these tests in understanding research outcomes.

Moreover, we applied the Pedroni cointegration test (Pedroni 2004) which is used to establish a long-term correlation between fertility rate and its chosen determinants.

Table 6. The results of the Pedroni cointegration test.

	Statistics	Probability
$\mathbf{W}\mathbf{i}\mathbf{t}\mathbf{h}\mathbf{i}\mathbf{n}$
$\mathrm{v}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	−9.2927	0.0027
$\mathrm{r}\mathrm{h}\mathrm{o}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	2.6308	0.0108
$\mathrm{P}\mathrm{P}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	1.5682	0.0000
$\mathrm{A}\mathrm{D}\mathrm{F}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	−25.4261	0.0000
$\mathbf{B}\mathbf{e}\mathbf{t}\mathbf{w}\mathbf{e}\mathbf{e}\mathbf{n}$
$\mathrm{r}\mathrm{h}\mathrm{o}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	−4.8900	0.0096
$\mathrm{P}\mathrm{P}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	−4.0436	0.0123
$\mathrm{A}\mathrm{D}\mathrm{F}-\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}$	−4.8554	0.0115

shows that v, PP, and ADF statistics in the “within dimension” analysis are statistically significant at the 1% level while rho statistics were considered at the 5% level. Based on the findings in the “between dimension” analysis, PP and ADF statistics are considered significant at the 5% level, indicating a strong association between variables over a long period.

5.1. GDP (Gross Domestic Product)–Fertility Rate

Based on the empirical findings presented in

Table 7. The results of long-run panel regressions.

	(1)	(2)	(3)
VARIABLES	FMOLS	DOLS	Robustness Check (Pooled OLS)
lnGDP	0.141 *** (0.009)	0.094 *** (0.024)	0.147 *** (0.020)
lnAGE	−2.675 *** (0.125)	−2.389 *** (0.333)	−2.649 *** (0.186)
lnGHGs	−0.080 *** (0.006)	−0.064 *** (0.014)	−0.083 *** (0.010)
lnINTERNET	0.029 *** (0.003)	0.036 *** (0.006)	0.028 *** (0.007)
lnLEX	−1.801 *** (0.198)	−2.073 *** (0.518)	−1.946 *** (0.349)
Constant	19.617 *** (0.528)	19.744 *** (1.257)	20.119 *** (1.335)
R-squared	0.483	0.964	0.936
Observations	159	157	160

Note: *** p < 0.01; (the numbers in parentheses represent standard errors).

, the results obtained from FMOLS, DOLS, and Pooled OLS (Robustness check) estimation approaches show that GDP has a positive effect on fertility rate at the 1% significance level. This means that a 1% increase GDP per capita results in 0.141%, 0.094%, and 0.147% respectively. This suggests that economic growth enables governments to boost healthcare expenditures, maintain financial stability, and provide affordable medical treatments, thus increasing fertility rate. The Aral Sea basin countries are facing significant ecological deterioration, which needs improved medical assistance to alleviate the health consequences on the local people. The findings align with previous research (Ashraf et al. 2013; Karra et al. 2017).

5.2. AGE (Population 15–64 Ages, Female)–Fertility Rate

The coefficients of AGE (population 15–64 in age, female) in FMOLS, DOLS, and Pooled OLS depict that women aged 15–64 have a strongly negative correlation with fertility rate at a 1% significance level, making up 2.675, 2.389, 2649, respectively. Generally, several factors, like increased economic participation and educational attainment in women in this age group, are linked to reducing fertility rates in the Aral Sea basin. Employment offers women financial autonomy and professional ambitions, potentially reducing the likelihood of marriage and childbirth, and enabling them to have fewer children (Sonfield et al. 2013).

5.3. Greenhouse Gas Emissions (GHGs)–Fertility Rate

Our empirical approaches demonstrate that greenhouse gas emissions (GHGs) significantly and adversely impact the fertility rate of the Aral Sea basin countries investigated. Precisely, a 1% increase in the amount of GHGs in the Aral Sea basin countries results in a 0.080%, 0.064%, and 0.083% reduction in women’s fertility rates. GHGs in the Aral Sea region have been shown to lower fertility rate due to their correlation with changes in vulnerable climates and environmental catastrophes. Our empirical estimates are in line with the findings of previous research (Matthew et al. 2018; Majeed and Ozturk 2020).

5.4. Internet Use–Fertility Rate

Based on the findings of Table 7, the usage of internet has a positive impact on women’s fertility rates in selected regions. More specifically, women’s fertility rates increased by 0.029%, 0.036%, and 0.028%, respectively, for every 1% rise in internet usage in the Aral Sea basin countries. Generally, the internet provides comprehensive information on reproductive health, family planning, and fertility treatments, aiding couples in making informed decisions and improving reproductive outcomes. Our results align with a previous study by (Billari et al. 2019).

5.5. Life Expectancy–Fertility Rate

The outcomes derived from FMOLS, DOLS, and Pooled OLS indicate that there is a negative relationship between life expectancy and women’s fertility rates in the selected region. Based on the results of Table 7, women’s fertility rates decrease by 1.801%, 2.073%, and 1.946% when life expectancy increases by every 1% significance level. This suggests that higher life expectancy improves living conditions and expenses, especially in urban areas, leading to fewer children based on economic stability in Aral Sea basin countries. Our findings align with those who confirm that increased life expectancy leads to reductions in fertility rates (Cheng et al. 2022; Low et al. 2013).

5.6. Panel Causality Test Results

Finally, we applied the Dumitrescu and Hurlin (2012) test which utilized a causality for heterogenous panels to examine the direction of connectivity between variables.

Table 8. Results of Dumitrescu and Hurlin (2012) panel non-causality test.

Null Hypothesis	W-Stat.	Zbar-Stat.	Probability
$lnFR$ homogenously does not affect $lnGDP$	3.5678	4.8039	0.0000
$lnGDP$ homogenously does not affect $lnFR$	9.9881	16.8340	0.0000
$lnFR$ homogenously does not affect $lnAGE$	56.7367	104.2737	0.0000
$lnAGE$ homogenously does not affect $lnFR$	9.0285	15.0199	0.000
$lnFR$ homogenously does not affect $lnGHGs$	1.4038	0.7555	0.4499
$lnGHGs$ homogenously does not affect $lnFR$	1.6182	1.1565	0.2475
$lnFR$ homogenously does not affect $lnINTERNET$	5.8423	9.0591	0.0000
$lnINTERNET$ homogenously does not affect $lnFR$	6.1596	9.6527	0.0000
$lnFR$ homogenously does not affect $lnLEX$	4.0190	5.6480	0.0000
$lnLEX$ homogenously does not affect $lnFR$	5.3729	8.1809	0.0000

depicts the results of the Dumitrescu and Hurlin (2012) non-causality test. This method uses $Wbar$ and $Zbar$ statistics, which are statistical methods that calculate the average of test statistics and standard normal distribution (Shafique et al. 2021). This test indicates a bidirectional impact from FR, GDP, and INTERNET to LEX, while GHGs does not cause FR, while showing the unidirectional causality (Table 8).

6. Conclusions

Many studies suggest that all nations in the Aral Sea basin should reduce environmental degradation and understand the impact of climate change and other environmental changes on reproduction rates to effectively mitigate these issues. Hence, this paper evaluated the impact of economic growth, the age of populations, life expectancy, internet use, and greenhouse gas emissions on fertility rates in Aral Sea basin countries from 1990 to 2021. This research utilized several econometric approaches, including FMOLS, DOLS, and Pooled OLS with the Driscoll–Kraay estimation method. Moreover, to check the causal relationships among variables, we applied Dumitrescu and Hurlin non-causality test.

Based on the empirical findings, increasing the age of the population (women aged 15–64, of total population) and increasing life expectancy lead to a reduction in fertility rate among women in Aral Sea basin countries. Similarly, greenhouse gas emissions (GHGs) have a negative impact on fertility rate. On the other hand, economic growth and internet usage can contribute to increasing the fertility rate in the region. Moreover, this research suggests that there is a causal relationship between the selected variables and fertility rate, except greenhouse gas emissions (GHGs). According to empirical findings, we advocate several policy implications for developing strategies in the Aral Sea basin region.

First, based on the results, greenhouse gas emissions significantly contribute to climate change, impacting the environment and human health, potentially affecting women’s fertility rates in the region. Therefore, governments should implement strategies to reduce greenhouse gas emissions by utilizing renewable energy sources, improving energy efficiency, and promoting sustainable transportation.

Second, the population of different age groups, especially the population of women aged 15–64 years, can contribute decreasing the fertility rate in the region. The region’s economic shift from agriculture to services and industries could increase women’s employment opportunities and their involvement in the labour market. Moreover, the Aral Sea’s ecological catastrophe has led to significant health issues, including respiratory and waterborne illnesses, which could potentially decrease reproduction rates by affecting overall reproductive health. Therefore, governments should implement inclusive initiatives that offer comprehensive financial, healthcare, and educational support to families, including housing, education, and healthcare services.

Third, based on empirical findings, higher life expectancy tends to decrease the fertility rate among women in the selected region. In this sense, the government should provide economic measures to ensure the long-term sustainability and mitigate the economic impact of an increasing ageing population, including new pension reforms and investment in technology and innovation.

Finally, the internet can contribute to increasing the fertility rate among women in Aral Sea countries by enhancing access to reproductive health information, economic opportunities, and social support. Therefore, these countries should focus on allocating funds for the expansion of broadband infrastructure to ensure that all communities, especially rural and difficult-to-reach areas, have access to reliable internet services.

Our research also has some limitations; as authors focused on general fertile age (15–64), subcategories could be focused on, such as primary (15–30) and secondary fertile age (30–64) groups. While on the one side this is related to limitations of data availability, on the other side it represents a future avenue to explore.