This study employs a quantitative research design to investigate the relationship between climate variables (rainfall, temperature, and CO₂ emissions), population growth, and net migration in Somalia. The study utilizes annual time series data spanning 30 years (1990–2020) for Somalia. The dataset includes five variables: net migration, rainfall, temperature, CO₂ emissions, and population growth, sourced from the World Bank and Climate Change Knowledge Portal. This 30-year dataset ensures robust temporal coverage and reliable statistical analysis of migration drivers, using time series econometric methods to capture both short- and long-term dynamics. Given the diverse integration orders of the variables (I (0) and I (1)), the Autoregressive Distributed Lag (ARDL) model is applied, as it accommodates mixed integration levels and allows for robust analysis of cointegrated relationships.

Data for this study are sourced from the World Bank and the Climate Change Knowledge Portal (CCKP), ensuring reliable and consistent coverage across the study period. Table 1 provides an overview of the variables, their measurements, and sources. The variables include net migration (NM), average annual precipitation (RF) in millimeters, average annual temperature (TP) in degrees Celsius, CO₂ emissions (CO2) measured in kilotons (kt), and annual population growth (PG) as a percentage.

These variables are chosen based on their relevance to the study’s aim of understanding migration dynamics. Net migration, obtained from the World Bank, is measured as the difference between inbound and outbound migrants annually, reflecting migration trends. Average annual precipitation, sourced from the CCKP, serves as a proxy for agricultural productivity and water availability. Temperature, also from CCKP, represents environmental stability, with significant temperature changes potentially impacting livelihoods. CO₂ emissions, obtained from the World Bank and expressed in kilotons, indicate levels of environmental degradation, where high emissions may correlate with resource depletion and restricted migration capacity. Population growth, sourced from the World Bank, is included as an indicator of demographic pressures, where rapid population growth in a fragile environment can intensify competition for resources, influencing migration.

The ARDL model is chosen for this analysis due to its flexibility in handling variables with mixed orders of integration, I (0) and I (1), without requiring all variables to be of the same order or cointegrated. This feature is crucial for this study, as the unit root tests (ADF and PP) revealed a mix of stationary variables at level (I (0)) and first difference (I (1)). The ARDL bounds testing approach allows for testing the existence of a long-run relationship even in the presence of variables with different integration levels, making it ideal for analyzing time series data with varied integration properties. Additionally, the ARDL model enables the estimation of both short- and long-term effects, providing a comprehensive view of how climate and demographic factors impact migration.

The key variables in this study include net migration as the dependent variable, and rainfall, temperature, CO₂ emissions, and population growth as independent variables. Net migration was selected as the dependent variable because it provides a holistic measure of population movement, capturing both immigration and emigration trends. However, in the Somali context, detailed data on immigration and emigration separately are not available. As a result, net migration primarily reflects emigration driven by climate stress, conflict, and economic instability, as immigration into the country has been minimal over the study period. Rainfall, measured as average annual precipitation (mm), is used as a proxy for agricultural productivity and water availability, where favorable rainfall conditions can support livelihoods, reduce displacement pressures, and affect migration patterns. Temperature, measured as average annual temperature (°C), represents climatic stability and resilience to environmental changes. CO₂ emissions, expressed in kilotons (kt), are used as a proxy for environmental degradation. High emissions are linked to climate-related resource depletion, which may limit people’s capacity to migrate due to reduced economic means. Population growth, expressed as a percentage, indicates demographic pressures on resources, where rapid population growth can increase competition for resources and intensify migration pressures. The general functional forms of such models are as follows: (1) InN M t = β 0 + β 1 I n R F t + β 2 I n T P t + β 3 InCO 2 t + β 4 I n P G t

InN M t is logarithms of dependent variable net migration at t time, RFt, TPt, CO2t, are logarithms of rainfall, temperature and carbon dioxide at time t, while PGt, is logarithms of population growth at t time. β 0 is the intercept, β 1 , to β 4 are coefficients. Similar models have been used in studies like Afifi and Warner (2008), on environmental degradation in Egypt and (Marchiori et al., 2012) on rainfall and migration in sub-Saharan Africa, showing the effectiveness of this approach for climate-migration analysis. The relationship is expressed in Equation 1 as: InNM = β 0 + ∑ i = 1 ρ φ 1 Δ R F t − i + ∑ i = 1 ρ φ 2 Δ InTP t − i + ∑ i = 1 ρ φ 3 Δ InCO 2 t − i + ∑ i = 1 ρ φ 4 Δ InPG t − i + δ 1 InRF t − 1 + δ 2 InTP t − 1 + δ 3 InCO 2 t − 1 + δ 4 InPG t − 1 + δ 7 E C t − 1 + ε t

β 0 is the intercept, φ is the coefficient for the long run, δ represents the short run, ρ is the number of lags, Δ is the operator of first difference, ECt-1 is the speed of adjustment parameter and ℇt is the error term.

To ensure the validity of the ARDL model, stationarity tests were conducted using the Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests. These tests helped identify the integration order of each variable, confirming that some variables were stationary at level (I (0)) and others at first difference (I (1)), which justifies the use of the ARDL model for this study. The bounds testing approach within ARDL is applied to determine the existence of a long-run relationship between the variables, while short-term dynamics are assessed through an Error Correction Model (ECM) derived from the ARDL results.

The ARDL model is ideal for this study due to its flexibility in handling variables with mixed integration orders (I (0) and I (1)). Unlike Vector Autoregressive (VAR) models, which assume uniform stationarity, ARDL accommodates datasets with different integration levels. Furthermore, it enables the simultaneous estimation of short- and long-term relationships, providing a comprehensive analysis of migration dynamics in Somalia.

To ensure the reliability and robustness of the ARDL model, various diagnostic tests were performed. The Breusch-Godfrey LM test was used to check for serial correlation, while the Breusch-Pagan-Godfrey test assessed homoscedasticity. The Jarque-Bera test verified the normality of residuals, and model stability was evaluated using the CUSUM and CUSUMQ tests. Additionally, stationarity tests (ADF and PP) were conducted to confirm the integration properties of the variables. The results of these tests, including unit root analysis and model diagnostics, are provided in the following section.

The ARDL model estimation and diagnostic tests are conducted using EViews. This study relies exclusively on secondary data from public sources, minimizing ethical concerns related to data collection. Proper acknowledgment of data sources is maintained to ensure academic integrity and transparency.

The descriptive statistics presented in Table 2 summarize the key variables used in the study. Net migration (NM) has a mean of −23,449.07, indicating that more people leave Somalia than enter, though the high standard deviation of 143,025.6 highlights substantial variability over the study period. Rainfall (RF) averages 277.69 mm, reflecting Somalia’s reliance on irregular precipitation for agriculture. Temperature (TP) is relatively stable, with a mean of 26.95°C, while CO₂ emissions (CO2) average 0.166 kilotons per capita, underscoring Somalia’s low industrial activity but high vulnerability to environmental degradation. Population growth (PG) exhibits variability, with a mean of 2.87%, reflecting Somalia’s youthful and rapidly growing population. The Jarque-Bera test reveals non-normality for net migration and population growth, which may be attributed to outliers or extreme migration events, such as drought-induced displacements.

The correlation matrix in Table 3 reveals important relationships among the variables. Net migration (NM) shows a strong positive correlation with population growth (PG) (0.723), indicating that demographic pressures are a significant driver of migration in Somalia. Rainfall (RF) is positively correlated with NM (0.2994), suggesting that improved rainfall conditions are associated with reduced migration pressures. Conversely, NM has a weak negative correlation with CO₂ emissions (−0.136), hinting at the role of environmental degradation in constraining migration capacity. Population growth (PG) shows a moderate negative correlation with CO₂ emissions (−0.309), emphasizing the interplay between demographic and environmental stressors in influencing migration dynamics.

The results of the Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests, presented in Table 4, indicate a mix of stationary variables at level (I (0)) and first difference (I (1)). Net migration (NM), rainfall (RF), and population growth (PG) are stationary at level, while temperature (TP) and CO₂ emissions (CO2) become stationary after first differencing. This mixed integration order justifies the use of the ARDL model for further analysis.

The Bounds Test for cointegration (Table 5) confirms a long-run relationship among the variables, with an F-statistic of 40.99 exceeding the critical upper bounds at all significance levels. This establishes the suitability of the ARDL model to capture both short- and long-term effects of climate and demographic factors on migration in Somalia.

The long-run coefficients in Table 6 provide important insights into the determinants of migration in Somalia. Rainfall (lnRF) exhibits a positive coefficient (0.17009) with a p-value of 0.0562, suggesting a marginally insignificant effect on migration. However, its positive sign indicates that increased rainfall could stabilize agricultural productivity, indirectly reducing migration pressures by improving rural livelihoods. Temperature (lnTP) has a negative coefficient (−4.00638) but is not statistically significant (p-value = 0.1014). This suggests that temperature alone may not play a central role in driving migration, although its effects could interact with other variables like rainfall and agricultural productivity. CO₂ emissions (lnCO2) show a significant negative impact on migration, with a coefficient of −0.06603 (p-value = 0.0181). This highlights the role of environmental degradation, such as deforestation and soil erosion, in reducing the financial and logistical capacity of households to migrate, potentially creating a “trapped population” effect. Population growth (lnPG) exerts a strong positive and highly significant influence on migration, with a coefficient of 1.858317 (p-value <0.0001). This reflects the demographic pressures driving migration, as growing populations increase competition for limited resources, pushing individuals to migrate in search of better opportunities. The constant term (C) is also significant, with a coefficient of 21.48173 (p-value = 0.0092), which likely captures unobserved factors influencing migration dynamics in the long run.

The short-run dynamics, presented in Table 7, reveal important insights into migration determinants in Somalia. Rainfall (D (lnRF(−1)) exerts a positive and significant effect on net migration, with a coefficient of 1.631248 (p-value = 0.0223). This indicates that favorable rainfall conditions have an immediate positive impact on migration by stabilizing agricultural productivity and rural livelihoods, reducing migration pressures. Temperature (D (lnTP(−1)) has a negative coefficient (−72.7576) but is not statistically significant (p-value = 0.172), suggesting that temperature variability does not have a substantial short-term impact on migration. This may reflect Somalia’s adaptation to temperature fluctuations or the stronger influence of other variables like rainfall. CO₂ emissions (D (lnCO2(−1)) have a negative but statistically insignificant coefficient (−0.29809, p-value = 0.9532). This suggests that while environmental degradation is significant in the long run, its immediate effects may not strongly influence migration decisions in the short term. Population growth (D (lnPG(−1)) remains a significant driver of migration in the short run, with a negative coefficient of −0.35705 (p-value = 0.0023). This underscores the immediate resource pressures created by a growing population, pushing individuals to migrate in search of better opportunities.

The error correction term (ECM(−1)) has a negative coefficient (−0.877089) and is marginally significant (p-value = 0.077). This indicates that about 87.7% of short-term deviations from the long-run equilibrium are corrected each period, highlighting the system’s ability to adjust toward long-run stability.

Table 8 provides diagnostic test results to assess the validity and robustness of the model. The Heteroskedasticity Test (Breusch-Pagan-Godfrey) shows an F-statistic of 1.296595 with a p-value of 0.3129, indicating no evidence of heteroskedasticity, as the p-value is above the common significance levels (1, 5, and 10%). This suggests that the residuals have constant variance, satisfying one of the key assumptions of regression analysis. The Breusch-Godfrey Serial Correlation LM Test yields an F-statistic of 0.405148 and a p-value of 0.675, indicating no serial correlation in the residuals. This is essential for ensuring unbiased and efficient estimators in time series models. Additionally, the Normality Test has an F-statistic of 7.121661 with a p-value of 0.616564, suggesting that the residuals are normally distributed, as the p-value is well above the 0.05 threshold.

Figure 1 illustrates the CUSUM and CUSUM of Squares tests for model stability. The CUSUM plot shows the cumulative sum of residuals over time, and it remains within the 5% significance boundaries, indicating that the model parameters are stable over the sample period. Similarly, the CUSUM of Squares plot also stays within the 5% critical bounds, reinforcing the stability of the model. Together, these diagnostic tests and stability checks confirm that the model is well-specified, with no issues related to heteroskedasticity, serial correlation, or parameter instability. This reliability is crucial for making accurate inferences from the model’s results.

The findings of this study underscore the need for comprehensive policies to address the drivers of migration in Somalia. Implementation strategies should prioritize investments in education, vocational training, and sustainable economic development to empower the growing youth population and reduce migration pressures.

A key area of focus is improving access to education in rural areas, where many families face barriers to schooling. This can be achieved by building and equipping schools in underserved regions, training and retaining qualified teachers, and introducing community-based education programs. Scholarships and incentives for girls’ education can further enhance enrollment, ensuring that more young people gain the skills needed to contribute to their communities. Collaborating with local leaders and Islamic scholars can promote educational initiatives in culturally appropriate ways, fostering community buy-in.

Vocational training tailored to Somalia’s economic context is equally critical. Programs should focus on practical skills that align with local needs, such as sustainable farming techniques, livestock management, and small business development. Mobile training units could bring vocational programs to remote areas, ensuring inclusivity. Public-private partnerships can support these efforts by linking training programs to market opportunities, such as creating pathways for trained individuals to access microloans, cooperatives, or apprenticeships. This would provide young people with tangible opportunities to start businesses or join the workforce, reducing the economic pressures that drive migration.

To support these educational and training initiatives, investments in rural infrastructure are essential. Building irrigation systems, marketplaces, and reliable transportation networks can amplify the impact of training by enabling rural communities to access resources and sell their products more efficiently. Community-driven development models can be employed to ensure infrastructure projects reflect local priorities and needs. For example, water user associations could manage small-scale irrigation systems, ensuring equitable access and sustainability.

Strengthening local economies through entrepreneurship programs and support for small businesses can further reduce migration pressures. Establishing business incubators in urban and semi-urban areas can help young entrepreneurs turn ideas into viable enterprises. These incubators could provide mentorship, funding, and technical assistance, particularly in sectors like renewable energy, sustainable agriculture, and technology. Partnerships with international organizations and Somali diaspora networks could bring in expertise and funding to bolster these initiatives. Finally, a coordinated approach involving government agencies, NGOs, and international partners is essential for implementing these policies effectively. A national task force could oversee the integration of education, training, and infrastructure programs into Somalia’s broader development strategy. Regular monitoring and evaluation would ensure accountability and help refine programs based on lessons learned, ensuring long-term success.

While this study provides valuable insights into the relationship between climate change, population growth, and migration in Somalia, it is not without its limitations. One key limitation is the reliance on secondary data, which may not fully capture the nuances of local migration drivers, especially in the context of political instability or social factors. Moreover, the available data on climate variables and migration trends, although comprehensive, may not reflect real-time changes or short-term fluctuations in local communities that experience the most direct impacts of climate shocks.

Another limitation is the focus on national-level analysis, which may overlook regional variations within Somalia. Different regions of the country experience varying degrees of climate stress, population growth, and migration pressures, which are not fully captured in this study. A more localized analysis would provide a deeper understanding of how regional differences in environmental and socio-economic factors influence migration.

Additionally, the study focuses primarily on environmental factors and population growth, but other critical determinants of migration, such as political instability, conflict, and economic conditions, are not explicitly analyzed. While this study isolates the effects of climate variables, migration is a multidimensional issue, and future research should explore how these additional factors interact with environmental drivers to influence migration patterns.

While the ARDL model effectively captures short- and long-term relationships between climate variables and migration, we acknowledge its limitations in fully reflecting the complex opportunity-constraint dynamics influencing migration decisions. The quantitative approach used here provides valuable insights into the statistical relationships between variables. However, migration decisions are also shaped by socio-political and cultural factors that are difficult to quantify.

Future research could benefit from incorporating qualitative methods, such as interviews, surveys, or ethnographic studies, to explore migration drivers in greater depth. These approaches would complement the quantitative findings by providing a more holistic understanding of the motivations behind migration decisions and the lived experiences of those affected by climate change and migration pressures.

Qualitative data would also offer richer insights into the complex interplay between climate stress, political instability, and economic constraints in shaping migration outcomes. Such studies would allow for a deeper exploration of the opportunity-constraint dynamics that influence individual and community decisions to migrate, which are difficult to fully capture through quantitative methods alone.

Finally, as Somalia continues to face new climate challenges, future studies should assess the effectiveness of policy interventions aimed at reducing migration pressures, particularly in the context of climate adaptation and population management. Evaluating the impact of specific programs, such as disaster preparedness initiatives or reforestation efforts, would provide valuable information for refining policy strategies in the future.