This gap is critical: Malawi’s rising educational migration rates and its distinct sociocultural and economic realities suggest that global findings may not apply. By centreing Malawian students, this study addresses three unresolved questions in the literature: (1) How do Malawi-specific cultural and economic structures mediate returnees’ reintegration? (2) What role do localised spiritual/community practices (e.g., Ubuntu frameworks) play in coping strategies? (3) How can policy interventions be adapted to low-resource contexts? Existing research, while foundational, lacks this contextual nuance, underscoring the need for the present inquiry.
This study employs a convergent mixed methods design to provide complementary insights into the challenges faced by returnees. The quantitative analysis, including descriptive statistics, regression, and SEM, identifies broad patterns and relationships (Banda et al. 2023), while the qualitative data from in-depth interviews offer a deeper, contextual understanding (Banda et al. 2024) of these issues. The integration of both data sets through triangulation enhances the validity of the findings by cross-verifying results (Banda and Liu 2025) and providing a comprehensive view of the reintegration process, resulting in more robust and dependable conclusions.
While the quantitative research highlights key determinants such as community judgement and financial issues, the qualitative interviews explore the underlying narratives, revealing additional themes and issues not captured by the survey.
The study targeted Malawian students who returned from international higher education destinations after failing their degree programmes. The data was collected between June 10, 2022 and September 3, 2022. Given the challenges of accessing this specific population, snowball and purposive sampling techniques were employed to recruit 143 participants. Snowball sampling, where seed subjects recruit future subjects from among their acquaintances (Banda et al. 2024), was particularly effective for reaching distant or “hidden” populations (Heckathorn 1997).
The inherent limitations of snowballing, including homophily bias and restricted network reach, were mitigated through the recruitment of multiple seeds strategically selected to maximise demographic and geographic diversity (Banda and Liu 2025). Seeds were drawn from distinct subgroups to broaden referral chains and reduce the overrepresentation of homogeneous networks. This approach enhanced sample heterogeneity, ensuring the inclusion of perspectives from marginalised or socially isolated individuals often excluded in single-seed designs.
While residual bias remains inherent to the method, the use of diversified seeds strengthened the sample’s representativeness, aligning with best practices for studying hard-to-reach populations. This approach was complemented by purposive sampling to ensure the inclusion of participants who could provide rich, detailed insights into their experiences, thereby enhancing the depth and validity of the findings (Banda and Liu 2025; Palinkas et al. 2015). In-depth semi-structured interviews were conducted with a purposive sample of 30 returnees. This sample size was deemed sufficient to achieve data saturation, where no new themes or insights emerged from additional interviews (Guest et al. 2006; Banda et al. 2024).
A structured questionnaire was developed to collect quantitative data, drawing on validated instruments from previous studies on international student experiences and academic failure (Bodycott 2009). To assess the post-return experiences of Malawian returnees, relevant items from Bodycott’s (2009) instrument were integrated with those from other validated scales. Specifically, items addressing psychological adjustment, such as feelings of emotional overwhelm and stress following return, were included to complement the GAD-7 and PHQ-9 in measuring anxiety and depression. Additionally, questions on identity loss and cultural reintegration were selected to capture the challenges of adjusting to home culture after studying abroad, aligning with the broader themes of social stigma and identity issues measured in other instruments. Social reintegration was assessed with items related to community isolation and relationship changes, which were combined with measures of social support from the SSQ-6 and FSS. This mixed approach ensures comprehensive measurement (Banda et al. 2025) of the psychosocial and cultural challenges returnees face, while seamlessly aligning with the economic and financial constructs captured in the study.
Item reduction was undertaken to minimise respondent burden, mitigate survey fatigue, and enhance participation rates while preserving the psychometric integrity and theoretical coherence of the instrument. By streamlining scales, the study prioritises statistical power through higher response rates without compromising construct validity.
Items were removed based on redundancy (duplicate measurement of identical constructs), weak factor loadings in prior validation studies (e.g., items explaining <5% variance in exploratory factor analyses), and low contextual relevance to the study's core focus on post-return reintegration. Clinically validated tools (e.g., PHQ-9, GAD-7) were retained in full to ensure diagnostic comparability, while multidimensional scales (e.g., Perceived Stigma Scale, Social Identity Scale) were truncated to retain items with the highest discriminant validity and alignment with the study's theoretical framework. Cross-loading items were critically reviewed and excluded unless they contributed unique variance to overlapping constructs (e.g., cultural dissonance versus identity loss). This approach balanced brevity with methodological rigour, ensuring the instrument remained robust, culturally relevant, and statistically interpretable. Pretesting was conducted to ensure the reliability and validity of the study. The final set of items is in Table 1.
The study examines several types of support mechanisms that play distinct roles in the psychological well-being and successful reintegration of returnees, including familial support, institutional support, and community support. Familial support was measured using the Family Support Scale (FSS), which captures the emotional and practical assistance provided by family members. Institutional support was assessed through the Institutional Support Scale (ISS), which evaluates the effectiveness of government policies and programmes designed to support returnees. Community support was measured using the Community Integration Questionnaire (CIQ) and the Social Identity Scale (SIS), with a focus on community intervention and reintegration experiences. These instruments collectively define the operational concept of "support mechanisms" in this study, highlighting how each source contributes differently to the emotional resilience and overall well-being of the returnees.
The differentiation between these forms of support is essential to understanding their unique roles in shaping the returnees' reintegration process. Familial support primarily addresses the emotional and logistical assistance offered by family members, while institutional support refers to the policies and interventions implemented by governments or organisations to facilitate reintegration. Furthermore, community support, assessed through the CIQ and SIS, emphasises the role of community participation and the sense of belonging that returnees experience upon re-entering their society.
To assess the psychological and emotional well-being of returnees, the study incorporated the Generalised Anxiety Disorder 7 (GAD-7) and Patient Health Questionnaire-9 (PHQ-9) to measure anxiety and depression. These instruments were supplemented with items from Bodycott's instrument to capture emotional distress, social alienation, and cultural dissonance, which may not be fully represented by the GAD-7 and PHQ-9 alone. These additional items, focusing on cultural reintegration and identity loss, were essential to understanding how returnees perceive their sense of belonging and cultural adjustment, complementing the insights gained from the SIS and CIQ.
In addition to these instruments, the study used the Social Support Questionnaire (SSQ-6) to measure the basic perceptions of social support, addressing the first study question. The Perceived Stigma Scale was employed to capture social stigma, assessing key dimensions without overwhelming respondents. Finally, the use of these validated scales ensured that the data collected on financial stability, job insecurity, and economic hardship were robust, reliable, and relevant to the study's focus on the economic dimensions of reintegration. These instruments, renowned for their brevity and high reliability, provided a comprehensive understanding of the psychosocial challenges faced by returnees during the reintegration process.
To address the second research question, the Financial Well-being Scale (FWB) was employed to assess returnees' perceptions of financial stability. The FWB evaluates individuals' ability to meet their financial needs and their overall sense of financial security. Key items selected from the FWB, as indicated in Table 1 were included to capture participants' sense of economic security and their confidence in managing daily financial demands. This scale was chosen for its brevity and strong reliability, offering a clear indicator of financial stability that aligns with the study's focus on economic challenges.
To assess employment insecurity, the study incorporated the Job Insecurity Scale. This scale measures returnees' perceptions of job stability, a crucial factor influencing economic reintegration. The Job Insecurity Scale was adopted for its established psychometric reliability and its relevance to understanding the employment-related struggles of returnees, particularly in the context of reintegration. In addition, the study utilised the Economic Hardship Index to evaluate the broader economic challenges returnees face, including financial strain, insufficient income, and debt. The Economic Hardship Index was chosen due to its well-documented validity in measuring financial difficulties, making it an ideal tool for understanding the economic exclusion experienced by returnees.
To effectively answer research question 3, the study utilised the Generalised Anxiety Disorder 7 (GAD-7) and the Patient Health Questionnaire-9 (PHQ-9) to assess the psychological well-being of returnees, specifically focusing on anxiety and depression. These instruments were chosen for their brevity, reliability, and strong psychometric properties, making them well-suited for large-scale surveys where efficiency and precision are critical. The GAD-7 is widely recognised for its effectiveness in measuring anxiety symptoms, with robust validation across diverse populations. The PHQ-9 is the gold standard for assessing depression severity and is widely used in clinical and research settings due to its strong diagnostic accuracy and ease of use. These scales were preferred over alternatives, such as the Beck Depression Inventory or the Hospital Anxiety and Depression Scale (HADS), due to their shorter format, making them more feasible for survey-based research in a population of returnees, while still providing reliable and clinically meaningful assessments of mental health.
To ensure the reliability and validity of the survey instrument, Cronbach's Alpha, the Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy, and Bartlett's Test of Sphericity were conducted. Cronbach's Alpha demonstrated high internal consistency (α = 0.82), indicating the reliability of the measurement of the constructs (Cobern and Adams 2020). The KMO measure confirmed the sampling adequacy for factor analysis (KMO = 0.79) (Kaiser 1974), while Bartlett's Test of Sphericity supported the suitability of the data for structure detection (χ(105) = 652.34, p < 0.001) (Bartlett 1950).
An exploratory factor analysis with Varimax rotation was conducted to identify the latent constructs underlying the 43-item instrument. A four-factor solution emerged, explaining 62.4% of the total variance. Factor loadings ≥0.40 were retained, with cross-loadings ≤0.30 indicating discriminant validity (Tabachnick and Fidell 2019). Table 2 summarises the exploratory factor analysis.
A confirmatory factor analysis (CFA) was conducted to validate the four-factor measurement model (Economic Exclusion, Psychological Distress, Support Deficits, Cultural Reintegration) identified during exploratory factor analysis. The model exhibited excellent fit to the data: χ(224) = 278.15, p = 0.08; CFI = 0.96; TLI = 0.95; RMSEA = 0.04 (90% CI: 0.02-0.06); SRMR = 0.03 (Hu and Bentler 1999). All items loaded significantly (p < 0.001) on their respective factors, with standardised loadings ranging from 0.58 to 0.88. Composite reliability values (0.78-0.91) and average variance extracted (0.52-0.64) confirmed internal consistency and convergent validity. Discriminant validity was supported, as maximum shared variance (MSV < AVE) and interfactor correlations (r = |0.29-0.52|) did not exceed 0.70 (Fornell and Larcker 1981). Economic Exclusion correlated moderately with Psychological Distress (r = 0.52, p < 0.01) and Support Deficits (r = 0.38, p < 0.01), while Cultural Reintegration showed inverse relationships with Psychological Distress (r = -0.34, p < 0.01) and Support Deficits (r = -0.41, p < 0.01). These results validate the structural coherence of the measurement model, warranting its use in advanced analyses.
The interview guide was designed procedurally to encompass several essential elements that guarantee the instrument's durability and pertinence. A literature review was conducted on the experiences of overseas students, academic failures, and challenges of reintegration. This assessment provided an overview of the primary topics and issues faced by returnees, aiding in the development of the interview guide. Subsequently, experts in higher education, psychology, and qualitative research methods examined the draft of the interview guide. Their comments were utilised to enhance the questions, ensuring they were thorough, clear, and attuned to participant experiences.
The interview guide was pretested with several returnees who were not included in the study sample. The pilot phase was crucial in determining if the questions were confusing or challenging to answer, allowing for refinement. Participants in the pilot study were used to evaluate the clarity and coherence of the questions and to implement requisite modifications in the phrasing and sequence of the questions. The interview guide was modified iteratively based on the findings from pilot testing. This method involved multiple cycles of evaluation and adjustment of the instrument to reflect the complex nature of returnees' experiences accurately.
The survey link was disseminated through multiple channels, including emails, social media platforms, and community networks. To maximise accessibility and response rates, a bifurcated approach was utilised, combining online and offline distribution methods. Online responses were collected via an electronic survey platform optimised for accessibility on various devices. For participants with limited digital access, physical copies of the questionnaire were distributed in person.
This dual approach facilitated broad participation and minimised potential biases associated with a single mode of data collection (Creswell and Plano Clark 2017). The study achieved an overall response rate of 59.6% (N = 143/240), with offline surveys yielding a 66.7% response rate (80/120), aligning with AAPOR (American Association for Public Opinion Research) standards for face-to-face surveys, and online surveys a 52.5% rate (63/120). Participants in both modalities received standardised monetary incentives to bolster participation and mitigate non-response bias, a strategy aligned with best practices for enhancing engagement in mixed-method research. The higher offline response rate likely reflects the immediacy of incentive delivery and reduced attrition inherent to face-to-face administration, whereas the online cohort's moderate rate aligns with typical digital survey trends despite incentivisation. Both rates exceed the thresholds for statistical power in hypothesis testing (α = 0.05, β = 0.80) with the final sample (N = 143). Post hoc representativeness checks confirmed minimal demographic disparities between modes.
Qualitative data was collected via structured interviews with purposefully selected returnees. This sampling strategy facilitated the inclusion of participants capable of providing more extensive descriptions of their experiences. The interview guide was designed to elicit detailed accounts of the participants' educational experiences, challenges, coping techniques, and reintegration processes. Interviews were conducted in English and Chichewa to accommodate language preferences and to obtain more comprehensive responses. Each interview was done with the participant's consent and recorded audibly for further transcription and analysis.
Acquiring data from this group presented specific problems, as detailed below. Their circumstances may be humiliating, stigmatising, and emotionally distressing, rendering the returnees a difficult-to-access and vulnerable demographic. Recruiting participants proved challenging due to the stigmatised nature of their experiences. Many returnees faced social rejection and shame in their communities, which led to a reluctance to openly discuss their academic setbacks. To address these issues, the study team implemented the following measures to foster trust and maintain confidentiality.
Initial participants were identified through collaborative partnerships with the international student offices of overseas universities, which maintain administrative records of all enrolled students (including non-graduates), as well as with Malawian institutions such as embassies, consulates, and the Ministry of Education's scholarships and training division. These entities provided anonymised contact information — personal identifiers (e.g., names, national IDs) were removed or encrypted — for students who had enrolled in international programmes but did not graduate.
Complementing this, alumni networks of Malawian students abroad (including the Association of Malawian Students in China and the Association of Malawian Students in India) facilitated access to returnees through their membership registries. This dual approach balanced methodological rigour with ethical safeguards, as researchers could not link anonymised data (e.g., encrypted emails) to individual identities, thereby mitigating risks of unintended disclosure.
The key issue was establishing relationships with participants. The authors emphasised that the research was done anonymously, ensuring that participants' identities and comments remain confidential. This confidence was crucial to motivating participants to articulate their experiences. Interviews were conducted in a secure and comfortable environment, utilising a peaceful location and/or the participants' online platforms.
Moreover, researchers articulated the study's objective and the potential benefits of volunteer engagement, including the development of enhanced support for future repatriates. This method effectively diminished the participants' hesitance, resulting in more frank and cohesive comments.
Reflexivity and intercoder reliability enhanced qualitative rigour. Two researchers independently coded 30% of transcripts, achieving 85% consensus. Discrepancies were resolved through discussion and refining the codebook. Thick description and member checking were incorporated. Five participants reviewed the thematic summaries to validate the accuracy of the transcripts, while peer debriefing with external scholars critiqued the thematic interpretation. An audit trail tracked coding decisions, raw data, and analytical iterations. Table 2 summarises the validation techniques used in this study.
The knowledge produced and disseminated employed both quantitative and qualitative methodologies to elucidate the experiences of the returnees. Quantitative data were analysed using SPSS 26, employing descriptive statistics to examine demographic factors and critical variables. Chi-square tests, t-tests, and multiple regression analysis were used to determine the relationship between variables and the predictors of reintegration success (Field 2018).
T-tests are appropriate for comparing means between two distinct groups, such as employed versus unemployed returnees, to assess differences in psychological well-being. Multiple regression enables the examination of the influence of multiple independent variables on a dependent variable while controlling for confounding factors. Finally, SEM is ideal for testing complex relationships between latent variables, such as community integration and psychological well-being, and examining direct and indirect effects. Power analysis supports the use of SEM, ensuring valid results with medium effect sizes. These methods collectively provide a comprehensive analysis of the reintegration process among returnees.
The linearity assumption was confirmed through scatterplots showing a linear relationship between the independent and dependent variables. Homoscedasticity was validated with a residuals versus fitted plot, which displayed constant variance across all levels of fitted values. The normality of residuals was supported by a histogram and kernel density estimate (KDE), with a Shapiro-Wilk p value of 0.095, indicating normality. Outliers and influential points were assessed using Cook's Distance, with all values remaining below the threshold of 1, suggesting no undue influence from individual data points, as shown in Fig. 1.
The multicollinearity assumption was checked by calculating the Variance Inflation Factors (VIFs), which showed values ranging from 3.2 to 5, indicating no significant multicollinearity between the independent variables. These results suggest that the data is suitable for regression analysis.
The qualitative data were examined by thematic analysis, following Braun and Clarke's (2006) six-phase process. The initial phase involved open coding to identify the principal themes within the data, followed by axial coding to examine the interrelationships among these themes. The final stage involved selective coding to construct a narrative elucidating the participants' overall experiences. NVivo 12 facilitated the methodical and rigorous examination of the gathered qualitative data.
To handle missing data in this study, we used the Full Information Maximum Likelihood method (FIML), which is particularly suited to the SEM framework employed in our CFA and latent variable analyses. FIML was selected for its capacity to utilise all available data without listwise deletion, thereby preserving statistical power and maintaining the integrity of complex models. This approach is robust under missing-at-random (MAR) assumptions, making it ideal for datasets with small to moderate missingness, even though survey items were designed as compulsory to minimise incomplete responses. A key strength of FIML lies in its avoidance of imputation biases, thereby preserving the relationships between latent constructs, such as Economic Exclusion and Psychological Distress. Its direct integration with SEM software aligns with the study's CFA results, which demonstrate an excellent model fit (CFI = 0.96, RMSEA = 0.04). Additionally, its efficiency in retaining observed data points supports the analysis of interconnected variables, such as social support, stigma, and financial stability.
While multiple imputation is widely recommended for general regression contexts, FIML is superior in this SEM-focused design due to its inherent compatibility with latent variable structures. SEM requires methods that respect the covariance matrix and theoretical coherence between measurement and structural models, both of which FIML inherently addresses. FIML reinforces the methodological rigour required to explore the study's multidimensional constructs by ensuring consistency across the CFA validation process and subsequent structural analyses. This approach safeguards against biases introduced by ad hoc imputation and aligns with best practices for psychometric validation in mixed-methods research. Table 4 lists the variables to be measured.
