Front. Hum. Dyn.Frontiers in Human DynamicsFront. Hum. Dyn.2673-2726Frontiers Media S.A.10.3389/fhumd.2026.1761649Original ResearchHidden spatial inequalities in youth unemployment in Gauteng: need for place-based interventionsMoetiThabiso*ConceptualizationData curationFormal analysisInvestigationMethodologyResourcesSoftwareValidationVisualizationWriting – original draftMokheleTholangValidationWriting – review & editingFundisiEmmanuel†ValidationWriting – review & editingGeospatial Computing and Analytics, eResearch Knowledge Unit, Human Sciences Research Council, Pretoria, South AfricaCorrespondence: Thabiso Moeti, TMoeti@hsrc.ac.za
One of South Africa's most persistent socioeconomic challenges is youth unemployment with rates among young people consistently higher than the national average. This study examined the spatial variation of youth unemployment in Gauteng province using data from the Gauteng City-Region Observatory (GCRO) Quality of Life Survey 7 (2023/24). Geographically Weighted Regression (GWR) was used to show how the relationships between youth unemployment and significant socioeconomic factors vary across space, instead of assuming that these relationships are uniform across the province. Youth dissatisfaction with government job creation initiatives, educational attainment, perceived job difficulties in finding work, gender and living in informal dwellings are the main focus of the analysis. Significant spatial heterogeneity was noticed in the GWR results, with Ekurhuleni, Sedibeng and parts of City of Johannesburg showing stronger associations between youth unemployment and these factors. The determinants of youth unemployment are highly place-specific as indicated by the significant variation in local model performance across wards (local R2 ranging from 0.39 to 0.77). These findings demonstrate that highly variable local factors influence youth unemployment in Gauteng and that place-based, spatially focused policy interventions are required rather than uniform, province-wide solutions.
GautengGCRO quality of life surveygeographically weighted regressionlocal bivariate relationshipyouth unemploymentThe author(s) declared that financial support was not received for this work and/or its publication.section-at-acceptancePopulation, Environment and DevelopmentIntroduction
One of the most challenging socioeconomic issues faced by South Africa is youth unemployment and it continues to be a widespread issue both nationally and globally. According to (Statistics South Africa, 2024), in the first quarter of 2024, youth unemployment rate in South Africa stood at 45.5% which is more than the national average of 32.9%. According to national policy and legislation, youth in South Africa is defined as individuals between the ages of 14 and 35. This definition is broader than the United Nations standard (Republic of South Africa, 2009; South African Government, 2021). The 2025 Quarterly Labor Force Survey (QLFS) reported that unemployment among youth aged 15-34 years reached 46.1% for quarter one, reflecting a 9.2% increase since 2015 (Statistics South Africa, 2025a), which indicates that the situation has continued to intensify. The continued increase in unemployment highlights the declining opportunities for majority of youth and shows how urgent it is to address youth exclusion from the labor force. Youth unemployment is comparatively low globally with the International Labor Organization [International Labour Organization (ILO), 2024] reporting a rate of under 6% in 2023 however over 20% of youth worldwide were not in employment, education or training (NEET). This difference highlights the challenges faced in South Africa.
Long term youth unemployment has numerous negative effects. High unemployment rates among youth increases inequality, poverty and social instability. Research shows that unemployed youth have a high risk of experiencing vulnerability to substance abuse, crime, social exclusion, and dissatisfaction (Zahid et al., 2023). Persistent youth unemployment not only represents a loss of present and future productive labor in a nation like South Africa, where young people make up nearly one-third of the population, but it also undermines long-term economic growth, weakens social cohesion and raises the risk of social instability through increased dependency, poverty and social exclusion (World Bank, 2024). The National Development Plan (NDP) 2030 outlines targets which include reducing unemployment to 6% and raising per capita GDP to R110 000 by 2030 (Matyana and Thusi, 2023). However, the continuous rise in unemployment suggests that these goals are challenging to accomplish.
Numerous factors contribute to the rising levels of youth unemployment. Level of education and wide-ranging socioeconomic disadvantages including poor housing conditions, poverty, weak social networks, limited access to transport and spatial marginalization are among the strong predictors of exclusion from the job market (Pereira et al., 2024). Compared to young men, young females face additional barriers to employment due to lower absorption and labor force participation rates (Statistics South Africa, 2024). Education plays an important role, with individuals who obtained tertiary qualifications more likely to be employed compared to those without matric (Statistics South Africa, 2022). Yet even among matriculants, unemployment remains prevalent. These results are in line with previous research that identified labor market segmentation, low skills development and structural inequality to be the major causes of unemployment in South Africa (Kingdon and Knight, 2004; Yu, 2013).
Additionally, labor markets vary significantly across provinces, showing geographic disparities in opportunity. Provinces like KwaZulu-Natal and Limpopo show high rates of discouraged youth but urban areas like the Western Cape tend to have better absorption rates (Statistics South Africa, 2024). These geographical disparities emphasize the importance of examining youth unemployment not only as a national phenomenon but also through a geographic perspective.
Gauteng's urban landscape is characterized by informal settlements, which reflect the region's significant spatial and socioeconomic disparities. Restricted access to essential services and job opportunities often resulting from peripheral urban development, historical spatial marginalization and long-standing under investment in infrastructure together with overcrowding and inadequate housing are common characteristics of these communities, which together create cycles of poverty and unemployment. According to Simon and Ngereja (2025), informal settlements are areas where poor planning and rapid urbanization collide, making people more vulnerable and excluded from the formal economy. Similar issues have been noted in other South African towns where ongoing inequality causes informal settlements to grow (Turok and Borel-Saladin, 2018). Incorporating this spatial dimension offers important insights into how settlement types especially informal dwellings, contributes to the uneven geography of youth unemployment throughout Gauteng.
Spatial analysis offers robust approach to understanding youth unemployment and its determinants. Geographically Weighted Regression (GWR) provides a way to assess how the relationship between youth unemployment and its determinants vary across space. A study by Lewandowska-Gwarda (2018) in Poland shows that GWR accurately reflects local variations in unemployment determinants more successfully than global models. Using a GWR hedonic model, Pacheco, 2021 showed distinct spatial variation in housing prices between formal and informal settlements in California. Recently, Simon and Ngereja (2025) found that the importance of physical, economical and accessibility factors varied significantly across neighborhoods when using GWR to identify spatially varying drivers of informal settlement expansion in Dar es Salaam. Cheruiyot (2023) explored barriers to youth participation in development planning in South Africa using GWR revealing spatial heterogeneity that traditional regression techniques might otherwise overlook. Expanding on these insights, this study applied GWR to explore spatial variation of youth unemployment in Gauteng province. By examining localized relationships between youth unemployment and key explanatory variables including perceptions of job search difficulty, type of dwelling, gender, government dissatisfaction and education, this study contributes to understanding the spatial dynamics of youth unemployment and provides evidence that can guide interventions that are geographically tailored.
The majority of current studies either focus on national or provincial aggregates or do not particularly look at youth unemployment at a fine spatial scale, despite the expanding body of literature using spatial econometric and GWR techniques to unemployment. Few empirical studies have used contemporary microdata to investigate how the drivers of youth unemployment differ geographically at the ward level in South Africa, especially in Gauteng. In order to close this gap, the following questions were posed in this study: (1) How do selected socioeconomic factors in Gauteng relate to youth unemployment globally? (2) How do these interactions vary locally among Gauteng's wards in terms of their strength and direction?
This study makes three key contributions. First, it offers an updated and fine-scale ward level spatial analysis of youth unemployment in South Africa's most economically significant province using the most recent data from Gauteng City-Region Observatory (GCRO) Quality of Life Survey. This level of spatial detail is rarely found in existing studies. Second, the study goes beyond global average effects and shows how the factors influencing youth unemployment differ significantly across space by using GWR in conjunction with local bivariate relationship analysis. This reveals patterns that would be hidden by traditional non-spatial or purely global models. Third, the findings directly produce policy-relevant evidence that supports the design of locally differentiated and spatially targeted interventions rather than one-size-fits-all labor market policies by clearly identifying place-specific relationships between youth unemployment and important socioeconomic factors.
MethodsStudy area
Gauteng, which translates to “Place of Gold” in Sesotho (one of the official languages in South Africa), is the most populous and economically significant province in South Africa despite having the smallest land size at 18,178 km2 or 1.4-1.5% of the country's total area (Parliament of the Republic of South Africa, 2025). The province contributes more than one-third of the country's GDP and acts as the administrative and economic center of South Africa (Statistics South Africa, 2025b). Gauteng's status as the nation's main economic center is reflected in its high population density, rapid urbanization and significant infrastructure needs (Statistics South Africa, 2025b). The province is divided administratively into two district municipalities, Sedibeng (Emfuleni, Lesedi, Midvaal) and West Rand (Merafong City, Mogale City, Rand West City), as well as three metropolitan municipalities, City of Johannesburg, City of Tshwane and Ekurhuleni (Statistics South Africa, 2025b). Gauteng is a crucial setting for analyzing the spatial dynamics of youth unemployment because despite its economic significance, the province faces substantial socioeconomic challenges such as low economic growth (below 2%), fiscal consolidation pressures and high unemployment levels estimated at 31.9% (Gauteng Provincial Treasury, 2025). Figure 1 shows the location of Gauteng Province and its municipalities within South Africa.
Location of Gauteng Province and its municipalities in South Africa.
Map of Gauteng Province in South Africa divided into municipalities, with each labeled: City of Tshwane, Mogale City, Rand West City, Merafong City, City of Johannesburg, Ekurhuleni, Emfuleni, Midvaal, and Lesedi. Inset map shows Gauteng’s location within South Africa. A north arrow, scale in kilometers, and a legend indicating Gauteng Municipalities and Gauteng Province are present.Data
Data from the Gauteng City-Region Observatory (GCRO) Quality of Life Survey 7 (2023/2024) was used in this study. The goal of the Quality of Life (QoL) survey is to regularly gather data about the socioeconomic conditions, values, psycho-social attitudes, satisfaction with service delivery and other aspects of residents in Gauteng (Neethling, 2024). Youth between the ages 18 and 34 were the focus of this study, and they were identified using Stata's recoded age variable. The ward level was utilized as the unit of analysis and the shapefile with Gauteng's ward boundaries was used for spatial analysis after joining with the aggregated youth dataset exported from Stata version 15 (Stata Corp, 2017).
In Gauteng, there are 529 wards in total. A total of 4,880 youth respondents at ward level make up the analytical sample. In South African context, a ward is a smaller geographic area within a municipality that is delimited by Municipal Demarcation Board (MDB) for electoral purposes (Municipal Demarcation Board (MDB), 2021). Five hundred and twenty five wards were retained for analysis following data aggregation and cleaning as one ward had insufficient or missing observations and therefore was excluded. Descriptive statistics for every variable utilized in the analysis at ward level are shown in Table 1. Depending on the specific characteristic being measured, the average number of youth observations per variable across wards ranges approximately between three and eight. Instead of representing population standardized rates, the ward level variables used in this study were constructed by aggregating individual survey responses and therefore represent counts of surveyed youth respondents per ward.
Descriptive statistics of ward-level variables.
Variable
Obs
Mean
Std. dev
Min
Max
Unemployed
528
3.44
2.34
0
11
Female
528
4.67
2.42
0
13
Matric
528
3.98
2.19
0
11
Informal dwelling
528
1.53
2.33
0
11
Job search harder
528
7.58
3.29
1
24
Government dissatisfaction
528
7.25
3.24
0
21
All variables are ward-level counts constructed by aggregating individual survey responses. The minimum and maximum values indicate the smallest and largest number of surveyed youth respondents per ward with the specified characteristic.
Measures
The primary outcome variable, youth unemployment, was based on the question “Are you unemployed and looking for work?” with responses coded as 1 = yes and 0 = no. For spatial modeling, this variable was chosen as the primary outcome. Sex (female), education (matric), dwelling type (informal dwelling type), job search (perception that it has become harder to find a job) and government initiatives (dissatisfaction with government initiatives to create jobs) were among the explanatory variables. The matric variable excludes respondents with tertiary qualifications and only includes those whose highest level of education completed is Grade 12. Sex is recorded as a binary variable (male/female) in the GCRO survey; in this study the variable female represents the count of female youth respondents per ward, aggregated from individual level responses. Similarly, the GCRO survey records dwelling type in multiple categories (formal dwelling, informal dwelling and other); for this study, a binary indicator for informal dwelling was created at the individual level and then aggregated to ward level, representing the number of young people living in informal dwellings per ward. The government dissatisfaction variable is based on respondents' subjective assessment of government efforts to create jobs, as recorded in the survey questionnaire. Respondents were not asked about a specific policy, but rather about their general perception of government performance in job creation. All explanatory variables were aggregated to ward level as counts of youth respondents exhibiting each characteristic. These variables, in particular the impact of gender, perceptions of job accessibility, housing conditions, satisfaction with government efforts on employment outcomes and education, were chosen due to their theoretical and empirical significance to youth unemployment in South Africa.
Although the GCRO QoL survey contains variables related to household income and transport accessibility, these were not included in the final models for two reasons. First, there are endogeneity and circularity issues when modeling unemployment outcomes because income and employment status are conceptually closely intertwined. Second, transport accessibility measurements showed substantial multicollinearity with dwelling type indicating a strong overlap between transport conditions and settlement characteristics at ward level. The final specification therefore only included variables that were statistically robust and theoretically meaningful in order to keep the regression model used in this study simple and interpretable.
Spatial analysisOrdinary least squares
The global (non-spatial) relationship between youth unemployment and selected explanatory variables was investigated using Ordinary Least Squares (OLS) regression. OLS is a most well-known regression approach and serves as the basis for all spatial regression analysis (ArcGIS Pro Documentation, 2025a). Before proceeding with local spatial modeling, multicollinearity was evaluated and significant predictors were identified using OLS model.
GWR
GWR was used to take into account the spatial variation in youth unemployment across Gauteng. GWR reveals local differences in the strength and direction of relationships between youth unemployment and explanatory variables, compared to global OLS model which assumes spatial stationarity, that is that the relationships are constant across space (Charlton et al., 1998). In analogy to time-series stationarity, where relationships are assumed constant over time, spatial stationarity assumes relationships are constant throughout geographical space. The adaptive kernel bandwidth method with automated bandwidth optimization was used to implement the model in ArcGIS Pro 3.4.2. A more detailed understanding of spatial variations in youth unemployment is provided by this localized modeling technique. Additionally, Fundisi et al. (2023) demonstrated that a GWR local model performed better than a global model confirming GWR's applicability for socioeconomic phenomena that change spatially.
Local bivariate relationship
To explore local associations between youth unemployment and explanatory variables, local bivariate relationship analysis was conducted. By identifying whether the values of one variable have an impact or are dependent upon the values of another, as well as if these relationships vary over geographic space, the local bivariate relationship tool enables one to quantify the relationship between two variables on the same map (ArcGIS Pro Documentation, 2025b). Based on how effectively the explanatory variable parameter predicts the dependent variable parameter, the final results group each input feature into a one relationship category (Yaakub et al., 2022). All spatial analyses were implemented using ArcGIS Pro 3.4.2 (ESRI, 2024) and Stata 15.0 (Stata Corp, 2017).
ResultsOrdinary least squares
The distribution of standardized residuals from the OLS model across Gauteng is shown in Figure 2. The joint Wald statistic confirmed that the OLS model was highly significant (p < 0.001) and it explained almost 45% of the variation in youth unemployment (R2 = 0.45; adjusted R2 = 0.44). Dwelling type (informal dwelling), sex (female), job search perception and dissatisfaction with government initiatives all emerged as statistically significant predictors of youth unemployment, while matric completion showed a positive but non-significant relationship. For all explanatory variables, the variance inflation factors (VIF) were below 5 indicating no evidence of multicollinearity. Moving beyond the global model is justified since the Koenker BP statistic was significant (p < 0.001) indicating non-stationarity in the relationships across space. The Jarque-Bera statistic, on the other hand, was not significant (p = 0.384) indicating that the residuals were approximately normally distributed. All of these findings show that although the OLS model finds significant predictors of youth unemployment, the spatial variability highlighted by the diagnostics necessitates the use of GWR for more thorough investigation.
Distribution of standardized residuals from the Ordinary Least Squares (OLS) model across Gauteng districts.
Choropleth map of Gauteng, South Africa, showing standardized residuals by district and municipality, with values ranging from less than minus two point five in dark blue to greater than two point five in dark red, overlaid with district names including West Rand, City of Johannesburg, City of Tshwane, Ekurhuleni, and Sedibeng, and a scale bar indicating distances up to forty kilometers.GWR
With residuals generally close to zero, the deviance residuals map (Figure 3) demonstrates that the model performed well across most wards. However, significant spatial variation was noted. The outside wards of the City of Tshwane, the western parts of West Rand and parts of Sedibeng had larger positive residuals showing under prediction of unemployment. On the other hand, negative residuals which indicate over prediction were mainly located in central Gauteng, specifically in Ekurhuleni and City of Johannesburg. Compared to the OLS model, the GWR model provided an improved fit with local R2 values ranging from 0.39 and 0.78 showing substantial spatial variation in model performance across Gauteng. The model explained more variation in urban and peri-urban areas compared to peripheral ones according to the local R2 map (Figure 4). Ekurhuleni, Sedibeng and western parts of West Rand had higher local R2 values indicating that the model more accurately reflected the dynamics of unemployment in these districts. Several areas of Johannesburg and the northern wards of Tshwane, on the other hand, showed lower local R2 values suggesting that unmeasured local factors might have an impact on youth unemployment in those areas.
Spatial distribution of deviance residuals from the GWR model of youth unemployment in Gauteng.
Choropleth map of districts in Gauteng, South Africa, illustrating deviance residuals using a gradient from light to dark green. Districts such as City of Johannesburg, City of Tshwane, West Rand, Ekurhuleni, and Sedibeng are labeled. A legend in the top left corner indicates five residual value ranges, from negative three point seven zero to five point four four. A north arrow and distance scale in kilometers are included.
Local R2 values from the GWR model showing the spatial variation in model fit across Gauteng wards.
Choropleth map displaying local R squared values by district within a metropolitan region, using shaded gradients from light to dark green to indicate increasing R squared intervals; districts are outlined in black and labeled, with a north arrow and distance scale for reference.
The coefficient maps (Figures 5A–E) revealed clear spatial heterogeneity in the determinants of youth unemployment across the province. Being female was more positively correlated with unemployment in the West Rand, Ekurhuleni and Sedibeng where the effect of gender (female) was strongest. Across most of Gauteng, especially in City of Tshwane and City of Johannesburg, government dissatisfaction showed a strong positive correlation with unemployment showing areas where increased unemployment is correlated with dissatisfaction with government initiatives. Particularly in wards that have a dense settlement patterns, the impact of informal dwelling was most noticeable in Ekurhuleni and City of Johannesburg while the northern regions of the province showed less evident or negative effects. Strong positive coefficients for job search difficulty were identified in Sedibeng, Ekurhuleni and City of Johannesburg indicating that wards with greater unemployment rates were also those where job searchers had a difficult time finding work. Lastly, Sedibeng, City of Tshwane and City of Johannesburg showed strong effects of matric level education with greater proportions of matriculants linked to higher unemployment rates. This could indicate disparity between educational attainment and labor market absorption. Collectively, these findings demonstrate the geographical variation of the factors influencing youth unemployment and the significance of context specific and locally focused policy measures in addressing inequities throughout Gauteng.
Spatial variation in the local coefficients from the GWR model of youth unemployment in Gauteng. (A) Female (B) Government dissatisfaction (C) Informal dwelling (D) Job search harder (E) Matric.
Five-panel figure presents choropleth maps of a region, each with districts shaded from light to dark green by coefficient values for different variables: A, Female; B, Government Dissatisfaction; C, Informal Dwelling; D, Job search harder; E, Matric. Each panel includes a legend showing coefficient ranges, district boundaries, major city names, and a scale bar indicating distances in kilometers.Local bivariate relationship
The local bivariate relationship analysis showed spatial variation in the associations between youth unemployment and its determinants across Gauteng (Figures 6A–E). Only a few localized clusters, primarily in the southern Ekurhuleni, West Rand and Sedibeng, showed a strong correlation between the proportions of females and youth unemployment. With strong positive linear associations across Sedibeng, West Rand, City of Johannesburg and Ekurhuleni, dissatisfaction with government initiatives showed the most widespread significant relationships. Concave and convex relationships were observed in a few wards in the West Rand, suggesting localized non-linear dynamics. Contrary, matric and informal dwelling variables showed no statistically significant relationships throughout the province, suggesting that these variables did not vary spatially in their association with youth unemployment. Perceptions that job searching has become harder showed substantial positive linear relationships in various districts especially in the West Rand, Sedibeng and parts of City of Johannesburg, with a small cluster on the West Rand showing a convex relationship. Overall, these results show that there are differences in spatial relationships between youth unemployment and its determinants, with job search difficulty and government dissatisfaction emerging as the most spatially variable drivers of youth unemployment.
Local bivariate relationships between youth unemployment and independent variables (A) female (B) government dissatisfaction (C) informal dwelling (D) job search harder (E) matric.
Five-panel map series labeled A to E depicts the Johannesburg metropolitan area, color-coded by type of statistical relationship per region. Categories include positive linear, negative linear, concave, convex, undefined complex, and not significant. Panels A, B, and D show varying areas primarily shaded pink and a few orange or blue, representing significant relationships, while C and E are mostly gray, indicating not significant areas. Each map outlines key administrative regions such as City of Johannesburg, West Rand, Ekurhuleni, and Sedibeng, with a scale bar and north arrow for orientation.Discussion
By showing that the factors influencing youth unemployment in Gauteng are not spatially uniform and that significant local variations are masked by global models, this study adds to the expanding body of research on spatial unemployment. The analysis offers a detailed insights of how socioeconomic and perceptual factors shape unemployment differently across space by integrating OLS, GWR and local bivariate relationship analysis at ward level using data from GCRO QoL Survey. Place-specific drivers that would not be apparent with province wide or global models alone can be identified with multi method spatial approach.
Furthermore, the findings from this study highlight the spatial variation of youth unemployment in Gauteng province. Although the global OLS model showed that the explanatory variables as a collective explained the variation in unemployment, the GWR results showed that their influence varied across space. Particularly, the coefficients for gender indicated that being female was strongly linked with high levels of youth unemployment in several wards, supporting the data from national statistics that young women are significantly disadvantaged in the job market (Statistics South Africa, 2024). Similarly, there were geographic variations in the relationship between unemployment and education, namely matric level, with some areas showing a strong negative association than others. This reinforces the notion that having more education enhances one's chance of finding employment but also highlights that the benefit of matriculation is not uniform throughout Gauteng.
Further illustrating how contextual disadvantages influence employment outcomes are the spatial patterns linked with housing conditions. Wards with higher numbers of young people living in informal dwellings tended to have high unemployment rates, although this association was not statistically significant across all areas. This aligns with existing literature that links poor housing conditions and limited access to basic services to reduced opportunities in the labor market. Similarly, dissatisfaction with government initiatives indicated strong localized associations with unemployment, especially in wards where young people expressed dissatisfaction with government efforts. These results resonate with research that highlights how crucial government responsiveness is in addressing unemployment and fostering trust with youth (Cheruiyot, 2023).
The perception that job search is hard emerged as another important factor, particularly in wards outside the central business district. These findings show that barriers to entering labor markets are not limited to individual characteristics but are closely related to geographic factors such as accessibility of transportation, proximity to economic centers and local job density. This supports international research showing that geographical disparities increase labor market exclusion [International Labour Organization (ILO), 2024]. The findings suggest that there is no one size fits all approach for measures meant to reduce youth unemployment in Gauteng. Although education and gender are broad factors, their effects vary by ward in terms of impact and severity. Therefore, localized interventions are needed including focused training and placement programs for women, assistance for youth living in informal settlements and measures that increase the effectiveness of job search. Government programmes should be designed to help youth regain confidence in places where dissatisfaction with public initiatives is closely linked with unemployment.
Studies conducted in South Africa and other countries have shown that socioeconomic systems seldom function consistently across space, which is consistent with the observed spatial non-stationarity in Gauteng. The strength and direction of relationships between socioeconomic determinants and development outcomes are significantly influenced by the local context, as shown by similar uses of GWR in urban and regional research. The current findings support the theory that locally specific combinations of structural and perceptual elements, rather than a single uniform process, are responsible for youth unemployment.
From a policy standpoint, these results imply that consistent province wide initiatives are unlikely to be equally successful everywhere. Instead, interventions that are context-specific and spatially targeted or place-based are needed, especially in regions where youth unemployment is mostly strongly correlated with job search barriers and dissatisfaction with government programs.
This study provides valuable information into the spatial variation of youth unemployment in Gauteng, however it is crucial to acknowledge certain limitations. The analysis was limited to Gauteng province therefore the findings reflect patterns and relationships specific to the province and might not accurately mirror trends in other South African provinces. Furthermore, due to time and scope limits, only a small number of explanatory variables were included while other pertinent factors such as access to transportation and income were excluded. Additionally, the study used cross sectional data in order to avoid drawing conclusions about causality regarding the dynamics of youth unemployment over time.
Rather than using population standardized ward level rates, this study use aggregate survey counts. Therefore, rather than reflecting the actual population prevalence of youth unemployment, the spatial patterns found reflect the distribution of surveyed youth respondents. This limitation arises from the survey based nature of the GCRO QoL data and the lack of compatible ward level youth population denominators for the same spatial units and time period. The results should therefore be interpreted as indicating relative spatial variation within the sample, rather than exact unemployment rates in the underlying population.
Another drawback is the potential for reverse causality between youth unemployment and government dissatisfaction. It is equally likely that unemployment itself contributes to reduced life satisfaction and high dissatisfaction with government performance, even though the modeling framework treats attitudinal variables as correlates of unemployment. The data's cross-sectional nature makes it impossible to determine the causative direction. Therefore, the predicted correlations should be viewed as associational rather than causal and more research employing longitudinal data would be necessary to separate these dynamics.
Overall, the study shows that youth unemployment in Gauteng is formed not only by individual level characteristics but also by local spatial dynamics. The study's use of GWR revealed trends that would have remained undetected in global models, highlighting the importance of spatially focused policy in addressing South Africa's ongoing youth unemployment issue.
Conclusion
This study used GWR to examine the spatial heterogeneity of youth unemployment in Gauteng province. The results of the analysis showed that the factors influencing youth unemployment vary across space suggesting that a single province wide policy approach might not be sufficient to address the issue. The GWR model shed more light on how these relationships vary geographically whereas the OLS results identified a number of significant factors such as perception of job search, female, informal dwelling and dissatisfaction with government initiatives. Specifically, in districts such as Ekurhuleni, Sedibeng and parts of the City of Johannesburg, youth unemployment was strongly associated with higher proportions of informal dwelling, females, perceptions that job search has become hard and dissatisfaction with government initiatives.
The importance of spatially targeted interventions was further highlighted by the local bivariate analysis which showed that while certain relationships were stable throughout the province, other were highly localized. According to these results, the problem of youth unemployment in Gauteng calls for regional specific approaches that consider socioeconomic and geographic factors into account. The disparities shown among wards may be lessened with the support of policies that expand public employment programs, improve education to work transitions and increase job accessibility. Although this study contributes to a spatial understanding of youth unemployment, more variables such as transport accessibility, digital inclusion and local industry composition could be included in future studies. The results highlight the usefulness of GWR as analytical tool for guiding place-based policy responses to South Africa's challenges with youth unemployment.
Data availability statement
The data analyzed in this study is subject to the following licenses/restrictions: The GCRO Quality of Life Survey 2020/2021 dataset is not publicly available. Access to the data is restricted and requires a formal application to the Gauteng City-Region Observatory (GCRO). The dataset can be requested directly from GCRO, subject to approval and data-use agreements. Requests to access these datasets should be directed to https://www.datafirst.uct.ac.za/dataportal/index.php/collections/GCRO.
Ethics statement
The studies involving humans were approved by University of Witwatersrand Human Research Ethics Committee (Non-Medical), Johannesburg, South Africa under protocol number H19/11/19. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants' legal guardians/next of kin in accordance with the national legislation and institutional requirements.
Gauteng City-Region Observatory (GCRO) is acknowledged for data used in this study.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
ReferencesArcGIS Pro Documentation (2025a). Available online at: https://pro.arcgis.com/en/pro-app/latest/tool-reference/spatial-statistics/how-ols-regression-works.htm (Accessed September 20, 2025).ArcGIS Pro Documentation (2025b). Available online at: https://pro.arcgis.com/en/pro-app/3.4/tool-reference/spatial-statistics/learnmore-localbivariaterelationships.htm (Accessed September 20, 2025).CharltonM. E.BrunsdonC. F.FotheringhamA. S. (1998). Geographically weighted regression-modelling spatial non-stationarity. J. Royal Stat. Soc. D Statist. 47, 431–443. doi: 10.1111/1467-9884.00145CheruiyotK. (2023). A geographically weighted regression analysis of barriers to youth's participation in local development planning in Gauteng Province, South Africa. S. Afr. J. Geom.12, 112–128. doi: 10.4314/sajg.v12i.2.1ESRI (2024). ArcGIS Pro (Version 3.4.2) (Software). Redlands, CA: Environmental Systems Research Institute, Inc.FundisiE.DlaminiS.MokheleT.Weir-SmithG.MotolwanaE. (2023). Exploring determinants of HIV/AIDS self-testing uptake in South Africa using generalised linear poisson and geographically weighted poisson regression. Healthcare11:881. doi: 10.3390/healthcare1106088136981538Gauteng Provincial Treasury (2025). Budget pressures for Gauteng: analysis of the 2025/26 MTBPS. GCRO. Available online at: https://www.gcro.ac.za/data-gallery/interactive-data-visualisations/detail/budget-pressures-for-gauteng-analysis-of-202526-mtbps (Accessed January 7, 2026)International Labour Organization (ILO) (2024). Global Employment Trends for Youth 2024: Decent work, brighter futures (20th anniversary edition). Geneva: ILO. XVI.KingdonG. G.KnightJ. (2004). Unemployment in South Africa: the nature of the beast. World Dev. 32, 391–408. doi: 10.1016/j.worlddev.2003.10.005Lewandowska-GwardaK. (2018). Geographically weighted regression in the analysis of unemployment in Poland. ISPRS Int. J. Geo-Inform.7:17. doi: 10.3390/ijgi7010017MatyanaM.ThusiX. (2023). Unemployment and poverty in South Africa: assessing the National Development Plan 2030 predictions. Int. J. Dev. Sustain.12, 212–226.Municipal Demarcation Board (MDB) (2021). Ward Delimitation for Local Elections. Available online at: https://www.demarcation.org.za/wp-content/uploads/2021/06/MDB-WARD-DELIMITATION-BOOKLET.pdf. (Accessed January 27, 2026).NeethlingA. (2024). GCRO Quality of Life Survey 7 (2023/24): Guide To Weighted Analysis. Johannesburg: Gauteng City-Region Observatory (GCRO)PachecoH. V. (2021). Spatial heterogeneity of housing prices in formal and informal settlements: A GWR hedonic model for segmented markets in Cali. Documentos de trabajo – Alianza EFI: Bogotá, Colombia 19293.Parliament of the Republic of South Africa (2025). Gauteng Provincial Profile: 2025. Available online at: https://www.parliament.gov.za/storage/app/media/Pages/2025/07-11-2025_NCOP_Provincial_week/background/Gauteng_Profile.pdf (Accessed January 7, 2026).PereiraD. J.JoosubN.BassonP. (2024). The psychological factors influencing youth moving from unemployment to employment in South Africa. Afr. J. Career Dev.6:118. doi: 10.4102/ajcd.v6i1.118Republic of South Africa (2009). National youth development agency act no. 54 of 2008.Government Gazette No. 31780. Cape Town: Republic of South Africa.SimonO.NgerejaZ. (2025). Spatial determinants of informal settlement expansion in Dar es Salaam Metropolitan City, Tanzania: a geographically weighted regression approach. GeoJournal90:100. doi: 10.1007/s10708-025-11352-2South African Government (2021). National Youth Policy 2020–2030. Available online at: https://www.gov.za/sites/default/files/gcis_document/202103/nationalyouthpolicy.pdf (Accessed January 26, 2026).Stata Corp (2017). Stata Statistical Software: Release 15. TX: StataCorp LLC: College StationStatistics South Africa (2022). Labour Market Dynamics in South Africa, 2022. Pretoria: Statistics South Africa.Statistics South Africa (2024). Available online at: https://www.statssa.gov.za/?p=17266 (Accessed September 20, 2025).Statistics South Africa (2025a). Available online at: https://www.statssa.gov.za/?p=18398 (Accessed September 20, 2025).Statistics South Africa (2025b). Provincial Profile: Gauteng. Pretoria: Stats SA. Available online at: https://www.statssa.gov.za/publications/Report-03-01-76/Report-03-01-762022.pdf (Accessed January 7, 2026).TurokI.Borel-SaladinJ. (2018). The theory and reality of urban slums: Pathways-out-of-poverty or cul-de-sacs?. Urban Stud.55, 767–789. doi: 10.1177/0042098016671109World Bank (2024). South Africa.https://documents1.worldbank.org/curated/en/099613104052410927/pdf/IDU1ffc84f731c8e91470018c4311f71d03092e1.pdf (Accessed January 26, 2026).YaakubN. F.MasronT.MarzukiA.SodaR. (2022). GIS-based spatial correlation analysis: sustainable development and two generations of demographic changes. Sustainability14:1490. doi: 10.3390/su14031490YuD. (2013). Youth unemployment in South Africa revisited. Dev. S. Afr.30, 545–563. doi: 10.1080/0376835X.2013.830964ZahidF.DurraniK.ShahS.AhmedS.MuhammadB. (2023). Youth unemployment and social stability: investigating the linkages and possible solutions in the context of Pakistan. Bull. Bus. Econ.12, 477–484. doi: 10.61506/01.00154
Edited by: Wasseem Mina, Economic Research Forum, Egypt
Reviewed by: Naman Mishra, Bennett University, India