Who Gains from Higher Education? District-Level Development and Labor Market Inactivity Among Indonesian Young Adults

1. Introduction

1.1 The Paradox of Higher Education: Inactivity Among Indonesia’s Young Adults

Over the past two decades, Indonesia has substantially expanded access to tertiary education. By 2024, the country’s gross enrollment rate (GER) in higher education had reached 39.4%, reflecting a steady upward trend. This figure, however, still lags behind neighboring countries such as Singapore (91.1%), Thailand (49.3%), and Malaysia (43.0%) (Brodjonegoro, 2024). As of 2022, Indonesia was home to 4,481 higher education institutions—including universities, Islamic colleges, service academies, and the Open University—many of which are privately managed (Moeliodihardjo, 2024). The rapid growth in institutional capacity has been accompanied by a rising social demand for higher education, mirroring global patterns in which tertiary education is seen as a pathway to human capital accumulation, upward mobility, and improved labor market outcomes (Lim et al., 2023; Psacharopoulos & Patrinos, 2018).

Despite these gains, a growing paradox has emerged. While more young Indonesians attain tertiary education than ever, a sizable share remains disengaged from the labor market—classified as Not in Employment, Education, or Training (NEET). This disconnect challenges conventional human capital assumptions that link higher education directly to improved employment outcomes (Becker, 1993; Brown et al., 2011). In low- and middle-income countries especially, the promise of education as a shield against exclusion is increasingly contested (Ilie & Rose, 2018; UNESCO, 2020).

In the Indonesian policy discourse, the focus on youth inactivity has often centered on those aged 15–24 (Jessica & Arcana, 2024; Naraswati & Jatmiko, 2022; Pattinasarany, 2024), consistent with Sustainable Development Goal indicators. However, this age group often includes students who have not yet completed their education, making it a less precise lens for studying labor market entry. By contrast, the 25–34 age group better captures individuals who are more likely to have exited education and face employment decisions directly (Aina et al., 2021; Dinku, 2024).

Drawing on nationally representative SUSENAS data from 2019 to 2024, Figure 1 illustrates labor market inactivity trends among Indonesians aged 25–34, disaggregated by educational attainment. Inactivity rates have remained persistently high—ranging from 26.8% to 27.8%—suggesting deep structural frictions. Tertiary-educated individuals consistently show lower inactivity (17.1% to 21.7%), but the protection is partial and uneven. Notably, inactivity among this group peaked at nearly 22% in 2022 following COVID-19 disruptions before declining to 19.5% in 2024. This volatility indicates that higher education alone is not a reliable guarantee of labor market integration.

Figure 1: Labor Market Inactivity Among Young Adults Aged 25–34, by Educational Attainment, 2019–2024

Source: Author's calculation

Similar trends have been observed in other middle-income countries, where the expansion of tertiary education has outpaced the capacity of labor markets to absorb graduates, resulting in widespread mismatches and underemployment (Battu & Bender, 2020; Neycheva, 2021). These patterns underscore the need to move beyond aggregate enrollment figures and examine the structural, institutional, and regional constraints that limit the ability of even well-educated young adults to transition successfully into meaningful employment

1.2 Purpose of the Study and Research Questions

This study investigates the determinants of labor market inactivity among young adults aged 25–34 in Indonesia, specifically focusing on how regional socioeconomic contexts shape the effectiveness of higher education in preventing NEET (Not in Employment, Education, or Training) outcomes. Using nationally representative SUSENAS data from 2019 to 2024, the analysis examines how individual educational attainment and household economic status interact with district-level structural factors—namely the Human Development Index (HDI) and poverty rates—to influence labor market participation.

Grounded in human capital theory and contextualized by economic development literature, the study explores whether the protective effect of tertiary education is uniform across regions or contingent upon broader development conditions. By incorporating cross-level interactions between individual- and district-level variables, the research design aims to capture the multidimensional dynamics that shape labor market outcomes in a middle-income country context. A multilevel mixed-effects logistic regression model is employed to account for clustering at the district level and to model heterogeneity in regional economic environments.

The research addresses the following research questions (RQs):

• RQ1: To what extent is tertiary education associated with a reduced likelihood of labor market inactivity among young adults in Indonesia?

• RQ2: How does household economic status, as measured by log per capita expenditure (LogPCE), relate to the probability of being NEET?

• RQ3: Do regional contextual factors—specifically HDI and poverty rates—independently affect the likelihood of labor market inactivity?

• RQ4: Do HDI and poverty rates moderate the relationship between tertiary education and NEET status?

• RQ5: Does the effect of household economic status on labor market inactivity vary across districts with differing levels of HDI and poverty?

1.3 Importance and Contribution of the Study

This study contributes to the literature on labor market integration by focusing on an underexamined demographic group: young adults aged 25–34 in Indonesia. Unlike the more commonly studied 15–24 age group, which includes many individuals still in formal education, this older cohort has largely completed their schooling, making them a more appropriate population for analyzing post-education labor market transitions and the realized economic value of tertiary education. In doing so, the study addresses a critical gap in the evidence base on education-to-employment outcomes in emerging economies.

The study offers three key contributions. First, it introduces a multilevel framework grounded in human capital theory, linking individual-level factors—tertiary education and household economic status—with district-level structural conditions, namely the HDI and poverty rates. This framework allows for a more comprehensive understanding of how macro-contexts shape the micro-level effects of education. Second, by incorporating cross-level interaction terms, the analysis tests whether the benefits of higher education and household resources vary by the level of regional development, thereby highlighting spatial conditionality in educational returns. Third, the study uses nationally representative, multi-year data (2019–2024) and applies multilevel mixed-effects logistic regression to capture the hierarchical structure of the data and regional clustering, enhancing the robustness and policy relevance of the findings.

The findings have practical implications for regional development policy, education system accountability, and labor market planning. By identifying where and for whom tertiary education fails to translate into employment, the study provides actionable insights for targeting interventions—such as skills matching, employment services, or investments in lagging districts. Conceptually, it contributes to ongoing debates on uneven returns to education, structural barriers to inclusion, and contextual inequality in labor market outcomes in middle-income countries.

2. Materials and Methods

2.1 Conceptual Framework

This study draws on human capital theory, which posits that educational investments increase an individual’s productivity and employability, leading to better labor market outcomes (Becker, 1993). From this perspective, tertiary education should protect against labor market exclusion, as it endows individuals with skills in demand in a modern economy. Becker conceptualizes schooling as a capital investment, yielding returns through enhanced earnings and labor force participation. In the context of Indonesian young adults, this framework suggests that individuals with tertiary education should be less likely to be inactive in the labor market as their advanced skills and credentials increase their competitiveness.

However, the extent to which educational attainment leads to employment is uneven and highly dependent on structural and contextual factors. Regional disparities in development often mediate this relationship, either enabling or obstructing the translation of education into labor market opportunities. In many low- and middle-income countries, including Indonesia, challenges such as underdeveloped labor markets, high levels of informality, and spatial inequalities significantly limit these returns (Filmer & Fox, 2014). In South Africa, for example, young adults with higher education continue to face exclusion from the labor market due to job shortages, skills mismatches, and institutional barriers (Yu, 2013). Similarly, Yeung and Yang (2020) emphasize that globally, even tertiary-educated youth often confront precarious or informal employment, with labor market outcomes shaped by broader inequalities linked to socioeconomic status, gender, and ethnicity.

This study incorporates two key district-level indicators to capture these contextual dynamics: the Human Development Index (HDI) and poverty rates. HDI reflects overall regional capabilities, combining health, education, and income indicators, while poverty rates measure the prevalence of deprivation and economic vulnerability. Both serve as proxies for the local opportunity structure in which young adults attempt to convert educational attainment into labor market participation. Empirical evidence from Western African countries shows that individuals in regions with low development levels often face greater barriers to employment, regardless of their education level (Dimova et al., 2010). These findings support the view that the effect of education is not absolute but contingent upon the structural and institutional characteristics of the region.

This study adopts a multilevel analytical approach to examine how individual-level attributes—tertiary education and household economic status—interact with regional development conditions. It explores whether the benefits of education and household resources vary systematically across districts with differing levels of HDI and poverty. Specifically, the analysis considers whether the protective effect of tertiary education against labor market inactivity is stronger in districts with higher HDI and lower poverty, where institutional support and job opportunities may be more robust. It also assesses whether the influence of household economic status, proxied by log per capita expenditure (LogPCE), is shaped or constrained by broader regional development and deprivation contexts.

By integrating individual and contextual determinants, the conceptual framework offers a more comprehensive understanding of labor market inactivity among young adults in Indonesia. It contributes to broader discussions on the spatial conditionality of educational returns in developing and emerging economies.

2.2 Methodology

This study employs a multilevel mixed-effects logistic regression model to examine the determinants of labor market inactivity among young adults in Indonesia, guided by five research questions. The model incorporates explanatory variables at two levels: individual and household characteristics (Level 1) and district-level contextual factors (Level 2), with tertiary education attainment as the key independent variable.

The multilevel approach is appropriate given the hierarchical structure of the data, with individuals nested within districts, and the binary nature of the dependent variable (inactivity status). Random intercepts at the district level account for unobserved contextual heterogeneity and adjust for clustering effects. This framework allows for the simultaneous estimation of fixed effects while capturing variation in baseline inactivity risks. By integrating cross-level interactions, the model tests how education and household economic status are shaped by broader regional development conditions, such as HDI and poverty (Rabe-Hesketh & Skrondal, 2022).

Formally, the Baseline Model is specified within a latent response framework, serving to establish the baseline relationships between individual- and district-level predictors and labor market inactivity. Let  denote the unobserved propensity for individual i in district j to be NEET. This latent variable is modelled as:

where  represents individual- and household-level covariates;  denotes district-level contextual variables;  is a covariate vector associated with the random effects (in this case, a scalar of 1);  is the district-specific random intercept; and  is the individual-level error term, assumed to follow a logistic distribution with mean 0 and variance , independent of .

The observed binary outcome  is linked to the latent variable  through the measurement equation:

Building on the Baseline Model, the Interaction Model introduces interaction terms between tertiary education attainment and two continuous district-level contextual variables—HDI and poverty rates. This specification tests whether the effect of tertiary education on labor market inactivity varies with regional development conditions. To complement this approach, the Categorical Interaction Model replaces the continuous indicators with binary categorical variables based on district-level rankings. Districts are classified into two groups for each indicator: low (low-HDI or low-poverty) and high (high-HDI or high-poverty), based on the bottom and top 50th percentiles, respectively. This grouping strategy enhances interpretability and facilitates direct policy-relevant comparisons across contrasting development contexts.

Unlike the continuous Interaction Model, which assumes linear moderation, the categorical model captures potential threshold effects and non-linearities in how regional development shapes the returns to tertiary education. To avoid multicollinearity, the continuous HDI and poverty variables are excluded, as the binary groupings are constructed from the same underlying distributions. Interaction terms between the binary HDI and poverty variables and tertiary education are included to assess whether the protective effect of higher education differs significantly across regional development tiers.

2.3 Data

This study uses data from the National Socioeconomic Survey (Survei Sosial Ekonomi Nasional, SUSENAS), an annual household survey conducted by Indonesia's Central Statistics Agency (Badan Pusat Statistik—BPS). SUSENAS is designed to be nationally and subnationally representative, covering all 514 districts (kabupaten/regencies and kota/municipalities). As one of Indonesia's most comprehensive sources of socioeconomic data, it enables rigorous analysis of population-level trends. The study draws on pooled data from six survey waves, spanning 2019 to 2024, and focuses on a subset of 977 453 individuals aged 25 to 34. Concentrating on this age group—typically beyond formal schooling—allows for a more accurate assessment of post-education labor market outcomes and structural inactivity.

SUSENAS adopts a stratified two-stage sampling design to ensure representativeness at multiple levels. In the first stage, census blocks are selected using probability proportional to size based on household counts. In the second stage, a fixed number of households are randomly selected within each chosen block. Each year, SUSENAS surveys approximately 320 000 households, yielding data on about 1.2 million individuals, enabling robust estimates across provinces and districts.

For this study, NEET status is defined as young adults aged 25–34 who were not engaged in employment, education, or training during the reference week prior to the survey. This category includes individuals with unpaid domestic responsibilities (e.g., caregiving or household tasks), those involved in personal or social activities (e.g., religious participation or informal learning), or those reporting no specific activity. The definition excludes individuals who were temporarily not working but maintained formal employment or self-employment status. By adopting this refined classification, the study aims to capture the structural labor market inactivity phenomenon more accurately, distinguishing it from short-term absences or voluntary non-participation.

The descriptive statistics of the variables used in this study, including means and standard deviations disaggregated by year, are presented in Table 1.

Table 1: Mean and Standard Deviation of Data

Notes: NEET status, Female, Married, Age, and Tertiary Education are calculated using individual-level weights. Urban residence, household members aged 0–4 and 60+, and Log(PCE) are calculated using household-level weights. HDI and Poverty rates are reported as unweighted district-level values.

Source: Author's calculation

3. Results and Discussions

The results of the multilevel mixed-effects logistic regression are presented across three model specifications, with all estimates reported as odds ratios for ease of interpretation. As outlined in the Methodology section, these include the Baseline Model (Column [1]), the Interaction Model (Column [2]), and the Categorical Interaction Model (Column [3]). The complete results are summarized in Table 2.

Table 2: Multilevel Logistic Regression of Labor Market Inactivity Among Young Adults

Notes: Standard errors are in parentheses. *** statistically significant at the 1% level, ** 5%, * 10%

Source: Author's calculation

Prior to estimating the full models, a null model was specified to assess the degree of clustering in NEET outcomes across districts. This intercept-only model, which excludes explanatory variables but includes random intercepts at the district level, serves to quantify baseline contextual variation. The intra-class correlation coefficient (ICC) derived from the null model was 0.064, indicating that approximately 6.4% of the total variance in NEET status is attributable to differences between districts (Snijders & Bosker, 2011). In addition, a likelihood ratio (LR) test comparing the multilevel specification to a single-level logistic regression produced a chi-square statistic of 24 154.04 (p < 0.001), confirming that the multilevel model offers a significantly better fit. Although the results of the null model are not reported in Table 2, they provide critical justification for the use of a multilevel framework in the subsequent analysis.

Following the null model, a variance inflation factor (VIF) test was conducted to assess the presence of multicollinearity among the explanatory variables. Using an ordinary least squares regression that included the same set of covariates specified in the full model, the test results show that all independent variables have VIF scores below 2. This confirms that multicollinearity is not a concern in the data and that the estimated coefficients in the subsequent multilevel logistic regression can be interpreted with confidence.

3.1 Education and Economic Status as Determinants of NEET Status (RQ1 & RQ2)

The estimation results from the Baseline Model (Column [1]) demonstrate that tertiary education has a significant protective effect against labor market inactivity among young adults in Indonesia. On average, individuals with tertiary education are 56% less likely to be NEET compared to those without (OR = 0.445), highlighting the role of higher education in reducing early adulthood exclusion from the labor market. This finding is consistent with international research showing that tertiary education improves employability by enhancing skills, expanding access to formal employment, and increasing resilience during economic shocks. For instance, Agranovich and Dreneva (2022) demonstrate that in Russia, individuals with advanced tertiary qualifications—particularly master’s degrees—are more likely to remain economically active even during periods of crisis. These cross-national patterns underscore the transformative impact of tertiary education on employment outcomes. In the Indonesian context, the results provide robust support for RQ1 by confirming that tertiary education serves as a powerful protective factor against NEET status.

Household economic status, measured by log per capita expenditure (LogPCE), emerges as a key predictor of labor market outcomes. The regression results indicate that each unit increase in LogPCE is associated with a 32% reduction in the odds of being NEET (OR = 0.684), suggesting that greater household resources provide a protective buffer against labor market inactivity. Improved economic conditions may enable young adults to invest more in job search activities, overcome spatial or financial barriers to employment, and access social networks that facilitate job entry. This finding is consistent with prior evidence; for instance, Naraswati and Jatmiko (2022) show that lower household income significantly increases NEET risk in Indonesia, particularly in contexts where poverty limits access to education, transportation, and job networks. Together, these results affirm RQ2 by highlighting the protective role of economic resources in reducing NEET status.

Robustness checks were conducted by estimating alternative models that incorporated quadratic, cubic, and quartic forms of LogPCE to assess potential nonlinearities in the relationship. However, none of these higher-order terms improved model fit or substantially altered the interpretation of results. As a result, the linear specification was retained for clarity and parsimony. This approach is consistent with broader empirical evidence on youth labor market outcomes in low- and middle-income countries.

The analysis extends beyond odds ratios by incorporating predicted probabilities derived from the multilevel logistic regression model to better illustrate how education and household economic status shape NEET risk. While odds ratios convey the direction and statistical significance of associations, predicted probabilities offer a more intuitive understanding of how the likelihood of labor market inactivity changes across combinations of LogPCE and educational attainment. This approach clarifies the substantive magnitude of differences in NEET risk and facilitates visual comparison across relevant subgroups. Examining predicted probabilities over a continuous range of LogPCE values and disaggregating by tertiary education status allows for a more nuanced understanding of how economic and educational factors jointly influence the probability of youth disengagement from the labor market.

Figure 2: Predicted NEET Probability By Education And Economic Status

Source: Author’s calculation 

The predicted probabilities (Figure 2) reveal a pronounced inverse relationship between LogPCE and NEET status. As LogPCE rises from 12.0 to 17.0, the predicted probability of being NEET among the general population declines from 38.0% to 12.5%, reinforcing the central role of household economic resources. When disaggregated by educational attainment, tertiary education consistently lowers NEET probabilities across the entire income spectrum. The effect is most pronounced among lower-income individuals: at LogPCE 12.0, the NEET gap between those with and without tertiary education is 12.8 percentage points, narrowing to 6.7 points at LogPCE 17.0. These patterns indicate that while rising household wealth reduces the overall risk of inactivity, tertiary education adds a critical layer of protection—particularly for individuals from economically disadvantaged backgrounds. The findings offer direct evidence in support of both RQ1 and RQ2. Tertiary education substantially reduces the likelihood of labor market inactivity, and the effect of economic status on NEET risk is clearly moderated by educational attainment (Rahmani & Groot, 2023). The interaction observed—wherein the protective benefit of tertiary education is greatest at lower levels of LogPCE—underscores the compounded disadvantages faced by uneducated youth in poor households.

3.2 Regional Development and the Risk of Labor Market Inactivity (RQ3)

Based on estimates from the Baseline Model (Column [1]), the results indicate that the Human Development Index (HDI) is positively and significantly associated with the likelihood of labor market inactivity among young adults. Specifically, a one-point increase in HDI (on the 0–100 scale) corresponds to a 2.7% increase in the odds of being labor market inactive (OR = 1.027), after controlling for individual and household characteristics, including gender, marital status, age, education, urban residence, household composition, and log per capita expenditure. While this may seem counterintuitive—given that higher HDI is typically associated with stronger development outcomes and institutional capacity—it may reflect labor market dynamics in more developed districts. In such areas, employment systems may be more formalized and competitive, making it harder for individuals without strong credentials or social capital to secure work (Pylypenko et al., 2023). Alternatively, higher expectations around job quality may lead young adults to delay employment, increasing temporary inactivity (ILO, 2024). These patterns are consistent with prior findings that development can sometimes widen gaps in access to quality employment when institutional pathways do not keep pace with educational expansion.

In contrast, the poverty rate is negatively and significantly associated with labor market inactivity. A one-percentage-point increase in district-level poverty is associated with a 1.6% decrease in the odds of being NEET (OR = 0.984). Although this direction appears counterintuitive, it may reflect necessity-driven labor force participation. In economically disadvantaged areas, young adults may be compelled to take up any form of employment—regardless of job quality, pay, or formality—to meet household needs. As a result, fewer individuals may fall into the NEET category, not because of stronger labor market integration but due to a lack of viable alternatives to participation. This survival-based dynamic has been observed in other developing contexts, where high poverty rates are correlated with higher informal sector absorption and lower rates of reported inactivity (Oviedo et al., 2009).

These findings directly address RQ3, confirming that regional contextual factors—HDI and poverty—independently influence the likelihood of labor market inactivity among young adults. However, the directions of the effects point to complex underlying mechanisms. Higher HDI may coincide with elevated barriers to entry or delayed employment choices, while higher poverty may suppress inactivity through necessity-based employment. These results underscore the need to interpret regional development indicators not as linear predictors of opportunity but as context-specific factors whose effects are shaped by local labor markets, social norms, and survival strategies.

3.3 Regional Moderation of Educational Effects on Labor Market Inactivity (RQ4)

The results provide strong evidence that the Human Development Index (HDI) and poverty rates significantly moderate the relationship between tertiary education and NEET status among young adults in Indonesia, based on estimates from the Interaction Model (Column [2]). The main effect of tertiary education is highly protective, with individuals who attained tertiary education having substantially lower odds of experiencing inactivity compared to those without such education (OR = 0.071). However, this protective effect is not uniform across districts. The interaction term between tertiary education and HDI is positive and statistically significant (OR = 1.022), indicating that in districts with higher HDI, the advantage conferred by tertiary education diminishes. Specifically, for each one-point increase in HDI, the odds of being NEET among tertiary-educated individuals increase by approximately 2.2% relative to their non-tertiary-educated peers. This suggests that labor market saturation or a mismatch between job expectations and available opportunities may erode the employment benefits typically associated with higher education in more developed districts. Similar trends have been observed in India, where structural barriers and education-employment mismatches have undermined the expected returns to tertiary education despite rising attainment rates (Chakraborty, 2024).

Likewise, poverty rates are also found to moderate the education–inactivity relationship significantly. While the main effect of district-level poverty is associated with lower odds of NEET status (OR = 0.981)—a counterintuitive finding—it becomes less protective when considering individuals with tertiary education. The positive and significant interaction term (OR = 1.023) implies that the benefits of higher education are constrained in poorer districts. This may reflect structural economic weaknesses, underdeveloped labor markets, or limited job-matching mechanisms in impoverished areas. The total effect calculations reinforce this interpretation: for tertiary-educated individuals, the combined effect of HDI yields an odds ratio of approximately 1.046, while the combined effect of poverty approaches 1.003, indicating a neutral to slightly adverse influence. These findings align with prior studies that emphasize the spatial heterogeneity in education-to-employment transitions and the role of contextual disadvantage in undermining the returns to education (Parsons, 2022).

These interaction effects confirm that HDI and poverty levels moderate the relationship between tertiary education and NEET status, thereby affirmatively answering RQ4. The analysis indicates that regional disparities significantly influence how tertiary education protects young adults from labor market exclusion. This underscores the need for policy responses that are not only education-focused but also spatially targeted, considering the varying capacity of local labor markets to absorb educated youth. Without such attention to place-based inequality, national efforts to promote higher education as a pathway to economic participation may fail to deliver inclusive outcomes.

3.4 Moderating Role of Regional Development in the Economic Status–Inactivity Link (RQ5)

District-level HDI and poverty rates were each grouped into two categories to examine whether regional development and deprivation moderate the relationship between household economic status and NEET status. The first group comprises districts in the bottom 50th percentile of the national distribution—classified as low-HDI or low-poverty—while the second group includes those in the top 50th percentile, categorized as high-HDI or high-poverty. These classifications were constructed separately for each survey year to reflect shifts in the distribution of development indicators over time. This relative grouping approach allows for consistent year-on-year comparisons. It facilitates the analysis of whether the protective effect of tertiary education varies systematically across more and less developed contexts (hereafter referred to as low- vs. high-HDI and low- vs. high-poverty districts). The resulting binary variables were used in Interaction Models and margin analyses to assess moderation effects.

The regression results from the Categorical Interaction Model (Column [3]) indicate that tertiary education remains a strong protective factor. Individuals with some higher education are about 60% less likely to be NEET than those without (OR = 0.398). However, the main effects of regional context are more complex. Living in a high-HDI district does not significantly alter NEET risk (OR = 0.974, p = 0.289). In contrast, residence in high-poverty districts is associated with slightly lower odds of being NEET (OR = 0.928, p = 0.001)—a pattern that may reflect economic necessity driving participation in informal work. More notably, the interaction terms reveal that the protective effect of tertiary education is significantly weaker in both high-HDI (OR = 1.130, p < 0.001) and high-poverty districts (OR = 1.093, p < 0.001), suggesting a partial erosion of returns to education in more developed and more deprived regions. These patterns may be explained by credential inflation, job saturation, or structural barriers such as weak labor demand and poor institutional support (Brown et al., 2011; Thurow, 1975).

Figure 3 further illustrates these dynamics using predicted probabilities. In the top panel, NEET probabilities decline with rising LogPCE across low- and high-HDI districts. At LogPCE 12.0, the NEET rate among non-tertiary individuals in low-HDI areas is 0.409, compared to 0.271 among those with tertiary education (a 13.8-point gap). A similar gap exists in high-HDI districts (0.405 vs. 0.285), though the benefit of education is slightly smaller. This pattern persists across the income range, with the education gap narrowing modestly at higher income levels. At LogPCE 14.0, the gap is 11.9 points in low-HDI and 10.3 in high-HDI areas, indicating that the returns to education are marginally less pronounced in more developed districts.

Figure 3: Predicted NEET Probability By Education, Economic Status, Human Development Index, and Poverty Rates

Source: Author’s calculation

The bottom panel of Figure 3 presents predicted NEET probabilities by poverty status. Like the HDI patterns, NEET probabilities decline with income, and tertiary education remains protective. At LogPCE 12.0, non-tertiary individuals in low-poverty districts face a NEET probability of 0.413, compared to 0.277 among their tertiary-educated counterparts—a gap of 13.6 points. In high-poverty districts, the gap is slightly smaller (0.401 vs. 0.279). This difference persists at higher income levels, but the gap narrows: at LogPCE 14.0, the NEET rate is 0.293 for non-tertiary and 0.175 for tertiary-educated youth in low-poverty districts, compared to 0.281 and 0.177 in high-poverty areas. These results suggest that while the poverty context modestly influences the benefits of education, household income and individual education remain the primary determinants.

Together, these findings provide a clear response to RQ5, which asks whether the effect of household economic status on labor market inactivity varies across districts with differing levels of HDI and poverty. While income consistently reduces NEET risk, the degree to which this effect is enhanced by tertiary education varies by regional context. The interaction terms and predicted probabilities indicate that education's protective value is marginally weaker in high-HDI and high-poverty areas (ILO, 2020). This context-sensitive effect reflects local labor market saturation, institutional weaknesses, or mismatches between graduate skills and available jobs. Although education and income remain central to reducing NEET status, their effectiveness is shaped by the opportunity structures in which young adults are embedded.

Robustness checks were conducted to confirm the consistency of these results. One test used fixed groupings based on 2024 values of HDI and poverty rates, applied uniformly across all survey years. Another reclassified HDI and poverty into terciles—low, middle, and high—rather than binary categories. The interaction patterns and effect sizes in both cases remained consistent with those reported above. These findings confirm that the moderating influence of regional development and deprivation on the relationship between tertiary education and NEET status is robust to alternative classification strategies, reinforcing the credibility of the analysis.

3.5 Estimates for Additional Individual and Household Predictors

The Baseline Model (Column [1]) also estimates the effects of additional individual and household-level characteristics on NEET status among young adults. Gender shows the strongest effect: women are over 14 times more likely than men to be NEET (OR = 14.140), highlighting persistent gender inequalities in labor market access—often shaped by caregiving burdens and gender norms (Kang & Youn, 2024). Married individuals are about 15% less likely to be NEET (OR = 0.854), potentially reflecting greater financial responsibilities or higher employment stability among those who marry (Mehta & Awasthi, 2025). The association between age and NEET status declines steadily from age 26 onward, with odds ratios falling to 0.641 by age 34. This age gradient reflects life-course transitions into more stable employment (Pattinasarany, 2024).

Urban residence is associated with slightly higher NEET risk (OR = 1.056), suggesting urban labor market exclusion despite greater job density—likely due to heightened competition or segmentation (Lindblad et al., 2025). Household composition also matters. The presence of young children (aged 0–4) increases the odds of being NEET (OR = 1.149), as does having elderly members (aged 60+), though more modestly (OR = 1.016), likely due to caregiving constraints (Parida & Pattayat, 2024). The year-fixed effects likely reflect the impact of the COVID-19 pandemic on youth labor market outcomes. Compared to 2019, the odds of being NEET increased significantly in 2020 and peaked in 2022 (OR = 1.159) before declining slightly in 2023 and 2024. This pattern is consistent with broader evidence that the COVID-19 pandemic disrupted school-to-work transitions and reduced job opportunities for young adults (ILO, 2022), leading to a temporary but notable rise in inactivity.

4. Conclusions

4.1 Summary of Findings

This study investigated the determinants of labor market inactivity among young adults aged 25–34 in Indonesia, focusing on how regional socioeconomic contexts shape the effectiveness of tertiary education in reducing NEET (Not in Employment, Education, or Training) outcomes. Drawing on nationally representative SUSENAS data from 2019 to 2024, the analysis examined how educational attainment and household economic status interact with district-level development conditions—specifically the Human Development Index (HDI) and poverty rates—using multilevel mixed-effects logistic regression to account for contextual heterogeneity.

The results confirm that tertiary education and household economic status are strong, independent predictors of NEET status. Individuals with tertiary education are significantly less likely to be inactive, and higher household income—as measured by LogPCE—is consistently associated with lower NEET risk. Predicted probabilities further reveal that these effects are interdependent: while income reduces NEET risk across all education levels, the protective effect of tertiary education is especially pronounced among individuals from lower-income households.

Regional context plays a moderating role in these relationships. The protective effect of tertiary education is weaker in high-HDI and high-poverty districts, suggesting that structural constraints—such as labor market saturation or limited job creation—diminish the employment benefits of education in these areas. Although rising income and education generally reduce inactivity, their effectiveness depends on the absorptive capacity of local economies. These findings underscore the need for spatially targeted strategies that align education investments with regional labor market conditions to ensure meaningful employment outcomes for young adults.

4.2 Policy and Economic Implications

The evidence that the effect of tertiary education on labor market inactivity varies across districts with different levels of HDI and poverty—classified using year-specific median splits— holds significant implications for economic theory and policy. It challenges the assumption embedded in classical human capital theory (Becker, 1993) that higher education yields uniformly positive returns by increasing individual productivity. Instead, the findings support a more context-dependent interpretation, in which regional labor market structures and institutional capacity shape the economic value of education.

In districts classified as high-HDI, where education infrastructure and access are relatively strong, the weakening of the protective effect of tertiary education suggests signs of labor market saturation and credential inflation. These outcomes are consistent with job competition theory (Thurow, 1975), which posits that education acts more as a screening device than a productivity enhancer in contexts of limited job creation. Similar dynamics have been observed in urban labor markets across Southeast Asia and Sub-Saharan Africa, where rising educational attainment has not been accompanied by a commensurate expansion of high-skilled employment opportunities. In the Indonesian context, a World Bank (2021) study highlights that while tertiary education does improve employability on average, challenges related to underemployment and informality—particularly among younger cohorts—persist, indicating a disconnect between educational expansion and labor market absorption.

In districts classified as high-poverty, the reduced effectiveness of tertiary education reflects demand-side constraints—a hallmark of dual labor market theory (Doeringer & Piore, 1985). These regions typically lack a diversified economic base, which limits the number of formal sector jobs available to absorb skilled labor. Even when education levels improve, structural weaknesses such as poor infrastructure, low private sector activity, and limited institutional support mean that individuals cannot capitalize on their qualifications. Recent studies reinforce this interpretation. Montenegro and Patrinos (2023) highlight that in many developing countries, the labor market returns to education are heavily moderated by local job conditions, with returns being systematically lower in economically lagging regions.

These patterns underscore three critical economic policy implications. First, better alignment between education and regional labor markets is essential. Policymakers must invest in graduate tracking systems and strengthen local labor observatories to inform curriculum reform and skill development programs tailored to district-level disparities. Second, the weakened returns to education in both high-HDI and high-poverty districts call for spatially targeted employment policies. In urban centers of Java and Sumatera, where skilled labor markets are saturated, more effective Active Labor Market Policies (ALMPs) are needed. Indonesia’s ALMPs—delivered through vocational training centers (BLKs), SIAPKerja platforms (Setyawan et al., 2024), and apprenticeship schemes—remain fragmented and uneven. Strengthening their integration with education systems and tailoring them to local labor demand can ease school-to-work transitions. In poorer regions, particularly Eastern Indonesia, ALMPs should be paired with investments in infrastructure, local enterprise support, and digital access to create demand for skilled labor and reduce inactivity among educated youth. Third, the study points to the need for integrated public spending strategies beyond investing in education alone. Without coordinated improvements in local economic conditions, rising educational attainment risks producing a class of overqualified but underemployed youth. This undermines the efficiency of human capital investment and may fuel frustration and disengagement among young adults, especially in peripheral regions.

4.3 Study Limitations and Future Extension

While the study offers important insights into how education, household economic status, and regional development influence youth labor market inactivity, it is not without limitations.  The analysis is based on repeated cross-sectional SUSENAS data, which hinders the assessment of long-term transitions into and out of NEET status. Future research using panel data is strongly recommended to gain a more comprehensive and dynamic understanding of young adult trajectories (Karma, 2024). Furthermore, the binary measure of tertiary education fails to consider variations in field of study, institutional quality, or completion status, all of which can significantly affect employment outcomes.

In addition, the use of HDI and poverty rates as proxies for regional development does not fully capture local labor market structures or institutional conditions. Including complementary indicators—such as job informality, industrial composition, or access to local employment programs—would provide a more complete picture of regional constraints (ILO, 2020). Finally, it is crucial to emphasize the need for future studies to adopt mixed-method approaches that integrate survey data with qualitative insights. This urgency in research direction will help to uncover how young people experience barriers to employment across different regional contexts (Sánchez-Soto & León, 2020). Such extensions would enrich the findings and offer deeper policy relevance.

Author Contributions: The sole author designed, analyzed, interpreted and prepared the manuscript.

Funding: The author received no financial support for the research, authorship, and/or publication of this article.

Conflicts of Interest: The author declares no competing interests.

Informed Consent Statement/Ethics approval: Informed consent/ethical approval was not required as the study did not involve human participants performed by the author.

Declaration of Generative AI and AI-assisted Technologies: During the preparation of this work, the author utilized ChatGPT (OpenAI) and Grammarly to enhance language clarity, readability, grammatical accuracy, and consistency through proofreading. After using this tool/service, the author reviewed and edited the content as needed and took full responsibility for the content of the publication.