Does the Smart Indonesia Program Have an Impact on Spatially Inclusive Economic Growth?

Improving the quality of human resources through education is a key prerequisite for inclusive economic growth. The Smart Indonesia Program (SIP) is a government policy instrument aimed at expanding access to education for children from poor and vulnerable households, with the objective of reducing structural barriers to school participation. By providing direct cash transfers to eligible students, the SIP not only seeks to reduce school dropout rates but also forms part of a long-term human capital development strategy aimed at enhancing productivity and promoting equitable welfare outcomes. Furthermore, the SIP represents a form of Conditional Cash Transfer (CCT) designed to alleviate the financial burden of education for low-income households and to increase school participation as a key indicator of the effectiveness of national education policy [1].

In theory, the effectiveness of the SIP can be explained through the human capital theory and capability approach. CCT programs have been shown to stimulate greater investment in education and health, which in turn helps strengthen human capital as a key driver of sustainable economic growth. These results align with evidence showing that the CCT program consistently increases school enrolment rates and lowers the likelihood of student dropouts in developing countries [1]. Thus, SIP can be understood as a tangible form of inclusive development that focuses on equal access to education and improving individual capabilities. Conditional assistance programs in Indonesia have a significant effect on improving school participation and public health, although their long-term influence on economic growth has not yet been fully optimized [2]. Meanwhile, the SIP program has been shown to be effective in reducing disparities in access to education in Indonesia [3].

Students receiving PIP have a lower risk of dropping out compared to those who do not, confirming the program’s effectiveness in supporting educational continuity, although challenges in targeting and monitoring persist [4]. A report by the Ministry of Education and Culture also confirms that PIP has successfully expanded access to education; however, it continues to face obstacles related to the uneven distribution of assistance across regions.

These indicate that educational assistance policies, such as SIP, play a significant role in promoting inclusive economic growth by improving access to education. SIP beneficiaries are less likely to drop out, highlighting the program’s effectiveness in supporting educational continuity, despite persistent challenges in targeting and monitoring [5]. A report from the Ministry of Education and Culture further confirms that SIP has succeeded in expanding access to education; however, it continues to face obstacles related to the unequal distribution of assistance across regions.

SIP is considered to be in line with the needs of the community and has a clear legal basis and objectives in supporting equitable access to education. In terms of input, the mechanism for distributing and receiving assistance has been carried out in accordance with the applicable technical guidelines. In terms of process, the procedures for disbursement, distribution, utilization, and supervision of funds have been implemented in accordance with regulations. However, in terms of product, the program results have not been fully optimal because the dropout rate is still increasing, which demonstrated that the goal of 12 years of compulsory education has not been fully achieved [6].

The SIP in vocational schools has been shown to support disadvantaged students in continuing their education and reducing dropout rates; however, its implementation faces challenges such as delayed funding, insufficient socialization, as well as administrative and coordination issues. Enhancing data systems and governance is necessary to improve the program’s effectiveness and the accuracy of beneficiary targeting [7]. These local findings underscore the importance of good governance to ensure that SIP can contribute optimally to promoting inclusive economic growth through education.

 In addition to the direct assistance provided through the SIP, several structural variables are used as control factors in determining the effectiveness of the SIP in improving access to education and supporting inclusive economic growth. One crucial structural variable is the availability of schools and educational infrastructure. The number and quality of school facilities determine enrollment capacity and the overall learning environment. SIP assistance may not achieve optimal outcomes if the distribution of schools across regions remains uneven and limited.

In rural areas, basic facilities such as access to clean water and waste management systems also contribute positively to learning outcomes, although not all types of infrastructure investment exhibit consistent effects. The study concludes that the effectiveness of educational investment depends not only on the size of the budget but also on the type and quality of facilities that directly support the learning process. Therefore, education policies should be more targeted and evidence-based in order to maximise their impact on student achievement [7].

In addition, conditional assistance programs such as SIP have the potential to enhance educational aspirations among beneficiary households. Empirical evidence from Indonesia indicates that participation in CCT programmes can raise parents’ aspirations for their children’s education, thereby reinforcing human capital investment and supporting long-term improvements in the quality of human resources [7].

The availability of road infrastructure is a critical factor influencing the accessibility of SIP. Regions and municipalities with limited transportation networks face significant challenges in enabling students to attend school, even when they receive SIP assistance. Improvements in rural road infrastructure have been shown to substantially enhance children’s access to education in remote areas by reducing travel time and transportation costs, thereby promoting higher school attendance, particularly among low-income households.

Nevertheless, these benefits are not evenly distributed, as communities with underdeveloped social and economic infrastructure continue to encounter barriers in fully utilising such improvements. Overall, rural road development generates substantial social benefits by expanding educational opportunities and supporting long-term human capital development [8].

Furthermore, the net enrolment rate (NER) serves as an initial indicator of the education system’s performance. A high NER indicates that a large proportion of school-age children are attending school at the appropriate level, enabling the SIP to support the continuity of education. Conversely, in regions with a low NER, SIP must be complemented by additional interventions to prevent early school dropout. The prevalence of poverty is also a critical factor affecting both the urgency and effectiveness of SIP. A high poverty rate signals an increased need for educational assistance to ensure that children from economically disadvantaged households maintain access to and continuity in education. The number of poor residents serves as a key basis for designing educational assistance policies. The SIP is therefore specifically designed to expand access to education for children from poor and vulnerable families.

Empirically, the relationship between poverty and access to education has been widely documented in international studies. The research shows household economic conditions are a crucial determinant of children’s educational participation. The study shows that although Ghana experienced expanded access to basic education and declining poverty rates, economic disparities continued to shape educational trajectories [5]. Children from poorer households were more likely to face barriers in progressing to higher levels of education, limiting their ability to benefit from the stronger economic returns associated with secondary and post-basic schooling. This aligns with broader empirical evidence indicating that poverty increases the risk of school dropout or non-enrollment, while improvements in household welfare enhance educational continuity. Therefore, the Ghana case provides concrete international evidence that reducing poverty and strengthening educational support policies are essential to ensuring sustained participation in formal education among vulnerable children.

Furthermore, the multidimensional poverty approach emphasizes that poverty is not solely related to income limitations but also associated with restricted access to basic services, including education. Research on multidimensional poverty shows that poor households often face structural barriers such as limited ability to afford education costs, lack of supporting learning facilities, and economic pressure that encourages children to work rather than continue their education [7]. Other studies also indicate that household economic constraints influence school choices and the quality of education accessible to children from low-income groups, thereby reinforcing the importance of inclusive education policy interventions [8]. Therefore, educational assistance SIP has strong justification as a form of social protection aimed at reducing inequality in educational access while simultaneously breaking the intergenerational cycle of poverty. Thus, the higher the poverty rate in a region, the greater the urgency for providing educational assistance to improve school participation and strengthen inclusive human capital development.

Educational cash transfer programs in Indonesia have demonstrated a cumulative effect in improving school participation rates, particularly among low-income households [3]. In this regard, the persistently high level of poverty not only indicates a substantial need for such interventions but also provides an important basis for evaluating the effectiveness of SIP as a social protection mechanism to support human capital development and advance inclusive economic growth.

This study empirically examines the impact of the Program Keluarga Harapan (PKH) on the welfare of low-income households in Indonesia by using household consumption expenditure as a proxy for well-being and applying econometric methods to identify causal effects. The findings indicate that program participation leads to a positive and statistically significant increase in household consumption, suggesting improvements in purchasing power and short-term welfare. However, the effects vary across wealth groups, with households in the lowest wealth quantile experiencing comparatively smaller gains. This uneven impact reflects structural constraints, including limited access to complementary services, administrative barriers, and reduced capacity of the poorest households to utilize transfers effectively. Overall, while the program demonstrates measurable welfare benefits, enhancing targeting accuracy and implementation quality is essential to ensure more inclusive and equitable outcomes [9].

An evaluation of the implementation of SIP in vocational high schools assessed both its effectiveness and the key challenges encountered [10]. The findings demonstrate that SIP has generally supported students from underprivileged families in meeting their educational needs and preventing potential school dropouts. Nevertheless, some recipients reportedly use aid funds for purposes beyond education, such as personal expenses and transportation. However, the study did not find strong evidence that such deviations are particularly prevalent among households with low consumption levels. These results highlight the need for enhanced socialization, supervision, and guidance to ensure that aid funds are utilized optimally in accordance with program objectives, thereby supporting both access to and quality of education. Collectively, these studies indicate that SIP plays a critical role in expanding educational access and fostering human capital development in Indonesia, including in West Java. At the same time, the gap between program design and field-level implementation underscores a persisting research gap.

Research linking the SIP to indicators of inclusive economic growth at the regional level remains limited. In addition, the quality of education for aid recipients has not been extensively examined, despite its importance in assessing the long-term impacts on productivity and welfare. Therefore, further research is required to evaluate SIP’s contribution not only in terms of school participation rates but also in relation to regional economic transformation towards inclusive growth.

West Java Province is one of the regions with the largest absolute number of poor residents in Indonesia, despite having a relatively lower poverty rate compared to other provinces. This demographic feature highlights the critical role of the poor population in supporting the implementation of the SIP. The greater the number of children from economically disadvantaged households in the region, the higher the demand for educational assistance to prevent school dropouts. Data from Statistics Indonesia (2023) indicate that over 3.7 million individuals in West Java live below the poverty line, making school-age children from poor households the primary target beneficiaries of SIP.

Existing studies have attempted to assess the effectiveness of Indonesia’s social assistance programs, including the Family Hope Program (FHP) and SIP, in reaching low-income populations; however, the evidence raises concerns regarding their targeting performance. Analyses of national datasets and quantitative evaluations consistently indicate that, despite the expansion in programme coverage, effectiveness remains limited due to persistent targeting inaccuracies. These inaccuracies are manifested in both inclusion errors where benefits are received by non-poor individuals and exclusion errors where eligible poor households are systematically left out suggesting structural weaknesses in beneficiary identification and validation mechanisms, as well as uneven implementation across regions. Moreover, empirical findings based on logit model estimations reveal that targeting outcomes in programmes such as CCT, education scholarships, and in-kind transfers are not primarily determined by poverty status. Instead, they are disproportionately influenced by urban–rural disparities, the possession of identity documentation, and inconsistencies in data quality. This misalignment indicates that existing welfare databases and eligibility criteria are insufficiently robust to accurately capture the intended beneficiaries, thereby undermining the redistributive function of these programmes and limiting their effectiveness in alleviating poverty [11].

These limitations point to a more fundamental issue, that the current architecture of social assistance delivery in Indonesia remains vulnerable to systematic exclusion and misallocation. Without substantial improvements in integrated data governance, verification protocols, and monitoring mechanisms, the expansion of social assistance risks becoming nominal rather than transformative. Therefore, strengthening data integration systems, enhancing verification processes, and improving distribution oversight are not merely administrative adjustments, but critical prerequisites for ensuring that social assistance programmes, including education support schemes, can achieve their intended objectives of reducing poverty and advancing equitable and sustainable access to education.

3.1 Research location

This study uses 26 cities/regions from 27 cities, excluding Banjar region in West Java Province as the unit of analysis. Banjar is excluded in the sample unit because Banjar does not have complete IEDI and supporting macro data. West Java Province is divided into four development coordination agencies, as follows: Regional Coordination Agency Bogor, which covers Bogor, Bekasi, and Sukabumi; Regional Coordination Agency Purwakarta, which covers Purwakarta, Subang, and Karawang; Regional Coordination Agency Cirebon, which covers Cirebon, Indramayu, Majalengka, and Kuningan; and Regional Coordination Agency Priangan (Garut), which covers Bandung, Sumedang, Garut, Tasikmalaya, Ciamis, and Banjar. This study covers the 2017–2023 period and therefore employs a balanced panel dataset for the analysis. Figure 2 shows the cities and regencies map in West Java.

Figure 2. Cities/region map in West Java

3.2 Geographically Weighted Panel Regression  method

Spatial analysis plays a crucial role in the examination of the distribution and effectiveness of educational assistance programs, as it facilitates the identification of spatial patterns, clustering, and disparities across geographic areas. Empirical evidence indicates that poverty is not randomly distributed; rather, it exhibits significant spatial clustering, suggesting the presence of spatial dependence among neighbouring regions.

This evidence underscores the importance of incorporating spatial analytical approaches into the evaluation of social assistance programs, as their effectiveness and targeting accuracy are inherently linked to the spatial distribution of poverty. By identifying geographic clusters and spatial disparities, policymakers are better positioned to design more targeted and equitable interventions that are responsive to localised socio-economic conditions [12].

The relationship between the dependent variable y and the independent variables x1, x2, …, xn in the Geographically Weighted Regression (GWR) model for location can be written as follows [6]:

$\begin{gathered}y i=\beta 0(u i, v i)+\sum \beta k(u i, v i) x i k+\varepsilon i, i=1,2, \ldots, n \\ k=1\end{gathered}$

Table 1 presents the definitions of the variables used in the above equation.

Table 1. Variable definition

Conventional econometric regression frameworks are grounded in the assumption that a single, global model can be consistently estimated conditional on a given sample of data. However, this assumption may be overly restrictive, particularly in the presence of spatial heterogeneity, where relationships between variables are unlikely to be constant across geographic units. In such contexts, global models risk producing biased or misleading estimates by imposing parameter homogeneity and neglecting localised variations.

In contrast, the Geographically Weighted Panel Regression (GWPR) model relaxes this assumption by incorporating location specific weighting schemes and subsampling procedures for each cross-sectional observation. This approach enables the estimation of spatially varying parameters, thereby capturing local heterogeneity in the underlying data-generating process. Consequently, GWPR not only facilitates a more nuanced understanding of spatial dynamics but also allows for systematic comparison with global model estimates, highlighting the extent to which spatial non-stationarity influences empirical results.

GWPR has evolved to incorporate the temporal dimension by assuming that observations proximate in both space and time exert greater influence on the estimation of local parameters at location i. This extension focuses on econometric techniques based on repeated observations for each location, with particular emphasis on panel data models with fixed effects. GWPR differs from conventional geographically weighted models in that, once a bandwidth is selected, all temporal observations for each subsampled location must be assigned the same weight in order to locally replicate the pooled structure of a global model across different time periods. For notational simplicity, two types of independent variables are considered, namely x and z.

The Locally Weighted Regression (LWR) model uses a weighting function (kernel function) to approximate (•). At each target location $i$, the prediction is obtained by weighting the differences between the normalized values of the independent variables at point $i$ and those at the target point, $x j-x i$ and $z j-z i$. The kernel function (see below) determines the weight that observation $j$ receives in estimating the value of $y$ at target point $i$. This can be generically represented as: $\mathrm{y}=\mathrm{f}(\mathrm{x}, \mathrm{z})+\mathrm{u}$.

The Kernel Regression model uses the kernel function based on $x j-x i$ and $z j-z i$ to weight the observations. However, only the values of $y j$ are used to predict $y i$. This can be represented as:

$y=f(y)+u$

The Conditional Parametric Regression (CPAR) model is a special case of the LWR model which assumes that the set of explanatory variables $(x, z)$ can be divided into portions that are fully nonparametric $(x)$ and conditionally parametric $(z)$. For fixed values of $z$, the model is a linear equation in which the coefficients vary with $z$ :

$y=\alpha(z)+\beta(z) x+u$

The CPAR model becomes spatial when the conditionally parametric variables are the geographic coordinates of each point, latitude and longitude:

$y=\alpha(l a, l o)+\beta(l a, l o) x+u$

GWR model is a special case of the CPAR (LWR) model where the coordinates are replaced by straight-line distances (d) among observations and are typically omitted from the list of explanatory variables.

$y=(d)+\beta(d) x+u$

In economic analysis, geographic coordinates may have limited intrinsic meaning; however, spatial distance often serves as a proxy for variables such as the number of schools, road network length, and NER. Conventional GWR captures spatial heterogeneity by allowing regression coefficients to vary across locations based on geographic proximity, while its panel extension, GWPR, applies location-specific kernel weights across time periods to integrate both spatial and temporal dimensions.

However, economic interactions are not determined solely by physical distance but also by similarities in regional characteristics. To address this limitation, Similarity-based GWR extends the framework by incorporating both spatial proximity and attribute similarity into the weighting mechanism. This approach enhances the estimation of local parameters and provides a more comprehensive representation of complex and heterogeneous spatial economic processes [13].

GWPR extends conventional panel data models by incorporating spatial heterogeneity through location specific parameter estimates that vary across cross sectional units while accounting for temporal dynamics. Within this framework, the kernel function determines how observations are weighted according to their spatial proximity to a given location.

In empirical applications, however, the choice of kernel function generally exerts only a limited influence on estimation outcomes, as commonly used kernels such as the Gaussian and bi-square functions share the fundamental property of assigning monotonically decreasing weights as distance increases. Hence, differences among kernel types typically reflect alternative functional forms of spatial decay rather than substantive changes in the underlying weighting mechanism as studied by Yacim and Boshoff [14].

In contrast, bandwidth selection plays a far more decisive role in shaping model performance. The bandwidth determines the spatial extent of the local neighbourhood and governs the rate at which weights decay, thereby directly influencing the bias–variance trade-off and the stability of local parameter estimates. An adaptive bandwidth, often referred to as a variable window size, assigns a distinct bandwidth hi to each location, ensuring that an equal number of nearest neighbours is incorporated into every local estimation.

This approach is particularly advantageous in contexts where spatial units are unevenly distributed, as it preserves a consistent local sample size across estimation points while allowing the spatial influence structure to vary. In this study, an adaptive bi-square kernel function is employed to weight observations in each local panel estimation, providing a flexible and robust mechanism for capturing spatially and temporally varying relationships.

Methodological challenge in the application of GWR lies in the selection of an appropriate bandwidth, or, equivalently, in determining the optimal number of nearest neighbours to be included in each local estimation. This issue becomes even more complex within GWPR framework, where bandwidth optimisation procedures remain an area of ongoing methodological development. Furthermore, the application of model selection criteria such as the Akaike Information Criterion (AIC) in a GWPR setting must account for the substantial computational burden associated with processing time-series observations at each spatial location.

Another important limitation of GWR-based analyses concerns their heavy reliance on visual interpretation through spatial maps. Beyond the well-documented limitations of choropleth mapping, particularly when applied to spatial units of heterogeneous size, the interpretation of mapped local parameter estimates remains inherently relative unless accompanied by measures of statistical significance. As such, without appropriate inferential support, spatial variations observed in parameter estimates may be prone to misinterpretation [15].

This study employs the GWR method, a localized regression approach that enables the detection of spatial variations in the relationships between dependent and independent variables across the study area [15]. The fundamental principle of this model is the incorporation of geographic or locational information as weights in the estimation of model parameters. This approach is grounded in Tobler’s First Law of Geography [15], which posits that spatial units located in close proximity tend to exhibit stronger relationships and greater similarity than those that are farther apart.

The data from the study area indicate the presence of spatial dependence between measurement values and geographic location. The weighting scheme in GWPR model constitutes a critical component in estimating location-specific parameters based on the spatial positions of individual observations. Consequently, the resulting parameter estimates vary across observation locations. The magnitude of the assigned weight increases as the distance between observation locations decreases, indicating that geographically proximate observations exert greater influence on local parameter estimation. The GWPR method is a form of local regression in which data are repeated across regional points for each spatial observation. In this framework, it is assumed that the temporal sequence of observations at a given geographic location represents a realization of a smooth spatiotemporal process; that is, observations that are closer in space and time are more closely related than those that are further apart [12].

GWPR is an extension of spatial regression techniques that allows regression parameters to vary across geographical locations, thereby capturing spatial heterogeneity in the relationships between independent and dependent variables. Unlike global regression models that assume constant parameter estimates across all regions, GWPR estimates local parameter values for each observation point, enabling the identification of spatially varying responses within a single study area.

This methodological approach is particularly useful for evaluating regional policy impacts, as socioeconomic variables often demonstrate different levels of influence depending on local geographic and structural characteristics. GWPR integrates spatial heterogeneity and temporal dynamics by estimating local regression coefficients at each geographic location, enabling examination of how social assistance variables interact with regional characteristics in shaping an inclusiveness index across cities and regions. Recent applications of GWPR demonstrate its ability to identify spatially varying relationships between outcome indicators and explanatory variables in panel data settings, such as poverty dynamics in West Java [16] and multidimensional local influences on food security indices in Indonesian provinces [12].

A previous empirical study [17] has demonstrated the robustness of GWPR in analyzing spatio-temporal socioeconomic data. For instance, research applying GWPR to poverty modeling in West Java shows that the influence of socioeconomic indicators on poverty varies across regions and time periods, highlighting the importance of localized parameter estimation in regional policy evaluation. Similarly, methodological developments in spatio-temporal modeling emphasize that GWPR improves analytical accuracy by incorporating spatial weighting functions and kernel-based estimations to account for geographical variations in panel datasets [18]. Furthermore, studies combining GWPR with efficiency and performance evaluation techniques reveal significant spatial heterogeneity in socioeconomic determinants across regions, demonstrating that local characteristics strongly influence development outcomes and policy effectiveness. These findings confirm that GWPR provides a comprehensive analytical framework for assessing the spatial variation of social assistance and regional characteristics in shaping inclusive economic development outcomes.

This study also includes a GWPR analysis to examine the significance and magnitude of the contribution of social assistance variables and regional characteristics to the inclusiveness index in each city/region in West Java. GWR is a statistical technique that calculates independent variables showing different reactions in different places within a single research area. GWPR calculates the parameter value for each location point so that it has different significance and regression parameter values [12]. This study uses R Studio software to process the GWPR results.

The GWPR model used is as follows:

Pillar $\_1_{i t}=\beta 0\left(u_i v_i\right)+\beta 1\left(u_i v_i\right)$ SIP $_{i t}+\beta 2\left(u_i v_i\right)$ School $_{i t}+\beta 3\left(u_i v_i\right)$ Teacherit $+\beta 4\left(u_i v_i\right) \operatorname{Road}_{i t}+\beta 5\left(u_i v_i\right) \operatorname{Poor}_{i t}+\beta 6\left(u_i v_i\right) \operatorname{Cons}_{i t}+\beta 7\left({u i v i}\right) \operatorname{NER}_{i t}+\varepsilon i t$

Pillar_2 ${ }_{\mathrm{it}}=\alpha_0\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right)+\alpha_1\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right)$ SIP $_{\mathrm{it}}+\alpha_2\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right)$ School $_{\mathrm{it}}+\alpha 3\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right)$ Teacher $_{\mathrm{it}}+\alpha 4\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right) \operatorname{Road}_{\mathrm{it}}+\alpha 5\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right) \operatorname{Poor}_{\mathrm{it}}+\alpha 6\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right) \operatorname{Cons}_{\mathrm{it}}+\alpha 7\left(\mathrm{u}_{\mathrm{i}} \mathrm{v}_{\mathrm{i}}\right) \operatorname{NER}_{\mathrm{it}}+\varepsilon_{\mathrm{it}}$

Pillar_ $3_{\text {it}}=\gamma_0\left(u_i v_i\right)+\gamma_1$ (uivi) SIP $_{\text {it}}+\gamma_2\left(u_i v_i\right)$ School $_{i t}+\gamma_3\left(\right.$ uivi $\left._{\text {i}}\right)$ Teacher $_{\text {it}}+\gamma_4\left(u_i v_i\right) \operatorname{Road}_{\text {it}}+\gamma 5\left(u_i v_i\right) \operatorname{Poor}_{i t}+\gamma 6\left(\right.$uivi) Cons $_{\text {it}}+\gamma_7\left(u_i v_i\right)$ NER $_{\text {it}}+\varepsilon_{\text {it}}$

EDI $_{i t}=\delta_0\left(u_i v_i\right)+\delta_1\left(u_i v_i\right)$ SIP $_{i t}+\delta_2\left(u_i v_i\right)$ School $_{i t}+\delta_3\left(u_i v_i\right)$ Teacherit $+\delta_4\left(u_i v_i\right)$ Road $_{i t}+\delta_5\left(u_i v_i\right)$ Poor $_{i t}+\delta_6\left(u_i v_i\right)$ Consit $+\delta_7\left(u_i v_i\right)$ NER $_{i t}+\varepsilon_{i t}$

The explanation of the above equation is as follows:

Pillar 1: Economic Growth Index

Pillar 2: Poverty and Inequality Index

Pillar 3: Access and Opportunity Index

IEDI: Inclusive Economic Growth Index

Β₀, β₁, … β₇: Constanta and Koef variabel in city/region i

Pillar 1 α₀, α₁, … α₇: Constanta and Koef variable in city/region i

Pillar 2 γ_o, γ₁, … γ₇: Constanta and Koef variabel city/region i

Pillar 3 δ₀, δ₁, … δ₇: Constanta and Koef variabel city/region i

IEDI u_iv_i: Coordinate Longitude, Latitude city/region i

SIP: Smart Indonesia Program in city/region i in year t

School: Number of schools in city/region i in year t

Teacher: Number of teachers in city/region i in year t

Road: Road length in city/region i in year t

Poor: Number of poor people in city/region i in year

Cons: Level of community consumption in city/region i in year

NER: Net enrolment rate for junior high schools

Εit: Error

NER measures the percentage of children of a specific official school-age group who are enrolled in the education level appropriate for their age. It reflects the degree of age-appropriate participation in education. The formula of NER is number of children of official school age enrolled at the corresponding education level divided the total population on the same official school age group. This indicator emphasizes age-appropriate participation, thereby better reflecting targeted access to education than other participation indicators. Conceptually, NER is used to assess the extent to which the education system is able to absorb the school-age population in accordance with the national education level structure.

Meanwhile, the poor variable refers to poor or vulnerable households that meet certain requirements (e.g., having a pregnant mother, toddlers, school-age children, elderly members, or members with disabilities). Assistance is provided on a conditional basis (CCT), whereby the poor must fulfil obligations such as ensuring that children attend school and utilize health services.

The GWPR method begins with a multicollinearity test to examine the independence of variables in order to avoid inconsistencies in variable significance and inefficient parameter coefficients. The next step is to select the bandwidth weight to determine the location weight function and test for spatial heterogeneity to ascertain whether there are elements of heterogeneity in the research data used.

The SIP has proven to have a different influence on each region in West Java when viewed from each Pillar in the IEDI.

1. Pillar 1 (Economic Growth)

SIP is statistically significant in the Bogor Area (Bogor Regency and Depok City), the Purwakarta Area (Purwakarta Regency), and the Bandung Area (Bandung Regency, West Bandung Regency, Bandung City, Cimahi City, and Sumedang Regency). The significance of these regions is primarily influenced by their proximity to major economic centres (Jakarta and Bandung) and the availability of adequate educational infrastructure. SIP operates effectively in these areas due to sufficient school and teacher resources, allowing recipients to fully utilize educational funds to support school continuity and enhance human capital quality. In turn, this contributes to strengthening the Economic Growth Pillar.

1. Pillar 2 (Poverty and Income Inequality)

SIP is statistically significant in Cianjur Regency, Cirebon Regency, Bekasi City, and Kuningan Regency. This significance is attributable to the high absolute number of poor residents in these areas, which substantially increases the risk of school dropout. SIP functions as a critical intervention to alleviate financial barriers to education and to reduce disparities in educational access between children from poor and non-poor households. Consequently, SIP in these regions contributes directly to poverty reduction and the mitigation of socio-economic inequalities.

2. Pillar 3 (Expanding Access and Opportunity)

SIP is statistically significant Cianjur, Sukabumi, Purwakarta, Bandung, and Cimahi. This significance can be attributed to the high absolute number of poor residents in these areas, which markedly increases the risk of school dropout. SIP serves as a critical intervention to alleviate financial barriers to education and to reduce disparities in educational access between children from poor and non-poor households. Consequently, SIP in these regions contributes directly to poverty reduction and the alleviation of socio-economic inequalities.

3. Inclusive Economic Development Index

Overall, SIP is statistically significant only in Sukabumi City and Regency, Kuningan Regency, Majalengka, Garut, Banjar City, and Sumedang Regency. This significance arises from a combination of factors, including high poverty levels (Kuningan, Garut, Banjar), a large school-age population (Sukabumi), limited access to education (Garut, Sukabumi), and socio-economic transformation resulting from infrastructure development (Majalengka, Sumedang). In these areas, SIP not only alleviates financial barriers to education but also prevents disadvantaged groups from being further marginalized in the development process, thereby contributing substantively to all three Pillars of the IEDI.

Poverty has been shown to exhibit a clustered spatial pattern with positive spatial autocorrelation, where regions with high poverty levels are often concentrated near other impoverished areas due to limited infrastructure, inadequate access to education and healthcare, geographic isolation, and uneven regional development [26]. This spatial concentration underscores the importance of geographically targeted poverty alleviation strategies, as uniform policy interventions may not effectively address the distinct structural challenges faced by high-poverty regions. Spatially responsive policies that prioritize infrastructure development, educational access, public services, and regional connectivity are therefore essential for reducing poverty more effectively and equitably [26].

In a similar context, the School Indonesia Program (SIP) in West Java demonstrates that social assistance effectiveness also varies significantly across regions depending on local socio-economic and spatial conditions. Under Pillar 1 (Economic Growth), SIP is more effective in areas proximate to major growth centers such as Bogor, Depok, Bandung, Cimahi, and Purwakarta, where stronger educational infrastructure supports better outcomes. Under Pillar 2 (Poverty and Income Inequality), SIP plays a more significant role in regions with larger poor populations, including Cianjur, Cirebon, Bekasi, and Kuningan. Meanwhile, under Pillar 3 (Expansion of Access and Opportunity), SIP has greater influence in regions characterized by large school-age populations and unequal educational access, such as Cianjur, Sukabumi, Greater Bandung, and Purwakarta. Overall, SIP is most significant in Sukabumi, Kuningan, Majalengka, Garut, Banjar, and Sumedang due to the combined presence of high poverty levels, limited educational opportunities, and ongoing economic transformation. These findings demonstrate that educational assistance programs can function not only as poverty reduction tools but also as strategic instruments for addressing spatial inequality, strengthening human capital, and promoting more inclusive economic growth when implemented through regionally adaptive policy frameworks.