The research employs district budget data from the Realization of the local government budget from 2014 to 2019, provided by the Ministry of Finance, and toll road data is obtained from the Indonesia Toll Road Authority. As controls, social-economic indicators are sourced from Statistics of Indonesia, BPS.

We collected data at the subdistrict level from 2015 to 2019 and merged the data from different sources. The number of toll road data observations is limited as not all districts have toll roads. Since not all data is available in each district, thus we have unbalanced panel data capturing more than 500 districts from 2015 to 2019. The study also excludes the districts in Special Capital Region of Jakarta province since the districts did not receive the fiscal transfer during the study period, especially the general allocation fund.

Several indicators measure fiscal decentralization, but since the study is concerned with local fiscal independence, we employ the autonomy indicator from 2015 to 2019, when the road infrastructure developed massively. The autonomy indicator is defined by dividing the district's own revenue by the total district revenue, called the fiscal decentralization ratio. The district revenues consist of the district's own revenue, transfer funds, and miscellaneous revenues. The fiscal decentralization ratio ranged from zero to one; the higher the ratio, the more independent the district.

Additionally, we employ alternative measurements of fiscal decentralization that still reflect fiscal autonomy or independence. The first alternative is calculating the ratio of the district's revenue and total transfer funds, called the independence ratio. The total transfer funds include national transfer, provincial transfer, and special autonomy funds. The ratio could be more than one if the district's own revenue is higher than the total transfer funds. A higher independence ratio indicates a more independent district.

Secondly, we calculate the ratio of total district transfer fund to total district revenue, called the dependency ratio. The higher the dependency ratio, the more dependent the district. It is important to highlight that the study does not account for village fund transfers as the funds go directly to the villages, not the districts. Hence, we expect that the effect on the district is limited. We also exclude grants when calculating total transfer funds due to the incomplete information on the grants' sources in the public budget financial report. Yet, the proportion of district grants to total district revenue is insignificant, ranging from 0.75 percent to 3 percent compared to total other transfers (including national and provincial transfers), ranging from 70 percent to 73 percent during 2015-2019.

Our interest variable, road quality, is a share of the length of the good quality road, including district highways and toll roads located in each district, to the total length of district roads and toll roads available in each district (Laborda & Sotelsek, 2019). In this study, highways are free highways under district government authority. We employ district highways as the highways connect the district and local activity centers, including local economic activity, which is expected to affect district independence. We also categorize toll roads as good quality since the Indonesia Toll Road Authority maintains the roads. The higher the value of road quality, the better road quality in the district. Further, the study includes socioeconomic factors as control variables, such as poverty rate, population density, and human development ratio at the district level from 2015 to 2019. The socioeconomic data are obtained from Statistics of Indonesia, BPS. Table 1 summarizes the measurement of variables.

When examining fiscal decentralization and road quality linkage, the research employs a fixed-effect model in Eq.1, following (Gertler et al., 2024) . The study employs a fixed-effects model to address the possibility of a time-invariant unobserved variable. The fixed-effect method could also reduce the endogeneity problems caused by reverse causality when estimating the nexus of road quality and local development, such as fiscal independence (Feng & Wu, 2018).

Yjt = α0 + α1roadqualityjt + α2Xjt + ηj + μj + γt + δjt + ujt (1)

(2)

(3)

The fiscal decentralization ratio of district j in year t is Yjt. The ratio ranges from zero to one, with one being the highest ratio of fiscal decentralization and zero otherwise. Yjt is obtained by dividing the district's own revenue by total district revenue (Eq. 2). roadqualityjt is the share of good quality roads, including district highways and toll roads, to the total roads in district j and year t (Eq.3).

Coefficient α1 estimates the magnitude of the relationship between the share of good quality roads and the fiscal decentralization ratio. Xjt is a vector of potential control variables that might explain fiscal decentralization. Xjt consists of a percentage of poor households or poverty rate, human development ratio, and population density following (Alexeev & Mamedov, 2017); (Canavire-Bacarreza & Martinez-Vazquez, 2012). Further, the study adds several fixed-effect controls, such as district fixed-effect, ηj; provincial fixed-effect, μj; year fixed-effect, γt; and island-time fixed-effect, δjt to isolate the time-invariant unobserved variable.

The study applies the fixed-effect method because of its superior advantages in reducing bias from constant unobserved variables compared to the simple ordinary least square. Including several fixed-effect controls also reduces the endogeneity issue. The endogeneity issue might be sourced from reverse causality, where the district with a higher fiscal decentralization capability could provide better road quality than the district with a lower fiscal decentralization capability.(Alder, 2023) (Mettetal, 2019) The district fixed-effect, provincial fixed-effect, and year-island fixed-effects are advantageous for absorbing districts' different socioeconomic and geographical characteristics. They could also handle the potential local and national policies affecting the relationship between road improvement and fiscal decentralization. The year-fixed effect captures the time trend effect of fiscal decentralization and road quality. All standard errors are clustered at the district level to deal with the potential autocorrelation from external disturbance across the districts that are uncorrelated with road quality but correlated with the fiscal dependency ratio. Therefore, the model estimates potentially close to the causality effect of road improvement on fiscal decentralization.

The study initially estimates the relationship between road quality and fiscal decentralization ratio using an ordinary least square approach with several fixed-effect controls for baseline. We also check the possibility of various effects between regions and islands as regions may have diverse policy-related budget allocation and infrastructure development.

Additionally, the study examines the per capita district's own revenue and per capita district's own revenue annual growth as a potential mechanism to explain the link between road quality and fiscal decentralization ratio. We compute the per capita district's own revenue yearly growth as the difference between the log per capita district's own revenue growth next year and this year for each district. Further, we conduct several robustness checks to test the baseline model consistency. Firstly, the study applies other dependent variables. We use fiscal independence and fiscal dependency ratios. The first is calculated by dividing the district's own revenue by total fiscal transfers, and the latter by dividing the district transfer fund by total district revenue. The study also tests the relationship of each type of road, i.e., the share of the toll road and good quality highways, to the fiscal decentralization to ensure the model's robustness.

Table 3 explains the descriptive statistics of variables. The average fiscal decentralization and independence ratios are minor, meaning most districts still rely on central government transfers. In the meantime, the share of good quality roads is higher than 58 percent. Districts with a relatively higher share of good quality roads also have a higher number of fiscal decentralization and independence ratios or, in other words, more independent than districts with a lower share of good quality roads.

Before we go to the estimation results, we display the relationship between the share of good-quality roads and the fiscal decentralization ratio graphically in Figure 3. Figure 3 and the Pearson correlation coefficient indicate a positive relationship between the share of road quality and the fiscal decentralization ratio. In other words, the district with a higher percentage of good-quality roads has a higher fiscal decentralization ratio.

Our results align with a previous study conducted by (Banerjee et al., 2020) and (Asher & Novosad, 2020), where the road development has no impact on more remote areas. Infrastructure might benefit the whole economy, hence the regional independence. However, the benefits might be limited in regions with no mobility, such as non-Java regions. The absence of a significant disparity between well-connected (city) and poorly connected areas (county) does not eliminate the possibility that infrastructure benefits all regions. However, limited factor mobility hindered the concentration of those benefits in the more well-connected areas.

Road quality also has a significant impact on Java. The dense population, advanced economic activities, and better integration into national and global markets in Java regions augment the benefit of infrastructure. Well-connected transportation networks and facilities in Java enhance resource distribution, support local industries, and attract investments, enabling regions to become more self-sufficient. In contrast, regions outside Java often lack the critical mass of economic activity and population density needed to fully leverage infrastructure, particularly roads. Weaker integration into national and global markets also reduces the capacity of roads to foster self-reliance in these areas.

The dominant effect in Java regions also might be due to the district characteristics and market structure. Road quality improvement may induce the district's competitiveness in the short term. Initially, developed districts have higher productivity, perhaps due to more advanced technology and human capital than less developed districts, thus making them more competitive than less developed districts. The higher connectivity through road quality improvement will increase their market dominance in their district and spread to the less productive districts. Therefore, the developed regions have higher economic benefits than the less developed regions. However, there is potential for long-term technology and human capital convergence because of better connectivity across districts. The existence of convergency will improve the economic integration and distribution between developed and less developed districts.

Second, the size of roads also matters when assessing the relationship between road improvement and the fiscal decentralization ratio. From Table 2, we see that although the percentage improvement of toll roads outside of Java is more significant than in the Java region, the size or length of the total road is much smaller than in Java's districts. Besides, outside Java, the development of the district highway roads connecting the local activity is limited, only growing around 2.5 percent from 2015 to 2019. As a result, the magnitude is higher in the Java region.

Moreover, despite the benefit of toll roads through connectivity, the pricing system of toll roads also potentially creates economic distractions, such as an increase in consumer prices, reduction of demand for input factors, location changes, and an increase in economic and accessibility inequality ((Kleist & Doll, 2005);(Dijk et al., 2015)). Those adverse impacts could crowd out the positive ones, causing a longer time and bigger size for the districts outside Java to yield the net benefit of toll roads.

The study found that road quality links to regional fiscal independence in Indonesia. However, road quality alone does not guarantee economic growth and fiscal independence for regions outside Java. The results underscore the need for integrated development approaches that combine infrastructure with investment in other fields, such as boosting local industries and enhancing factor mobility.

The infrastructure investments must consider local socio-economic contexts. In regions outside Java, where agriculture or resource extraction dominates, good quality roads may have limited association unless paired with efforts to diversify local economies and enhance market access. Better roads might reduce transportation costs, but the benefit remains marginal without sufficient demand for goods and services or the capacity to produce high-value products.

Limited factor mobility, such as labor and capital, in areas outside Java further restricts the impact of road improvements. Improve roads may facilitate travel, but if labor markets are not integrated, skilled workers cannot relocate to regions with better infrastructure, limiting economic growth and regional fiscal independence. Without enabling policies to encourage the flow of resources, the economic benefits of infrastructure improvements remain underutilized.

We want to thank the editor and two anonymous referees for their valuable comments and suggestions, which helped us improve the manuscript's quality.