Carrying the Weight of Autonomy? Growth and Spatial Inequality Among Municipalities in Colombia
- Carlos Felipe Holguin Isaza
- Jul 26
- 4 min read
Carlos F. Holguín-Isaza & Carlos A. Mesa-Guerra
Relevance of this study.
Although decentralization reforms have been a prominent policy instrument over the past fifty years, their actual impact remains contested—particularly in developing countries, where structural deficiencies often obscure the effects of decentralization on local living conditions and development outcomes.
Theoretically, both political and fiscal decentralization are expected to enhance welfare. Drawing on microeconomic choice theory and endogenous growth theory, the fiscal federalism literature argues that local governments are best positioned to identify and respond to citizens' needs. Provided that these governments have sufficient autonomy to raise revenues and allocate expenditures, the provision of public goods should become more efficient (Tiebout, 1956; Musgrave, 1959; Oates, 1999).
The fiscal federalism theory has been tested primarily through cross-country studies that examine the relationship between various measures of fiscal decentralization and national economic growth. While this approach offers valuable insights supporting the theory's predictions, it falls short in assessing whether these effects hold at the subnational level—where, in theory, the benefits of decentralization should be directly observable.
In light of Colombia's significant fiscal decentralization over the past three decades, this study aims to assess whether these reforms have impacted municipal growth dynamics. Theoretically, we build on the foundations of fiscal federalism and integrate them into the endogenous growth model. Empirically, we construct a novel panel dataset that combines municipal fiscal indicators (including both tax revenues and public expenditures) with value-added estimates generated through a machine learning model trained on night-time lights, terrain features, and socio-demographic characteristics.
¿How did we do it?
Measuring fiscal decentralization isn't as straightforward as one may think. It entails a complex process of transferring decision-making powers to subnational governments over the collection and allocation of resources.
To analyze fiscal decentralization in Colombia, we focused on how much autonomy local governments have over raising and spending public funds. Using detailed municipal data compiled since the 1980s, we constructed two main indicators:
Decentralization through higher own revenues: the share of locally raised taxes in total revenue (excluding royalties, grants, and loans), which reflects each municipality's fiscal capacity.
Decentralization through expenditure autonomy: the share of total expenditures funded by Freely Disposable Current Income—a measure published annually by the Comptroller General—indicates the degree of discretion over spending decisions.
Higher values in both reflect stronger effective decentralization. As robustness checks, we used three complementary indicators:
Non-earmarked tax revenues.
Transfer independence (the share of revenues not tied to national transfers or royalties).
Municipal independence, a binary variable indicating whether a municipality is certified to manage education and health funds directly.
How we measured population-adjusted income in Colombian Municipalities
To estimate municipal value added (income) across Colombia from 1992 to 2020, we trained an artificial neural network (ANN) using satellite-based Night-Time Lights (NTLs) data and other geo-referenced inputs. This method builds on recent literature showing the strong predictive power of machine learning models—especially ANNs—in capturing non-linear relationships between light emissions and economic activity. By leveraging a harmonized NTLs series from DMSP-OLS and VIIRS satellites (Li et al., 2018), alongside annual data on land cover, oil production, and population, the model estimates each municipality's share of national value added.
The ANN was trained using the only available years with official municipal income data (2011–2018), and its predictions were transformed into per capita value added by redistributing national income based on predicted shares. To ensure accuracy, the model excluded departments with bottom-coded satellite pixels or high data noise. Despite limitations—such as the inability to fully capture structural changes before 2011—the model achieved high predictive performance (R² ≈ 0.99), enabling a consistent estimation of municipal income over nearly three decades.
Our empirical strategy
To assess the impact of fiscal decentralization on municipal income growth in Colombia, we estimate a growth regression where the key variable is the initial level of per capita income—capturing the idea of conditional convergence. The model incorporates variables from endogenous growth theory, including national income, public consumption, and human capital (proxied by average years of schooling). To address endogeneity—especially the possibility that growth affects decentralization—we use three instrumental variables: (1) four-year lags of decentralization measures, (2) the share of exogenous intergovernmental transfers determined by national formulas, and (3) the timing of cadastral updates, which affects municipal revenue via property taxes. We exclude municipal fixed effects to avoid upward bias in the convergence rate, and control for a range of demographic, geographic, and institutional factors. Finally, because value added is estimated using satellite-based proxies, we instrument current income with its four-year lag to reduce measurement error.
Key results
Our findings show that revenue-based fiscal decentralization significantly accelerates income convergence. Annual convergence rates are higher than Barros's "iron law of convergence", ranging from 2.7% to 4.2%, while measures of decentralization are associated with a direct increase in per capita income growth. Based on these estimates, it takes roughly 30 years for the poorest municipality—based on the levels observed in 2000—to close half of its income gap with a richer one. Now, expenditure-based decentralization measures—especially those reflecting greater autonomy over spending decisions—are less robust and show no significant effect. The convergence-enhancing effect of fiscal decentralization is strongest among poorer municipalities, where public capital is likely to be more productive. These results provide new empirical evidence on the developmental role of fiscal decentralization in a highly unequal, decentralized context. Our findings suggest that effective local revenue autonomy, rather than mere delegation of spending responsibilities, is key to fostering spatial equity and long-run convergence.
Read the full paper
Comments