How does government expenditure impact sustainable development? Studying the multidimensional link between budgets and development gaps

Abstract

We develop a bottom-up causal framework to study the impact of public spending on high-dimensional and interdependent policy spaces in the context of socioeconomic and environmental development. Using data across 140 countries, we estimate the indicator-country-specific development gaps that will remain open in 2030. We find large heterogeneity in development gaps, and non-linear responses to changes in the total amount of government expenditure. Importantly, our method identifies bounds to how much a gap can be reduced by 2030 through sheer increments in public spending. We show that these structural bottlenecks cannot be addressed through expenditure on the existing government programs, but require novel micro-policies intended to affect behaviors, technologies, and organizational practices. One particular set of bottlenecks that stands out relates to the environmental issues contained in the sustainable development goals 14 and 15.

Subnational sustainable development: The role of vertical intergovernmental transfers in reaching multidimensional goals

“…We develop an agent-based model of the policymaking process that allows linking public spending to policy outcomes (measured through development indicators) under a causal framework. The model is based on Guerrero and Castañeda [21]; who study the feasibility of the SDGs worldwide. Guerrero and Castañeda [22,23] provide a thorough motivation of a closely related model, and an application to the national case of Mexico.…”

“…Thus, the parameters that need calibration are α 1 , …, α N , β, where N is the total number of indicators. The calibration is performed for each state independently by using an algorithm developed by Guerrero and Castañeda [21]. Broadly speaking, the algorithm seeks to find a vector α 1 , …, α N that minimizes the difference between the final value of each empirical indicator, and the ones expected from a set of Monte Carlo simulations.…”

“…Heuristic methods such as differential evolution, and even Bayesian approaches such as the Tree-structured Parzen estimator also fail. The method developed in Guerrero and Castañeda [21] employs a multi-output gradient descent algorithm. The calibration is simultaneous for all parameters, so the method is efficient and minimizes indicator-specific errors; and Guerrero and Castañeda [21] show that it yields high fitness scores.…”

“…The method developed in Guerrero and Castañeda [21] employs a multi-output gradient descent algorithm. The calibration is simultaneous for all parameters, so the method is efficient and minimizes indicator-specific errors; and Guerrero and Castañeda [21] show that it yields high fitness scores. gap is smaller than the retrospective one, i.e.…”

“…where i denotes the indicator and T is the number of years. 21 Thus, the SDG gap for T = 12 can be interpreted as the gap-in percentage terms-that has not yet been closed in 2030 for indicator i in a given state.…”