This study aims to identify the main areas and objectives of smart city procurements and to determine the spatial and socio-economic factors influencing smart city development. Using data on all successful procurements in Poland in years 2017-2021, we propose a novel methodological approach to identifying smart city investments based on neural topic modelling (BERTopic).

We found that smart city development in Poland remains at an early stage, with a predominant focus on hardware and software purchases that support public services. That smart city investments are driven by municipalities’ financial capacity and size, residents' age and education levels, and the quality of the institutional environment. Although smart city investments are concentrated in major cities, we observe diffusion into surrounding metropolitan areas and into more peripheral regions.

This study provides valuable policy insights and demonstrates the potential of using micro-data and text-mining techniques to analyze local public investments.

We examine whether municipal childcare expansions generate net growth or merely reallocate childrearing households across municipalities. We study Japan’s Akashi City, which sharply expanded child medical-care and childcare-fee subsidies in the 2010s, widening benefit gaps with nearby municipalities. Using a Difference-in-Differences design—Akashi as the positive treatment group, adjacent wards in Kobe as the negative treatment group, and Himeji (a nearby city in a different commuting zone) as the control—we find preschool-related outcomes rise in Akashi but fall on the Kobe side; when the two treated areas are pooled, the average effect is indistinguishable from zero. A Geographic Regression Discontinuity design at the Akashi–Kobe boundary reveals a post-reform discontinuity, driven mainly by declines on the Kobe side near the border. Overall, the evidence points to cross-border spillovers and suggests that, at the broader Akashi–Kobe scale, Akashi’s gain may reflect a reallocation rather than a net increase.

This article explores the application of machine learning models to identify financial difficulties in Polish local governments from 2010 to 2023. Poland's current public finance monitoring is reactive and lacks effective forecasting, complicating early detection of fiscal risks. The study examined 2,375 municipalities using a dataset of 33 financial, demographic, and macroeconomic variables. Four classification models were compared: logistic regression, LASSO regression, Random Forest, and XGBoost. Results indicated that XGBoost performed best, achieving an ROC AUC of 0.837 and an average precision of 0.723. SHAP value analysis identified operating profit per capita from the previous year and debt indicators as key predictors. The study highlights that these machine learning models can aid audit institutions, such as Regional Audit Chambers, in developing modern early-warning systems, enabling auditors to concentrate resources on entities in genuine financial distress, despite the challenges posed by sudden regulatory changes.

This paper analyzes the local economic impacts of military troop deployments. We exploit variation from the historic large-scale US troop withdrawal from Germany triggered by the end of the Cold War, to estimate the effect on local labor markets and local public finances. We use administrative data to precisely quantify the size of the troop withdrawal at the municipal level. Using a synthetic difference-in-differences estimator, we find negative effects on local labor markets. The decrease in economic activity results in a reduction of revenues for affected municipalities. To balance decreased revenues, affected municipalities lower their expenditures, while increasing business and property tax multipliers. We estimate the cost per job to be USD 110,400 of US military spending. Our worker-level analysis reveals that workers displaced by the closure of a US military base have persistently lower employment rates 15 years after the withdrawal.