Karina Acosta (p)

Inmaculada Alvarez

Colombia is characterized by its regional disparities. The wealthiest territories have income per capita around six times that observed in the least developed territories. One strategy implemented by Colombia to reduce such economic disparities is through an asymmetric decentralization system implemented since the Constitution of 1991. The national government uses a typological classification of the municipalities in the country to assort financial allocations across the country. The aim of these monetary transfers is to gain local autonomy and minimize the social and economic disparities between areas. Yet, within the last decades, regional disparities are still wide. We argue the current norm of decentralization has been sufficient and justified to reduce the important inequalities observed within the country. This paper takes advantage of new advances in machine learning and an enhanced information system in Colombia to propose an interpretable clustering strategy to bundle municipalities based on more than fifty annual territorial variables. We follow the proposal of Bertsimas et al. (2020), which proves to be superior to the standard k-means strategies.

Cristiano Ricci (p), Angela Parenti, Davide Fiaschi

Karina Acosta

In this paper, we analyse the spatial distribution dynamics of the population density of Italian municipalities over the period 1984-2019. Firstly, we refine the standard Moran-based classification, using as additional dividers the bisector and the estimated nonparametric Moran's I. The proposed taxonomy resembles other classifications of municipalities into urban and rural but has the advantage of being based on a very limited amount of information that is only population density and the definition of Local Labour Areas (LLA). This allows us to get a taxonomy for each year and study its dynamic over time. Moreover, we are also able to study the evolution of municipalities in continuous state space without relying on discrete taxonomy, therefore providing a more comprehensive understanding of the historic track that led to the current configuration. Our findings show the presence of three dynamic attractors, an urban attractor, a suburban attractor and a rural attractor, where all municipalities and their LLA are converging.

Yigong Hu (p)

Cristiano Ricci

Spatial heterogeneity is a typical and common form of spatial effect which refers to the uneven distribution in geographical entities and their relationships. Geographically Weighted Regression (GWR) and its extensions, including Multiscale GWR (MGWR), are important local modelling techniques in exploring spatial heterogeneity based on data borrowing. When dealing with spatial data of overlapping samples, GWR-based models would encounter several problems, such as tremendous variations in bandwidths. As data of this characteristic have spatial hierarchical structures (i.e., they have group-level and sample-level variables), Hierarchical Linear Modelling (HLM) is suitable to deal with them. But a problem would occur that for some position-related fixed effects, spatial heterogeneity is missing in their estimations. In this paper, we are going to propose a model combining GWR and HLM, called HGWR. It divides coefficients into three types: local fixed effects, global fixed effects, and random effects. Correspondingly, we also propose a back-fitting maximum likelihood solution to it. Results of a simulation experiment and a robust check experiment show that this model could successfully distinguish local fixed effects from others. Furthermore, the spatial heterogeneity is reflected in estimations of local fixed effects, together with the spatial hierarchical structure in other effects. Compared with others, which are only good at fitting some types of effects, HGWR produces estimations of lowest deviations no matter which coefficients are active. For big data, although MGWR could produce comparable results, HGWR is more efficient.

Inmaculada Alvarez (p), Luis Orea, Alan Wall

Yigong Hu

We use a stochastic frontier analysis (SFA) approach to model the propagation of the COVID-19 epidemic across geographical areas. The proposed models permit reported and undocumented cases to be estimated, which is important as case counts are overwhelmingly believed to be undercounted. The models can be estimated using only epidemic-type data but are flexible enough to permit these reporting rates to vary across geographical cross-section units of observation. We provide an empirical application of our models to Spanish data corresponding to the initial months of the original outbreak of the virus in early 2020. We find remarkable rates of under-reporting that might explain why the Spanish Government took its time to implement strict mitigation strategies. We also provide insights into the effectiveness of the national and regional lockdown measures and the influence of socio-economic factors in the propagation of the virus.