Nikolas

Abstract. Identifying most influential sets (MIS) – size-k subsets whose removal maximally changes a target estimand – is typically infeasible because it requires searching over n choose k subsets. We show that, for a broad class of estimands whose leave-set-out effect admits a linear-fractional form, the MIS problem reduces to a one-parameter sequence of top-k selections. Using Dinkelbach's method, we obtain an efficient algorithm that runs in O(n) per iteration and terminates in finitely many steps. We show that our approach returns globally optimal sets for univariate settings, such as average treatment effect estimation in randomized experiments. For partial linear models, we establish selection consistency under Neyman orthogonality and mild first-stage stability. We validate our method through simulations and real-world applications – recovering MIS that were previously computationally inaccessible.

Abstract. Small subsets of data with disproportionate influence on model outcomes can have dramatic impacts on conclusions, with a few data points sometimes overturning key findings. While recent work has developed methods to identify these most influential sets, no formal theory exists to determine when their influence reflects genuine problems rather than natural sampling variation. We address this gap by developing a principled framework for assessing the statistical significance of most influential sets. Our theoretical results characterize the extreme value distributions of maximal influence and enable rigorous hypothesis tests for excessive influence, replacing current ad-hoc sensitivity checks. We demonstrate the practical value of our approach through applications across economics, biology, and machine learning benchmarks.

Abstract. Mining operations in Africa are expanding rapidly, creating negative externalities that remain poorly understood. In this paper, we provide causal evidence for the impact of water pollution from mines on vegetation and agriculture across the continent. We exploit a natural experiment, where mines cause a discontinuity in water pollution along river networks, comparing vegetation health in upstream and downstream locations. We find that mines significantly reduce peak vegetation indices downstream by 1.3–1.5%, and impair the productivity of over 74,000 km² of croplands. Impacts may reach farther downstream, and are particularly strong in fertile regions and areas where gold mining predominates. Our findings highlight substantial externalities of mining and an urgent need for enhanced regulation and oversight to mitigate and monitor them.

Figure: Illustration of the vegetation discontinuity from mine sites along river basins.

Abstract. Mining provides crucial materials for the global economy and the climate transition, but can have severe environmental and social impacts. Current analyses of these impacts are limited by a lack of data on mining activity, particularly in the regions most affected. In this paper, we present a novel panel dataset mapping mining sites along the tropical belt from 2016 to 2024. Our approach uses a machine learning model, trained on over 25,000 mining polygons from the literature, to automatically segment mining areas in high-resolution (<5 m) satellite imagery. The dataset maps over 145,000 mining polygons covering an average area of 65,000 km² annually, with an accuracy of 87.7% and precision of 84.1%. Our approach allows for accurate, precise, and consistent delineation, and can be scaled to new locations and periods. The dataset enables detailed analyses of local environmental, social, and economic impacts of mining in regions where conventional data is scarce or incomplete.

Figure: Segmentation of the Toka Tindung gold mine in Indonesia over time (Planet/NICFI).

Abstract. Brazil once set the example for curtailing deforestation with command and control policies, but, in the last decade, these interventions have gone astray. Environmental research and policy today are largely informed by the earlier successes of deforestation interventions, but not their recent failures. Here, we investigate the resilience of deforestation interventions. We discuss how the recent trend reversal in Brazil came to be, and what its implications for the design of future policies are. We use newly compiled information on environmental fines in an econometric model to show that the enforcement of environmental policy has become ineffective in recent years. Our results add empirical evidence to earlier studies documenting the erosion of the institutions responsible for forest protection, and highlight the considerable deforestation impacts of this erosion. Future efforts for sustainable forest protection should be aimed at strengthening institutions, spreading responsibilities, and redistributing the common value of forests via incentive-based systems.

Abstract. Growing demand for minerals continues to drive deforestation worldwide. Tropical forests are particularly vulnerable to the environmental impacts of mining and mineral processing. Many local- to regional-scale studies document extensive, long-lasting impacts of mining on biodiversity and ecosystem services. However, the full scope of deforestation induced by industrial mining across the tropics is yet unknown. Here, we present a biome-wide assessment to show where industrial mine expansion has caused the most deforestation from 2000 to 2019. We find that 3,264 km2 of forest was directly lost due to industrial mining, with 80% occurring in only four countries: Indonesia, Brazil, Ghana, and Suriname. Additionally, controlling for other nonmining determinants of deforestation, we find that mining caused indirect forest loss in two-thirds of the investigated countries. Our results illustrate significant yet unevenly distributed and often unmanaged impacts on these biodiverse ecosystems. Impact assessments and mitigation plans of industrial mining activities must address direct and indirect impacts to support conservation of the world's tropical forests.

Abstract. Bayesian approaches play an important role in the development of new spatial econometric methods, but are uncommon in applied work. This is partly due to a lack of accessible, flexible software for the Bayesian estimation of spatial models. Established probabilistic software struggles with the specifics of spatial econometrics, while classical implementations do not harness the flexibility of Bayesian modelling. In this paper, I present a layered, objected-oriented software architecture that bridges this gap. An R implementation in the bsreg package allows quick and easy estimation of spatial econometric models, while remaining maintainable and extensible. I demonstrate the benefits of the Bayesian approach and using a well-known dataset on cigarette demand. First, I show that Bayesian posterior densities yield better insights into the uncertainty of non-linear models. Second, I find that earlier studies overestimate spillover effects for distance-based connectivities due to a scaling error, highlighting the need for tried and tested software.

Abstract. Vector autoregression (VAR) models are widely used for multivariate time series analysis in macroeconomics, finance, and related fields. Bayesian methods are often employed to deal with their dense parameterization, imposing structure on model coefficients via prior information. The optimal choice of the degree of informativeness implied by these priors is subject of much debate and can be approached via hierarchical modeling. This paper introduces BVAR, an R package dedicated to the estimation of Bayesian VAR models with hierarchical prior selection. It implements functionalities and options that permit addressing a wide range of research problems, while retaining an easy-to-use and transparent interface. Features include structural analysis of impulse responses, forecasts, the most commonly used conjugate priors, as well as a framework for defining custom dummy-observation priors. BVAR makes Bayesian VAR models user-friendly and provides an accessible reference implementation.

Abstract. Deforestation of the Amazon rainforest is a threat to global climate, biodiversity, and many other ecosystem services. In order to address this threat, an understanding of the drivers of deforestation processes is required. Spillover effects and factors that differ across locations and over time play important roles in these processes. They are largely disregarded in applied research and thus in the design of evidence-based policies. In this study, we model connectivity between regions and consider heterogeneous effects to gain more accurate quantitative insights into the inherent complexity of deforestation. We investigate the impacts of agriculture in Mato Grosso, Brazil, for the period 2006–2017 considering spatial spillovers and varying impacts over time and space. Spillovers between municipalities that emanate from croplands in the Amazon appear as the major driver of deforestation, with no direct effects from agriculture in recent years. This suggests a moderate success of the Soy Moratorium and Cattle Agreements, but highlights their inability to address indirect effects. We find that the neglect of the spatial dimension and the assumption of homogeneous impacts lead to distorted inference. Researchers need to be aware of the complex and dynamic processes behind deforestation, in order to facilitate effective policy design.

Abstract. In a recent study, Chaves et al. find international consumption and trade to be major drivers of ‘malaria risk’ via deforestation. Their analysis is based on a counterfactual ‘malaria risk’ footprint, defined as the number of malaria cases in absence of two malaria interventions, which is constructed using linear regression. In this letter, I argue that their study hinges on an obscured weighting scheme and suffers from methodological flaws, such as disregard for sources of bias. When addressed properly, these issues nullify results, overturning the significance and reversing the direction of the claimed relationship. Nonetheless, I see great potential in the mixed methods approach and conclude with recommendations for future studies.

Abstract. Harvested biomass is linked to final consumption by networks of processes and actors that convert and distribute food and nonfood goods. Achieving a sustainable resource metabolism of the economy is an overarching challenge which manifests itself in a number of the UN Sustainable Development Goals. Modeling the physical dimensions of biomass conversion and distribution networks is essential to understanding the characteristics, drivers, and dynamics of the socio-economic biomass metabolism. In this paper, we present the Food and Agriculture Biomass Input–Output model (FABIO), a set of multiregional supply, use and input–output tables in physical units, that document the complex flows of agricultural and food products in the global economy. The model assembles FAOSTAT statistics reporting crop production, trade, and utilization in physical units, supplemented by data on technical and metabolic conversion efficiencies, into a consistent, balanced, input–output framework. FABIO covers 191 countries and 130 agriculture, food and forestry products from 1986 to 2013. The physical supply use tables offered by FABIO provide a comprehensive, transparent, and flexible structure for organizing data representing flows of materials within metabolic networks. They allow tracing of biomass flows and embodied environmental pressures along global supply chains at an unprecedented level of product and country detail and can help to answer a range of questions regarding environment, agriculture, and trade. Here we apply FABIO to the case of cropland footprints and show the evolution of consumption-based cropland demand in China, the E.U., and the U.S.A. for plant-based and livestock-based food and nonfood products.

Abstract. Spillover effects between regions are common in deforestation and environmental economics. Yet, data on the networks behind them is elusive, and empirical analyses rely on proxies and assumptions. In this paper, I present a hierarchical approach to jointly model spillovers and the latent networks driving them. I apply this approach to investigate the deforestation impact from blacklisting municipalities in the Brazilian Amazon. I find large positive spillovers from the intervention that are underestimated considerably when using conventional spatial proxies. Results further suggest that endogenous spillovers may lead to upward bias when assessing deforestation interventions. My approach is widely applicable, and its flexibility can improve our understanding of spillovers by revealing the networks behind them and enabling targeted, effective interventions.

Figure: The Brazilian 'Legal Amazon' and areas that were cleared over 2004–2022.

Figure: Estimates of centrality from a model with individual specific connectivity.

Abstract. The sensitivity of econometric results is central to their credibility. In this paper, we investigate the sensitivity of regression-based inference to influential sets of observations and show how to reliably identify and interpret them. We explore three algorithmic approaches to analyze influential sets, and assess the sensitivity of a number of earlier studies in the field of development economics to them. Many results hinge on small influential sets, and inspecting them can provide crucial insights. The analysis of influential sets may reveal omitted variable bias, unobserved heterogeneity, lacking external validity, and informs about technical limitations of the methodological approach used.

Figure: Illustration of influential sets, joint influence, and masking.

Figure: Application of our approach and an illustration for the seeming differential effects of ruggedness in Africa.

Abstract. Demand for agricultural products is a major driver of deforestation in the Brazilian Amazon. However, the extent of their deforestation impact is contested, as deforested land is relatively unproductive, and many products are barred from agriculture supply chains. In this paper, we quantify the deforestation impacts of expanding agricultural production, differentiating it from other channels with different implications for economic and environmental policy. We use a shift-share design, exploiting international changes in beef consumption to causally identify the deforestation impact of agricultural demand. We find that pasture and cattle herd expansions are major direct drivers of deforestation. Their direct impacts diminished during the recent deforestation boom, suggesting that land speculation motives have become more important. Our findings indicate that intensification and improved land tenure security could help decrease land pressure, but also highlight that deforestation interventions need to target the dominant role of agricultural production.

Abstract. Malaria places a great burden on the health and prosperity of many and occupies a great number of scientists and policymakers. The dynamics of the disease are tightly interwoven with economics — incidence is both tied to economic circumstances and impacts them. Economic research plays an important role in understanding and supporting the fight against malaria. The economic literature, however, features a number of peculiarities that can hamper accessibility and has been slow to approach interdisciplinary issues. In this chapter, we explain the economic perspective and summarise the literature on the economic impacts of malaria. Malaria has severe impacts on individual and aggregate economic outcomes, including mortality and morbidity, but also indirect burdens that materialise with a delay. The fight against malaria is not an economic policy per se, but may provide beneficial economic spillovers and can be vital in establishing an environment that allows for prosperity. Economic insights can make a difference in the design and implementation of effective and efficient eradication and control strategies. This is critical in the light of increasing disease (re-)exposure due to climate change and the emergence of resistant vectors and pathogens.