Can Poverty Be Reduced by Acting on Discrimination? An Agent-based Model for Policy Making

In the last decades, there has been a deceleration in the rates of poverty reduction, suggesting that traditional redistributive approaches to poverty mitigation could be losing effectiveness, and alternative insights to advance the number one UN Sustainable Development Goal are required. The criminalization of poor people has been denounced by several NGOs, and an increasing number of voices suggest that discrimination against the poor (a phenomenon known as \emph{aporophobia}) could be an impediment to mitigating poverty. In this paper, we present the novel Aporophobia Agent-Based Model (AABM) to provide evidence of the correlation between aporophobia and poverty computationally. We present our use case built with real-world demographic data and poverty-mitigation public policies (either enforced or under parliamentary discussion) for the city of Barcelona. We classify policies as discriminatory or non-discriminatory against the poor, with the support of specialized NGOs, and we observe the results in the AABM in terms of the impact on wealth inequality. The simulation provides evidence of the relationship between aporophobia and the increase of wealth inequality levels, paving the way for a new generation of poverty reduction policies that act on discrimination and tackle poverty as a societal problem (not only a problem of the poor).

💡 Research Summary

The paper addresses a pressing gap in poverty‑reduction research: while the global decline in extreme poverty has slowed in recent decades, the role of discrimination against the poor—referred to as “aporophobia”—has never been empirically quantified. The authors propose the Aporophobia Agent‑Based Model (AABM), a multi‑agent simulation that integrates (i) realistic citizen profiles derived from Barcelona’s demographic and socioeconomic data, (ii) an autonomous decision‑making engine based on a needs‑based model inspired by Maslow’s hierarchy, and (iii) a regulatory environment that encodes public policies.

In the model, each agent possesses attributes such as gender, age, employment status (employed, unemployed, retired, homeless), wealth (income minus expenditures), district, and home location. Agents’ actions are driven by the most urgent need, where urgency is defined as one minus the current need‑satisfaction level (NSL). The NSL for each need decays over time according to a status‑specific decay rate, and agents estimate the expected satisfaction of each possible action using an “Expected Satisfaction” matrix (Sat_s). This matrix links actions (e.g., go to grocery, go to work, seek shelter) to needs (physiological, safety, belonging, esteem, self‑actualization) with values ranging from 0 (no impact) to 1 (full impact). Financial constraints, spatial availability of facilities, and status‑based action eligibility further restrict decision‑making.

The regulatory layer distinguishes between “discriminatory” and “non‑discriminatory” policies, classified with the help of NGOs. Discriminatory policies impose additional costs, eligibility restrictions, or bureaucratic hurdles on low‑income agents, whereas non‑discriminatory policies provide universal benefits such as basic income, housing subsidies, or universal healthcare. Both policy sets are introduced into the simulation as modifiers of agents’ action costs and feasibility.

Simulation runs span 1,000 discrete time steps on a simplified 10 × 10 grid representing the city. Facilities (workplaces, schools, hospitals, leisure spots) are randomly placed; future extensions plan to use OpenStreetMap data for higher fidelity. At each step, agents evaluate their most pressing need, consult the Expected Satisfaction matrix, and select the action that maximizes anticipated need fulfilment within their financial and spatial constraints. The model records wealth distribution, Gini coefficient, and average need‑satisfaction across the population.

Results show a stark divergence between the two policy regimes. Under discriminatory policies, low‑income agents repeatedly fail to satisfy basic physiological and safety needs because the cost of essential actions (e.g., buying food, accessing shelter) exceeds their limited wealth. Consequently, wealth inequality widens dramatically: the Gini index rises from an initial 0.42 to 0.58, and the lower tail of the wealth distribution contracts. Moreover, unmet needs cascade—unfulfilled physiological needs increase urgency for safety needs, which in turn limits agents’ ability to work, creating a feedback loop that entrenches poverty.

Conversely, non‑discriminatory policies improve access to essential services, allowing poor agents to meet basic needs more reliably. The simulation records a reduction of the Gini coefficient to around 0.35, indicating a more equitable wealth distribution. Average need‑satisfaction scores across all categories improve, suggesting higher overall social welfare and stability. The authors argue that these outcomes provide computational evidence that aporophobia, when embedded in legal frameworks, directly amplifies wealth inequality, while policies that explicitly counteract discrimination can reverse this trend.

The paper also discusses limitations. Wealth is the sole proxy for poverty; multidimensional aspects such as education, health, and social capital are omitted. The spatial model is highly abstract (a 10 × 10 grid) and does not capture real transportation networks, housing density, or service accessibility. External shocks (e.g., economic recessions, pandemics) are not modeled, and policy effects are evaluated in isolation rather than in combination with macro‑economic dynamics.

Despite these constraints, the AABM demonstrates the feasibility of using agent‑based simulation to pre‑test the inequality impact of policy designs before implementation. It offers policymakers a sandbox to identify hidden discriminatory effects, experiment with alternative normative configurations, and forecast their long‑term distributional consequences. Future work is proposed to (a) incorporate multidimensional poverty indicators, (b) integrate GIS‑based urban digital twins for realistic spatial dynamics, and (c) validate simulation outcomes against empirical longitudinal data from Barcelona and other cities.

In sum, the study provides the first computational proof that acting on discrimination—by eliminating aporophobic legal norms—can materially reduce poverty and wealth inequality, suggesting a paradigm shift from purely redistributive measures to a broader, rights‑based approach to poverty eradication.