Poverty traps are rare, but trappedness isn't

Abstract

The persistence of poverty is not well explained by who is poor. We argue the relevant object of measurement is trappedness--expected escape time from deprivation--which varies systematically across institutional environments and is invisible to standard poverty indices. Using Markov chains estimated on twenty years of longitudinal data from 27 European countries, we show that countries with identical deprivation rates differ in escape times by up to fourfold. These differences are not explained by household characteristics alone: exogenous shocks reshape welfare landscapes differently across countries, with divergence tracking welfare regime architecture rather than household composition. The mechanism is behavioural: health constrains a household's capacity to convert income gains into durable welfare improvement. Income transfers without health improvement fail to reduce poverty-return risk; combined interventions are super-additive across 28 countries, and the gap widens with transfer size. These findings dissolve the long-running poverty trap debate--studies that rejected traps measured the wrong dimension; studies that found them captured one projection of a multidimensional dynamic process. Trappedness is continuous, multidimensional, and institutionally shaped.

Figures (13)

• Figure 1: From deprivation traps to trappedness. Static and unidimensional approaches treat traps as fixed objects of the environment -- i.e. the returns to investment. Static multidimensional indices capture joint deprivation -- with somewhat arbitrary thresholds -- but often see persistence as binary. Mobility matrices capture dynamics, often intergenerational income mobility, but only in a single dimension. A dynamic, multidimensional view instead represents welfare as an evolving landscape, where barriers and escape times vary continuously across income, health, and education—motivating trappedness paired with deprivation as the relevant object of measurement.
• Figure 2: Multidimensional social mobility landscapes in the Netherlands.(a) Potential landscape over income showing multiple equilibria: stable (filled) and unstable (open) fixed points. Shaded band: bootstrap uncertainty. (b) Two-dimensional landscape over income and health. Darker regions are deeper basins; contour lines denote equipotential boundaries for MFPT calculation. White arrow: stylised optimal pathway. (c) Expected escape time from poverty versus AROPE rate (2019, 27 EU countries). Highlighted pairs share near-identical deprivation rates yet differ in escape times by up to 4×—the two measures capture distinct properties.
• Figure 3: COVID-19 reshapes social mobility landscapes differently across countries.(a)--(c) Pre-COVID (2003--2019, solid) and during-COVID (2019--2024, dashed) landscapes over income and health for the Netherlands, Poland, and France. Contour lines: equipotential boundaries for MFPT calculation; shaded regions: deepest basins. Basin contraction and displacement show that COVID differentially altered mobility pathways—lower-income groups faced steeper barriers while higher equilibria remained broadly stable. (d) Expected recovery times to pre-pandemic steady state, obtained by perturbing the pre-COVID steady state with the COVID-era transition matrix. Countries in white lacked data for the full period. (e) Normalised net mobility change during COVID; negative values indicate net downward mobility.
• Figure 4: Health constrains the returns to income transfers.(a) Five-year poverty-return risk reduction $(\Delta R )$ for income-only, health-only, and combined interventions across 28 European countries. Households are moved from the lowest welfare state to the 25th percentile in each dimension. Countries ordered by super-additivity (grey tick): the degree to which the combined effect exceeds the sum of its parts. (b) Retention probability above median welfare after income boosts of varying size, stratified by starting health quintile. Better initial health yields higher retention at every transfer size, and the gap widens as transfers grow. Shaded bands: $\pm$1 s.d. across countries.
• Figure 5: A 2-dimensional economic landscape We define an economic landscape as a multidimensional, dynamic representation of the distribution and transitions of welfare states within a population over time, integrating key dimensions such as income, education, and health. It is conceptualized as a topographical map of social and economic opportunity, where stable welfare equilibria (red cross) constitute “basins,” and barriers or poverty traps form ridges and valleys. Transitions between these states, modelled as stochastic processes via Markov chains, capture the temporal and synergistic interactions shaping upward and downward mobility. Navigating between different fixed points (red crosses) raises the question of optimal paths. Alleviation pathways (red arrows) out of poverty focusing on only one dimension might neglect more cost-effective pathways. The methodological approach: longitudinal survey micro-data is binned into discrete states, transition matrices are estimated and subsequently analysed for resulting dynamics.

Theorems & Definitions (1)

• proof