Asymmetric Housing Co-adaptation Reshapes Neighborhood Recovery Trajectories after Hurricanes

Abstract Post-hurricane housing co-adaptation is a collective process in which household socioeconomic conditions, housing damage, and past activities interact with household-to-household influences to shape repair and/or sell decisions and further change community recovery trajectories. Existing research provides limited process-level understanding, constraining when and where recovery resources can be effectively targeted. We develop the Household-level Decision Diffusion (HDD) framework to reconstruct and forecast dynamic housing co-adaptation processes for the first time. We evaluate HDD by constructing a first-of-its-kind longitudinal dataset linking damage, housing and household conditions, time-series repair permits and time-series property transactions for more than 1.3 million households in nearly 2,800 communities after Hurricanes Harvey, Ida and Ian. HDD recovered latent neighborhood structure with higher edge precision than distance-based or classical Hawkes reconstructions, and forecasted both future repair and sale with AUC value of upto 0.94 and 0.82, respectively, significantly improving both decision prediction accuracy compared to state-of-the-art models and providing a process-level basis for effective recovery policy design. The revealed dynamics show that same-type neighborhood associations were positive for both decisions in all three events, but repair reinforcement was bounded and early while sale reinforcement accumulated across the observation window. Cross-type coupling was asymmetric in both magnitude and sign: the sale-to-repair pathway was more strongly coupled than the reverse in 69--82\% of communities, yet its realized effect was negative, with a greater sale history among connected neighbors associated with a lower probability of repair in almost every community after Ian and in most after Ida. Neighborhood influence is therefore better described as a two-channel process than as a single social multiplier: within-channel histories reinforce continued activity, whereas the exit-to-investment channel inhibits it. Neighborhood influences extended to selling among undamaged households as well. Recovery policy should therefore concentrate repair support in the early window, sustain exit-related monitoring long after it closes, avoid delivering buyouts and repair assistance into the same community at once, and assess eligibility at community rather than parcel scale.