Means-Tested Social Assistance and Domestic Violence Among Older Adults: A Triple-Difference Analysis from Taiwan

Background and Purpose: Existing evidence on whether social assistance reduces family violence has focused mainly on child protection and intimate partner violence, while older adults remain largely overlooked. Important gaps remain in understanding the effects of cash transfer support on older adults. This study leverages a unique opportunity to use administrative data linking program eligibility to officially reported incidents of violence at the individual level. It addresses two research questions: Does improved access to social assistance change the risk of reported domestic violence among older adults in Taiwan? If so, through which subtypes of violence are these effects observed?

Methods: We construct an individual-level longitudinal administrative dataset covering 2014–2022 by linking Taiwan's population registration records, tax-based income and asset files, official low-income assistance rosters, and domestic violence reporting records. Our identification strategy exploits a rule in Taiwan's Social Assistance Act: for adults deemed able to work, the means test imputes earnings regardless of actual employment, but this imputation is lifted at age 65, improving eligibility conditions for households in need. We implement a stacked triple-difference design. The first difference compares the same individual before and after turning 65; the second compares eligible households with ineligible near-poor households whose income and assets fall within twice the eligibility thresholds; The third difference contrasts cohorts that cross the age-65 threshold during the observation period with a younger cohort that reaches age 60, but not 65, over the same period. This design helps account for differential aging effects across eligible and ineligible households that might otherwise confound the estimated effects. All models include individual and calendar-year fixed effects and control for lagged household covariates.

Results: The age-65 eligibility expansion significantly increases realized program receipt, by 5.255 percentage points (s.e. = 0.126) In reduced-form estimates, reported domestic violence victimization rises by 0.057 percentage points (s.e. = 0.020). This increase is not driven by physical or psychological abuse, for which estimates are small and statistically insignificant, but is concentrated in neglect: 0.068 percentage points (s.e. = 0.005). Effects are concentrated among male victims, with overall victimization rising by 0.117 percentage points (s.e. = 0.024) and neglect by 0.113 percentage points (s.e. = 0.009). Female victimization effects are not statistically significant, and perpetrator-side estimates are close to zero.

Conclusions and Implications: We find limited evidence that improved social assistance eligibility changes the risk of physical or psychological abuse among older adults in Taiwan, but we document a robust increase in reported neglect, concentrated among older men. Rather than suggesting that social assistance causes harm, this pattern is consistent with a detection channel: the eligibility transition may connect previously isolated older adults to the welfare and care system, providing a pathway through which preexisting neglect is identified and formally reported. Once they become social assistance recipients, older adults can also access state-subsidized care. These findings suggest that social assistance programs can serve a dual function—alleviating poverty while facilitating the integration of vulnerable older adults into the care system.