Past research suggests that spouses influence one another to vote, but it relies almost exclusively on correlation in turnout. It is therefore difficult to establish whether spouses mobilize each other or tend to marry similar others. Here, we test the dependency hypothesis by examining voting behavior before and after the death of a spouse. We link nearly six million California voter records to Social Security death records and use both coarsened exact matching and multiple cohort comparison to estimate the effects of spousal loss. The results show that after turnout rates stabilize, widowed individuals vote nine percentage points fewer than they would had their spouse still been living; the results also suggest that this change may persist indefinitely. Variations in this “widowhood effect” on voting support a social-isolation explanation for the drop in turnout.