Using a panel dataset from Burundi where information on protection payments during the 10 year civil war were collected, we test the relationship between payments, the nature of extraction by the rebels, and welfare outcomes. We ask, does payment to rebels insure against future welfare loss and does the nature of payment matter? Specifically, does the level of institutionalisation of extraction within the rebel governance structure provide a form of insurance for future welfare? No less than 30% of the interviewees made at least one payment. Rebels extract these taxes through one of two routes: an‘institutionalised’ and regular cash-with-receipt method or an ad hoc and unpredictable labour extraction. Using matching methods we find that payment through the institutionalised route increases household welfare between 16 and 25%. Ad hoc extraction has no effect. We situate our findings in the empirical literatures on contributions to mafia- type organisations and rebel governance.