The underlying theory behind child support guidelines implies that child support orders should change when the incomes of noncustodial parents change. This paper documents changes in noncustodial fathers' earnings over a five-year period and examines the relationship between the changes in earnings and modifications in child support orders. Using detailed longitudinal administrative data from Wisconsin, the authors examine the history of orders and earnings for fathers in couples who had their first child support ordered in 2000. A substantial proportion of fathers experience large changes in earnings, but relatively few of the associated child support orders are modified. Using discrete-time multinomial event history models that consider time-varying variables and control for censored observations, we find some evidence of changes in earnings being associated with changes in orders, all else equal, but the relationship is relatively weak and order changes are not proportional to earnings changes. The findings highlight the challenges and importance of developing policies that result in child support orders being more responsive to changes in fathers' incomes. (Contains 2 figures, 3 tables and 11 footnotes.)