The task of agencies is often considered to be an important determinant of agency design, autonomy, governance and control. The evidence for these expectations has, however, been limited so far. Moreover, there are several problems with the ‘task-variable’. There is no clear definition, agencies often perform multiple tasks, and as task is a nominal variable there are restrictions on the use of statistical techniques. Two questions arise: does task matter and does it matter how task is measured? Using survey data on Dutch agencies (N = 206), several expectations are tested with different techniques. Overall, some tasks have some effect on agency autonomy and control, however, often only indirectly and not sustained in multivariate analyses. Formal autonomy and budget are more decisive than task in explaining agency autonomy and control. Analysis with dummy variables offers better opportunities to test task effects than non-parametrical tests. Researchers should use multiple task categories in future analyses instead of simple dichotomies, not only because agencies perform multiple tasks but also because specific tasks have specific effects.

Points for practitioners

Agencies have multiple tasks in different combinations. This should be taken into account when designing agencies' degrees of autonomy and control. There are no straightforward relations between task and autonomy or control. The type of agency and the size of its budget are most decisive to agency autonomy and control.