From an ancient (but readworthy) JPE article by Christopher Sims:
Economists must inevitably try to sort out systematic patterns from random variations in the past-if only because, unassisted, policymakers would do the same thing more naively. In doing so economists will need probabilistic models and statistical methods of inference. Like historians, though, they must accept that a single agreed view of the causal structure of the record they examine will never emerge [Sims, C. A., 1981, What Kind of a Science Is Economics? A Review Article on Causality in Economics by John R. Hicks, The Journal of Political Economy, 89 (3), pp. 582-582].