Causal Contribution

Abstract

Causal language in everyday life and even in policymaking contexts encourages a certain kind of mistake: given some quantifiable outcome (say, the total number of years of life lost in a certain community, over a certain time period), it may be taken for granted that it makes sense to ask, How much of this outcome was due to its various causes? But only in very rare circumstances—when the causal factors in question act additively—is this question well posed. This chapter explains what this additivity requirement amounts to, why it is almost never met, and why an alternative that some have found beguiling—draw on the game-theoretic concept of “Shapley values”—provides no refuge. In discussions of the global burden of disease, the question of what percentage of a given outcome each of its causes “contributed” to that outcome should be rejected as meaningless.