Suppose that we are worried about local government corruption by mayors. The President of Brazil has been reading some behavioral economics literature and thinks that asking mayors to sign a pledge that they will not engage in corruption could reduce corruption levels. Half of mayors are randomly asked to take the pledge. About half of mayors that are asked choose to sign the pledge. Suppose we have good measures of actual corruption for all mayors but only after the program is introduced. The President proposes the following strategy: Take the set of mayors that were asked to take the pledge, and compare corruption of those that signed and those that didn't. His economic advisor tells him that this will not provide an accurate estimate of the causal effect of signing the pledge. Explain why not (2-3 sentences)