Policy should be built on the real cause of the problem, not the preferred guess.

… otherwise it won’t work, but unfortunately social problems are too complex to isolate the pure outcome of the intervention (all else equal). Is the result (success or failure) due to the policy itself or some factor else (or some combination of the factors)? The complexity of interconnections is probably why policy based on preferred guess is so popular in social sciences.

One good example is the policy to deal with the current global recession. New Keynesians argues for the fiscal stimulus while the neoclassical economists oppose it, although both haven’t provided a persuasive hypothesis (preferably validated with formal model) about why the global recession occur in the first place. Many policies proposed by economists are more ideology-driven than addressing the real cause.