Slippery Slope June: Cost-Lowering Slippery Slopes, the Costs of Uncertainty, and Learning Curves

[This month, I'm serializing my 2003 Harvard Law Review article, The Mechanisms of the Slippery Slope; in last week's posts, I laid out some examples, definitions, and general observations, and turned to a specific kind of slippery slope mechanism—cost-lowering slippery slopes. This week, I'll elaborate on that, and shift to some other mechanisms.]

The example in Friday's post involves the cost of tangible items: cameras. But another cost of any new project is the cost of learning how to implement it properly, and the related risk that it will be implemented badly.

People are often skeptical of new proposals (from Social Security privatization to education reform) on these very grounds. Broad change B—for instance, an across-the-board school choice program—might thus be opposed by a coalition of (1) people who oppose it in principle (for instance, because they don't want tax money going to religious education or because they want to maintain the primacy of government-run schools), and (2) those who might support it in theory but suspect that it would be badly implemented in practice. This lineup is similar to what we saw in the camera example. {As before, I express no view here on the merits of this particular B. My question isn't whether particular policies make sense in the abstract; rather, it's how people who do oppose B should act to better implement their preferences.}

But say that someone proposes a relatively modest school choice program A, for instance one that is limited to nonreligious schools or to children who would otherwise go to the worst of the government-run schools. Some people might support this project on its own terms. But as a side effect of A, the government and the public will learn how school choice programs can be effectively implemented, for instance what sorts of private schools should be eligible, how (if at all) they should be supervised, and so on.

If A is a total failure, then voters may become even more skeptical about the broader proposal B. But if after some years of difficulty, the government eventually creates an A that works fairly well, some voters might become more confident that the government—armed with this new knowledge derived from the A experiment—can implement B effectively.

A will thus have led to a B that might have been avoided had A not been implemented. In the path dependence literature, this is described as a form of "increasing returns path dependence" that focuses on "learning effects": "In processes that exhibit … characteristics [such as learning effects], a step in one direction decreases the cost (or increases the benefit) of an additional step in the same direction, creating a powerful cycle of self-reinforcing activity or positive feedback." And because of this path dependence, "decisions may have large, unanticipated, and unintended effects."

For those who support broad school choice (B) in principle, this is good: the experiment with A will have led some voters to have more confidence that B would be properly implemented, and thus made enacting B more politically feasible.

But, as in the cameras example, those who support A but oppose B in principle might find that their voting for A has backfired. Some of A's supporters might therefore decide to vote strategically against A, given the risk that A would lead to B. The government, they might reason, ought not learn how to efficiently do bad things like B (bad in the strategic voter's opinion), precisely because the knowledge can make it more likely that the government will indeed do these bad things.