A New Use for Federalism? The Benefits and Constitutionality of Randomness in Federal Policymaking

Determining whether progressives should pursue change through the legislatures or the courts depends on our understanding both of what each of these institutions should do and of what these institutions are capable of doing... and they might be more versatile than we've come to assume. As Adam Chandler explains, social science methods point to some interesting uses Congress could make of federalism.

Where laws and regulations differ across state borders, researchers are provided with natural tests of the impacts of those policies. For instance, folks with statistical training can use geographical panel data techniques to discover the effect of a law that is enacted in multiples states at staggered times. Such studies have been done on the deterrent impact of capital punishment and the impact of right-to-carry laws on crime rates to give just two examples. These analyses, however, are necessarily retrospective and constrained by inference techniques. Extensive and careful effort must be used to control for, among other variables, the underlying reasons some states enacted the laws and others did not. More often than not, the resulting answer is that there is not enough evidence to draw a conclusion.

Consider, in contrast, a federal law designed to apply only to randomly selected states (or congressional districts, etc.). Controlled randomized experiments are often described as a “gold standard” in social science research. Adapted from clinical trials, they attempt to isolate the effects of some intervention — say, a new sex ed program — from the environment’s chaotic soup of natural influences and trends. That’s done by comparing a randomly constituted “experimental” group’s experience under the intervention to the natural, everyday changes that a second randomly constituted group experiences when left alone (this second group is the “control” group). These comparisons can help us measure the causal link between a policy and an outcome. And where do the groups come from? I suggest we randomly assign geographical regions, like states, into one or the other.

Perhaps the law could grant twenty random states the funding for a new sexual education curriculum. Then some years later, we could determine the new curriculum’s impact on teen pregnancy rates by comparing the twenty “experimental” states’ teen pregnancy rates to the rates in the thirty “control” states. In this way, such a law could provide one of the first nationwide experimental tests of a policy’s effectiveness. That is, perhaps our country’s federalist structure could allow us to use the states as policy laboratories. Could this be a new use for federalism?

It’s rarely the case that government policies are purposefully applied to some people and not others for the sake of comparison. Even more rarely are policies randomly applied to some and not others, but there are a few impressive examples. In the fall of 1994, the Department of Housing and Urban Development began an experimental housing mobility program in five urban centers based explicitly on random assignment. Families were randomly selected to receive assistance in moving to wealthier neighborhoods. As a result of the randomization, and to the surprise of many, this “Move to Opportunity” program was convincingly deemed to be much less successful than had been previously argued. The National Job Corps Study in 1993 convinced a skeptical Department of Labor that the Job Corps, a training program for disadvantaged 16- to 24-year-olds, is effective for increasing earnings, increasing educational attainment, and decreasing criminality. The random assignment at the core of the study essentially saved the Job Corps from elimination. More recently, the “No Child Left Behind” Act called for the use of “scientifically based research” as the basis for many education programs, indicating at least some appetite for randomization in a recent Congress, if only implicitly.

Inevitably in a geographically-randomized scheme like this, there are concerns about state sovereignty to consider. Of course, the federal government regularly discriminates among the states in funding and regulation, but it rarely does so randomly. In the criminal sentencing arena, there could be Eighth Amendment arguments about “unusual punishments” if people committing identical crimes are subjected to different sentencing guidelines because of their state of residence. And indeed, the U.S. Supreme Court’s primary basis for selecting its cases is to smooth out differential interpretations of federal law among Circuits.

There are also potential Equal Protection complaints about such a randomization scheme. Because geography does not constitute a subject class and as long as no fundamental rights are implicated, such “randomization of application” laws would probably only have to pass the “rational basis” bar. But does the randomization built into the laws make them by definition "arbitrary" or "capricious," undermining their rationality? Or alternatively, when would the laws be rationally related to a legitimate government purpose? Is determining the effectiveness of a certain policy or program, like a new sex ed curriculum, “legitimate”? Such an inquiry has the potential to put courts in a role analogous to research funding bodies, answering the question: Does the potential result of this trial intervention justify its cost in arbitrariness and unfairness (for courts) or in dollars (for funders)?

Setting aside constitutional and ethical objections for the moment, the value of such a nationwide randomized study is easy to see. Political candidates often campaign on the promise of eliminating programs that do not work and expanding those that do. If they truly seek to know which is which, more randomization in federal policymaking is a powerful solution.

What objections to this randomization scheme, constitutional or otherwise, can you see?  Drop a comment below. Or, if you email me at Adam.C2020@gmail.com, I might highlight them in a future post.