The puzzle is why judges review the constitutionality of (some) laws deferentially. Last week I considered one of the most often provided reasons for deference, that the judges are “unelected,” and so, in a republican political system, should be careful when striking down the enactments of popularly elected legislators. Yet in the American constitutional system, national-level courts are fully republican institutions. They are immunized from direct electoral supervision to make them and the system more republican, not less republican. They are selected by the people’s representatives, and serve only during “good behavior.” This makes courts republican institutions according to the “rigid definition” that James Madison provides in Federalist No. 39. So, too, at the state level, many judges are elected directly, or at least undergo retention elections. It’s hard to maintain at either the state or national level that the judiciary is a non-republican branch of government.
But what about functionalist reasons for judges to defer to legislative enactments? In particular, legislatures (and elections more generally) do not simply aggregate political preferences, they can aggregate information as well. While often overshadowed by the theme of controlling faction, the authors of The Federalist, and the framers more generally, devoted a fair amount of attention to the topic of attracting, producing, and using policy-relevant information in the legislative process.
A lot of this discussion gets overlooked today because it reflected a time prior to systematic data-collection efforts of modern national and state governments. Publius, for example, discusses legislators with knowledge of their own local regions getting together in the House of Representatives, and aggregating their local knowledge in deliberation together to craft better tax measures than could be written with fewer representatives. Judges do not engage in this information-sharing, deliberative process, the thought goes, and so, functionally, should defer to legislatures.
Yet there are obvious challenges to the notion that legislatures are informationally superior to courts. On the one hand, some argue that judges are, on average at least, intellectually superior to legislators. Louis Lusky, Justice Stone’s law clerk (and originator of Carolene Products’ famous Footnote 4), wrote, “By and large, officeholders are an undistinguished lot. For intellect, insight, and moral courage, few of them compare with most of the Justices.”
Putting intellectual capabilities to one side, it is easy to think that judges often make decisions in a richer informational environment than legislators do. Legislative process is generally prospective; judicial process is generally retrospective. Legislation generally looks forward to predicted or anticipated outcomes. Judges consider cases in which a harm has already occurred, or is so imminent as to confer legal standing on a litigant. Litigants can, and often do, file thousands of pages of information about actual consequences of legislation. Legislators can legislate without holding a single hearing, or making a single finding of fact. Litigants can provide ream of information to judicial decision makers about the actual impact of a law.
But all of the advantage does not lie with courts. Legislatures can have countervailing informational advantages, advantages that stem from their plural structure. There is the possibility of measured deliberation, however rare that might be. The plural nature of American legislatures also help them, under certain circumstances, to take advantage of voting as an information-aggregation mechanism. Assuming critical conditions are met, the result of a majority vote can generate informationally more accurate predications than any one of the voters individually. As the number of voters increases, the accuracy of the voting outcome increases as well. Indeed, a group of almost ignorant legislators might generate a judgment informationally superior to a judge who intellectually far outshines each of the legislators individually.
Here’s a simple example. Let’s say there are two policies, A and B. Without confusion, let’s say that policy A is best in state-of-the-world A, and policy B is best in state-of-the-world B. Let’s assume state A is the true state of the world, but the legislators don’t know that for sure.
There are three legislators. Each legislator observes the correct state of the world with (only) 60 percent probability. That means that there is a sixty percent chance any one legislator observes that state A is the state-of-the-world, and a forty percent chance of error, that is, of a legislator thinking that B is the true state of the world instead of A being the true state of the world.
Each legislator votes for the policy that legislator thinks is best in the state of the world he or she observes. The probability that all three legislators vote for policy A Is therefore (0.6)*(0.6)*(0.6). Or 0.216. The probability that the first two legislators vote for policy A, and the third legislator votes for policy B is ((0.6)*(0.6)*(0.4), or 0.144. The probability that all three legislators vote for policy B is (0.4)*(0.4)*(0.4), or 0.064. Adding up all of the possible voting configurations, the probability that a majority of the legislators (two or three of them) vote for policy A is 0.648. That is, when voting, the legislators will vote for the appropriate policy about one out of twenty times more often than they would if they were individually to select what they believed to be the right policy.
This is with only three legislators. As the number of legislators increases the probability that a majority selects the state-appropriate policy increases. Adding two additional legislators increases the odds that the correct policy is selected to 0.68256. Chambers of 30 or 50 or 100 legislators would see significantly improved outcomes. (Plus then aggregating across the chambers in bicameral legislatures as well.)
The upshot is that legislators individually could be significantly less informed, even less capable than any single judge, as Louis Lusky suggested yet, as a collective body, have a significantly greater probability of selecting the right policy for the circumstances. This advantage would hold for large plural legislatures relative to small groups of judges as well. The jury theorem, however, also depends on some strong assumptions that won’t always be met in legislatures. Nonetheless, the aggregation result does sketch a justification why judges might themselves choose to defer to the decisions of plural legislatures. They would do so because non-deferential intervention might risk creating worse policy outcomes rather than improving policy outcomes. Interestingly as well, this informational rationale for deferring to plural legislatures would not extend to deference toward decisions by unitary executives, and would not necessarily apply to bureaucratic decisions.