This Essay suggests the necessity of a co-evolutionary process among empirical and theoretical advances in law and economics. Empirical work alone is suggestive, but should not be taken too seriously. The weaknesses in empirical work, and by this I mostly mean regression-based work which has come to dominate law and economics, lead to a kind of virus that begins with over-statements and misapprehensions, and then spreads as more scholars copy the mistakes and engage in empirical work as a means of entry into the field. Regression-based work will become suspect as its current assumptions are questioned, and as replication failures reveal its weaknesses. Empirical work in law and economics looks very different when underlying distributions are not easily probed with regressions but are understood as reflecting power-laws, or as simply random. Once inconvenient distributions are acknowledged, the key question is why observations might be distributed in this fashion. This is likely to be a task for theorists as law and economics enters its next phase. On the other hand, empirical work has been important and has made law and economics a respectable science. The claim here is that good empirical work—especially in law and economics—is hard to produce, and it is important not to overvalue its products. Moreover, it is more useful when combined with good theory.
The focus in this Essay is on three weaknesses of empirical work, though in a larger sense most of the problems come from omitted variables and, in some cases, insufficiently large data sets. First, much of the empirical work in law and economics is driven by models that rely on error minimization techniques, and these techniques are unreliable when errors are surprisingly and unevenly distributed (that is, when they suffer from heteroscedasticity). Second, it is likely that when empiricists connect data with a model, the process is flawed because there might be a hidden transition to a second distribution. Discovering multiple distributions is likely to require theoretical work. These and other problems are exacerbated by the likelihood that conclusions are based on the tail end of data sets, inasmuch as scholarly journals only bring to light statistically significant results. In addition, empirical work in law and economics suffers from the absence of sizeable data sets. Without such sets it is difficult to test conclusions and to escape the omnipresent challenge of omitted variables. Reversal paradoxes are yet another serious problem, and especially so in the absence of large data sets.
The larger and more optimistic claim is that data and theory can and must work together. Regressions have come to play a critical role in law and economics, and econometric methods have improved over time. It has become apparent that data can suggest theories, and theories can be tested, to a degree, with data. But some theoretical insights are so convincing that data testing, though comforting even when flawed, may be unnecessary— and it may, in any event, be tainted by the spread of the theory. It is likely that empirical work in law and economics will find itself in retreat, even as its quality improves because of renewed attention to theory.