How plausible is it that models seeking to explain variation in bivariate litigation outcomes (e.g., success/failure) among lawsuits brought by various (and standard) party types (e.g., individuals, corporations, insurance companies, government) might need to account for the quality (or strength) of legal claims pushed or arguments advanced? Put slightly differently, is there any reason to assume, ex ante, that lawsuit quality will systematically vary across party types? (My query assumes that a lawsuit’s quality informs the probability of its success.) If so, and to the extent that endogeneity might lurk as a methodological issue, what are prudent responses?
At one level, data selection can perform some crude filtering tasks, at least at the margins. For example, by limiting a dataset to only cases fully litigated and generating a written, published opinion researchers functionally screen lawsuits that do not survive the preliminary motion stage. (This implies, rightly or wrongly, that lawsuits kicked at the preliminary motion stage possess less legal “quality” than those lawsuits that survive preliminary motions. Limiting a dataset to written, published opinions, however, swaps one source of potential bias for another.) If one aspires a more finely grained proxy for lawsuit quality, however, do other variables make sense?--keeping in mind that any such variable must be objective, reliable, replicable, etc. On a related though distinct matter, suggestions about how researchers might instrument for party types in this context are also welcome.