A new study out of BYU attempts to answer the question.  It’s summarized at TaxProf and the full article is here. From the abstract on SSRN:

What, if any, is the relationship between speed and grades on first year law school examinations? Are time-pressured law school examinations typing speed tests? Employing both simple linear regression and mixed effects linear regression, we present an empirical hypothesis test on the relationship between first year law school grades and speed, with speed represented by two variables: word count and student typing speed. Our empirical findings of a strong statistically significant positive correlation between total words written on first year law school examinations and grades suggest that speed matters. On average, the more a student types, the better her grade. In the end, however, typing speed was not a statistically significant variable explaining first year law students’ grades. At the same time, factors other than speed are relevant to student performance.

In addition to our empirical analysis, we discuss the importance of speed in law school examinations as a theoretical question and indicator of future performance as a lawyer, contextualizing the question in relation to the debate in the relevant psychometric literature regarding speed and ability or intelligence. Given that empirically, speed matters, we encourage law professors to consider more explicitly whether their exams over-reward length, and thus speed, or whether length and assumptions about speed are actually a useful proxy for future professional performance and success as lawyers.

The study raises important questions of how we structure exams. I know of colleagues who impose word count limits (enforceable thanks to exam software), and I think I may be joining the ranks. More broadly, are our high-stakes final exams truly measuring what we want them to?