UPDATE: Some wondered about the scale used for the visualization below, and I respond with some thoughts in a subsequent blog post.
On the heels of the February 2016 multistate bar exam (MBE) scores reaching a 33-year low, including a sharp drop in recent years, and a small improvement in the July 2016 test while scores remained near all-time lows, we now have the February 2017 statistics, courtesy of Pennsylvania (PDF). After a drop from 136.2 to 135 last year, scores dropped another full point to 134. It likely portends a drop in overall pass rates in most jurisdictions.
This is the lowest February score in the history of aggregated MBE results. (The test was first introduced in 1972 but, as far as I know, national aggregate statistics begin in 1976, as data demonstrates.) The previous record low was 134.3 in 1980.
It's worth noting that the February 2017 test had a small change in its administration: rather than 190 question that were scaled into the score and 10 experimental questions, the split in this exam was 175/25. It's unlikely (PDF) this caused much of a change, but it's worth noting as a factor to think about. And it's not because the MBE was "harder" than usual. Instead, it primarily reflects continued fall-out from law schools accepting more students of lower ability, then graduating those students who go on to take the bar exam. Given the relatively small cohort that takes the February test, it's anyone's guess what this will portends for the July 2017 test.
Visualization note: the non-zero Y axis is designed to demonstrate recent relative performance of bar scores, not absolute scores.