The Trouble with “The Curve”

BPPbranden-lsat-blog-curve-explained

When people online talk about how a particular LSAT administration is scored, they typically talk about the number of questions you can get wrong to get a 170. This isn’t surprising, as a lot of LSAT students fixate on 170 as a goal.

Years ago, I used to teach the GMAT, too. The GMAT is scored from 200 to 800 in 10-point increments, or in other words a lot like the LSAT’s 120 to 180 scale. What score did my GMAT students fixate on? 700.

Makes sense, right? Actually, it doesn’t. A 700 on the GMAT puts you in the 88th percentile, whereas a 170 on the LSAT puts you in roughly the 98th percentile. Compared to other test takers, a 700 on the GMAT is roughly equivalent to a 163 on the LSAT. The 170 is much more impressive.

But there’s nothing extra special about a 170, or a 700. Since we count in a base-10 system, people just want to be in the top 10 or 100. This goes for the LSAT score conversion table, or “curve,” too. There’s nothing special about a 170, and focusing on the number of questions you can get wrong to get a 170 might lead you to ignore other differences between tests.

Every LSAT has easy, medium, and hard questions, and there’s a reason for that. Imagine for a moment that a particular LSAT had only medium questions. That test would do a great job at telling test takers in the middle of the score range from each other. It would do a great job, better than usual, at determining whether a slightly-better-than-average test taker deserves a 154 or a 156.

But such a test would fail at telling whether a test taker deserves a 170 or a 172. If you can get a 172, you mostly get the medium questions right. The same goes for people who get a 170. Or a 168. On a test with only medium questions, all these test takers would get a lot of questions right, and the differences in their skills wouldn’t really be tested. Likewise, a test with only medium questions wouldn’t do well at telling someone with 128-level skills from someone with 132-level skills. Both test takers would just get a lot of stuff wrong.

So the “curve” to get a 170 mostly reflects how hard the hardest 10 or 15 questions on the test are. If you’re getting a 150, you’re probably getting those questions wrong anyway, so that isn’t terribly relevant to you. Likewise, one test might have slightly harder medium questions than another. If you’re scoring a 154 that’s a big deal, but if you’re scoring a 172, you’re probably getting those questions right anyway.

As my colleague pointed out the other day, the “curve” at the 170 level isn’t all that predictive of the score conversion at other levels. In June 2018, you could miss 10 questions for a 170. A year earlier in June, you could miss 9. Not a huge difference. But the June 2018 LSAT, despite having a more forgiving “curve” at the 170 level, was less forgiving at the 150 level than was the June 2017 LSAT. In June 2017, you could get 46 wrong for a 150. In June 2018, you could only miss 42. That’s less room for error.

What does this mean? It means that while the two tests had similar difficulties at the top end, the June 2018 test had slightly easier medium questions. That means that medium scorers had less room for error in June 2018, but the questions were easier so in theory you wouldn’t need as much room for error. Whatever.

So how should we measure the curve? The curve is actually more consistent at the 160 or 150 level than at the 170 level, so looking at the 170 number isn’t the worst thing in the world. But there are exceptions, like June 2018. Or October 2004, which had a normal -10 for a 170 but a really tough -40 for a 150. There’s no real way to measure the whole curve without looking at the whole score conversion table, which is a pain.

Really, don’t worry about the curve. If the “curve” is stricter at a given score level, that just means the questions at that level are slightly easier. There may be some slight advantage or disadvantage depending on your personal strengths or weaknesses. For example, I’m pretty confident on the hardest questions but I make careless mistakes. So I want a harder test with a more forgiving curve, if I get to choose. But all in all, the “curve” is out of your control and hence not worth obsessing over.

But I know you’re going to keep obsessing anyway, in which case at least remember: the number you can get wrong for a 170 doesn’t tell the whole story. Things might be different elsewhere in the score conversion scale. Also, after you take a test, don’t trust people’s subjective assessment of difficulty as an indication of what the “curve” might be, especially if you’re debating whether to cancel a score. Nobody is, in my opinion, good enough at this to judge the overall difficulty of an LSAT. I can’t really tell, and your average test taker really can’t tell.

Leave a Reply

Your email will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>