Why do score conversion charts differ? Forum

[deblaw]

While I was looking at 07' LSAT, I noticed that the conversion chart differ greatly from one to another.
Is it possible to know the socre conversion chart before taking the exam?

[soj]

While I was looking at 07' LSAT, I noticed that the conversion chart differ greatly from one to another.

Because, based on research methods that include the experimental section, LSAC has determined that some tests are harder than others. More precisely, some tests have more questions that high scorers get wrong than other tests. Those tests get more lenient curves so that you don't need to get quite as many questions correct to get the same score.

Is it possible to know the socre conversion chart before taking the exam?

No.

[incompetentia]

The scaled score (120-180) has an established relationship to percentile of test takers. 90th percentile is ALWAYS going to be around 163, and 80th percentile is ALWAYS going to be around 160.

It's essentially an elaborate curve; you're primarily competing against those you're taking the test with. However, if you look at the administrations that have been released as PTs, you'll notice that on most tests, the number of questions you can miss and still get a 170/165/160 doesn't change drastically.

[barnum]

While I was looking at 07' LSAT, I noticed that the conversion chart differ greatly from one to another.

Because, based on research methods that include the experimental section, LSAC has determined that some tests are harder than others. More precisely, some tests have more questions that high scorers get wrong than other tests. Those tests get more lenient curves so that you don't need to get quite as many questions correct to get the same score.

Is it possible to know the socre conversion chart before taking the exam?

No.

This is right.

The scaled score (120-180) has an established relationship to percentile of test takers. 90th percentile is ALWAYS going to be around 163, and 80th percentile is ALWAYS going to be around 160.

It's essentially an elaborate curve; you're primarily competing against those you're taking the test with. However, if you look at the administrations that have been released as PTs, you'll notice that on most tests, the number of questions you can miss and still get a 170/165/160 doesn't change drastically.

This is very very wrong. The percentiles are NOT fixed to the scores. It is NOT a curve. You are NOT essentially competing against those you are taking the test with. The scale IS pre-determined based on previously administered experimental sections.

[510Chicken]

This is very very wrong. The percentiles are NOT fixed to the scores.

Equating, blah blah, not a curve, I know.

That said, are you sure the percentiles aren't "always going to be around" the same score? Your LSAT doesn't seem like it would be particularly useful to anybody if the score did not reflect the (relatively) same percentile bracket as a previous or future test. That is, you cannot compare two 170's if one of them represents the 98th percentile and the other the 95th without actually starting to monitor when each person took the test etc etc, which would defeat its purpose.

I'm not saying that the score is directly linked to the percentile, but it seems like it would be at least indirectly linked. If equating is done properly, there should be proportionally the same amount of 170 scorers in each testing cycle, give or take a little variation.

[barnum]

Yes, percentiles are always near the same score, but not as a result of a curve as the previous poster had commented. They are near the same score because the general pool from year to year of LSAT test takers is remarkably similar in LSAT ability. It also remains fairly stable because it is actually based on 3 years worth of test takers. So the percentiles are based on 12 LSAT administrations. This means with each new LSAT that group of test-takers only represents 1/12th of the group of takers that the percentiles are based on and so from administration to administration the percentile doesn't move very much. But if it were fixed to scores, it would never change, and it does. For example, a few years ago a 171 was 99th percentile, now it is 98th.

[tomwatts]

See also: score drift in the GRE. The GRE is equated in much the same way that the LSAT is, and the average/median was 500 on each section originally. Right now, because of the equating and the change in the test-taking population, the median Verbal is well below a 500 and the median Math is well above a 500. This could, in principle, happen on the LSAT, too. It never has, but it could.

If you're old enough to remember it, the same thing happened with the SAT. It was recentered in 1994, and for years afterward, your SAT score had an R on it.