Just sharing some comments in answers to questions posed by readers of the CommpassPrep Blog:
http://www.■■■■■■■■■■■■■■■/college_board_forum_update_2_20_2015/#comment-11435
“Art Sawyer (of Commpass Prep) says:
January 17, 2016 at 9:15 pm
National Merit Scholarship Corp will not release official cut-offs until September 2016. Last year’s cut-off was 215 for North Carolina, and we expect it to be close to that mark this year. It seems counter-intuitive because of the lowering of the maximum score from 240 to 228, but at the critical low 700’s on each section, scores have not changed much from 2014 to 2015.”
Bruce Reed says:
January 11, 2016 at 5:07 am
Hi John,
“Cut scores in many states are likely to fall at least slightly, but they could rise in a few states too. The highest possible Selection Index has dropped from 240 to 228 because of a new scoring system and index formula but that doesn’t mean cut scores will drop by 12 points. There are other statistical factors applying upward pressure on the index and it’s too early to make accurate predictions (although many are trying.) These competing forces will have slightly different effects state-by-state.”
I like “The CB know how to count, and their computers know how to count” and “They took an extra month to release the scores to make sure they got them right”
I think the concordance table is a broad approximation. They will be used by college admissions officers. For their purposes, they don’t need to nail the number down to a specific number. All they need is a range. So if a 214 is 209-211, and if that concords to a 2090-2110 on the old SAT, that’s not that big of a deal.
The concordance MIGHT be somewhat related to SIs, but not 1:1.
And one more thing. The people that tend to come on College Confidential tend to be obsessive, ambitious over-achievers. This can create the false impression that everyone has scored in the 99%, or an SI at or above 220/228. That leads people to stress hysterically and speculate that their 228/228 in Alabama might not be good enough. Those with very high scores tend to make it sound like anything lower than their own score is probably not good enough. The bias tends to be towards overstating the threshold. If you scored in the 99% both in the National Representative Sample, and in the User Group category, you should feel very good about your chances, and no matter what, you can be truly proud of your achievement. At a bare minimum, you are going to be commended, and likely more.
@mantua wrote:" The CB know how to count, and their computers know how to count. They are not going to suddenly tell a bunch of people’ “oops, our bad, you are actually in the 97 percentile”
I am guessing they’re going to tell us that we misinterpreted their table (well they won’t tell us anything, but you know what I mean).
Possible scenario:
We’ll say “I got a 214. The table said that means that I scored higher than 99.5% of the 1.7 million kids who took the test.”
Maybe they’ll say, “No, it means you scored the same or better than 99.5% of the 3 million kids in the US”.
Everyone says that college admissions officers don’t look at your PSAT score. So has the PSAT EVER concorded to the SAT? College admissions officers won’t be looking at the PSAT score for 2015. So, technically they don’t NEED to concord. Although it’s probably nice if they are ANYWHERE near the ballpark.
Admissions officers will be looking at the SAT scores and the concordance will be approximate.
Yes @likestowrite - this discrepency is exactly the thing that doesn’t make sense. For the percentile definition difference, you could compensate for that by taking the percentile for 214 from the 213 row (which would match the previous years definition of percentile). But obviously that only makes a small difference.
@micgeaux We are looking for concordance to the last year PSAT to see how this years scores might qualify for last year’s NMSF cutoffs. We are not looking to concord it with the SAT
@thshadow I have more questions. I don’t understand why we would assume that 214 would necessarily go down to 213 when you apply the old definition. I think that we do not know the actual numbers of test takers at the score level of 214. Thus, how can we definitively know how far to move the score down? Why are you sure it is moved from 214 to 213 with the definition change? Couldn’t it move from 214 to around 210 depending upon the number of students who scored at that 214 mark? The PSAT doesn’t usually have a score distribution in a perfect bell curve, does it? And if it does, wouldn’t the amount of scores to move down depend upon where it was in the bell curve?
Yes. But I don’t think CB cares if they concord exactly. The only people who would need to concord PSAT scores would be sophomores. The juniors won’t be taking it again. So if it’s off, even by 5 points, does it really matter to a sophomore? Maybe they will study harder if they think they scored lower? Am I missing something?
@DoyleB - Thanks for the great analysis. I might add that we are not an Ivy League bound small private school. Public school with about 2500 students. Most students take ACT, very few take SAT.
Indeed superb analysis- however after all of this analysis is anybody in this thread willing to predict the cuff off scores for states ? Are we at point where we can venture to make prediction for all states similar to the ones test prep sites have predicted or has it already been posted somewhere else in this thread?
micgeaux" “The only people who would need to concord PSAT scores would be sophomores.”
Can you explain this further? What would sophomores be using the concorded scores for? Comparing their sophomore scores with their freshman scores? Something else?
I don’t really understand why anyone needs the PSAT concordances, actually. Maybe school counselors could concord their average reading, writing, and math scores, and see if, say, doing a PSAT/SAT course at school had increased PSAT scores over the last year’s scores. I guess. Or maybe a school would want, as I said earlier, to convert its old minimum PSAT requirement for AP courses for the new PSAT.
But, other than folks trying to figure out SI cutoffs for NMSF, I can’t actually see much reason why anyone would be desperate to have PSAT concordance tables.
SAT concordance tables are a different thing, obviously.
@CA1543 those percentiles in the 2014 “Understanding Scores” are actually based on 2013 data. Beginning in 2012 CB didn’t release the percentiles for the test year until the following year. 2014 percentiles (which would have been released this year had the test not been redesigned) were never released.
The percentile tables I provided earlier (sorry can’t call the post # I’ll try to find that and re-post) actually lined up the percentiles consistently with the SI’s by year.
Changing the topic somewhat back to the scoring curves. As I’ve stated, CB can set the scoring curves to distribute the scores however they want. But they would need some target to equate to. That could be historical data, research sample, closing the cutoff gap, etc.
The idea they want the cutoffs to be similar to last year could mean they are equating to historical data, 2014’s results in particular. But then why even have a research study? They already know 2014’s test takers, score distributions, etc. I’m still hung up on the idea they would want to use the research study as their target and set the curves to model that distribution. Maybe the research study gives similar data to 2014’s results? 2014 results and the research study did both happen in 2014 - but with different test formats.
Using 2014 state cutoffs as 2015 state cutoffs, means New Jersey has to have 225 / 228 vs 225 / 240 last year. It would mean only perfect scores or nearly perfect scores will make SF in NJ and DC. That seems quite remarkable to me. CA at 223, TX at 220, wouldn’t need perfection, but close to it.
We usually have around 100 psat perfect scores nationwide. So DC and NJ might have 5 of those perfect scores at most. The 2 states combined have around 600 SF slots allocated.
I’m having trouble accepting 2015 cutoffs = 2014 cutoffs.
I agree Speedy that there isn’t enough room at the top to have high scores from last year map over directly. The concordance tables result in lowering the high cutoffs, leave the ones in the high teens basically the same, and raise the lower ones. I can see why CB would like it like that, and target that with their scoring tables. Testmasters estimated cutoffs based on concordance show this effect.