At least give students & families the info they are giving GCs – it could help families better understand the current PSAT scores & their implications re NM. Basing reports on a sample of students who did not take the test is just not meaningful to most (any?) people. Was that done in the past reports?
Hopefully the state summaries will be forthcoming soon. But if in September, the cut offs turn out to be very “harsh” / different - higher - than current ones in many cases, I do think trust in the CB will be eroded much further (is there far to go though?). Hope that’s not the case.
@DoyleB I hear you.
Spoke to GC and she thinks the percentiles are ‘inflated’ because they did not use real student testers, instead used ‘representative testers’. But she was referring to the %'s that people see on the report (National, not SI). I didn’t ask too many questions because we just met and I didn’t want to be ‘that parent’. Anyway, I don’t think a 2014 225 = a 2015 218 or 219 like some of the predictions based on %. The question is, how off are they and why? I think the highest cutoffs will be 221-222. Hopefully 221 (since that is what my kid got, LOL). The concordance charts predict 221-228.
I don’t understand - why CB would inflate the SI % so much. I get they want more kids to take the test, but inflating them a huge amount will only trick people for the first sitting, eventually it will get out and no one looks at these percentiles after PSAT season (so not until next year).
Would love to hear some more ‘insider info’. No GC’s on this site? How about more tutors and test prep guys? @Plotinus - thoughts?
There have been many people thinking the SI % tables are wrong and not based on actual 2015 testers. While that may be true, it may not be true.
Jed Applerouth ADMITTED in his email to me (posted on CC) that the 2015 Guide did not state where the SI % table came from. He assumed it came from the “user” sample - but that is not a known fact at this time.
We really need CB to state if the SI % Table came from a research sample or from real 2015 test takers. I’ve emailed them, but haven’t received a response. Maybe others CCers can also email them, ask on twitter, etc? Basic question is where the SI % Table came from - research or actual testers.
For now, until prove otherwise, I’m betting the SI % Table IS actual 2015 testers. Hey, why not use actual testers, that data is sitting in CB’s database.
I also have a thought on so many parents, GCs and others stating a larger than normal number of students at a high school are in the 99% range. Remember, the “national” sample is inflated because it gives percentile as if everyone in the nation took it. If GCs are using the National numbers and saying there are too many 99%, that is totally understandable. If they are using the User numbers and saying there are too many 99%, that would be a concern. From many of the postings here, which numbers the GCs use is almost never stated.
I bet many GCs are using National because that higher number jumps out at you. Maybe many of those students are 99% National, but 98% or even 97% on User - but GCs see that 99% National number and say, “too many 99%ers”.
This is only a question that can be asked on College Confidential 
Here goes:
What is 99+ percentile? (Forget the PSAT for a second). Does 99+ mean any percentile greater than 99.5%ile?
@suzyQ7 Yes, at or above 99.5%
There was a post that someone in the past confirmed it.
I suppose CB did not want to round up and write 100% or introduce decimal point in the table
Here are my thoughts:
-
It looks to me that over the years College Board has become less and less interested in the task of distinguishing among the students at the upper part of the curve, competing for Ivy League level schools and NMSF, and more interested in the task of testing whether students have the basic skills required to attend to a mid- or low-level college.
-
This is probably because there is a lot more money in testing the millions of students in the middle/bottom than in testing the few tens of thousands at the top.
-
The result of the successive curve “recenterings” (1995) and test redesigns (2005, 2016) has been to boost the scores of people in the middle, to shore up scores of certain minority groups and females, and to pack many, many more students into the top part of the curve (the Ivy League range).
-
The 2016 SAT follows this general trend, so that I expect there will be more people in the upper scoring range than there were in the upper range of the 2015 test.
-
An NMSF score is a score in the top 16,000 juniors nationally. Since there were more juniors taking the test in 2016 than in 2015, the top 16k students will correspond to a higher percentile in 2016 than in 2015. The number of juniors taking the test in 2016 has not been published, so we don’t know how much higher.
-
In 2015, 1.6 million juniors took the test, so 16k/1.6 million = 1%= 99th percentile.
Assuming, for example, 2.3 million juniors took the test in 2016 (just guessing here), then 16k/2.3 million =0.7%=99.3th percentile. This is just an example. We don’t have the real numbers. One parent said these numbers will be published with the state reports in February. Then there should be enough data to make reasonable predictions. -
I have no idea what SI corresponds to 99.3th percentile. Further, this number is just a guess at the national average. NMSF is by state. In some states the cutoff will be above the above the national average, and in some states the cutoff will be below the national average. I think the cutoffs in states with relatively strong student populations and schools might be higher compared to the 2015 cutoffs, and the cutoffs in the states with relatively weak student populations and schools might be lower compared to the 2015 cutoffs. The reason is that I think the new test is easier for strong students and harder for weak students. However, there are only a few NMSF awards for each state, so if there are a few hundred or a couple of thousand students who are very strong in a state that is weak on average, the NMSF cutoff might be high even in a weak state.
-
What I really think is that there is no way to predict what is going to happen with the NMSF cutoffs because there is not enough data and NMSF is not a top priority for CB any more. I would not be surprised if some weird things happen at that end of the curve. For example, it could be that the cutoffs turn out really, really high, especially in some states, and really low in others (if there are states in which few schools are teaching Common Core). Or they might be really high in all states. I think there are going to be more high scores, so I think the cutoffs may be higher on average.
-
My own view is that the SAT should be junked as an NMSF adjudicator and substituted with a much harder test that reliably determines the relative merit of excellent students. But that is not going to happen.
I now predict that the cutoff will be 226 in high cutoff states, 222 in mid, and 218 in low. The test was too easy, plain and simple. Hope I am wrong.
All, I am relatively new to this topic but have followed this thread with some interest. I hate to make myself unpopular on a Sunday morning but unfortunately agree with plotinus’ thoughts. The cutoffs are going to be way higher than people expect, at least in the high score states.
First my disclosures:
- I am not a math person per se but have some experience with statistics.
- We live in NJ, consistently among the highest cutoff states. I have therefore focused at the top of the tables.
- I have a horse in this race, so to speak. My D got a 224, consisting of Reading 37 (2 wrong), Writing (1 wrong) and Math 38 (1 wrong).
- D said the test was pretty easy. However, she had already taken an SAT in September and re-took in December. Her scores were entirely consistent with the 224 from the PSAT with a 740 Reading, 720 Writing and 770 Math is 2230.
Next my method:
- I know the concordance tables are labeled preliminary and have been much debated, but they should tell us something about what the CB thinks will happen.
- I looked at the NJ state report from last year and lined up the data about the actual curve with the concordance tables. The results are shown below.
Reading
2014 US # US % NJ # NJ % 2015
75-80 12,269 0.77% 595 0.72% 37-38
70-74 40,435 2.53% 1,946 2.36% 36
65-69 50,059 3.14% 2,309 2.80% 34-35
60-64 132,960 8.33% 6,016 7.29% 32-34
55-59 151,282 9.48% 7,257 8.80% 30-31
50-54 262,519 16.45% 12,594 15.27% 27-29
45-49 241,246 15.12% 12,263 14.87% 25-26
40-44 277,848 17.41% 14,750 17.89% 22-24
35-39 205,506 12.88% 11,459 13.89% 20-22
30-34 123,731 7.76% 7,176 8.70% 18-19
25-29 65,274 4.09% 4,116 4.99% 17-18
20-25 32,357 2.03% 1,990 2.41% 15-17
1,595,486 82,471
Writing
2014 US # US % NJ # NJ % 2015
75-80 14,474 0.91% 962 1.17% 37-38
70-74 30,945 1.94% 1,740 2.11% 36-37
65-69 58,912 3.69% 3,129 3.79% 34-35
60-64 80,202 5.03% 3,803 4.61% 32-34
55-59 164,768 10.33% 7,686 9.32% 30-32
50-54 180,999 11.34% 8,132 9.86% 28-29
45-49 277,129 17.37% 13,518 16.39% 25-27
40-44 258,258 16.19% 13,212 16.02% 23-26
35-39 258,214 16.18% 13,942 16.91% 20-22
30-34 133,110 8.34% 7,654 9.28% 17-19
25-29 75,906 4.76% 4,436 5.38% 16-17
20-25 62,569 3.92% 4,257 5.16% 14-16
1,595,486 82,471
Math
2014 US # US % NJ # NJ % 2015
75-80 27,200 1.70% 1,824 2.21% 37-38
70-74 39,427 2.47% 2,440 2.96% 36
65-69 99,827 6.26% 5,374 6.52% 34-35
60-64 117,266 7.35% 5,964 7.23% 32-34
55-59 190,357 11.93% 9,322 11.30% 30-31
50-54 238,624 14.96% 11,548 14.00% 27-29
45-49 300,608 18.84% 14,816 17.97% 25-26
40-44 206,440 12.94% 10,599 12.85% 22-24
35-39 205,350 12.87% 11,017 13.36% 20-22
30-34 92,785 5.82% 5,030 6.10% 18-19
25-29 52,906 3.32% 3,030 3.67% 17-18
20-25 24,696 1.55% 1,507 1.83% 15-17
1,595,486 82,471
Last year the cutoff for NJ was 225. The concordance tables suggest that the CB thought the top 1% of scores for each section would fall in the 37-38 bracket (at least for NJ). In math, you probably need a 37.5 or even a 38 to be in the top 1%. That puts the score somewhere around 224.
Now maybe not many kids score in the top 1% of each, some are relatively more mathy and some more verbal. So maybe the SSI cutoff falls by a point or two. But probably not much.
This does call into question the SSI percent table, so be sure. But the accuracy of that table has also been much debated.
I would love to be proved wrong in my thinking! My D is probably on the cusp according to this analysis and I have already spent WAY too much time on this topic. She hasn’t given it a second thought and I should do the same! Nevertheless please feel free to poke holes in my post so I can put this down until we hear more from the CB.
The only comment I have is, why would the CB do this? Because they really don’t care about NMS? I thought NMC was a big funder of PSAT.
Was the CB “fooled” by these research study kids who did not do as well on this test (which is why “real” SI percentiles are inflated? )
The CB knows perfectly well that NMS is a contest for the top 1% and they made the test so easy that 1 question differentiates one person to another by 2 SI points, which is ridiculous.
I have no idea but probably they are trying to make it more applicable to the kids across the spectrum. Unfortunately this makes it harder for the kids at the top to really distinguish themselves. The only way is to get almost perfect scores.
The other thing to consider is that they have compressed the score ranges at the very top, probably to accommodate the lack of penalty for wrong answers. Look at the concordance tables for reading and math. The previous range of 70-74 is captured by one score, of 36. All the other brackets of 4 or 5 pre-2015 scores are covered by a new range of two scores (e.g. 34-35).
My guess on this is that they are lining up the new score so it is pretty equivalent to an old score. Meaning, to use my D as an example, her score of 224 out of 228 is actually pretty darn close to the 2014 scale of 224 out of 240.
One question could previously differentiate by more than 2 points @suzyQ7 . My son got one wrong in math his sophomore year, and it cost him 4 points. Another wrong in Writing cost him 5. That’s NINE points for just two questions. I do agree that the scores will end up similar to last year, with a squash job on the top end.
If the scores end up similar to last year (but squished at the high end) then the SI table CB put out last week isn’t accurate. But maybe we’re looking at it wrong.
Let’s say the table is now research based, and now the percentile is based not on scores of everyone who took the test, but scores of every student in the US, whether they took the test or not. Then assume that students who didn’t take the test before typically score low (at least not at the top) and thus don’t impact the top end of the curve.
That causes the effective distance from the 100th percentile to half (for example, if you were the top scorer out of 100, you were at the 99%. Add 100 low scoring students, and you’re the top out of 200 or the 99.5%).
So you can unwind the table at the top end by doubling the distance from 100. The 99.5 percentile becomes the 99th percentile. A 99 becomes a 98. Do that to the published SI table, squish the high end down to 228, and last year’s table and this year’s table look a lot more alike at the high end
This may be a silly question but is there another outside indicator of NM potential for a student - say, Cal Tech sends you an e-mail or something? I realize that UChicago blankets everyone but most of the other top 10’s don’t. If your kid checked the box for the College Search Service and you start getting communication from some of the Ivy’s and other tippy top schools what does that mean, exactly?
The reason I bring this up is that CB said they developed some system “to work behind the scenes to help colleges and universities during the transition year”. (page 21 of the Understanding Scores report).
Still hard to believe that the cutoffs will be as high as 226 in the high cutoff states, 222 in the middle and 218 in the low as: 1) I have heard some jump onto CC and talk about higher scores, but not that many in reality - still a small sample group, 2) CB would gain nothing by administering a PSAT test that has no relative value to the colleges they serve - kids will abandon the SAT and flood the ACT if the PSAT results are discounted (we already decided the ACT was best for them when we knew a change was coming and are happy with our twin DD’s single-sitting 34’s from the September test - they are done as testing is way overrated/ they can and should be spending their time doing other stuff to make themselves a great college candidate, 3) NMSQT participation is interesting to the CB, but not enough kids/revenue to move the needle so why would they bend the PSAT curve to boost the SI cutoffs, 4) if these preliminary results are correct I can envision an ACT driven NMSQT which would be like the CB purposefully driving the car off the bridge
Net, net I don’t believe that the CB could have been that wrong in their release of preliminary PSAT percentiles and we will wait; DD’s have already moved on, like the next morning, but yes they would like to learn in October that a 219 is an NMSF - but they won’t worry about it until then.
"One question could previously differentiate by more than 2 points @suzyQ7 . My son got one wrong in math his sophomore year, and it cost him 4 points. Another wrong in Writing cost him 5. That’s NINE points for just two questions. I do agree that the scores will end up similar to last year, with a squash job on the top end. "
True - but this year you can have the same amount of questions wrong as someone else, the same TOTAL score, and 1 person will make it and the other 1 won’t. This will likely happen to us since my son has 1480/ 221, and the other top scorer in his school has a 1480/222. Likely, that other student will make it and he won’t. So is life.
Hopefully CB will make the March SAT as easy as the PSAT (at least for the first sitting) so my kid can have a nice inflated total score to present to colleges - who probably aren’t spending hours and hours doing this type of analysis and will just take it at face value (I know this is unfair, but what CB has done to class of 2017, also not fair). Maybe he can’t get some scholarships based on that.
The one question I have is what advantage will CB gain if the scores are significantly inflated? Wouldn’t they face even bigger public opinion problems long term? They may gain more test takers for Mar and perhaps May, but when those test scores and NM are released, they wouldn’t they lose even more credibility if PSAT scores were radically different from the SAT scores? Conversely, how would colleges respond if SAT scores are inflated? Seems lose-lose for CB to me.
ETA: Typing while you were @Chembiodad I agree with your analysis.
That would suck @suzyQ7 , I hope it doesn’t come to that. But again, that was true in previous years, just not as highlighted by the new way of scoring the actual SAT. Math errors have always been more costly than verbal ones.
Well, my D just got her fist email from CalTech, just after PSAT scores came out. I don’t think that makes me feel any better though.
I agree with some of the other posts about the CB shooting themselves in the foot with inflated scoring. But I can’t get over the concordance tables and what they imply. I think all this will make things complicated for the March SAT test takers. Glad we are happy enough with her current SAT scores and can put down the #2 pencil! And yet, here I am, following this thread.
I am a bit more optimistic than the most pessimistic here. For the 2014 test, in the most stringent areas (D.C., internationally), students needed to average a scaled score of just 75 (because 75x3=225). Just from glancing at the “Understanding 2014 PSAT/NMSQT Scores” report, one can see combinations of 6 or 7 wrong that would meet or exceed that average. (I did not attempt to determine the range of wrong answers that could yield at least a 225 in 2014.) 75 per subtest corresponded to two 99th (not 99th+) and one 98th percentile, and that’s for a 225. NJ only required a 224. CA and MA required a 223. These were reasonably forgiving scores with respect to corresponding subscale percentiles and number of errors permitted.
Yes, the new test is easier for high-scoring kids. Yes, the guessing penalty has been eliminated. Yes, students choose among 4 rather than 5 answers. Yes, there may be more high scorers simply because more students took the test. But there are 13 additional questions. Even one error on the Writing subtest on Oct. 14 eliminates the possibility of a 228 or 227. The number of kids with perfect Writing scores will not exceed the number of NMSF slots per state. And those perfect-Writing kids would need to make no more than one error each on Math and Reading to get a 226. (Again, there are several permutations that would yield 226, but all are unforgiving.) One error on Writing requires perfect 38s (1 Math error and 1 Reading error allowed) on the other sections for a 226. These are rare outcomes. In the 2nd-highest-scoring states last year, a LOT more errors than that met the cutoff.
So I believe that the addition of 13 questions, the extremely stringent standards for extremely high scores, and the proven imperfection of even the very high-scoring kids argues that cutoffs at the high end will not exceed last year’s. In addition, if one converts a 38-point scale to an 80-point scale (forget the concordance tables; just use pure arithmetic), a 37 is almost a 78. My guess is that this year’s cutoffs will be a bit lower than last year’s.
If you manipulate the 2015 published SI table as I suggest, it yields the bottom of the 99s as a 214, the bottom of the 98s a 205, and the bottom of the 97s as a 202. Last year’s table values for those points? 213, 206, 202.
Well not exactly. 98 becomes a 96, not a 97, which is a 198. But I can understand the rationale behind trying to keep scores somewhat comparable.
Hmm. Guess I’ll just wait for September.