[Anonymous]

I think it would be interesting to see what people think is the most curved and the least curved. So, based on your experience, rank the sections from most curved to least curved.

I’ll go first….

FAR

AUD

BEC

REG

[OnMyWayToCPA]

Ummm… For me it's a toss up between FAR and AUD at the top. I kinda want to put AUD first, because there are so many 99's, but since I've heard of people walking out of AUD feeling like they got a 99, but never heard of anyone feeling the same for FAR, I'll put far first.

I'd put REG before BEC.

So final tally.

FAR

AUD

REG

BEC

B - 79 PASSED (1/4 there!)
A - 77 PASSED (1/2 there!)
F - 79 PASSED (3/4 there!)
R - 83 PASSED (DONE!!!!!!)
Review Course: Lambers , Wiley and Ninja

[Anonymous]

The CPA exam is not curved.

https://www.aicpa.org/BecomeACPA/CPAExam/ForCandidates/FAQ/Pages/computer_faqs_3.aspx

“The CPA Examination is NOT curved. Every candidate’s score is entirely

independent of other candidates’ Examination results.

The CPA Examination is a criterion-referenced examination which means that it rests upon pre-determined standards. Every candidate’s performance is measured against established standards to determine whether the candidate has demonstrated the level of knowledge and skills that is represented by the passing score. Every candidate is judged against the same standards, and every score is an independent result.”

That's why there is a low passing rate.

Remember under the paper-and-pencil exam the passing rate was 15-20% for the entire thing. Now we're up to 45%-50% for each section.

[Anonymous]

A lot of people believe that statement is incorrect. Many people here swear that these exams are curved.

[Anonymous]

I also think BEC is the least “curved”.

[Minimorty]

Yeah, but the “established standards” were determined based on candidates' performances. The questions are weighted based on their difficulty, ability to be guessed correctly, etc., etc. Long story short – the exam is curved.

[Anonymous]

STATISTICALLY MANIPULATED is a better word than curved. Why else do they have “practice questions” on the exam and why else wouldn't you be able to walk out of prometric with score in hand? They KNOW before they ask every question approximately how many candiates will get it right or wrong and they use those results to configure the testing is such a way that only ~50% will actually pass the entire test. So while they might not curve the results against other candidates, they most certainly are engineering the exams in such a way that they keep the overall passing rates where they want them to be…..

[Anonymous]

@dtomasello It sounds like splitting hairs, but curving and statistically manipulating sound very similar. If you're saying that a certain percent are suposed to pass, it has the same effect as curving.

I can cook bacon in a microwave or on a skillet, either which way the bacon is cooked.

[Anonymous]

What people are basically saying is that each question is worth a certain # of points based on difficulty. How do they know the difficulty of the questions? Based on the performances of other candidates!! That’s why they have pretest questions. They are trying to see how many people get the questions right. From there, they decide whether to toss out the question for future exams or to put it in for a future exam. If they decide on the latter, they either place it in the Medium Difficult category or the Difficult category. And then afterwards, they decide on how many points the question is worth based on difficulty.

How was all this determined? By looking at how other candidates performed on the questions. And that is essentially the same as curving.

[Anonymous]

@bobkorz – “Curved” to me implies the scores are adjusted after the window closes based on how everyone else did. That's not true, your results are statically predetermined before you even sit down. Everyone COULD pass in a window if they really, really wanted too. It's not like the AICPA will fail you anyway after the fact to hit that 50%. They just know its incredibly unlikey based on past statisical research of the questions they have asked.

Yes, what they are doing results in a way to curving but curving implies what another person does on an exam in the same window will affect your result and that's absolutely not true. It will however affect people in future windows…

@CPAMan – The key difference is your results will affect candidates in future windows not the window your sitting in. I don't consider that curving, maybe other people do. If so then we're just splitting hairs.

[katiekanton]

@dtomasello

You may very well be right, but unless you have an inside contact at the AICPA who has told you this, there's no evidence for it. They have not released ANY details about how the score you earn is mapped to the score you receive. I personally believe it's a based on a percentile with a few adjustments.** I think it would be great if the AICPA established a baseline of points that each candidate has to accumulate by answering MCQs correctly to be deemed knowledgeable enough to pass, and if 10% of the candidates achieve that baseline, then only 10% pass, and if 85% pass, then 85% pass. However, that isn't backed up by any released information, and certainly isn't supported by the pass rate that's freakishly and unrealistically consistent.

**e.g. if the candidates scores are ranked lowest to highest, the 45% mark is found, and if that mark is found at say, 162 points, then everyone who scored 162 points and above passes. I suspect they do this three times for each window, giving us the waves. This would account for the “pass rate” being a little different every time but always hovering around the same mark.

AUD - 88
FAR - 90
REG - 85
BEC - 88

[Anonymous]

@KatieKanton – They have released some details about the scoring process. I attached a few links below. What you described in your foot note is curving. The AICPA has DEFINATIVELY said they do not curve. Your scores are based SOLELY on how you perform, which they can generally already “predict” based on past statistical models. That's how they keep the pass rates at 50%.

These aren't all encompassing but worth reading to give you a general idea of the model they use:

https://www.aicpa.org/BecomeACPA/CPAExam/PsychometricsandScoring/ScoringInformation/DownloadableDocuments/How_the_CPA_Exam_is_Scored.pdf

https://www.aicpa.org/BecomeACPA/CPAExam/ForCandidates/FAQ/Pages/computer_faqs_3.aspx

[katiekanton]

I've read those. I didn't see any information whatsoever that reveals how they tie your earned score to what they call your “reported score”. I also disagree that what I'm describing is curving. My understanding of curving is, you take the highest score, add however many points you need to to get that to 100, and add the same points to everyone else's score. They obviously aren't doing that, therefore they are technically telling the truth when they say they don't curve.

AUD - 88
FAR - 90
REG - 85
BEC - 88

[Minimorty]

dtomasello and I don't always agree, but he and I have been preaching the exact same thing on scoring a while now. I couldnt agree more with his interpretations of the scoring method. The AICPA has released enough information to make a logical deduction on how it is scored. @dtom has described it quite well.

[katiekanton]

Followup: My theory of how they do it may or may not be how it's done. One thing I know for certain is that it is definitely not based on a percentage correct. I know that from my own testing experience, and I know that because it says so in the score report. I know how many bizarre questions I got that I didn't know the answer to and still walked out with high scores.

For the record, I'm not disagreeing about weighting of questions. I believe there's a way that weighted questions could be used to establish an array of many different possible accumulations of points, and say 200 of such points earns you a 75. I also think this would be a better way to do it and more effective at protecting the public.

AUD - 88
FAR - 90
REG - 85
BEC - 88

[Minimorty]

I'll just add that you have to remember that the AICPA has years and years of data with thousands upon thousands of candidates taking the exam. They have plenty of data to make appropriate weighting decisions that would result in a 42%-50% pass rate. That range is not exactly “freakishly and unrealistically consistent.”

[Author]

Replies