New Jersey should be proud to have a resident as expert and dedicated to math education as Eric Milou weighing in on PARCC. Â I suspect I will like him when I meet him. Â He has more experience in the field than I do.
I am afraid, however that his position on standardized testing will be used to lead the state on a path away from the demonstrated excellence it has achieved in public education.
Milou wrote a cogent argument against PARCC you can read here.Â He makes sixÂ assertions that are worthy of discussion
- The PARCC approach is a step forward from â€œbubble tests,â€ but only a minor one.
The use of the term â€œbubble testsâ€ is pejorative on its face. Â It describes a test instrument in which a question has one and only one correct response, which in days past, a test taker would indicate by filling in a bubble with a number two pencil. Â In the internet world, it is known as a â€œradio button,â€ a single selection that can be made to the exclusion of all others. Â I took the PARCC Algebra I test.Â There were a bunch of items in response to which I had to select one to six correct responses. Â Instead of say, four mutually exclusive bubbles to color, the student had 26 (64) possible responses to the test item, only one of which, apparently, was correct. Â In the Algebra I test first unit, three of fourteen questions were to be hand graded, including an evaluation of a studentâ€™s work process.
- There are approximately 40 total questions on the PARCC exam in any given grade.
The implication is that a student cannot be fairly evaluated on a test of only forty questions. Â This is a matter for statisticians to evaluate.Â Is a forty question test a reliable and valid measure of student achievement? Â I look forward to the day that PARCC is an adaptive test, one that tailors the selection of easier or harder questions based on the studentâ€™s previous answers. Â With an adaptive test, the number of questions a student addresses is lower and the exam can be completed quicker.Â It is also capable of discriminating performance along a longer spectrum of abilities.
DescribingÂ the test as having forty questions is disingenuous. Â When you take one you will see that a â€œquestionâ€ has multiple â€œparts.â€Â For each part you can get points or not. Â That sure sounds like a question to me. Â Each part may have as many as seven options as described above. Â Each requires a decision by the test-taker.Â Itâ€™s not as if the student selects forty correct answers and is done. Â At some later date, Iâ€™ll retake the test and count the number of decisions it takes to complete what is advertised as a forty question test.
- â€œNo student who has taken PARCC is now in collegeâ€ so predictions of college success and comparison with the SAT are invalid.
Thatâ€™s true, as far as it goes. Â It wonâ€™t be true for long though. Â Itâ€™s a specious argument. Â In controlled studies PARCC has been compared to the SAT, the Massachusetts test suite (MCAS), and high schoolÂ GPA as a predictor of college success. Â GPA is far and away the best predictor.Â The correlation of high school GPA to college GPA is .845, an almost perfect predictor (Belfield 2012).Â For mathematics, the correlation with college grades for scores on the two PARCC integrated math components (0.43) is also statistically indistinguishable from the association for MCAS math test scores (0.36) (Nichols-Barrer 2015).
Those were studies of a sample population. Â In a couple of years, weâ€™ll have the big numbers, from classes of high school graduates. Â Thatâ€™s long before the 2021 use of the PARCC for New Jersey graduation.
- Scoring 1, 2, or 3 on a NJ PARCC test was no indication that remedial classes were required in college.
Milou is absolutely correct.Â The 650-850 scores on the NJ PARCC tests may be reliable, but the classification into five levels is way off. Â Besides discouraging students and parents, it brings doubt on the whole New Jersey system. Â There is no way that 65% of NJ seniors are not ready to graduate.Â Getting the cut score right needs to be job one for the NJ BOE, and quickly.
- There is insufficient disclosure about the conversion of raw scores to the 650 to 850 point scale and thence to the five levels.
Milou is absolutely correct.Â If the process is secret, there can be no public trust in the results. Â For all parents know, the scoring mechanism is a random number generator. Â Itâ€™s perfectly OK to have some non-scored questions inserted as tests. Â Itâ€™s not unreasonable to have some questions weighted more heavily.Â Those details and the test bank itself probably need to be secret. Â The process, though, needs to be visible.
- Milou goes on to assert that â€œstandardized tests like PARCC will never raise the bar.â€
Milou misses the point in his conclusion. Â PARCC need not raise the bar. Â It needs to equalize the bar for all.Â Track and field has a rigid protocol for measuring the bar: from the ground to the lowest point on the top surface of the bar with a steel or fiberglass tape (laser in some competitions). Â Without a defined protocol, any record report is subject to question. Â Likewise for the PARCC.Â It provides a common test for all Algebra I students in New Jersey. Â It should eliminate bias from teachers who inadvertently provide tests easier or harder than the standard curriculum.
When we issue a New Jersey diploma to a student we owe it to that student to ensure that it means something, not just a certificate for showing up in school. Â Itâ€™s not a kind gesture to let a student slide through an Algebra class only to have him fail in the world because he cannot figure out whether to accept a more attractive job offer at a lower salary rate.