Thursday, July 17, 2008

How to massage data. ( KIPPing it real. )

Let me say at the start that I don't disagree with the KIPP schools. I believe they think they are doing the right thing with their students. I am not criticizing their methods or policies. I am focusing here on a statistic and the misuse of that statistic in making public school policy and curriculum decisions.

There's an interesting article in Newsweek, by John Alter, in which he says
"The irony is, we know what works to close the achievement gap. At the 60 KIPP (Knowledge Is Power Program) schools, more than 80 percent of 16,000 randomly selected low-income students go to college, four times the national average for poor kids."
That's an impressive statistic until you look around. "80%" applies only to the original couple hundred students, not the current population of 16,000, and even then the sample is skewed.

Even KIPP agrees ... from their website, "This national college matriculation percentage only includes students who attended the original two KIPP schools in Houston and New York. At this time, those are the only KIPP schools that have been in operation long enough for alumni to progress from eighth grade to college freshmen."

I get annoyed when facts like these are thrown into the public view where "education experts" (i.e. big mouth, no current teaching experience) and administrators (i.e., big mouth, no teaching experience) take them and make really bad decisions.

That line "We know what works to close the achievement gap" is very powerful to these people. I have had a principal say, "Longer school days work for KIPP and we should have that here." His proof was an Internet article similar to this one; it was enough for him.

This is not a double-blind study with treatment & control groups. No one is controlling for confounding factors. There is no analysis of correlation / causation. The sample size is small and biased, applying to the "randomly selected" (from out of KIPP schools, not the city as a whole) from a selective group (those who made it through 4 years +) chosen from yet another sub-group (those admitted to KIPP schools).

So, specific issues ...

(1) The students are not chosen randomly from the city population: they are screened by an application process. Before the program even has a chance to work or not, this weeds out many who won't succeed -- in KIPP or in college.

(2) It ignores those who might have attended successfully but didn't for other reasons -- such as an unwillingness to spend 10 hours a day doing heavy-duty drill and practice for standardized tests. "KIPP schools relentlessly focus on high student performance on standardized tests and other objective measures. " Again, I don't have a problem with KIPP here, just Alter's assumption that this is the gloriously perfect solution.

(3) It ignores all those who do not stay in college after the freshman orientation.

(4) The students have completed 5th through 8th in a KIPP program AND four years of high school. From KIPP "As for how we track students, we do this through the KIPP to College alumni program at both schools. The KIPP to College program aims to empower each middle school graduate to continue using the scholarly habits, knowledge, and skills learned at KIPP schools as they continue on the path to college (and beyond.)"

How about that? Anyone who isn't part of the "KIPP to College alumni program", isn't included in the statistic!
  • Can you imagine anyone going through eight years of this (see below for some more details of the program) while not having the proper attitude for college?
  • Can you imagine continuing KIPP to College program in high school if you have no intention of going to college all along?
Frankly, I'm amazed that number isn't closer to 95% if they're going to calculate it that way.

Remember how scientists get temperatures close to absolute zero? Conventional methods get you close but one of the final steps is this: you have a pool of atoms, some hot and some cold. Hot ones bounce higher and are shot off with a laser, leaving a cooler overall substance. After a while, it's not surprising that you can take averages of the remaining ones and claim a good success rate.

There are so many places and ways to leave that attrition is more common than graduation. There is no way to know whether KIPP is teaching particularly well, but it sure can winnow them out. The fact that a student has the parents, patience, willingness and dedication to tolerate the work and the schedule may be the reason that he or she goes to college. I would certainly think so. Whether the KIPP style of teaching is effective has not been introduced, studied or discussed.

Please don't assume that it will work for the rest of the educational system based on this data.

Some other information from the KIPP site ...
Time in school is 60% greater than that of a normal PHS and is a big factor: 7:25 a.m. til 5:00 p.m. (4:00 p.m. on Friday.) Saturdays 9:15 a.m. til 1:05 p.m. Summer school.

The student contract ...
"I will always work, think, and behave in the best way I know how, and I will do whatever it takes for me and my fellow students to learn. This also means that I will complete all my homework every night, I will call my teachers if I have a problem with the homework or a problem with coming to school, and I will raise my hand and ask questions in class if I do not understand something.
"I will always make myself available to parents and teachers, and address any concerns they might have. If I make a mistake, this means I will tell the truth to my teachers and accept responsibility for my actions.
"I will always behave so as to protect the safety, interests, and rights of all individuals in the classroom. This also means that I will always listen to all my KIPP teammates and give everyone my respect.
"I will follow the KIPP dress code.
"I am responsible for my own behavior, and I will follow the teachers' directions.


  1. Much of what you have written, is, ironically in a post about Alter's incorrect use of statistics, incorrect.

    To your issues:

    1) KIPP schools do not have a screening process. They have a lottery for their admissions. You made the screening stuff up, or believed the lies presented on other blogs. Much worse than Alter, who didn't say anything that was factually incorrect.

    2) KIPP schools in most cases get a population that very closely resembles their district population.

    3) It does not ignore those who don't stay in college after freshmen orientation. This is a complete fabrication. KIPP defines its alumni as kids who finished 8th grade with them.

    4) Every student who stays with KIPP through 8th grade is in the KIPP through College program. So once again, your conclusions are founded on misinformation.

    In sum, KIPP gets typical district kids and gets them excited about school and invested in college, so they are much more willing to work the extra hours. That's what makes KIPP much more successful than district schools, despite all the lies that you made up about them.

  2. Anonymous poster takes issue with some of my points, calling them lies and fabrications. I don't appreciate being called a liar, but here's an explanation. If it's not good enough, too bad.

    "1) KIPP schools do not have a screening process. They have a lottery for their admissions."

    Perhaps the problem is in my phrasing and not making it clear that I'm speaking from a statistic perspective. If one were to expect a study to predict behavior of a population, as Alter does in his piece, then one must randomly select from the population. KIPP uses a lottery to choose from people who have applied. At this point your sample is biased to the 2000 or so (a guess) who applied, rather than the 4 million or so New York students. Which people will apply? Those who agree to a much more rigorous academic situation. Those who agree that this program will work. Those who agree with 9-10 hours a day plus 4 on Saturday, plus summers, plus dress code, and contract and all the rest.

    It seems to work for those who get in and who go along, but is it "we know what works to close the achievement gap." No necessarily for any students who don't agree with your methods.

    The lottery on admissions does not effect this aspect of the bias, but it doesn't improve it either.

    "2) KIPP schools in most cases get a population that very closely resembles their district population." Visually, economically and racially, yes. Academically, no. Remember, only a tiny portion of the population as a whole is willing to accept the program.

    "3) It does not ignore those who don't stay in college after freshmen orientation. This is a complete fabrication. KIPP defines its alumni as kids who finished 8th grade with them."

    First off, there hasn't been a KIPP student who's finished college yet. So maybe my point should have been worded better.

    My point should have been that I dislike people who rate a school by how many show up at college rather than by how many finish it.

    In (4), you say that KIPP alumni stay in the KIPP to college program. If you aren't counting those who drop from your program after 8th grade, then you've introduced another bias into the sample since it's logical to expect that kids will stay in KIPP to College only if they feel they still have a chance to go to college.

    For this to be more applicable, you would have to compare the college matriculation to the number of kids who entered the program way back in 5th grade. Reasonably, maybe limit that to those who made it through at least a year.

    It still doesn't make KIPP anything for PS123 (made up school) to emulate, because PS123 isn't working with the group KIPP is.

    It doesn't rule it out either. Longer hours may make a difference. "relentlessly focus(ing) on high student performance on standardized tests and other objective measures" might indeed be the way to go. Longer school year may make a difference.

    Your "80% success rate" isn't the right statistic.

    And, sorry to say, none of that's a lie, no matter how much you want to think so.

    Have a nice day.