Go back to transcript of Jimmy
Mr. Herr: Good morning. My name is Norm Herr from Cal State Northridge,
a professor in science education. It's my pleasure this morning to introduce three
speakers on the implementation of standards and how we implement them. The first
speaker will be Sandy Horn. Sandy just arrived from Tennessee, she will be going
back tomorrow for a triathalon so we need to allow her time to catch her plane. She
will be talking about value added teaching in Tennessee. She has done research for
the University of Tennessee and also is a high school teacher. Come on up, Sandy.
Ms. Horn: Hi. Yesterday I did a full day's work as Head Library Media Specialist
at South Doyle High School in Knoxville, Tennessee before catching my flight to here.
The reason I say that is because I want you to understand that the stuff I do at
UT is very much informed and is part of the loop that my experience in the classroom
informs which is informed, in turn, by my experience with value added assessment.
I don't know how many of you have heard of the Tennessee Value Added Assessment System
developed by William Sanders at the University of Tennessee. I worked with Bill for
about the last 7 years, since TVAAS was put into the Education Improvement Act in
Tennessee as part of the accountability measures that went along with a bunch of
extra funding because, as you know, these days when you get extra money you are expected
to be accountable for it. Fortunately, in Tennessee they picked Bill's method.
I'll tell you a little bit about TVAAS, but first I'm going to state some things
that may be obvious or self evident to all of us. The first is "The measure
of whether education is successful is whether a student learns." The second
is "Parents have a right to expect that their child will progress at a normal
rate, AT LEAST, every year." Thirdly, "It is not excessive or unreasonable
to expect that a teacher will produce normal academic gain in his or her students
over the course of a year."
It should go without saying that educational programs must be able to demonstrate
that they produce gains in academic learning of students.
But teachers are evaluated on how they look. Programs are evaluated on how good they
feel. Data are met with suspicion. Education cries "Spare me the facts."
And for these reasons it's been a very interesting 7 years with TVAAS. What TVAAS
is is a system for estimating the effects of schools, school systems and teachers
on the academic gains of students. It is a statistical mixed model methodology fitted
to a massive, longitudinally-merged data base of educational data. It's unique in
the field of education. You are looking at a woman very much in love with this data
base. It's currently composed of about 6 million records, longitudinally merged,
including annual test scores for 5 subjects administered to all the students in Tennessee
in grades 2-8 and high school subject area tests, some of which are now on line and
some of which are in developmental stages. The new tests will be brought online in
the next few years.
We keep these data for several years, employing new data to revise past estimates
as they become available. All of our estimates are based upon at least 3 years of
data and sometimes as many as 5.
In TVAAS, each child serves as his or her own control, which makes it possible to
partition educational, socioeconomic, and environmental effects that confounded prior
attempts to use data for educational assessment. Conceptually, TVAAS models a growth
curve for each student. Although we don't expect these curves to be smooth any more
than we expect a child to grow the same amount in height each year, the deviations
in a child's growth curve can tell us some things. The deviation in ONE child's growth
curve can tell us very little. That child may have had a stomach virus the morning
of the test. That child may have had a divorce in the family. That child may have
had a wonderful event happen that made learning much easier for him that year. But
if you aggregate the deviations in the growth curves of many students over several
years, and if those deviations occur when those students happen to be in the classroom
of a certain teacher or in a certain school, then we can say something about the
educational effects of that particular teacher or school or system.
Now, I only have a few minutes to talk to you, but I'm going to talk to you mostly
about some of the things we have found out because we have this methodology. I'm
going to tell you first of all and foremost of all, it's the teacher. It doesn't
matter about a lot of other things. We have done research to check on classroom heterogeneity--how
homogeneous is the academic level in the classroom. We have looked at class size.
We have looked at ethnicity and socioeconomic levels and what matters is the effectiveness
of the teacher. Period.
Everything else is trivial in comparison to the teacher effect. I'm not using that
term statistically. Trivial, though in my estimation as a teacher in the school and--when
the effects of other variables we have studied are not trivial statistically, we
have found that they are, consistently, of far less importance than the teacher effect.
Here are things we thought we knew about education and I'm speaking very broadly
here when I use the word we: "Poor kids are poor students." We've looked
at the distribution of gains in schools across the State of Tennessee, and based
on this huge data base, we do not see any correlation between the socioeconomic level
of those students and the likelihood that they will achieve normal or above normal
gains or below normal gains. It's simply not correlated.
Secondly, we thought we knew "minority kids can't learn." At least that's
the unstated assumption of many people including some teachers who think if they
have poor or minority kids they can't expect much from them. That is simply not the
case. That is not correlated with academic gain in Tennessee.
"Special education kids can't learn either." We've found that our special
ed kids made gains quite comparable to the other children in the classroom. As a
matter of fact, the only real deviation related to academic level we've found is
that the higher performing kids don't score as large gains as the lower performing
students. We've found that only the most effective teachers can bring those kids
We've found that teachers at the median of effectiveness can produce normal gains
in most of their kids, but they still fall behind in producing gains in their higher
level kids. Another thing that we found is that the more effective the teacher, the
less likely they are to be in minority predominant schools.
What does this say? It says perhaps it's true what all of those deconstructionists
were saying--that we do perpetuate this gap in education by assigning less-effective
teachers to those schools. We have the data to back that up, in Tennessee at least
in two of our largest metropolitan areas.
We thought we knew that if it feels good, it is good. Right now we are looking at
the New American Schools models in effect in Memphis, Tennessee, which is your largest
school district. There are 14 different models testing out now. Only one or two are
actually effective. You wouldn't know that and you wouldn't know which ones if you
weren't looking at the student outcomes.
"If it feels good, it is good." Maybe that's why they call it educational
theory. (Laughter). I don't know, but a lot of what we learn from our colleges of
education and a lot of what we were told when we were being trained taught us to
look with suspicion at any kind of scientific data. I hope that that is changing
now but I do not see that as fact. Teachers are not taught how to look at data. They
don't know how to use student test scores and to find their own student gains. Some
of them are puzzled by scatter graphs.
"If it looks like a good teacher, it is a good teacher." Well, in Tennessee
many of us are trained by the Madeline Hunter model. That used to be how we were
trained--how we provided the proper set and closure and what came in between. But
we who deal with value-added assessment do not pretend to know what you should look
like when you teach. Personally, I think if you want to stand on your head and whistle
Dixie in the corner, if your students learn and consistently learn and produce wonderful
gains, I'm going to come learn why from you.
We also thought the good teachers can make up for bad teachers. We know now through
research that Dr. Sanders and June Rivers have conducted that that is not true. The
effects of a bad teacher do not go away over time. Children who have two consistently
ineffective teachers will never achieve up to the potential they would have if they
had been assigned to better teachers.
We also thought that bright kids would get it on their own. They don't. I already
told you that. We also thought in the past that high test scores meant a good school.
That is not true. In Tennessee we have brain trust schools. We have Oak Ridge, Tennessee,
where the rocket scientists are. In another section of the state, there are enclaves
of highly educated professional people who work for other large companies. There
was a large difference between the Oak Ridge gains and those of some of the other
comparable systems. In Oak Ridge, the students are producing excellent gains even
though they are one of the highest scoring schools in the State. In other equally
privileged systems, they are just now catching up because they were resting on their
genetically-endowed laurels. The kids were in the 98 percentile. They were born that
way, were economically and educationally advantaged, and the schools got them that
way. Their kids are now making exceptional gains, too, at least in part because of
the information they received from their value-added reports. We mustn't ignore those
children. They do not get it on their own at all. It's the teachers.
These are things that you can learn if you want to know it. You can set standards
all you want, but unless you actually look at the data, you don't even know whether
those are the right standards. You have to find out whether you are making a difference.
You need to look at the difference. First, what did you get before the Standards?
Then you need to see what the difference is after the Standards are implemented.
Now there are lots of ways to do that, but one of them should involve objective,
statistical measures of student outcomes.
Now why do I say that? Because we don't need snap shots in education. I cannot tell
you everything that a student knows on the basis of his or her scholastic test scores
any more than you can tell me what a student can do entirely based upon what is sometimes
termed authentic assessment. We need all of those things. We have to take every bit
of information that we can gather--very bit of reliable and valid information we
can gather to compose a holograph, so we can walk around this monster we are dealing
with and try to make it into something that can be of service to all of us. Thank
you for having me. (Applause)
Mr. Herr: We have a few minutes for some questions so please come up to the microphones
Audience member: I have a question. It was so good to hear you. In our district socioeconomic
status is the last column on the right. An excuse to how they perform. It's the parent's
education level. You are telling us that doesn't matter.
Ms. Horn: Unless you are looking at raw test scores. If you look at how much the
students learn it has no bearing whatsoever.
Audience member: Thank you for saying that.
Audience member: We've had a lot of discussion over the past years in California
about what kind of data we should be looking at. We went through norm reference test
and criteria reference tests. It's hard to get good longitudinal data. What I wanted
to ask was are your data and methodology available in a form-- are they on the web
in some form or can can you tell us how to look in more detail at the work you've
done. It sounds interesting and helpful.
Ms. Horn: We have published quite profusely in several -- there are a couple ofbooks
in particular. Jason Millman's books called "Grading Teachers -- Grading Schools"
details the methodology pretty clearly, I think. That's the first place I would send
you. We have several articles in "The Journal of Personnel Evaluation in Education."
The Tennessee data, of course, links student data to teachers to schools to systems.
We trace students as they move among teachers and schools and systems in Tennessee.
We can share general data. We cannot give you specific data. Individual teacher data
is entirely a matter of privacy in Tennessee. That's something you have to think
about because, in some states, all teacher records are of public record. In Tennessee,
by law, that is private matter. All of the school and system data is public and that's
available to everyone. By the way, we are very very open to research projects with
responsible people who are interested in learning from the data, because we don't
have the personnel or the time to do all of this ourselves, to do all the research.
I encourage you to contact us at the University of Tennessee Value-Added Research
and Assessment Center.
Audience member: My question is very simple. I would like to know which data you
use. Do you use percentile ranking. Grade equivalent. What data do you use.
Ms. Horn: In the past our State testing data have come from the Tennessee Comprehensive
Assessment Program, from which we used the CTBS/4 part. We now use Terra Nova. We
take the scale scores, but we use the gains that students make from one year to the
Audience member: Thank you.
Audience member: I'm a teacher for 31 years. What I have seen and this is what concerns
me from what you say. I respectfully listen to what you say and buy three quarters
of it. I have seen, because of the pressure put on principals, the teachers are pressured
to manipulate the test scores so that their schools look particularly good. There
are many ways that this is done and I've seen it at a variety of schools and I'm
concerned when decisions are made simply based on test scores, they do not take into
account this little problem. I was wondering how you deal with that.
Ms. Horn: First of all, as specified in the law, Tennessee uses fresh, equivalent,
non-redundant tests every year. You can't teach to the test precisely because everything
changes from year to year. We also have ways of detecting footprints. It just doesn't
happen that often. Thirdly, we use gains, and because we use gains, everybody can
be a winner. We don't have a median line where half the schools are below and half
are above. What we have is a line that is the normed gain. We think every school
can achieve that normed gain. And if teachers achieve that normed gain year after
year, some people have said that is pretty low standards. Well it's not. If your
school achieved a normed gain every year, we wouldn't be sitting here worrying about
Audience member: Thank you. (Applause)
Audience member: Tell me if you can't understand me. I have laryngitis. How do you
determine whether your tests are equivalent because we have that in Texas and have
good reason to believe they are not equivalent. Number two, have you used this value
added system to play into a teacher incentive program based on merit because I'm
always told there is no way to identify master teachers and I don't believe that.
Three, Have you encountered
Ms. Horn: Wait. Virginia, can we do these one at a time because I won't remember
them all. The first, about the equivalency of the tests, when we were using the CTBS/4,
that was a part of the contract with McGraw-Hill. The original tests were correlated
to the curricula in Tennessee with the assistant of teacher input. There were meetings
across the State to develop questions and then they were analyzed according to Item
Response Theory by the testing company and whatever else they do to ensure validity
and reliability. Your second question?
Audience member: I think this is what we did in Texas. There were certain test items
that aren't common. That's not enough to determine if tests are equivalent. You would
need a test, retest comparing the tests from one year.
Ms. Horn: Well, yes. Let me put it this way. Our gain scores across the board show
an equivalency. As a matter of fact at one time we did have a real problem with reading
one year in a way that could not--I mean that was state-wide. We went immediately
back to McGraw-Hill. We said there is a problem with this test in particular, because
we have this data on a state-wide level from all the kids including special ed.
Audience member: It couldn't be true.
Ms. Horn: We knew it couldn't be true. We are keeping an eye on the tests for them.
Mr. Herr: I'm sorry I think we need to move on.
Audience member: That teacher question is very important here in California. Has
this test been used to identify master teachers and in essence begin merit pay?
Mr. Horn: No and I'll tell you why. The only people who have access to the teacher
data are the teacher and appropriate administrator. So if we ask for volunteer teachers,
we can do that analysis, but we cannot have a sample that is not self-selected in
order to do that. We do have some people who have come into the State who are doing
that research, but they are doing it with self selected teachers. I still think it's
Go to transcript ofJustine Su