Welcome to the Profit of Education website. Continuing the conversation begun in the book Profit of Education, we discuss the latest economic evidence on education reform.

Grumpy academic and assessing first-year teachers

The New Teacher Project has released findings, Leap Year, from their program for assessing first-year teachers. The TNTP report provides some great information that can certainly help school districts work with first-year teachers.

I found one statement especially exciting.

Multiple measures tend to point to the same conclusion about a teacher’s potential. Teachers who do well on one ACE measure earn generally high scores overall.

If you look at value-added data, i.e. test score data, as I do, it’s important to know that such data matches up reasonably well with more subjective measures. So the claim is great news…except it ain’t true.

If you read the whole report you find the actual correlations between different measures.

Correlation of ACE measures

The fact is that none of the measures is very highly correlated with any of the others. And note that the value-added measures have particularly low correlations.

I’m not saying there’s anything wrong with TNTP’s overall evaluation method. But please don’t tell us that different measures all tell us the same thing when your own calculations show the opposite.

Share
This entry was posted in Uncategorized and tagged , , . Bookmark the permalink.

2 Responses to Grumpy academic and assessing first-year teachers

  1. Jesse Rothstein says:

    The MET data showed similarly low correlations (though I’m not sure they are exactly comparable — MET reported mostly “disattenuated” correlations that abstracted from year-to-year variability in value-added). And MET drew similar conclusions from those low correlations. See the review I wrote here:

    http://nepc.colorado.edu/files/ttr-final-met-rothstein.pdf

    I read the evidence the same way you do. But what to make of it? I’m not sure. I don’t see any way to decide which of the measures is/isn’t getting at something important except by validating them against each other. So is there a better way to learn about this than simply to try implementing policies based on various combinations of the measures, then evaluating those policies?

Leave a Reply

Your email address will not be published. Required fields are marked *