And of course there is this problem:
http://www.washingtonpost.com/wp-dyn/content/article/2011/01/18/AR2011011804873.html
Thursday, August 4, 2011
Consistency of ratings in DC's IMPACT
So DC has just fired 206 Teachers due to ineffective ratings.
http://www.washingtonpost.com/local/education/206-low-performing-dc-teachers-fired/2011/07/15/gIQANEj5GI_story.html
Lets look closely at this data.
566 teachers were rated minimally effective year one
528 were rated minimally effective in year 2.
141 of those teachers were rated minimally effective both years
60% of teachers rated as minimally effective in year one were rated higher in year 2. (That would be about 339 teachers)
That should leave 89 teachers that were rated as minimally effective in year 1 and didn't return, or were rated as ineffective in year 2.
and here is the point... that leaves
387 teachers who were newly identified as ineffective in year 2. That is roughly 73% of those teachers rated as minimally effective were either new teachers, or were actually rated effective in previous years. This raises two issues that are not well address by this system or article.
First, that the assumption in firing teachers is that there is a ready source of effective teachers waiting to be hired. If new teachers fill the ranks of the minimally effective ranking, this assumption is not supported.
Second, if a large number of teachers rated as minimally effective were actually rated as effective in the past, we are presented with the hard to accept conclusion that past rating is not necessarily a good predictor of future rating. So either actual teacher quality jumps around a considerable amount, in which case we can't be guaranteed that firing today's ineffective teacher would have actually been ineffective tomorrow, or if there is a large amount of error in these ratings it exlains the large percentage of teachers moving from ineffective to effective as one due to expected chance rather than actual improvement.
To demonstrate this in another way, notice that we have a total of 3341 teachers in year 1, where 641 earned ratings of minimally effective or lower, which is about 20%. If we identified teachers at RANDOM we would expect 4% of teachers to be identified two years in a row. 4% of 3341 is ... drum roll please ... 132.
So when we consider the 141 that were rated as ineffective twice, we should remember that we would have expected 132 to be identified by random.
http://www.washingtonpost.com/local/education/206-low-performing-dc-teachers-fired/2011/07/15/gIQANEj5GI_story.html
Lets look closely at this data.
566 teachers were rated minimally effective year one
528 were rated minimally effective in year 2.
141 of those teachers were rated minimally effective both years
60% of teachers rated as minimally effective in year one were rated higher in year 2. (That would be about 339 teachers)
That should leave 89 teachers that were rated as minimally effective in year 1 and didn't return, or were rated as ineffective in year 2.
and here is the point... that leaves
387 teachers who were newly identified as ineffective in year 2. That is roughly 73% of those teachers rated as minimally effective were either new teachers, or were actually rated effective in previous years. This raises two issues that are not well address by this system or article.
First, that the assumption in firing teachers is that there is a ready source of effective teachers waiting to be hired. If new teachers fill the ranks of the minimally effective ranking, this assumption is not supported.
Second, if a large number of teachers rated as minimally effective were actually rated as effective in the past, we are presented with the hard to accept conclusion that past rating is not necessarily a good predictor of future rating. So either actual teacher quality jumps around a considerable amount, in which case we can't be guaranteed that firing today's ineffective teacher would have actually been ineffective tomorrow, or if there is a large amount of error in these ratings it exlains the large percentage of teachers moving from ineffective to effective as one due to expected chance rather than actual improvement.
To demonstrate this in another way, notice that we have a total of 3341 teachers in year 1, where 641 earned ratings of minimally effective or lower, which is about 20%. If we identified teachers at RANDOM we would expect 4% of teachers to be identified two years in a row. 4% of 3341 is ... drum roll please ... 132.
So when we consider the 141 that were rated as ineffective twice, we should remember that we would have expected 132 to be identified by random.
Subscribe to:
Posts (Atom)