Dismissal threats increased the voluntary attrition of low-performing teachers and improved the performance of those who decided to remain

While the effects of teacher quality on student development, achievement, and
later outcomes have been widely studied, there is no agreement on how to
systematically drive improvements in the quality of teachers. Teacher salaries
are traditionally based only on experience and credentials. However, these
traits may not have consistent links to teacher quality. In a push toward "pay
for performance," teacher compensation is increasingly measured by teacher
performance evaluations, such as the IMPACT policy that was introduced in
District of Columbia public schools in 2009.

In Incentives, Selection, and Teacher Performance: Evidence from IMPACT (NBER Working Paper No. 19529), Thomas Dee and James Wyckoff examine the effects of the IMPACT program on the retention of high- and low-performing teachers and on the subsequent performance of
teachers who were retained.

The IMPACT program established several explicit measures of teacher performance
with a particular emphasis on structured classroom observations of teachers'
instructional practices. Overall measured performance implied both large
financial incentives for high-performing teachers as well as the threat of
dismissal for persistently low-performing teachers.

The IMPACT program used thresholds to determine the effect of measured
performance on both pay and dismissal, so the authors were able to compare the
retention and performance outcomes among teachers whose prior-year performance
scores placed them just below or just above the threshold values for receiving a
permanent increase in base salary or a dismissal threat. They argue that
teachers who score just above or just below such a threshold are quite similar
and that the large disparities in the consequences of their scores provide an
opportunity to study the incentive effects of the IMPACT program.

The results indicate that dismissal threats increased the voluntary attrition of
low-performing teachers and improved the performance of those who decided to
remain. Moreover, financial incentives further improved the performance of
high-performing teachers. Interestingly, most of the action comes in the second
year of the program, when it was clearer that the program was politically
durable. In the second year, the dismissal threat increased the attrition of
low-performing teachers by 11 percentage points, an increase of over 50 percent.
The performance gains among remaining teachers were equivalent to moving a
teacher from the 10th to the 15th percentile of the performance distribution.
--Claire Brunel