Linking Evidence to Practice: Data Systems That Can Be Used to Improve Teaching and Learning

1
Linking Evidence to Practice Data Systems That
Can Be Used to Improve Teaching and Learning

• Dan Goldhaber
• University of Washington and Urban Institute

2
Why Focus on Teachers?

• Teacher quality is the key to school success
• Its impact is far larger than any other
  quantifiable schooling input (Coleman et al.,
  1966 Goldhaber, 2002 Rivkin et al., 2005)
• The estimated impact of teacher quality is large
• It can explain more than a full grade-level
  equivalent in test performance (Hanushek, 1992)

3
Central Thesis

• K-12 education should be a continuous learning
  system, but its not
• Key missing ingredients data and analysis
  capacity
• Combined, these could be used to convincingly
  answer key policy questions
• In the absence of such a system decisions are
  made based on fads, institutional inertia, and
  politics

4
Weve Actually Learned A Lot from Longitudinal
Data in Select States

• Data Quality Campaign specifies key
  characteristics of state longitudinal systems
• The most important schooling research is going on
  based on data from a select number of states (see
  CALDER)
• Three lessons about schooling
• Teacher quality is the most important factor over
  which schools have control
• Teachers effectiveness varies considerably
• Teacher credentials are weak, at best, predictors
  of teacher quality

5
Teacher Quality Appears to be Primarily
Unobservable
back
6
Experience Levels
back degrees NBCTs
7
Degree Levels
back experience NBCTs
8
NBPTS Certification Status
Difference in means 0.19 sd of teacher quality
back degrees experience
9
Bad Data Limited or Bad Research Wrong Answers

• In the absence of this linkage one cannot assess
  most teacher-based policies
• Teachers and students are not randomly matched to
  schools or classrooms
• Miss impacts on subgroups
• Aggregation bias
• Matching process produces erroneous estimates of
  the impact of teacher credentials

10
Hurdles Associated with Student-Teacher Linkages

• FERPA
• Cost
• Politics

11
Conclusions

• We could learn a great deal more about schooling
  than we do based on the data that are already
  collected
• Data systems need to be able to talk with one
  another
• We cant afford not to create data systems that
  allow us to emprical test the efficacy of
  schooling policies and practices
• The student achievement costs of not having K-12
  education be a learning system are not readily
  apparent, but they are real