Skip to main content

Selection Bias

Go Search
Home
About
New Atlas
Atlas, A-Z
Atlas Maps
MPP/MPA Programs
Subjects
Core Topics
Illustrative Courses
Topic Encyclopedia
Concept Dictionary
Competencies
Career Tips
IGOs
Best Practices Project


 
PPGPortal > Home > Concept Dictionary > S > Selection Bias
 
Selection Bias

A bias that is present when choosing between the test group and the control group due to preexisting differences between the two groups.

(OECD, 2007, p. 7)

---------------------------------

The beneficiaries of an intervention may be selected by some criteria (or select themselves), which is correlated with the observed outcome. For example, entrepreneurs being selected for microcredit or for a business development scheme may have done better than those who did not bother to apply, even in the absence of the support. Hence comparing outcomes of beneficiaries and non-beneficiaries can give misleading results. Where these criteria are not observed (i.e. there are no data on them), then there is a bias in the impact evaluation findings (this point is discussed further below). However, where the determinants of participation are observed, then the bias can be removed using quasi-experimental methods.

References

Organisation for Economic Co-operation and Development (OECD) (2007). “Outline of Principles of Impact Evaluation”. Retrieved on June 25, 2008, from
http://www.oecd.org/dataoecd/46/16/37671602.pdf.
     


Selection Bias

A bias that is present when choosing between the test group and the control group due to preexisting differences between the two groups.

(OECD, 2007, p. 7)

---------------------------------

The beneficiaries of an intervention may be selected by some criteria (or select themselves), which is correlated with the observed outcome. For example, entrepreneurs being selected for microcredit or for a business development scheme may have done better than those who did not bother to apply, even in the absence of the support. Hence comparing outcomes of beneficiaries and non-beneficiaries can give misleading results. Where these criteria are not observed (i.e. there are no data on them), then there is a bias in the impact evaluation findings (this point is discussed further below). However, where the determinants of participation are observed, then the bias can be removed using quasi-experimental methods.

References

Organisation for Economic Co-operation and Development (OECD) (2007). “Outline of Principles of Impact Evaluation”. Retrieved on June 25, 2008, from
http://www.oecd.org/dataoecd/46/16/37671602.pdf.

Approved for glossaryposting by Ben Eisen on December 5, 2010


Important Notices
© University of Toronto 2008
School of Public Policy and Governance