EXPERIMENTS IN THE
PUBLIC SECTOR:
GETTING THE BASICS
RIGHT
Piret Tõnurist
Innovation Specialist, Observatory
of Public Sector Innovation, OECD
June 27, 2017
ADAPTIVE POLICY MAKING: WAITING OR GOING?
27-June-2017 Governing better through evidence-informed policy making 2
 What is evidence under uncertainty?
 Golden decision rule in PA: careful attention to detail,
‘good’ professional or well-considered judgment
 Accurate versus complex decisions – ‘good’ decisions
vary greatly
 Don’t be paralyzed by uncertainty!
WHAT ARE EXPERIMENTS?
27-June-2017 Governing better through evidence-informed policy making 3
 Procedures to support, refute, or validate a hypothesis usually by creating
an intervention which is observed, measured and evaluated
Aerial view of the Hawthorne Works, ca. 1925.
PROCESS OF EXPERIMENTATION
27-June-2017 Governing better through evidence-informed policy making 4
 Frame the problem
 Identify the intended goals
 Identify possible solutions
 Set an hypothesis
 Select a method
 Define the levels of success
 Test
 Review and communicate
results
 Make a decision: pull out, re-
test or scale?
 Launch
 Monitor and evaluate
WHEN TO EXPERIMENT?
27-June-2017 Governing better through evidence-informed policy making 5
 When you don’t know the right answer
 If you know, do! Don’t get caught in the call
for action and ignore well-grounded prior
evidence!
 What is the source of uncertainty
(substantive, strategic)?
 When there is room for act
 When proof-of-concept can lead to broader
positive changeDon’t love the method, love the problem!
WHY EXPERIMENT?
27-June-2017 Governing better through evidence-informed policy making 6
 A ‘controlled small-scale’
 A ‘phase-wise policy design tools’
 ‘Variation promoters’
 ‘Alternative evaluation modes’
 ‘Looking glasses to the future’
 ‘Adaption devices’ or ‘learning
accelerators’
GO-TO THINGS WHEN PREPARING EXPERIMENTS
27-June-2017 Governing better through evidence-informed policy making 7
 Randomized assignment of subjects whenever possible
 Size of the sample
 Time-frame of the experiment
 Are control conditions needed and what they should be
 Examine limits to causality and indirect effects
 Examine external validity
 What are the success factors (when to stop, re-test or scale)
 Transparent publication of results
 What are the next steps after an experiment
MAIN CHALLENGES: DON’T HIT THE TARGET, BUT MISS
THE POINT!
27-June-2017 Governing better through evidence-informed policy making 8
 Underlining cultural, organizational shift
 Financial risks & future discounting
 How many experiments are allowed to be scaled up?
 Politics of policy experiments (donor influence, accepting uncertainty,
reputational stake, risk of manipulation, institutional barriers etc.)
 Evaluation: attribution and causality
 Bias towards prototyping
 Obtaining ‘evidential support’ for pre-determined policy solutions
 Diffusion: ‘craft rather than science’
 Ethics of experimentation
THANK YOU FOR
THE ATTENTION!
LETS
EXPERIMENT!
27-June-2017 Governing better through evidence-informed policy making 9

OECD Piret Tonurist - Evidence informed policy making - 27 June 2017

  • 1.
    EXPERIMENTS IN THE PUBLICSECTOR: GETTING THE BASICS RIGHT Piret Tõnurist Innovation Specialist, Observatory of Public Sector Innovation, OECD June 27, 2017
  • 2.
    ADAPTIVE POLICY MAKING:WAITING OR GOING? 27-June-2017 Governing better through evidence-informed policy making 2  What is evidence under uncertainty?  Golden decision rule in PA: careful attention to detail, ‘good’ professional or well-considered judgment  Accurate versus complex decisions – ‘good’ decisions vary greatly  Don’t be paralyzed by uncertainty!
  • 3.
    WHAT ARE EXPERIMENTS? 27-June-2017Governing better through evidence-informed policy making 3  Procedures to support, refute, or validate a hypothesis usually by creating an intervention which is observed, measured and evaluated Aerial view of the Hawthorne Works, ca. 1925.
  • 4.
    PROCESS OF EXPERIMENTATION 27-June-2017Governing better through evidence-informed policy making 4  Frame the problem  Identify the intended goals  Identify possible solutions  Set an hypothesis  Select a method  Define the levels of success  Test  Review and communicate results  Make a decision: pull out, re- test or scale?  Launch  Monitor and evaluate
  • 5.
    WHEN TO EXPERIMENT? 27-June-2017Governing better through evidence-informed policy making 5  When you don’t know the right answer  If you know, do! Don’t get caught in the call for action and ignore well-grounded prior evidence!  What is the source of uncertainty (substantive, strategic)?  When there is room for act  When proof-of-concept can lead to broader positive changeDon’t love the method, love the problem!
  • 6.
    WHY EXPERIMENT? 27-June-2017 Governingbetter through evidence-informed policy making 6  A ‘controlled small-scale’  A ‘phase-wise policy design tools’  ‘Variation promoters’  ‘Alternative evaluation modes’  ‘Looking glasses to the future’  ‘Adaption devices’ or ‘learning accelerators’
  • 7.
    GO-TO THINGS WHENPREPARING EXPERIMENTS 27-June-2017 Governing better through evidence-informed policy making 7  Randomized assignment of subjects whenever possible  Size of the sample  Time-frame of the experiment  Are control conditions needed and what they should be  Examine limits to causality and indirect effects  Examine external validity  What are the success factors (when to stop, re-test or scale)  Transparent publication of results  What are the next steps after an experiment
  • 8.
    MAIN CHALLENGES: DON’THIT THE TARGET, BUT MISS THE POINT! 27-June-2017 Governing better through evidence-informed policy making 8  Underlining cultural, organizational shift  Financial risks & future discounting  How many experiments are allowed to be scaled up?  Politics of policy experiments (donor influence, accepting uncertainty, reputational stake, risk of manipulation, institutional barriers etc.)  Evaluation: attribution and causality  Bias towards prototyping  Obtaining ‘evidential support’ for pre-determined policy solutions  Diffusion: ‘craft rather than science’  Ethics of experimentation
  • 9.
    THANK YOU FOR THEATTENTION! LETS EXPERIMENT! 27-June-2017 Governing better through evidence-informed policy making 9