evidence based policy: presenting statistical information for policy makers
DESCRIPTION
Evidence based policy: Presenting statistical information for policy makers. 14 March 2014 Sandra Pattison & Patrick Lim. What is the VET policy environment?. Australian Constitution National, State, Regional, Industry needs Government funding 2012 $8428.5 million - PowerPoint PPT PresentationTRANSCRIPT
Evidence based policy: Presenting statistical information for policy
makers14 March 2014
Sandra Pattison & Patrick Lim
What is the VET policy environment? Australian Constitution National, State, Regional, Industry needs Government funding 2012
► $8428.5 million – $4 333.4 million states and territories– $2 325.4 million Aust government
Private markets, market failure, substitution.
What are the key policy questions? Objective
► A productive and highly skilled workforce enabling all working age Australians to develop the skills and qualifications needed to participate effectively in the labour market and contribute to Australia's economic future; and supports the achievement of increased rates of workforce participation.
Outcomes► the skill levels of the working age population
are increased to meet the changing needs of the economy;
► all working age Australians have the opportunity to develop skills;
► training delivers the skills and capabilities needed for improved economic participation for working age Australians.
How do policy makers like their information presented to them?
How do policy makers like their information presented to them?
How do policy makers like their information presented to them?
Effect β SE df t Pr > |t|
Intercept 2.383 0.099 325 24.01 <0.0001
Student SES 0.082 0.070 4381 1.17 0.2418
Student academic achievement
1.220 0.077 4381 15.83 <0.0001
School quality 0.470 0.091 325 5.14 <0.0001
Student SES by student academic achievement
-0.045 0.056 4381 -0.81 0.4159
Student SES by school quality
-0.150 0.053 4381 -2.84 0.0046
Student academic achievement by school quality
-0.132 0.063 4381 -2.09 0.0366
vs
How do policy makers like their information presented to them?
vs
Characteristics
Courses/units
Intentions
Outcomes
Qualifications
Characteristics
Employer views
Activity
Funding
VET ProviderApprentice & TraineeStudent Outcomes SurveyEmployer SurveyVET Financial Data
Plus other relevant data (eg: ABS Survey of Education and Training, Census, HILDA)
Data Sources?
Data Quality Data quality is vital. Quality is an on-going and continuous
process1. Import and collection.
2. Robust quality assessment when reporting.
3. Internal and external review of reports and publications.
Data Quality on collection Administrative Data:
► Governed by data standards (AVETMISS) and data input tools that ensure consistency with the data standard.
► Quality checks undertaken by NCVER upon receiving of the data.
Survey Data:► Close cooperation with the market research companies► Define ranges of acceptable responses for quantitative
data.► Quality checks undertaken upon receiving of the data.
Data Quality for reporting Usual exploratory data analysis. Cross-checking of tables against known
bottom-lines. Research papers - traditional academic
referee process:► 2 referees (either internal or external).
Defining the question? Often NCVER is asked by Government to
assess a broad general statement:► eg: Has the Productivity Places Program
(PPP) worked? What does this actually mean?
We spend considerable effort in helping refine questions.
Data is collected with no specific policy questions in mind.
Defining the question? Often, there is no robust, reliable data to adequately
answer a question. It is “our” role as the key agency in VET to identify
weaknesses in data collections and to advise on how to collect the required data/information.
Policy questions can and should inform data development & collection. However, most data collections and surveys exist longer than policy.
Thus, changes to standards/questionnaires needs to be undertaken systematically.
Evaluation needs to be considered when setting up a program, not once it has been delivered.
Informing policy? Very difficult to measure the impact of
research, even more-so when trying to determine if it informs policy.
Difficult to attribute a research project to a particular policy development.
Academic journals struggle to inform policy. Often - unpublished reports have the biggest
impact.
Informing policy?
Informing policy?
Layering the dissemination
Journal Articles
Synthesised into a very short briefing notes for Ministers, Advisors or other parliament committee
Very complicated piece of research may often be reduced to two or three lines in a ministerial briefing
Robust and accurate statistical methodology using reliable data is vital!
Policy makers don’t care that you have fitted a particular type of model.
They need to have faith in the data and methodology.
It is up to us to highlight shortcomings, but this needs to be done succinctly and clearly.
Informing policy?
Apprentice and trainee incentive changes Completion rates Indigenous systematic review Part-time work and study School Effects Youth Allowance
Informing policy?
100
35
Completion rate = 35%
Problems
Estimating course completion rates
Commencing Continuing
Quit Completed
Commencing Continuing
Quit Completed
?
?? ? ?
?
11
Commencing Continuing Continuing Completed
Markov Chain model
0.20.2
0.3
0.60.1
0.4
0.4
Commencing Continuing
Quit Completed 11
Probability of eventually completing
31
commence continue
quit complete
commence continue
quit complete
or
p1
p3
q1
q3
3
131 1
)course completing eventuallyPr(q
qpp
Key Message:
Our research won’t influence anyone if policy makers can’t understand what it is we are presenting!
Informing policy?
Discussion Evidence base policy – does it exist? Presenting complicated statistical models (yes, the most basic
regression is complicated to others)? Preserving statistical rigour - how do we ensure that the information
presented remains accessible? How do we balance the need for providing timely and relevant
information/data when:► Data sources may be scarce or unreliable► We need to maintain independence
Data Quality Statements – Fit for purpose? Accepting the fact that your analysis or work may in fact be
misrepresented or misunderstood and be prepared for it?