how we know what works (and what doesn’t) in wp...university. 2. to promote empiricism and...
TRANSCRIPT
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
How we know what works (and what doesn’t) in WP
Miriam Styrnol and Vanessa Todman (in spirit)
What Works, King’s College London
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Objectives of the WWD
1. To contribute to understanding of what works in enabling people to access and succeed at university.
2. To promote empiricism and innovation in widening participation.
3. To support King’s and the sector to think differently about designing and evaluating their initiatives.
The What Works DepartmentGeneration
• Working in partnership on new research to fill the gaps in existing knowledge and validate international insights in the UK context
Dissemination
• Developing reports, briefings and training on cutting-edge research findings and methods
Synthesis
• Drawing together existing research from the UK and elsewhere on supporting access to and success in university
@KCLWhatWorks 3
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
4
Find us!
@KCLWhatWorks
https://www.kcl.ac.uk/study/assets/PDF/widening-participation/What-works-project-report.pdf https://blogs.kcl.ac.uk/behaviouralinsights
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
• We’re talking about impact evaluations – generating evidence of the causal impact of an intervention in changing the outcomes of target participants.
❖There are other types of evaluation (e.g. process evaluations), and other types of research.
❖These are all valid and useful; just out of scope for this presentation.
• It’s important to talk about ‘what works’ and impact evaluations because this is an area that is less covered and researched in an otherwise rich and diverse research field.
Before we begin
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon DisplayWe don’t need to find out: we’re
experts and we already know the answer.
Results can be surprising, and even if we’re right
about the direction, what is the effect size?
We have robust findings because
we apply the intervention and
measure impact before and after.
How do we know other factors didn’t play a part?
What if an improvement would have happened
anyway?
How do we know that the groups aren’t also
different in other important ways?
We show the added benefit of our
programme by comparing different
individuals, groups or areas
RCTs are a “black box” into which
inputs disappear and out of which
outcomes emerge with no
explanation.
A good RCT is led by a hypothesis, informed
by theory and primary or secondary – often
qualitative – research.
Interviews, surveys and focus
groups fit our complex context
better.
Are participants are telling the truth? Do they
really know what they want, and how their
behaviour will change in a new situation?
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Example: Encouraging Volunteering
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Example: Encouraging Volunteering
http://www.behaviouralinsights.co.uk/wp-content/uploads/2016/09/BIT-Update-Report-2015-16.pdf © Behavioural Insights Ltd
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
The evaluation cycle
@KCLWhatWorks
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Diagnose: Research the topic
Sense of
Belonging
Visibility
One purpose of this intervention was to make tasks such as studying for exams achievable. For
example, our study skills video, gave students quick, easy to apply tips and ended with a student
holding a sign saying “you’ve got this”.¹
Knowledge
Self-Efficacy
Videos were designed to tackle a persistent sense of difference among students. By hearing people
talk about what it means to be a King’s student, or asking them to reflect on this, we hypothesise
we could increase student’s sense of belonging.²
In this task we asked students to think about what they may have in common with their personal
tutor, in line with research suggests similarities across various dimensions can increase
compliance and pro-social behavior.³
We designed a video setting out key study skills for students who do not feel prepared. This video
was designed to be basic but uplifting, with BAME students providing key information and the
aim to boost knowledge of study tactics.4
We identified that an intervention designed for all students that targeted BAME issues might be effective
¹Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191–215.
²Johnson et al. (2007) Examining sense of belonging amongst first-year undergraduates from different racial/ethnic groups. Journal of College Student Development, 48(5), 525-542.
³Gehlbach et al. (2016). Creating birds of similar feathers: Leveraging similarity to improve teacher-student relationships and academic achievement. Journal of Educational Psychology,
108 (3), 342-352. 4 Miller, M (2016) The Ethnicity Attainment Gap: Literature Review
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Plan: Identifying Research Questions
Causal impact of your
evaluation
Did [scheme] increase [main
outcome] among [group]?
Did YKYS improve second year
attainment among
ethnic minorities?
Primary Research Question
These are overarching question that your evaluation will seek to answer.
They will determine the scope and approach of your evaluation.
Secondary Research Question Process Evaluation Question
Focus on specific groups or
intermediate outcomes
Did [scheme] increase [main
outcome/ secondary outcome]
among [group/subgroup]?
Did YKYS improve second year
attainment among ethnic
minorities in specific faculty?
Focus on implementation
and efficiency of your set-up
✓ Was the initiative delivered
the way we expected?
✓ Are we targeting the right
students?
✓ What was the cost-
effectiveness of the initiative?
@KCLWhatWorks
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Outcome measures
“I’ll know [outcome reached] when I see [indicator]”
Observable indicators are
those we can build into
the evaluation and control
out
Example: Demography
@KCLWhatWorks
Unobservable indicators
are those we can’t observe
and therefore can’t build
into the evaluative model.
Example: Social Capital
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
SecondaryAverage Attainment
after Year 2 of Studies
Measure: YKYS outcomes
Primary
SecondaryKing’s Re-/Enrolment
Questions
ProcessUsefulness of video(s)
and task(s)
75%67%
60%53% 52%
40% 37% 33% 31% 26%
% responding positively to the below statements
@KCLWhatWorks
Williams, D. 2006. On and off the ’net: scales for social capital in an online era. Journal of Computer-Mediated Communication, 11(2), 593–628.
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Data Collection
Objective
Comparable
Patterns
Structures
Tendencies
Measurement
Behaviour
Attitudes
QualitativeHow?Why?
QuantitativeHow much?
How many?
What people do or how
they act unconsciously
Self-reported
preferences, recognition
and memories
Subjective
Difference
Meanings
Context
Experiences
Perceptions
Documentary Analysis
Journey Mapping
Observations
Experiment
Causal Data Analysis
Correlatory Data Analysis
Simulations
Eye Tracking
Closed Survey
Structured Interview
Open Survey
Participatory Observation
Semi-structured InterviewUnstructured Interview
Prompts and visuals
Focus Groups
@KCLWhatWorks
v
v
v
vv
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Selecting the right research method
Level 1 - Monitor
Level 3 - Identify
Level 2 - Compare
RCT/ quasi-exp.
method
Dosage-response
Comparing with non-
participants
Qual research w/ participants and comp.
group
Rationale/ Theory of
Change
Secondary Research
Tracking
Pre/Post
Qual research w/ participants
only
@KCLWhatWorks
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Level 1: Monitor
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Level 1: ToCThere is a BAME attainment gap at King s
Reduce the BAME attainment gap using an intervention inspired from behavioural insightsAims
Rationale &
Assumptions
Students will sign up and complete tasks. The intervention won t negatively influence attainment. Faculties willing to take part. Tasks can increase attainment. We can afford the incentives.
• Enrolment survey
• Initial/evaluation
survey
• Staff time
• Incentives
• Video budgeting
• Stakeholder time
• Student advice
• Student time
• Director s input
• Commissioning
board input
• Research time
• Qualtrics
• Comms
development
• Ethical review time
ImpactInputs Activities Outputs Outcomes
• Analysing enrolment
data
• Analysing initial/
evaluation survey
• Sending out survey
• Recruiting students and
faculties
• Sending comms to
students/staff
• Ethical approval
• Creating materials
• Reporting findings
• Stakeholder
consultations
• Support staff recruitment
• Students completing
tasks
• Student opt-in
• Enrolment task
responses
• Movement between
initial/evaluation
survey
• Task responses
• Reports
• Blog posts
• Presentation to
commissioning board
Short:
• Increase in sense of belonging and self-efficacy
• Improved personal tutor interaction
• Increase in social capital
• What Works establishing itself with its 1st flagship project
• Faculty/stakeholder relationships built
Long:
• Increase in BAME students getting a good honours degree
• Improved King s
knowledge of
psychological
factors influencing
attainment
• Reduction in
BAME attainment
gap
Situation1
2
3 45 67
8
Process Impact
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Level 2: Compare
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Level 2: Confounding variables
Other studies:
• Wellness programmes: Jones, D., Molitor, D., & Reif, J. (2018). What Do Workplace Wellness Programs Do? Evidence from the Illinois
Workplace Wellness Study (No. w24229). National Bureau of Economic Research.
• Online advertising: Gordon, B. R., Zettelmeyer, F., Bhargava, N., & Chapsky, D. (2018). A comparison of approaches to advertising
measurement: Evidence from big field experiments at Facebook. Retrieved from https://ssrn.com/abstract=3033144.© Behavioural Insights Ltd
7.0
4.7
Naïve Naïve with covariates Post-consent randomisation
Effect of ‘Study Supporter’ on attendance rate8.4
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Level 3: Identify
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Level 3: ‘What Works’ (RCTs)Your King’s Your Success pilot Jan- March 2019
Control (not contacted)
2nd years invited to opt in
Belonging task & video
Belonging task
Belonging video
301 opt in to trial and complete
baseline survey
Personal Tutor video
Study skills video
214 fill out endlinesurvey
Personal Tutor task
Study skills task
Personal Tutor task & video
Study skills task & video
Change in attainment from 1st to 2nd year compared for each
group
February 2019 March 2019 September 2019
@KCLWhatWorks
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Consider the context of an evaluationResearch question
• For example, an experimental design can only tell you whether something works, not how or why (mixed methods can help here).
Sample size
• Ideally over 1,000 across treatment and control, 200ish if good baseline data on outcome.
Data
• Collected consistently and universally (administrative datasets are best).
@KCLWhatWorks 24
Fidelity and validity
• Try to maintain protocol consistency within your intervention; this can sometimes be difficult—and can sometimes reduce external validity.
Ethics
• If there is substantial, consistent, high-quality evidence that something is effective, it shouldn’t be withheld from anyone who could benefit.
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Other examples at King’s:
• Does sending a text message increase attendance at Study Abroad briefings?
• Do phone calls from 2nd and 3rd years reduce non-continuation among 1st
years?
• Do metacognition exercises increase metacognitive skills, and does this increase attainment?
• Do bursaries improve outcomes for low-income students?
@KCLWhatWorks 25
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Lessons from the field
1. Know the data journey
2. Get to know the system and the people
3. Look for SMART interventions
4. Communicate and monitor regularly
5. Use behavioural insights
6. There is only so much you can do
7. TASO is here to help
@KCLWhatWorks 26
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
Securing funding
Doyle, M., & Griffin, M. (2012). Raised aspirations and attainment? A review of the impact of Aimhigher (2004–2011) on
widening participation in higher education in England. London Review of Education, 10(1), 75-88.
Aimhigher
✓ 35% of HEIs attributed increased applications to their
institutions to Aimhigher
✓ Associated with rising GCSE results
BUT no evidence that it actually impacted
progression rates, particularly to selective universities
Discontinued in 2011
The Office for Students (and before that, HEFCE and
OFFA) has consistently signalled they don’t have
confidence in the evidence underpinning institutions’
access activities, both in public statements and in their
comments on Access and Participation Plans.
Policy developments like Augar put WP funding in a
perilous situation. Now is the time to put
methodological differences aside and to start sharing
insights to get our evaluative findings out there
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
‘What Works’ is a beginning, not an end
‘What works’
doesn’t have to be
‘one size fits all’
Focus on ‘what works’
doesn’t need to narrow
the scope of
interventions to what is
easily measurable.
Knowing what works, for whom
and under what circumstances is
easier with a good identification
strategy (e.g. randomisation)
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
@Taso_HE
Title 1: 26pt King’s Bureau Grotesque Three Seven white outTitle 2: 20pt King’s Caslon Display
@KCLWhatWorks
Thank You.