5.1 Data: Where to get it and how to assess it
Evaluation involves using performance information or data to answer the evaluation
questions you have developed. As detailed in Section 3.3, a program logic can help us
to identify the performance information we need. This section takes a closer look at
data sources and collection methods.
Before deciding on the data collection methods to use, it is important to consider:
- What are the key evaluation questions to be answered?
- What information sources and/or measurement instruments already exist?
- What are the most appropriate and practicable methods for collecting new data?
- What resources do different methods require and are there adequate resources available to collect and analyse the data?
- What risks and ethical issues need to be considered?
It is also worth thinking about how you will define and understand different data. For
example, it seems straightforward to count participants, but what do we mean by a
‘participant’? Is this a person who registers their interest in a program but does not
attend, someone who registers and pays but only attends once, or someone who
registers and attends at least 60% of the time? Any definition is possible but it is
worth creating a meaningful definition that all stakeholders can work with and that
you can use consistently throughout your evaluation.
This helps to provide consistency in the collection and use of data, make data easier
to analyse and preserve knowledge of what you did so it can be replicated or used
later on. A formal way of doing this is through a data dictionary which describes the
meanings and purposes of data elements within the context of a project.
Quantitative and qualitative data
Data can be quantitative or qualitative.
Quantitative data> refers to data that is measured in terms of numbers and counts
and can be sourced from administrative sources and survey questions.
Qualitative data is more narrative or text-based and collected through methods such
as case studies, observation, focus groups, and stories of change.
A mixed methodology is often recommended and includes both quantitative and
qualitative data and methods of analysis. For example, surveys often include
quantitative measures (e.g. scales of agreement or satisfaction, measures of income,
and wellbeing) as well as qualitative measures (e.g. open-ended questions about what
participants most liked about an activity).
Data sources and collection
Different types of data sources and analysis include:
-
Academic or published knowledge - published in research journals or collected
through systematic reviews. For instance, systematic reviews from the Campbell Systematic Reviews and the Cochrane Library.
-
Lived experience - the lived experience of community members, specific cohorts or beneficiaries.
-
Administrative data - collected within the local government, community services
you partner with, or as part of the program you are running.
-
Evaluation data collected - collected as part of the measurement
and evaluation methods. Depending on what you are measuring you may develop your own
methods or you may want to identify if there are existing available measurement
instruments for the outcome you want to measure—for example, the World Health
Organization’s Quality of Life Scale.
Data collection methods include:
- surveys
- interviews
- focus groups
- case studies
- analysis of existing data
- document reviews
Evaluation data can also be drawn from:
- program documentation
- visual records
- social media
- storytelling
- observation
- community meetings
Any information that is collected systematically (and, ideally, consistently too) that
you can make sense of in relation to the evaluation questions can be considered as
data. Often there are limitations (such as biases) with the data that an evaluator must
rely on. As an evaluation practitioner it is important that you are aware about such
limitations. Having a rich variety of sources can help to counter the limitations of any
one particular source.
Administrative data
Administrative data is information collected by government, business and other
organisations for purposes such as record keeping, registrations and completing
transactions.
Administrative data can be used in a number of ways. Data collected around certain
populations can be extracted for comparison with a participant group.
Examples of administrative data include hospital admissions, school
attendance, Medicare enrolments and police attendance.
Administrative data collected over a period of time can provide a basis for
longitudinal evaluation. Caution should always be taken due to the possibility of
missing data, data quality, access and data protection.
Be careful to understand how terms are defined as well—so you know what the measures mean!
Data quality
The ability of data to answer evaluation questions depends on the quality of the data.
Data quality is often assessed in terms of the following criteria:
- Relevance: the data must add meaning and meet the needs of the evaluation.
- Accuracy: the data must have fidelity to what it is supposed to represent, and be truthful.
- Timeliness: the data must be up-to-date and available.
- Coherence: the data must be comparable, reliable and consistent over time.
- Interpretability: the data must be able to be understood and utilised
- Accessibility: the data must be easily accessible for those doing the evaluation
- Validity: the data must precisely reflect what is intended.
Triangulation is when more than one method is used to collect data on the same
question. It is a way to reinforce or examine results to assure rigorous and valid findings.
e.g. Triangulation:
For example, an evaluation of a community garden initiative may include a survey of all
participants as well as in-depth interviews with a selection of participants.
Data from both methods is then compared to see if different sources of data confirm the same findings.
Validity and reliability of data
Validity refers to the extent an instrument or measure accurately measures what it
intended to measure. We can consider both face validity (does the measure appear
to adequately measure what it is intended to) and content validity (when thinking of
the agreed definition of a particular construct/idea such as ‘wellbeing’ would experts
and others agree that the measure captures the meaning of the construct).
Reliability> refers to the extent to which an instrument produces consistent results.
There are three types of reliability common in program evaluations:
- Test-retest reliability: does an instrument produce similar results with repeated testing?
- Inter-rate reliability: do two or more people administering the instrument produce similar results?
- Internal consistency reliability: do survey items that are intended to measure the same characteristic correlate?
External factors impacting outcomes
It is important to think about external factors that impact on the outcomes observed
in an evaluation, and collect some information about this as you go (even if the
information you collect is only notes). For example, when evaluating the effectiveness
of a breakfast club program to improve school attendance, it would be important to
know what other interventions are also targeting school attendance, as well as any
other factors which might have a bearing upon the results. An example of other
interventions might include an incentive program being run by a local sporting club
to encourage students to attend school.
It is important to consider external factors that might influence outcomes when
collecting data, so you can fairly assess the effectiveness of an activity.
At times your context may be quite complex with many factors influencing the
outcomes all at once, within a general environment of change.
Keeping notes about these changes as your program unfolds is called
situational analysis.
e.g. The 'Your Move' program:
During COVID-19, many factors in the external environment may have
influenced your program, so it is a matter of documenting all the factors
that may have contributed to the outcomes.
For example, the ‘Your Move’ program, which involves the City of
Stirling, looks at the outcome of encouraging people to use the train.
However, during the COVID-19 period there were many changing external
conditions that would have influenced outcomes, for example:
- fears of infection from public transport
- lockdown and closure of businesses in the CBD
- more people working remotely
- introduction of free parking in the city
"Data collection and analysis can take a long time, and this should be taken into account when planning interviews..."
5.2 Data collection: How do we measure it?
Once you have reviewed your program logic (and evaluation questions), you will have
a better idea of what existing data you want to access and use. You may also decide
that you need to collect further information as part of the evaluation.
Analysis of existing data and evidence
Data and evidence can be obtained from sources such as project documents,
government records and publicly available statistics. The Campbell Systematic
Reviews (Campbell Collaboration) and Cochrane Library provide access to systemic
reviews, which can be useful sources of knowledge.
Common data collection methods
A summary of some methods to choose from is provided here.
Remember: When collecting data, you need to consider informed consent.
At a minimum, people should be told why the information is being collected, how it
will be used, and how confidentiality will be protected.
Document reviews
Data and evidence can be collected from existing documents for the purpose of
background information.
Case studies
A case study is a detailed investigation of a person, group or event in its real-world
context, over a period of time.
Surveys
Almost all written surveys are now conducted via web-based platforms such as
Survey Monkey, which you can access for free.
Surveys are convenient and can be anonymous.
When surveys are used to gather information from a large number of people, they
can provide statistical findings that may be generalised to the broader population.
When undertaken at various intervals (e.g. prior to, during and after program
implementation) the data can capture change over time.
Surveys give limited opportunity to gather in-depth information, and low response
rates and small sample sizes will limit your ability to generalise findings, or do
meaningful quantitative analysis. (As a very general guide, generally aim for a sample
size of 30 people or more to make a survey worth doing.)
Interviews
Interviews can be conducted in person, by telephone or online. Consider which
setting will best ensure respondents are comfortable sharing information. The mutual
sharing that happens in an interview encourages the building of trust and can
facilitate information gathering from sources who may otherwise be difficult to
access.
In-depth discussions can uncover deep insights and unique perspectives and, as such,
interviews can help understand why and how an intervention has worked. Interviews
are suitable for use at any stage prior to, during, or after a program.
While in-depth views and opinions are easily accessed through this method,
gathering large quantities of data through interviews can be resource intensive. Data
collection and analysis can take a long time, and this should be taken into account
when planning interviews.
Focus groups
Focus groups are an informal way of interviewing groups of typically 6–12 people,
and usually last for around 90 minutes. They allow people to express their opinions
and ideas freely and encourage open expression due to dynamic conversation
created through multiple perspectives. Group discussion is rich and can complement
or build on quantitative survey. Focus group members can be presented with results
to be interpreted and emerging patterns to be explored.
Focus groups can help build understanding of why and how an intervention has
worked or might work, and are suitable for use at any stage prior to, during, or after
a program.
Given the short timeframe and nature of focus group discussions, there is little
opportunity to cover many topics in this setting. A broader range of ideas can be
discussed in a one-on-one interview setting.
Selecting a method
A number of factors will impact upon the types of methods used to conduct an
evaluation, including:
- The stage of the program being evaluated:
-
Is the program already underway? This gives you scope to access participants
with some degree of ease.
-
Is it already complete? Consider how you will access participants and how many
might respond to a request for data collection.
-
Is it about to enter the planning and design phase? This gives you the greatest
scope to design an evaluation that will provide the strongest evidence on
impact and implementation. Designing your evaluation prior to the start of the
program is the preferred approach as it gives you the best chance to establish
pre-post data and develop comparison reference points.
- Timeframe:
-
Is there scope to capture data at multiple points in time?
-
How quickly does the evaluation need to be developed in order to access
participants during program delivery? Methods such as interviews can be more
time consuming, while surveys can be turned around faster.
-
If the program operates in cycles, data collection can be designed around this,
and feed into iterations of the program.
- Stakeholders:
-
Are there accessibility considerations which might impact respondents’ ability to
participate? Certain formats may be more appropriate in certain instances. For
example, interviews might be more culturally sensitive in Indigenous
communities when approached in an appropriate manner. Include a broad set
of stakeholders in your evaluation planning and always include a role for
beneficiaries of programs.
- The type of data sought:
-
A mixed methods approach is ideal as both quantitative and qualitative data will
provide more robust evaluation findings.
- Prior evaluation experience:
-
It is okay to select methods based on what you feel most comfortable using.
A ‘good enough’ approach is accepted by expert evaluators: real world
constraints will always be present. Start simply and build capacity with
experience, while knowing that lack of prior evaluation experience is not a
barrier to conducting an evaluation.
"Data collection and analysis can take a long time, and this should be taken into account when planning interviews..."