You are on page 1of 9

NUMBER 16

1ST EDITION 2010

PERFORMANCE MONITORING & EVALUATION

TIPS
CONDUCTING MIXED-METHOD EVALUATIONS

ABOUT TIPS
These TIPS provide practical advice and suggestions to USAID managers on issues related to
performance monitoring and evaluation. This publication is a supplemental reference to the
Automated Directive System (ADS) Chapter 203.

INTRODUCTION MIXED-METHOD RATIONALE FOR


EVALUATIONS USING A MIXED-
DEFINED METHOD
This TIPS provides guidance on
using a mixed-methods approach A mixed-method evaluation is
EVALUATION DESIGN
for evaluation research. one that uses two or more There are several possible cases
Frequently, evaluation statements techniques or methods to collect in which it would be highly
of work specify that a mix of the data needed to answer one or beneficial to employ mixed-
methods be used to answer more evaluation questions. Some methods in an evaluation design:
evaluation questions. This TIPS of the different data collection
includes the rationale for using a methods that might be combined When a mix of different
mixed-method evaluation design, in an evaluation include methods is used to collect data
guidance for selecting among structured observations, key from different sources to
methods (with an example from informant interviews, pre- and provide independent estimates
an evaluation of a training post-test surveys, and reviews of of key indicators—and those
program) and examples of government statistics. This could estimates complement one
techniques for analyzing data involve the collection and use of another—it increases the
collected with several different both quantitative and qualitative validity of conclusions related
methods (﴾including “parallel data to analyze and identify to an evaluation question. This
analysis”)﴿. findings and to develop is referred to as triangulation.
conclusions in response to the (See TIPS 5: Rapid Appraisal,
evaluation questions. and Bamberger, Rugh and

1
Key Steps in Developing a Mixed-Method Evaluation Design and Analysis
Strategy
1. In order to determine the methods that will be employed, carefully review the purpose of the evaluation and the
primary evaluation questions. Then select the methods that will be the most useful and cost-effective to answer
each question in the time period allotted for the evaluation. Sometimes it is apparent that there is one method
that can be used to answer most, but not all, aspects of the evaluation question.

2. Select complementary methods to cover different aspects of the evaluation question (for example, the how and
why issues) that the first method selected cannot alone answer, and/or to enrich and strengthen data analysis
and interpretation of findings.

3. In situations when the strength of findings and conclusions for a key question is absolutely essential, employ a
triangulation strategy. What additional data sources and methods can be used to obtain information to answer
the same question in order to increase the validity of findings from the first method selected?

4. Re-examine the purpose of the evaluation and the methods initially selected to ensure that all aspects of the
primary evaluation questions are covered thoroughly. This is the basis of the evaluation design. Develop data
collection instruments accordingly.

5. Design a data analysis strategy to analyze the data that will be generated from the selection of methods chosen
for the evaluation.

6. Ensure that the evaluation team composition includes members that are well-versed and experienced in applying
each type of data collection method and subsequent analysis.

7. Ensure that there is sufficient time in the evaluation statement of work for evaluators to fully analyze data
Mabry
generated[2006]
from eachfor method employedricher
further analysis
and to realize the and canof conducting
benefits also Mixed-method evaluations
a mixed method evaluation.
explanation and descriptions of provide a better understanding often yield a wider range of
triangulation strategies used in of the context in which a points of view that might
evaluations.) program operates. otherwise be missed.

When reliance on one method There are a number of additional


alone may not be sufficient to benefits derived from using a mix DETERMINING
answer all aspects of each
evaluation question.
of methods in any given WHICH METHODS TO
evaluation.
USE
Using mixed-methods can
When the data collected from In a mixed-method evaluation,
more readily yield examples of
one method can help interpret the evaluator may use a
unanticipated changes or
findings from the analysis of combination of methods, such as
responses.
data collected from another a survey using comparison
method. For example, Mixed-method evaluations groups in a quasi-experimental or
qualitative data from in-depth have the potential of surfacing experimental design, a review of
interviews or focus groups can other key issues and providing key documents, a reanalysis of
help interpret statistical a deeper understanding of government statistics, in-depth
patterns from quantitative data program context that should interviews with key informants,
collected through a random- be considered when analyzing focus groups, and structured
sample survey. This yields a data and developing findings observations. The selection of
and conclusions. methods, or mix, depends on the

2
nature of the evaluation purpose survey 100 percent of the access to finance a problem? Did
and the key questions to be individuals who graduated from they conduct an adequate market
addressed. the training program using a analysis? Did some individuals
close-ended questionnaire. It start with prior business skills?
SELECTION OF DATA follows that a survey instrument Are there factors in the local
should be designed to determine economy, such as local business
COLLECTION if these individuals have actually regulations, that either promote
METHODS – AN succeeded in starting up a new or discourage small business
business. start-ups? There are numerous
EXAMPLE factors which could have
While this sounds relatively influenced this outcome.
The selection of which methods
straightforward, organizations are
to use in an evaluation is The selection of additional
often interested in related issues.
driven by the key evaluation methods to be employed is,
If less than 70 percent of the
questions to be addressed. again, based on the nature of
individuals started a new business
Frequently, one primary each aspect of the issue or set
one year after completion of the
evaluation method is apparent. of related questions that the
training, the organization
For example, suppose an organization wants to probe.
generally wants to know why
organization wants to know To continue with this example,
some graduates from the
about the effectiveness of a pilot the evaluator might expand the
program were successful while
training program conducted for number of survey questions to
others were not. Did the training
100 individuals to set up their address issues related to the
these individuals received actually
own small businesses after the effectiveness of the training and
help them start up a small
completion of the training. external factors such as access to
business? Were there topics that
should have been covered to finance. These additional
The evaluator should ask what questions can be designed to
more thoroughly prepare them
methods are most useful and yield additional quantitative data
for the realities of setting up a
cost-effective to assess the and to probe for information
business? Were there other
question of the effectiveness of such as the level of satisfaction
topics that should have been
that training program within the with the training program, the
addressed? In summary, this
given time frame allotted for the usefulness of the training
organization wants to learn not
evaluation. The answer to this program in establishing a
only whether at least 70 percent
question must be based on the business, whether the training
of the individuals trained have
stated outcome expected from graduate received a small
started up a business, but also
the training program. In this business start-up loan, if the size
how effectively the training
example, let us say that the of the loan the graduate received
equipped them to do so. It also
organization’s expectations were was sufficient, and whether
wants to know both the strengths
that, within one year, 70 percent graduates are still in the process
and the shortcomings of the
of the 100 individuals that were of starting up their businesses or
training so that it can improve
trained will have used their new instead have found employment.
future training programs.
skills and knowledge to start a Intake data from the training
small business. The organization may also want program on characteristics of
to know if there were factors each trainee can also be
What is the best method to outside the actual intervention examined to see if there are any
determine whether this outcome that had a bearing on the particular characteristics, such as
has been achieved? The most training’s success or failure. For sex or ethnic background, that
cost- effective means of example, did some individuals can be correlated with the survey
answering this question is to find employment instead? Was findings.

3
It is important to draw on other sources using a separate on its own measure of program
additional methods to help data collection method for each success, but also whether access
explain the statistical findings source. The first data source to finance contributed to either
from the survey, probe the includes the quantitative data success or failure in starting up a
strengths and shortcomings of from the survey of the 100 new business. The analysis of the
the training program, further training graduates. The data will be used to strengthen
understand issues related to evaluation designers determine the training design and content
access to finance, and identify that the second data source will employed in the pilot training
external factors affecting success be the managers of local banks course, and as previously stated,
in starting a business. In this and credit unions that survey perhaps to design a microfinance
case, the evaluation design could respondents reported having program.
focus on a sub-set of the 100 approached for start-up loans.
individuals to obtain additional In-depth interviews will be The last step in the process of
qualitative information. A conducted to record and designing a mixed-method
selected group of 25 people understand policies for lending to evaluation is to determine how
could be asked to answer an entrepreneurs trying to establish the data derived from using
additional series of open-ended small businesses, the application mixed-methods will be analyzed
questions during the same of those policies, and other to produce findings and to
interview session, expanding it business practices with respect to determine the key conclusions.
from 30 minutes to 60 minutes. prospective clients. The third
Whereas asking 100 people data source is comprised of bank ANALYZING DATA
open-ended questions would be loan statistics for entrepreneurs FROM A MIXED-
better than just 25 people, costs who have applied to start up
prohibit interviewing the entire small businesses. Now there are METHOD
group. three independent data sources EVALUATION –
using different data collection
Using the same example, methods to assess whether DESIGNING A DATA
suppose the organization has
learned through informal
access to finance is an additional ANALYSIS STRATEGY
key factor in determining small
feedback that access to finance is business start-up success. It is important to design the data
likely a key factor in determining analysis strategy before the
success in business start-up in In this example, the total mix of actual data collection begins.
addition to the training program methods the evaluator would use Having done so, the evaluator
itself. Depending on the includes the following: the survey can begin thinking about trends
evaluation findings, the of all 100 training graduates, data in findings from different sets of
organization may want to design from open-ended questions from data to see if findings converge
a finance program that increases a subset of graduates selected for or diverge. Analyzing data
access to loans for small business longer interviews, analysis of collected from a mixture of
start-ups. To determine the training intake data on trainee methods is admittedly more
validity of this assumption, the characteristics, in-depth complicated than analyzing the
evaluation design relies on a interviews with managers of data derived from one method.
triangulation approach to assess lending institutions, and an This entails a process in which
whether and how access to examination of loan data. The quantitative and qualitative data
finance for business start-ups use of mixed-methods was analysis strategies are eventually
provides further explanations necessary because the client connected to determine and
regarding success or failure organization in this case not only understand key findings. Several
outcomes. The design includes a wanted to know how effective the different techniques can be used
plan to collect data from two pilot training course was based

4
to analyze data from mixed- conducted. Then, a separate and evaluators use findings from
methods approaches, including independent analysis is the analysis of qualitative data.
parallel analysis, conversion conducted of qualitative data This method can provide a
analysis, sequential analysis, derived from, for example, in- richer analysis and set of
multilevel analysis, and data depth interviews, case studies, explanations affecting program
synthesis. The choice of analytical focus groups, or structured outcomes that enhance the
techniques should be matched observations to determine utility of the evaluation for
with the purpose of the emergent themes, broad program managers.
evaluation using mixed-methods. patterns, and contextual factors. Conversely, patterns and
Table 1 briefly describes the The main point is that the associations arising from the
different analysis techniques and analysis of data collected from analysis of quantitative data
the situations in which each each method must be can inform additional patterns
method is best applied. In conducted independently. to look for in analyzing
complex evaluations with qualitative data. The analysis
multiple issues to address, skilled Step 2: Once the analysis of the of qualitative data can also
evaluators may use more than data generated by each data enhance the understanding of
one of these techniques to collection method is completed, important program context
analyze the data. the evaluator focuses on how the data. This method is often used
analysis and findings from each when program managers want
data set can inform, explain,
EXAMPLE OF and/or strengthen findings from
to know not only whether or
not a program is achieving its
APPLICATION the other data set. There are two intended results, but also, why
possible primary analytical or why not.
Here we present an example of methods for doing this – and
parallel mixed-data analysis, sometimes both methods are WHEN FINDINGS DO NOT
because it is the most widely used in the same evaluation. CONVERGE
used analytical technique in Again, the method used depends In cases where mixed-method
mixed-method evaluations. This on the purpose of the evaluation. evaluations employ triangulation,
is followed by examples of how
it is not unusual that findings
to resolve situations where In cases where more than one
from the separate analysis of
divergent findings arise from the method is used specifically to
each data set do not
analysis of data collected through strengthen and validate
automatically converge. If this
a triangulation process. findings for the same question
occurs, the evaluator must try to
through a triangulation design,
resolve the conflict among
PARALLEL MIXED-DATA the evaluator compares the
divergent findings. This is not a
ANALYSIS findings from the independent
disaster. Often this kind of
Parallel mixed-data analysis is analysis on each data set to
situation can present an
comprised of two major steps: determine if there is a
opportunity to generate more
convergence of findings. This
Step 1: This involves two or nuanced explanations and
method is used when it is
more analytical processes. The important additional findings that
critical to produce defensible
data collected from each method are of great value.
conclusions that can be used to
employed must be analyzed inform major program One method evaluators use when
separately. For example, a decisions (e.g., end or extend a findings from different methods
statistical analysis of quantitative program). diverge is to carefully re-examine
data derived from a survey, a set
To interpret or explain findings the raw qualitative data through
of height/weight measures, or a
from quantitative analysis, a second and more in-depth
set of government statistics is
content analysis. This is done to

5
determine if there were any from incomplete recording or and, to profit from the use of
factors or issues that were missed incorrect coding of responses. those methods, evaluators must
when these data were first being (See TIPS 12: Data Quality use and analyze all of the data
organized for analysis. The Standards.) If the evaluators are sets. Through the use of mixed-
results of this third layer of still at the program site, it is method evaluations, findings and
analysis can produce a deeper possible to resolve data quality conclusions can be enriched and
understanding of the data, and issues with limited follow-up data strengthened. Yet there is a
can then be used to generate collection by, for example, tendency to underuse, or even
new interpretations. In some conducting in-depth interviews not to use, all the data collected
cases, other factors external to with key informants (if time and for the evaluation. Evaluators can
the program might be discovered budget permit). rely too heavily on one particular
through contextual analysis of data source if it generates easily
economic, social or political In cases where an overall digestible and understandable
conditions or an analysis of summative program conclusion is information for a program
operations and interventions required, another analytical tool manager. For example, in many
across program sites. that is used to resolve divergent cases data generated from
findings is the data synthesis qualitative methods are
Another approach is to reanalyze method. (See Table 2.) This insufficiently analyzed. In some
all the disaggregated data in method rates the strength of cases only findings from one
each data set separately, by findings generated from the source are reported.
characteristics of the respondents analysis of each data set based
as appropriate to the study, such on the intensity of the impact One way to prevent
as age, gender, educational (e.g., on a scale from very high underutilization of findings is to
background, economic strata, positive to very high negative) write a statement of work that
etc., and/or by geography/locale and the quality and validity of the provides the evaluator sufficient
of respondents. data. An overall rating is assigned time to analyze the data sets
for each data set, but different from each method employed,
The results of this analysis may weights can then be assigned to and hence to develop valid
yield other information that can different data sets if the evaluator findings, explanations, and strong
help to resolve the divergence of knows that certain data sources conclusions that a program
findings. In this case, the or methods for collecting data manager can use with
evaluator should attempt to rank are stronger than others. confidence. Additionally,
order these factors in terms of statements of work for evaluation
frequency of occurrence. This Ultimately, an index is created
should require evidence of, and
further analysis will provide based on the average of those
reporting on, the analysis of data
additional explanations for the ratings to synthesize an overall
sets from each method that was
variances in findings. While most program effect on the outcome.
used to collect data, or
professionals build this type of See McConney, Rudd and Ayres
methodological justification for
disaggregation into the analysis (2002) to learn more about this
having discarded any data sets.
of the data during the design method.
phase of the evaluation, it is
worth reexamining patterns from REPORTING ON
disaggregated data. MIXED-METHOD
Evaluators should also check for EVALUATIONS
data quality issues, such as the
validity of secondary data sources Mixed-method evaluations
or possible errors in survey data generate a great deal of data,

6
REFERENCES

Bamberger, Michael, Jim Rugh and Linda Mabry. Real World Evaluation: Working Under Budget,
Time, Data and Political Constraints, Chapter 13, “Mixed-Method Evaluation,” pp. 303-322, Sage
Publications Inc., Thousand Oaks, CA, 2006.

Greene, Jennifer C. and Valerie J. Caracelli. “Defining and Describing the Paradigm Issue in Mixed-
methods Evaluation,” in Advances in Mixed-Method Evaluation: The Challenges and Benefits of
Integrating Diverse Paradigms, Green and Caracelli eds. New Directions for Evaluation. Josey-Bass
Publishers, No. 74, Summer 1997, pp 5-17.

Mark, Melvin M., Irwin Feller and Scott B. Button. “Integrating Qualitative Methods in a
Predominantly Quantitative Evaluation: A Case Study and Some Reflections,” in Advances in
Mixed-Method Evaluation: The Challenges and Benefits of Integrating Diverse Paradigms, Green
and Caracelli eds. New Directions for Evaluation. Josey-Bass Publishers, No. 74, Summer 1997, pp
47-59.

McConney, Andrew, Andy Rudd, and Robert Ayres. “Getting to the Bottom Line: A Method for
Synthesizing Findings Within Mixed-method Program Evaluations,” in American Journal of
Evaluation, Vol. 3, No. 2, 2002, pp. 121-140.

Teddlie, Charles and Abbas Tashakkori, Foundations of Mixed-methods Research: Integrating


Quantitative and Qualitative Approaches in the Behavioral Science, Sage Publications, Inc., Los
Angeles, 2009.

7
TABLE 1 – METHODS FOR ANALYZING MIXED-METHODS DATA1
Analytical Brief Description Best for…
Method
Parallel Two or more data sets collected using a mix of Triangulation designs to look for
methods (quantitative and qualitative) are analyzed convergence of findings when the strength
independently. The findings are then combined or of the findings and conclusions is critical,
integrated. or to use analysis of qualitative data to
yield deeper explanations of findings from
quantitative data analysis.
Conversion Two types of data are generated from one data source Extending the findings of one data set, say,
beginning with the form (quantitative or qualitative) of quantitative, to generate additional
the original data source that was collected. Then the findings and/or to compare and potentially
data are converted into either numerical or narrative strengthen the findings generated from a
data. A common example is the transformation of complimentary set of, say, qualitative data.
qualitative narrative data into numerical data for
statistical analysis (e.g., on the simplest level,
frequency counts of certain responses).
Sequential A chronological analysis of two or more data sets Testing hypotheses generated from the
(quantitative and qualitative) where the results of the analysis of the first data set.
analysis from the first data set are used to inform the
analysis of the second data set. The type of analysis
conducted on the second data set is dependent on the
outcome of the first data set.

Multilevel Qualitative and quantitative techniques are used at Evaluations where organizational units for
different levels of aggregation within a study from at study are nested (e.g., patient, nurse,
least two data sources to answer interrelated evaluation doctor, hospital, hospital administrator in
questions. One type of analysis (qualitative) is used at an evaluation to understand the quality of
one level (e.g., patient) and another type of analysis patient treatment).
(quantitative) is used in at least one other level (e.g.,
nurse).
Data A multi-step analytical process in which: 1) a rating of Providing a bottom-line measure in cases
Synthesis
program effectiveness using the analysis of each data where the evaluation purpose is to provide
set is conducted (e.g., large positive effect, small a summative program-wise conclusion
positive effect, no discernable effect, small negative when findings from mixed-method
effect, large negative effect; 2) quality of evidence evaluations using a triangulation strategy
assessments are conducted for each data set using do not converge and appear to be
“criteria of worth” to rate the quality and validity of each irresolvable, yet a defensible conclusion is
data set gathered; 3) using the ratings collected under needed to make a firm program decision.
the first two steps, develop an aggregated equation for Note: there may still be some divergence in
each outcome under consideration to assess the overall the evaluation findings from mixed data
strength and validity of each finding; and 4) average sets that the evaluator can still attempt to
outcome-wise effectiveness estimates to produce one resolve and/or explore to further enrich the
overall program-wise effectiveness index. analysis and findings.

1
See Teddlie and Tashakkori (2009) and Mark, Feller and Button (1997) for examples and further explanations of parallel data analysis.
See Teddlie and Tashakkori (2009) on conversion, sequential, multilevel, and fully integrated mixed-methods data analysis; and
McConney, Rudd, and Ayers (2002), for a further explanation of data synthesis analysis.

8
For more information:
TIPS publications are available online at [insert website].

Acknowledgements:

Our thanks to those whose experience and insights helped shape this publication including USAID’s
Office of Management Policy, Budget and Performance (MPBP). This publication was written by Dr.
Patricia Vondal of Management Systems International.

Comments regarding this publication can be directed to:


Gerald Britan, Ph.D.
Tel: (202) 712-1158
gbritan@usaid.gov

Contracted under RAN-M-00-04-00049-A-FY0S-84


Integrated Managing for Results II

You might also like