Measuring Women’s

Economic
Empowerment in
Financial Sector
Deepening and
Financial Inclusion
programmes
Guidance Note
November 2023

Gareth Davies &

Joanna Ledgerwood
Acknowledgements

This Guidance Note is based on a set of guidance materials originally produced by the authors for
Financial Sector Deepening Kenya (FSDK) and Access to Finance Rwanda (AFR). The Guidance Note
builds on the seminal work of Naila Kabeer as well as the excellent guidance published by the Abdul
Latif Jameel Poverty Action Lab (J-PAL) and the International Center for Research on Women (ICRW),
among other sources. The authors would also like to thank Sabine Garbarino for her invaluable
comments.

1
1 Overview
For many programmes working in Financial Sector Deepening (FSD) or financial inclusion
programmes, women and Women’s Economic Empowerment (WEE) are increasingly at the
centre of programme strategies and Theories of Change. This makes it imperative that
programmes are able to report credible and robust results on WEE, and that monitoring and
evaluation systems generate timely and useful data and insights that can be used by staff to
continuously improve the WEE impacts achieved. Better gender data, including on WEE, can
also be effective in highlighting gender gaps and driving programme and partner commitment
to delivering results for women and girls.

The measurement of WEE is widely acknowledged to be complex and challenging.

Conceptually, WEE can be defined in different ways, and what counts as ‘empowerment’ can
vary across contexts, regions, and individuals. Robustly measuring subjective elements of
empowerment can be tricky, with direct survey questions often subject to reporting bias.
Features of the data collection process itself – such as the gender of the enumerator or the time
and place of the interview or survey – can also introduce measurement bias if not considered
carefully.1

Given these complexities, this Guidance Note presents a framework and methodology for
measuring the WEE outcomes of FSD projects and interventions. The Note can be used for
both monitoring and evaluation purposes. The Note also provides useful guidance for
undertaking research exploring concepts relating to empowerment (or any research where
collecting the perspectives and experiences of women as well as men is important).

The rest of the Note is organised as follows:

• Section 2 presents a simple framework for conceptualising WEE. Key concepts in

empowerment such as ‘access’ and ‘agency’ are introduced and defined, with
illustrative examples from FSD programming. This section provides the conceptual
foundation for the remainder of the Note.
• Section 3 presents a five-step methodology for measuring WEE. For any given project
or intervention, the first step is to map the empowerment pathway. In the second step,
indicators are defined for each element of the empowerment pathway. The third step
involves developing a measurement plan which sets out the data collection methods,
sources, samples, timing, and responsibilities for each indicator. The fourth step covers
the actual data collection process, including the design of data collection instruments,
pre-testing, enumerator recruitment, and so on. In the final step, data is analysed and
synthesised and used to both report progress and results to donors and other
stakeholders and to inform programme decision-making.
• Section 4 provides a list of external resources. Examples of survey questions and tools
for measuring WEE are also included in the Annex.

1
For an excellent summary of the challenges in measuring WEE, see: Glennerster, R. et al. “A
practical guide to measuring women’s and girls’ empowerment in impact evaluations”. J-PAL.

2
2 WEE Conceptual Framework
This section presents a simple framework for conceptualising WEE. Key concepts in
empowerment such as ‘access’ and ‘agency’ are introduced and defined, with illustrative
examples from FSD programming. This section builds on the extensive literature on WEE,
including Kabeer’s seminal 1999 article on empowerment2 and more recent guidance from
International Centre for Research on Women (ICRW).3

The 3A model of empowerment: Access, Agency, and Achievements

An economically empowered woman can be defined as a woman who is able to achieve the
economic and non-economic goals she has set for herself (and her family). Rather than seeing
empowerment as a binary outcome – ‘empowered’ or ‘disempowered’ – empowerment is best
thought of as a continuum: the more a woman is able to set her own goals and achieve those
goals, the more empowered she is. This definition of empowerment involves three interrelated
concepts: access, agency, and achievements – see Figure 1.

Figure 1: the 3A model of empowerment

Each element of the model is discussed further below.

2
Kabeer, N. (1999). “Resources, Agency, Achievements: Reflections on the Measurement of
Women’s Empowerment”. Development and Change, 30 (3): 435-464.
3
Golla, A. M. et al (2018). “Understanding & Measuring Women’s Economic Empowerment:
Definition, Framework & Indicators”. ICRW.

3
To achieve their goals, women require access to certain resources. This includes access to
goods, services, and assets, as well as more intangible resources such as access to information
and networks, and human and social capital. In the context of FSD programming, examples
might include:

• Access to formal and informal financial services;

• Access to technology and other enabling products and services, such as mobile
phones;
• Access to sources of knowledge, skills, and information that help build an individual’s
numeracy, literacy, and financial capabilities;
• Access to assets, such as land, which can serve as collateral for loan products;
• Access to social networks and social capital, for example through savings groups.

To be empowered, women also need the agency to set and pursue their own goals and make
their own choices. This includes active participation in decision-making processes and the
power women have vis-à-vis other decision-makers (such as husbands and partners), as well as
the ability and confidence women enjoy to set their own goals. One important factor in agency
is the division of family unpaid care responsibilities, including housework, meal preparation,
childcare, care of elderly relatives, and so on. Women with lower levels of agency often
shoulder a disproportionate share of the unpaid care burden within a household, which in turn
reduces their ability to set other goals and to pursue other options outside the home. Other
important features of agency include:

• Mobility, both physically – such as being able to leave the home without the permission
of a male family member – and economically – such as being free to enter occupations
that may be seen as ‘male occupations’;
• Voice, self-confidence, and leadership;
• Financial independence and autonomy, such as having control over one’s own savings;
• Freedom from violence, shame, or stigma that might otherwise restrict the real set of
choices available.

Achievements considers the extent to which women are able to achieve the goals they have set
for themselves and, as a result, enjoy meaningful improvements in well-being and life
outcomes. In terms of economic empowerment, this could include goals such as starting a new
business or growing an existing business, earning higher income, or securing employment in a
desired occupation. Achievements can also include the non-economic benefits that result from
these economic gains, such as improved health and education outcomes, as well as subjective
benefits such as feelings of wellbeing, life satisfaction, and self-esteem.

Note that there is not necessarily a simple linear pathway from access to agency to
achievements. Each of the three As can interact in different ways and there can be both
positive and negative feedback loops between the three different concepts. For example, to
access a financial service such as an e-wallet, a woman will first need a minimal level of agency
such as the ability to decide to open an account, the ability to negotiate successfully with her
husband (if she first needs her husband’s permission), the mobility to leave the home and meet
with the agent, and so on. She will also need access to other resources such as a mobile
phone, information about the product, and the requisite levels of literacy and financial
capabilities, which in turn may require additional starting levels of agency. If the woman is able
to access the financial service, this may further boost her agency, for example by providing a
safe place to save, thereby increasing her financial autonomy. It may also allow her to achieve
other goals, for example helping her to start a business. This may contribute to further gains in

4
her agency if, for example, it further improves her self-confidence and negotiating power within
the household.

There is also the potential for negative feedback loops or trade-offs. For example, access to a
financial service that allows a woman to expand her business may increase her income
(achievements) but also result in increased working hours which, without changes in the
division of unpaid care (agency), will lead to increased time poverty and potentially lower life
satisfaction overall. Similarly, gaining employment in a job that is informal, insecure, and with
poor health and safety may not feel empowering and result in worse life outcomes on some
metrics.

Underpinning the access, agency, and achievements of individual women are a broader set of
structural factors and institutions. This includes cultural, social, political, and economic
institutions, both formal and informal, that influence which resources women can access and
the levels of agency they can exercise. For example, there may be a set of gendered social
norms that favour men over women in the inheritance of land. This may be further reinforced
and reflected in the laws around inheritance and land ownership, and in the cultural practices
of community leaders, judges, and so on. This in turn shapes women’s control over land and
the ability to access collateralised loans, as well as their financial autonomy and decision-
making power in the household (agency).

The strength and influence of these structural factors can vary across time and space and can
interact with individual circumstances and characteristics in various ways. For example, women
with higher levels of education and better access to legal services may be better able to
challenge discriminatory informal norms around inheritance. In general, woman (and men)
may seek to circumvent or directly challenge institutions that create barriers to the achievement
of their goals, thereby contributing to societal change. However, depending on the context,
this can carry the risk of backlash or social sanctions for the individuals involved.

5
3 Five-step Methodology for Measuring WEE
This section presents a five-step methodology for measuring WEE outcomes for a given FSD
project or intervention, building on the conceptual framework presented in Section 2. The five
steps are summarised in Figure 2 below.

Figure 2: five-step methodology for measuring WEE

Each of the steps are described in more detail below. References for additional external
guidance can be found in Section 4.

3.1 Step 1: map the empowerment pathway

The first step is to map out the causal pathway illustrating how the intervention or project is
expected to contribute to improved WEE outcomes. From a project design perspective, this
step provides a systematic process for project teams to set-out the specific WEE results they
hope to achieve as well as the causal mechanisms involved. It also provides an opportunity to
systematically record the theories or hypotheses underpinning these causal mechanisms and
the assumptions or conditions required for these causal mechanisms to hold, and to test these
theories and assumptions against the available evidence and data. This is important for
targeted interventions with the primary objective of empowering women, and for
mainstreaming where implementers want to ensure results for women without WEE necessarily
being the main objective. From a measurement perspective, this process also provides the
project-specific framing for subsequent WEE data collection and analysis. Given that WEE is a
broad and complex concept, this helps to focus data collection on those aspects of WEE that
are most relevant to the given project or intervention.

The pathway should incorporate the three As of the WEE conceptual framework presented in
Section 2. As noted in Section 2, WEE is not a linear process. The sequencing of changes in
access and agency may therefore vary by project (although achievements will typically come
after improvements in access and/or agency). The pathway may also include various feedback
loops between the three As.

Note that the three As can appear as outcome boxes in the empowerment pathway and as
assumptions underlying the causal logic (if they are preconditions needed for the causal logic
to hold). To build on the example used in Section 2, the empowerment pathway for an

6
intervention around digital financial services might state that the launch of a revised e-wallet
savings product is expected to increase women’s access to financial services, which in turn
increases her agency (because she now has a safe, private space to save thereby increasing her
financial autonomy, and by allowing her to better spend her income in accordance with her
own preferences), which in turn increases her achievements (because she can better achieve
her economic goals, such as starting or growing a business). Underpinning this will be a set of
assumptions about pre-existing levels of access and agency that need to be in-place if the
pathway is to hold, such as access to a mobile phone, minimum levels of financial capability
and decision-making power, and so on. This example is illustrated in Figure 3.

Figure 3: hypothetical example of an empowerment pathway for a digital financial services

intervention4

The same thought-process can be used to identify project risks in terms of barriers to the
delivery of positive WEE outcome and the potential for unintended negative WEE outcomes.
For example, as noted in Section 2, the expansion of a business can lead to an increase in
women’s time poverty if not accompanied by a redistribution of unpaid care burdens.
Documenting the assumptions underpinning the empowerment pathway also helps to identify
potential barriers to economic empowerment (should the assumptions fail to hold in practice).
For example, in a context where many women have low access to basic education, some
women may not have the requisite financial capabilities to access an e-wallet. From a project
design perspective, this can help project staff to think of ways in which the product can be
modified in order to minimise these barriers, for example by working with the financial service
provider to ensure that the product interface is as simple as possible or offers an interactive
voice response platform to address barriers related to literacy or visual impairments, and that
agents are trained to provide the necessary customer support. From a monitoring perspective,

4
Note that in the interest of space, the assumptions or conditions have only been enumerated for
one link in the empowerment pathway. In practice, assumptions or conditions should be developed
for each link in the pathway.

7
this process helps to identify key assumptions to be tested and potential unintended negative
outcomes that should be monitored for. It can also help to identify people and questions to
investigate in the monitoring plan and data collection instruments, such as monitoring
husbands’ attitudes towards their wives accessing financial services like e-wallets (Step 4).

The empowerment pathway should be evidence-based, drawing on primary and secondary

data and research. This can include evidence from previous programme interventions,
interventions from other FSD programmes, and research and evidence from other countries
that share similar contexts. From a strategy perspective, this can help to focus programme
resources on projects and interventions that the evidence suggests will have the greatest
likelihood of delivering strong, positive WEE outcomes.

Note that if a Results Chain exists for the project or intervention, the empowerment pathway
can be incorporated into the Results Chain.5 If no Results Chain exists, a separate
empowerment pathway can be developed.

Learning Questions

Having constructed the empowerment pathway, project and monitoring staff have the option
of developing a set of Learning Questions. These are key questions that staff and partners want
answered through the monitoring and evaluation process. Learning Questions can be
developed systematically for each link in the empowerment pathway, or might focus on specific
linkages where the evidence base is currently weak. Following the hypothetical example
developed above, Learning Questions for the e-wallet intervention might include:

• What are the key barriers for women accessing digital financial services like e-wallets,
and how do these vary by characteristics – such as levels of education – and contexts –
such as rural versus urban settings?
• To what extent does access to digital financial services like e-wallets contribute to
improvements in women’s agency and what are the causal mechanisms?

Learning Questions such as these can provide useful prompts when developing measurement
plans (Step 3), Terms of Reference for evaluations or impact assessments, or when analysing
and synthesising data (Step 5).

Participatory, self-defined empowerment outcomes

When mapping out the potential WEE outcomes for a given project or intervention, several
WEE guidance notes recommend using participatory approaches whereby the women
targeted by an intervention are themselves able to define what would count as a positive
empowerment outcome.6 This recognises the fact that what counts as ‘empowerment’ can vary
across cultures, regions, and individuals, and avoids the imposition of an outsider’s view of
what women should want. It also helps to understand how women themselves would trade off
different outcomes, such as extended working hours that increase income but may reduce
leisure time and increase time poverty. One useful technique is the Participative Ranking
Methodology (PRM). This involves focus group participants identifying and discussing a
number of different possible responses to a question posed by the moderator, then ranking
the responses in order of importance. By analysing responses across a number of groups, the

5
For more guidance on constructing Results Chains, see the Donor Committee for Enterprise
Development (DCED) Standard for Results Measurement (www.enterprise-
development.org/implementing-the-dced-standard/).
6
See, for example, Glennerster, R. et al., op. cit. and Anand, M. et al, (2014). “Practical Tools and
Frameworks for Measuring Agency in Women’s Economic Empowerment”. The SEEP Network.

8
most frequently cited responses can be identified and used to develop locally-grounded WEE
outcomes and indicators. However, as noted by Glennerster et al., “we also need to keep in
mind that women’s preferences may reflect society’s views about gender rather than their own
true preferences. Even though women’s preferences are an important component of
empowerment, measuring preferences alone may not always fully reflect women’s ability to
make a meaningful choice.”

Project Example: participatory, self-defined empowerment outcomes

“The Arab Women’s Enterprise Fund (AWEF) developed a tailored qualitative research
process based on Focus Group Discussions (FGDs) conducted at the start of the project. It
allowed for the collection of specific agency characteristics and pathways directly from female
participants. To structure this process, an imaginary woman was introduced to the group. This
woman was constructed as a ‘typical case’ with whom participants could easily identify. The
facilitator told the story of the woman as she progressed through an intervention but stopped
short of providing any outcomes. Participants were asked to continue the woman's story and
describe the agency-oriented outcomes she would achieve. Once a sufficient number of
agency characteristics were identified, participants voted individually on which are most
important and likely to occur to the imaginary woman, resulting in a continuum of indicators
that can be used for quantitative agency measurement. During this process, participants were
also encouraged to share their own individual stories.”
Reproduced from Anand, M. et al, (2014). “Practical Tools and Frameworks for Measuring Agency in Women’s
Economic Empowerment”. The SEEP Network.

3.2 Step 2: develop WEE indicators

The next step is to develop a set of WEE indicators; at least one indicator should be developed
for each box in the empowerment pathway. This will allow project teams to track progress and
test the causal theories and assumptions underpinning the empowerment pathway. Should
the WEE outcomes not materialise as expected, the team will then be able to pinpoint where
the empowerment pathway is breaking down. By measuring agency as well as achievements it
will also be possible to assess whether any positive achievements observed actually reflect
women’s own priorities and choices (the true measure of empowerment).

Indicators can be quantitative (numerical) or qualitative (descriptive); in general, it is good

practice to combine quantitative and qualitative indicators. This is especially so in the case of
WEE given the complexity and nuance involved in measuring empowerment outcomes and the
subjectivity of many of the underlying concepts (such as self-confidence, wellbeing, and self-
esteem). Qualitative indicators can also provide a rich source of data regarding the barriers
facing women in accessing financial services and in achieving their goals, data which can be
used by project teams and partners to adapt and improve project design and delivery (see
Step 5).

Indicators can be defined to measure either the ‘breadth’ or the ‘depth’ of change. In the case
of WEE, typically we want to measure change in both dimensions. For example, if a digital
financial service project is expected to contribute to an increase in women’s financial
autonomy, an indicator of breadth might be: ‘the number of women reporting an increase in
savings that they control’. An indicator of depth might be: ‘the average value of savings that
women control’, or ‘qualitative changes in the extent to which women feel financially
autonomous’. Combining breadth and depth indicators allows project staff to assess both how
many women have been reached, and how substantive or consequential the outcome changes
are for the women concerned.

9
As noted above, WEE is a complex term and it is not possible to measure every aspect of
empowerment for every project or intervention; indicators should therefore focus on the
specific aspects of access, agency, and achievements that the project or intervention is
expected to bring about (as identified in Step 1). This also means that while a programme
might use some common or standard WEE indicators (to aid comparability and aggregation
across projects), these should be complemented by project-specific indicators tailored to the
particular empowerment pathway and relevant to the project context.

Given the real potential for unintended WEE outcomes – such as increased time poverty – and
the trade-offs many women have to negotiate – such as taking a job that provides increased
income but that increases the risk of sexual harassment – the set of indicators defined needs to
capture these different potential outcomes (to the extent that project M&E resources allow).
For example, an indicator relating to the number of women gaining formal employment might
be complemented by a qualitative indicator relating to job quality.

Whilst it is important to attempt to capture the subjective elements of WEE, these types of
indicators can be difficult to measure and subject to reporting bias and should therefore be
complemented with more ‘objective’ indicators. This is particularly the case with indicators
relating to gender norms which can be subject to social desirability bias whereby respondents
provide answers that they think the enumerator wants to hear or that are more in-line with what
is perceived to be socially acceptable behaviour (rather than the answers that best reflect their
reality). Some people may also be uncomfortable speaking freely about certain WEE-related
topics. It is therefore good practice, where possible, to combine subjective indicators with
objective or proxy indicators – see the box below for an example.

Project Example: combining ‘subjective’ and ‘objective’ indicators

In a randomised evaluation of commitment savings accounts in the Philippines, the evaluators
identified women’s influence in household spending decisions as their main outcome of
interest. They collected data on two indicators:
1) An index of women and men’s responses about who decides in nine common
household spending decisions, and
2) Household expenditures on what respondents identified as typically ‘male’ or ‘female’
goods.
This allowed the evaluators to triangulate the more subjective indicator on household
decision-making with a more objective measure of the actual outcome of the decision-making
process in terms of actual household expenditures. However, this approach only works if
women and men report different spending preferences, and individual preferences can be
subject to change over time.
See Ashraf, N. (2010): “Female Empowerment: Impact of a Commitment Savings Product in the Philippines”.
World Development 38 (3): 333-344. https://doi.org/10.1016/j.orlddev.2009.05.010

All WEE indicators should be SMART:

• Specific: indicators are clearly defined and unambiguous, reducing the scope for
subjective interpretation.
• Measurable: indicators can be measured or assessed (either by numbers or a
description); indicators are within the means of the programme to measure. Proxies
can be used when the underlying change is too complex or costly to measure given
programme resources.
• Attributable (or ‘contribution-sensitive’): indicators are tightly defined around the
change the project is trying to catalyse; indicators are not so broad that there are many

10
 external factors that could cause changes in the indicator such that the contribution of
the project to changes in the indicator cannot be plausibly established.
• Relevant: indicators are appropriate and applicable to the box in the empowerment
pathway being measured; indicators will provide useful information for management
decision-making.
• Time-bound: when relating to a flow (e.g. number of women accessing a new or revised
financial service) rather than a stock, indicators are defined over a specific timeframe
(e.g. number of women accessing a new or revised financial service in the last 12
months).

By the end of Step 2, the set of indicators defined should cover each of the boxes in the
empowerment pathway and provide a means for answering the key Learning Questions
identified in Step 1. Indicators can also be developed to monitor and test any key assumptions
or potential unintended negative outcomes.

Sex-disaggregation of indicators

In some cases a programme will already have defined indicators that are relevant to the
empowerment pathway developed in Step 1. For example, at the access level, indicators may
already exist for the number of individuals accessing and using new or revised financial
services, or the number of individuals reporting increased income. Also, often programmes
will want to measure certain outcomes for both women and men. In these cases, indicators
should be sex-disaggregated.

Whilst this is straightforward for indicators relating to individuals (such as ‘the number of
people accessing a financial service in the last 12 months’), sex-disaggregation of indicators
relating to households, enterprises, or other units or organisations is more complicated.
Afterall, a unit such as an enterprise does not have a sex, only people do. One seemingly
obvious solution is to look at the sex of the enterprise owner and disaggregate all enterprise-
related indicators by Women-Owned Enterprises. However, beyond micro enterprises and
sole-traders, an enterprise may have multiple owners, potentially of different sexes. Even in a
small family enterprise primarily operated by a woman, her husband may exercise control over
key business decisions. Conversely, an enterprise formally ‘owned’ by the husband may in fact
be primarily run and controlled by the wife. Similar issues arise for household-related
indicators, where sex disaggregation is usually based on who ‘heads’ the household. However,
households designated as ‘male-headed’ can hide a full spectrum of decision-making power
enjoyed by female household members. For this reason, when it comes to measuring WEE, it
is best to define indicators relating to individuals rather than units. This means the black box of
‘the household’ or ‘the enterprise’ can be unpacked, giving a truer picture of the agency and
achievements of women.

Similar issues arise when trying to define indicators relating to income. Measuring income at a
household level, while useful from a wider impact perspective, does not provide much insight
regarding WEE outcomes as it glosses over intra-household decision-making. Instead, income
could be measured at the individual level. However, care is needed when defining the
indicator as receipt of any additional income may not translate into control of that income, or
into realising the benefits of the additional income. Project staff therefore need to carefully
consider which concept of income they are trying to capture and also consider adding
indicators relating to decision-making and/or household expenditure (see box above).

Beyond sex-disaggregation, indicators can be disaggregated by other characteristics such as

age, marital status, education level, location (rural vs. urban), disability and so on. This builds
on the idea of intersectionality: ‘women’ are not a homogenous block, and what works for

11
women with certain characteristics might not work as well for women with completely different
characteristics. Which disaggregations will be most useful can be informed by the project
assumptions identified in Step 1. For example, if a key condition or assumption is that women
possess a minimal level of literacy and numeracy to access a new digital financial service,
disaggregating an indicator by education level will be a useful means of testing this assumption
and for identifying the threshold level of education women need. If the project includes
objectives related to disability inclusion, specialised tools such as those by the Washington
Group on Disability Statistics may be useful. These additional levels of disaggregation allow
more nuanced insights to be developed during the data analysis stage (see Step 5).

Example WEE indicators

The table below provides some example WEE indicators for each of the three As in the
conceptual framework.7

Table 1: example WEE indicators

Elements of WEE Example indicator

Access Number of women accessing and using a new or revised financial service
Number of women with access to a mobile phone
Average financial literacy scores of women (based on a standardised
financial literacy test)
Agency Number of women reporting improvements in household decision-making
power
Qualitative improvements in women’s reported voice, self-confidence, and
leadership
Number of households where household expenditure more closely
matches women’s preferences
Average number of hours spent by women per day on household chores
and unpaid care responsibilities
Achievements Number of women reporting improvements in their ability to achieve their
economic goals
Qualitative changes in the extent to which women report being able to
achieve their goals
Number of women reporting improvements in their feeling of wellbeing
and life satisfaction (using standardised psychometric scales 8)

7
Note that indicators are not necessarily the same as survey questions. For example, for the
indicator ‘Number of women reporting improvements in household decision-making power’,
directly asking women in a single question whether they have seen improvements in their
household decision-making power is unlikely to yield meaningful results. Instead, the indicator
could, for example, be calculated based on the responses to a set of more specific, relatable
questions regarding different types of concrete decisions. See Section 3.4 for more guidance on
designing WEE survey questions.
8
See Section 3.4 for examples.

12
3.3 Step 3: develop a WEE measurement plan
Having developed WEE indicators for each box in the empowerment pathway, the next step is
to develop a measurement plan. For each indicator, the measurement plan summarises:

• Any indicator definitions / calculations, including definitions for key terms used in the
indicators (e.g. ‘usage’ defined as using the financial service at least once in the
previous quarter) and the calculations used to populate the indicator (e.g. ‘increase’
calculated as a percentage increase from the baseline);
• The data collection method(s) to be used to populate the indicator (e.g. observation,
interviews, Focus Group Discussions, surveys, etc.);
• The sources / samples used to populate the indicator, including the sampling frame
and sampling method;
• Whether baseline data collection is required and the timing of any baseline data
collection;
• The frequency / timing of post-baseline data collection;
• The responsibility for collecting and verifying the data (or for overseeing data collection
by a third party).

The basic format of a measurement plan is reproduced in Table 2 below.

Table 2: project or intervention measurement plan

Empowerment Indicator Definitions / Data Sources / Baseline Frequency / Responsibility
pathway box calculations collection samples data timing
method collection
Box 1… Indicator
1.1…
Indicator
1.2…
Box 2…

Underpinning the detailed measurement plan should also be an attribution or contribution

methodology, detailing how the programme will assess the attribution or contribution of
programme activities to the observed changes in WEE outcome (versus other external factors).9

Each of these elements of the measurement plan are considered below, focusing on what is
unique about measuring WEE.10

Data collection methods

For each WEE indicator, the Measurement Plan specifies which data collection method will be
used to populate the indicator. A variety of measurement tools are available including:
observation, interviews, Focus Group Discussions (FGDs), surveys, and so on. Some tools are
more suited to collect either qualitative or quantitative data. For example, FGDs are usually
used to collect qualitative data whereas surveys are primarily used to collect quantitative data
(although some qualitative data can also be collected through open-ended survey questions,
and scoring within FGDs can yield quantitative data).

9
For more on attribution and contribution methodologies, see Sen, N. (2021). “Measuring
Attributable Change: Implementation Guidelines for the DCED Standard”. DCED.
10
For more general guidance on measurement planning, see the DCED Standard for Results
Measurement (www.enterprise-development.org/measuring-results-the-dced-standard/).

13
The choice of data collection method depends on the type of changes being measured, the
robustness required, the type of data needed (quantitative or qualitative), and the budget
available. Subject to project M&E resources, triangulating methods will provide a more robust
and insightful set of data than just relying on one or two methods. This is especially so in the
case of WEE given that common data collection methods like surveys and interviews can be
subject to reporting and measurement bias. If possible, combining these tools with techniques
such as direct observation, games, or experimental vignettes can provide more robust findings
(see below). These other tools can also be helpful for collecting data on elements of WEE that
are hard to measure, such as attitudes towards gender norms that respondents might not
report honestly or that they are not even aware of (e.g. subconscious gender bias).

Including qualitative techniques alongside quantitative methods is also important given the
complex and nuanced nature of the empowerment process and the real risk of causing
unintended negative WEE outcomes which need to be monitored for (such as increased time
poverty or sexual harassment). Qualitative data is also useful for identifying WEE barriers and
informing adaptations to project design (see Step 5).

Some of the main data collection methods for measuring WEE are discussed further below.

Direct observation. This involves directly observing and recording behaviours and actions of
individuals and organisations in the field. For example, project staff might directly observe
savings group meetings and record various quantitative and qualitative observations such as
how well-run the meetings are, how many people were in attendance, and whether and how
many of the steps in the group’s constitution were followed. Direct observation has the
advantage of providing more objective measures of phenomenon of interest. For example, the
number of times women speak in community meetings, and how long they speak, may provide
a more objective measure of ‘voice’ and ‘self-confidence’ (to complement more subjective self-
reported data from surveys and interviews). However, it may not be possible or may be too
costly to observe many of the decision-making processes and other events that a programme
may be interested in.

Secondary data. This involves obtaining data already collected by market actors, stakeholders,
and other development programmes. Client data collected by partner financial service
providers is one important example. The great advantage of secondary data is that, as it is
already being collected by third parties, it costs very little for the programme to obtain such
data. It can also be collected at a scale that would be prohibitively expensive for a donor
programme to duplicate. However, a big disadvantage is that, because the data is primarily
being collected for someone else’s needs, it may not be as robust as we would like, and may
not cover all of the things we are interested in. This is particularly so in the case of WEE:
beyond access, few financial service providers will have an interest in, or the capabilities to,
collect data on women’s agency and achievements. Even at the access level, many financial
service providers do not routinely collect accurate sex-disaggregated client data – see box.

14
 Challenges in collecting sex-disaggregated data from financial service providers
Financial Sector Deepening and financial inclusion programmes often expect and rely on
partner financial service providers to collect and share sex-disaggregated client data.
However, across the world many financial service providers do not routinely collect sex-
disaggregated data and, if they do, it can often be inaccurate.
For example, a study by the GBA and McKinsey found only 55% of banks interviewed reported
having the capability to disaggregate customer data by sex.11 Similarly, at the 2015 AFI Global
Policy Forum, half of respondents stated that the capacity to disaggregate by sex was a big
challenge. This included restricted functionality of basic ‘off the shelf’ Management
Information System (MIS) software. They also cited the challenge of assigning a ‘sex’ to an
enterprise – such as what counts as a Women-Owned Enterprise (discussed in Step 2) – and
how to treat joint accounts. Participants also reported problems with the quality of data: staff
were not always properly trained to understand the field categorisations in the MIS. Cultural
and social barriers can also affect the quality of data. For example, HBL in Pakistan performed
a random sampling exercise of its female deposit portfolio in 2014. It found that the actual
female control of accounts tagged as ‘woman-owned’ in rural areas was only around 50%.12
As a first step, programmes should include a requirement for financial service providers to
collect and report on sex-disaggregated client data (at least as it relates to the programme
support provided) in MoUs and partnership agreements. However, recognising that some
financial service providers may not have the systems and capabilities to do so, additional
financial and/or technical support may be required to enable financial service providers to
upgrade their data systems. Longer term, donor programmes can also work with regulators,
policy-makers, and other stakeholders to encourage or mandate the collection of sex-
disaggregated data across the financial sector.13

Where secondary data is available, consideration needs to be given to the reliability and
relevance of the data. This might include an assessment of the methodology used in the
collection of national statistics, or an assessment of the internal quality assurance processes
partners use to verify their internal data.

Semi-structured interviews. Semi-structured interviews are loosely structured interviews,

shaped around a pre-defined set of open-ended questions. They resemble a conversation,
allowing a free flow of ideas and information. Interviewers frame follow-up questions
spontaneously, probe for information, and take notes (which are elaborated on and organised
after the interview). Because they are loosely structured, they are ideal for identifying
unintended consequences of an intervention. They are also ideally suited to collecting
nuanced data and insights into concepts around agency and the more subjective elements of
achievements. Because they are conducted one-on-one (versus FGDs which are conducted in
a group setting), they are also a good tool for collecting data on more sensitive aspects of
WEE. However, the findings from semi-structured interviews are not always easily
generalisable, and the small sample sizes means the findings may not be as robust as surveys.
Subject to the available M&E resources, it is generally preferable therefore to combine
interviews with surveys.

Focus Group Discussions (FGDs). FGDs are similar to in-depth interviews except that they are
conducted with a group rather than an individual. They are primarily used to collect qualitative

11
“How Banks Can Profit from the Multi-Trillion Dollar Female Economy”. Global Banking Alliance
for Women (2014).
12
“The 2015 Afi Global Policy Forum Report”. Alliance for Financial Inclusion (2015).
13
For more ideas on how organisations can help close the gender data gap, see: “Transforming the
Data Landscape: Solutions to Close Gender Data Gaps”. Data 2X (2022).

15
data. They are useful when it is anticipated that a group dynamic will be useful to elicit fuller
and more complete answers from respondents. Participants are guided by a moderator or
facilitator who introduces topics for discussion and helps the group to participate in a lively and
natural discussion amongst themselves. The strength of FGDs relies on allowing the
participants to agree or disagree with each other so as to provide an insight into how a group
thinks about an issue, and to highlight any inconsistencies and variation that exists in a
particular group in terms of beliefs, experiences, and practices. Compared to interviews,
however, FGDs may not be suitable for exploring more sensitive WEE topics. There is also a
risk that one or two higher-status (more empowered) individuals dominate the discussion,
leading to biased data.

Surveys. Surveys are used to gather (predominantly) quantitative information from a large
number of respondents (although small amounts of qualitative information can also be
gathered using open-ended questions). The great advantage of surveys is that they allow the
collection of statistically robust data (assuming the questionnaire is well designed and biases
are not introduced in the administration of the survey – see Step 4). Because a standardised set
of questions is used, a larger number of respondents can be reached than through more time-
intensive forms of data collection like semi-structured interviews. On the flip side, however,
surveys can be quite reductive in the data collected, missing some of the important nuances
and insights provided by interviews and FGDs. As noted above, they can also be quite limited
in measuring certain elements of WEE, such as attitudes to gender norms.

Experimental vignettes. Vignettes are short descriptions of hypothetical situations. They are
commonly used in questionnaires and in interview guides as a way of asking questions about
concepts that may not be easy to grasp if asked directly. Experimental vignettes involve
changing a key detail in the story, with each version assigned randomly to respondents to test
whether the change makes any different to the responses given. In the case of WEE, changing
the sex of the protagonist in the vignette, for example, would provide an insight into hard-to-
measure things like subconscious gender bias. For example, to assess the effectiveness of
gender training for bank loan staff, two versions of a loan application could be constructed,
identical but for the sex of the fictitious applicant. Randomly assigning these two versions of
the loan application to a sample of bank loan staff, before and after the training, could provide
insights into whether the training had been effective in reducing gender bias.

Project Example: using experimental vignettes to examine unconscious gender bias

One evaluation of a local governance project in India looked at whether exposure to female
leaders in Indian village councils changed perceptions about women’s effectiveness as
leaders. As part of the survey, researchers played a short recording of a speech by a local
leader responding to a complaint from a villager. Respondents were randomly assigned to
hear the recording spoken by a man or woman. After the speech was over, they were asked
to rate the leader’s effectiveness. The vignette allowed researchers to measure whether there
was a subconscious bias that led people to rate female leaders as relatively less effective. The
study found that exposure to a female leader through the policy that reserved village council
head positions for women reduced men’s bias against female leaders.
See Beaman, L. et al (2009): “Powerful Women: Does Exposure Reduce Bias?.” The Quarterly Journal of
Economics 124, no. 4: 1497-1540.

Games. This involves constructing an artificial scenario in which an individual or group of

individuals are asked to make choices or decisions. They are designed to reveal something
about how participants might act in the real world. For example, the case in the box below
shows how a game involving small amounts of real money was used to measure the bargaining
power of women vis-à-vis their husbands. Games are particularly useful for complementing

16
survey or interview questions that may be subject to reporting bias or that cover issues that are
difficult for respondents to conceptualise or provide answers to, such as household power
dynamics. However, games need to be carefully designed to avoid the bias introduced by the
fact that people are being observed while playing the game, and how people act in the game
may not reflect how they actually behave in real life situations or where the stakes are higher.

Project Example: using games to examine differences in spending preferences

In a randomised evaluation in Kenya, Simone Schaner examined how differences in spending
preferences between husbands and wives affected demand for savings accounts. The
researchers used a game to measure intra-household bargaining power to test whether the
savings account programmes had different impacts on women with more or less bargaining
power. At the end of the survey, husbands and wives, who were being surveyed separately,
were asked to divide a small cash prize between themselves and their spouse. Each spouse
recorded his or her allocation separately on cards and placed the amount they allocated to
themselves in their tin and the amount allocated to the spouse in the spouse’s tin. Then they
came together to decide how to allocate the prize money between them and recorded it on
cards, which they added to each of their tins. To ensure respondents’ privacy, the husband
and wife also each added an additional envelope to each of their tins with a randomly selected
amount of money. The husband and wife then chose one card from each of their tins and were
immediately given the cash amount allocated to them on the card they chose.
This game allowed researchers to identify women with relatively low- or high-bargaining
power and test whether the impact of the intervention was different for women with different
levels of bargaining power.
See Schaner, S. (2017): “The Cost of Convenience? Transaction Costs, Bargaining Power, and Savings Account
Use in Kenya”. Journal of Human Resources 52 (4): 919-945. https://doi.org/10.3368/jhr.52.4.0815-7350R1

Sources and sampling

Having determined the data collection methods, the next step is to sketch out the data sources
and sampling approach. For methods such as interviews and FGDs, this means planning how
many interviews or FGDs to conduct and how participants will be selected. For surveys,
thought needs to be given to the sample size and sampling methodology. In each case, this
involves thinking through who possesses the information we need, and whose voices and
perceptions we need to capture. Note that from a WEE perspective this does not mean only
interviewing or surveying women. For example, to measure changes in intra-household
decision-making it is often useful to interview or survey both women and men in the household.
This will allow answers to be triangulated and for differences in perspectives and attitudes to be
identified.

There is no magic number for how many interviews or FGDs to conduct, or how many surveys
to administer. For interviews and FGDs it is often not until the interviews and FGDs are being
conducted that it is possible to determine exactly how many to do. If the time and resources
allow, it is good practice to repeat them until they no longer provide new insights. For surveys,
the exact sample size to be used depends on the level of robustness required, which in turn
depends upon the size of the population and the amount of variability in responses (measured
as the relative standard error). A larger sample size increases the robustness of the survey, but
costs more to administer.14 If questions relating to WEE outcomes are being included as part

14
Note that increases in the size of the population above 10,000 do not materially affect the size of
the sample required to achieve a given level of robustness. Robustness is measured in terms of the
confidence interval and the confidence level. As a rough guide, with a relative standard error of 0.5

17
of a larger mixed-sex survey, it is important to ensure that women are adequately represented
in the sample (for example by using random stratified sampling whereby the number of women
in the sample is proportional to the number of women in the overall population)15. Depending
on the attribution or contribution methodology, the sample may also need to include a
treatment and control group.

Timing and frequency of data collection

For each indicator, the Measurement Plan specifies when and how often the indicator will be
collected. In general, to calculate the change in an indicator, two observations are required: a
baseline, which records the value of the indicator before the effects of the intervention are felt,
and a second observation taken after the effects of the intervention are first felt. Often further
observations are taken to record trends over time and to monitor sustainability. Note that when
conducting a baseline assessment, ‘before the effects of the intervention are felt’ does not
necessarily mean ‘before the start of an intervention’. For example, it may take many months of
working with a partner before they launch a new or revised financial service, in which case it is
not necessary to conduct a baseline WEE assessment right at the start of the intervention. In
fact, it may be impossible to do so if it is not possible to identify the treatment group until they
have actually accessed the service. In some instances it might be possible to collect before and
after data in one data collection exercise by asking respondents to recall the baseline situation.
However, care should be taken to ensure the recall periods are realistic for respondents (see
also Step 4).

When thinking about the timing of WEE data collection, thought needs to be given to the
timeframes over which change is realistically expected to happen. In terms of the 3A model,
access level changes will typically happen the quickest. Changes to agency and achievements
can take longer to manifest, depending on the nature of the intervention and the
empowerment pathway (see Step 1). These timeframes need to be factored into the
Measurement Plan: if agency and achievement level changes are measured too soon, and
without further rounds of data collection, important WEE outcomes might be missed. In some
cases it might be necessary to continue to monitor changes a year or more after the end of the
project or intervention.

Although the full WEE outcomes of a project may unfold over a long time period, projects
should look for early signs of positive or negative change, to inform on-going project
adaptation and decision-making. If WEE outcomes are only measured at the end of a project,
or after a project has closed, this limits the ability of project teams to adapt their strategies and
tactics in response to what is and is not working on-the-ground. The measurement plan should
therefore include data collection exercises that will allow teams to test early on key assumptions
underlying the empowerment pathway and to check for any potential negative outcomes. For
example, FGDs and small surveys with female clients of a new or revised financial service,
carried out within a few months of the product launch, may reveal useful insights into women’s
experiences of accessing and using the new product (including any barriers encountered),
what they intend to use the product for, and whether they anticipate the product to contribute
to any changes to intra-household dynamics. These insights can then be used by the project
team to co-develop modifications to the product design or distribution strategy before the
product is rolled-out further.

and population of 10,000, a sample size of 96 will give a +/-10% confidence interval with a
confidence level of 95%.
15
Depending on the indicator disaggregation, it might also be necessary to identify additional strata
such as rural versus urban populations.

18
3.4 Step 4: collect WEE data
Having developed the Measurement Plan, the next step is to design the individual data
collection instruments (such as survey tools and questionnaires, or interview guides) and collect
the data. From a WEE perspective, getting this step right is critical given the potential for
reporting or measurement bias when collecting WEE data.

Designing data collection instruments

For surveys, semi-structured interviews, and FGDs, a well-designed questionnaire or interview /

FGD guide is important in minimising biases and measurement error. As discussed in Step 3,
questions relating to WEE are prone to bias and error. For example, people may be reluctant
to openly discuss or answer questions relating to sensitive gender topics; answers can also be
subject to social desirability bias. Poorly designed surveys can also lead to the minimisation of
women’s roles and voices, for example by forcing a binary choice between ‘male-headed’ and
‘female-headed’ households or enterprises. Involving WEE experts from within or outside the
programme when developing data collection instruments is recommended.

Glennerster et al. identify five good-practice principles for constructing survey (and interview)
questions that minimise potential bias and error:

• Specific. Each question should only ask one thing at a time.

• Neutral. The wording of the question should not bias respondents to give a particular
answer one way or another.
• Understandable. All survey questions should be relatively easy to comprehend by
anyone in the sample. This means avoiding unfamiliar terms or concepts that are not
clearly defined in the survey prompt. In the context of WEE, questions should typically
avoid abstract concepts like ‘empowerment’ or ‘agency’. For example, instead of asking
‘are you empowered to make decisions around accessing financial services?’, a better
question might be: ‘who in your household decides whether to open a new bank
account or other financial service?’.
• Clearly framed. Good questions have clear boundaries and well-defined timeframes. If
the responses are multiple choice, the list of possible answers should be mutually
exclusive and collectively exhaustive. Any question asking a respondent to recall
something from the past should include a clear timeframe and this recall period should
be short enough that the respondent can remember accurately.
• Relevant. It is important to be respectful of study participants’ time, so all the survey,
interview, or FGD questions should measure indicators that will actually be used by the
programme. Questions should also be relevant to the local context.16

Various techniques can be used to put the respondent at ease and increase their willingness to
answer potentially sensitive questions. A good way to start a survey or interview is with simple
questions such as basic demographic information. This is useful for disaggregating the analysis
(see Step 5); it also helps the enumerator to build rapport with the respondent before
proceeding to more sensitive topics. Another technique is to frame questions indirectly
through the use of vignettes or hypothetical scenarios. For example, instead of asking a direct
question about the respondent’s own life, the question could be framed by asking: ‘for
someone similar to you in your community…’. Vignettes can also be used to ask for people’s
opinions about a fictional situation, the answers to which reveal something about their real-life
opinions and viewpoints. To elicit insightful data, it is important that these vignettes are
grounded in the local context and therefore realistic and relatable for respondents (see the

16
Adapted from Glennerster et al., op. cit.

19
Annex for examples). Similarly, when asking about things like decision-making processes,
experience suggests that asking specific questions about specific decision-making processes
provide more reliable data than more general questions. For example, rather than asking a
single general question such as ‘who in your household makes decision about finances?’,
consider developing a series of more specific questions such as: ‘who in your household
decides whether to open a savings account (if you have one)?’, ‘who in your household decides
how much money to save each month?’, ‘if you have taken a loan in the last 6 months, who in
your household decided to take the loan’, etc. As noted by Glennerster and Walsh (2017),
asking more specific questions about a concrete scenario tailored to the choices women care
most about in the study context may be easier for respondents to answer accurately, and may
reveal more about whether they can make meaningful choices that matter to them.17

Project Example: measuring psychological components of empowerment

The agency and achievements element of the 3A empowerment framework includes various
subjective or psychological aspects such as feelings of wellbeing, self-esteem, and self-
confidence. There are various existing survey tools and questionnaires from psychology that
can be used to measure these aspects of WEE.
For example, the CARE evaluation of a financial inclusion and empowerment project in
Uganda used two different psychometric scales as a measure of empowerment (and
disempowerment). The Pearlin Mastery Scale asks respondents to rate responses from 1
(strongly disagree) to 7 (strongly agree) to questions such as:
• ‘There’s no way I can solve the problems I have’
• ‘Sometimes I feel I am being pushed here and there in life’
• ‘I have little control over things that happen to me’
• ‘I can do anything when I put my mind to it’
Similarly, the Kessler Phycological Distress Scale (K10) asks for responses rated on a scale from
‘none of the time’ to ‘all of the time’ to questions such as:
• ‘During the last 30 days, about how often did you feel tired out for no good reason?’
• ‘During the last 30 days, about how often did you feel hopeless?’
• ‘During the last 30 days, about how often did you feel depressed?’
The two scales were used in before and after surveys of three different groups: a treatment
group that accessed a Village Savings and Loan Association (VSLA) plus mobile wallet for
women; a second treatment group that accessed the VSLA, mobile wallet, plus a course of
seven household counselling sessions; and a control group. The evaluation found that for
married women, there was a significant positive improvement against both psychometric
scales for the second treatment group but not the first treatment group (vis-à-vis the control
group). The CARE study complemented these psychometric scales with more ‘objective’
proxy indicators such as the redirection of male income from alcohol toward family goals.
Another example of a psychometric scale used in WEE project evaluations is the Relative
Autonomy Index, which provides a measure of the extent to which people feel they have the
capacity to achieve their own goals, or are at the fate of external factors like social norms or
coercion.
Given the standardised nature of these scales and the questions they are based on, a risk is
that they are not well suited to the project’s local context or easily understandable or
answerable for the survey population. It is therefore important to pre-test the questions before
they are deployed in the full survey (see below).
See Scott, L. (2020). “Digital Subwallets and Household Dialogues: Final Report”. CARE.

See: Glennerster, R., and Walsh, C. (2017). “Is It Time to Re-Think How We Measure Women’s
17

Household Decision-Making Power?” J-PAL.

20
Pre-testing data collection instruments

Before embarking on a large-scale survey exercise it is important to pre-test the data collection
instruments in locations similar to the ones where the full survey will take place. This is
especially important for WEE data collection given the often complex and abstract nature of the
concepts we are ultimately interested in, the sensitivity of some topics, and the potential for
reporting bias. Pre-testing is useful for checking things such as:

• Whether respondents understand the questions and find them easy to answer. For
example, some terms may not translate well into local languages or be confusing for
respondents.
• Whether the survey is too long. Respondents can become fatigued during long surveys
and their answers may become less accurate towards the end of the survey.
• Whether recall periods are realistic. People may struggle to accurately answer
questions about decisions or behaviours that occurred too long ago.
• Whether the questions are sensitive enough to capture variation. For example, if asking
questions to assess a respondent’s level of financial capabilities, the questions should
not be too easy such that every respondent gets them right, or too hard such that every
respondent gets them wrong.
• Whether the timing and location of the interviews is right. For example, while it may be
more convenient for the enumerator to interview female workers at their place of work,
respondents may be too busy to answer the questions. Thought also needs to be given
to women’s unpaid care burdens. For example, female retailers may tend to open their
shops latter then men due to childcare responsibilities – this needs to be taken into
account if planning to interview women at their shops (see also below).

To gather respondents’ feedback on the survey instrument itself, two techniques can be used:

1. Respondent debriefings, whereby enumerators implement the survey and then collect
feedback and comments from respondents at the end of the survey;
2. Cognitive interviews, whereby respondents describe their reactions and thought
processes in addition to answering the survey question.

Based on the findings of the pre-test, revisions may be required to the data collection
instrument and/or the planned data collection process.

Data collection processes

During WEE data collection, careful consideration needs to be given to enumerator training
and recruitment and the where and when of data collection. All of these considerations have a
gender aspect which have the potential to bias results or introduce measurement error.

Enumerator recruitment. When collecting WEE data from respondents it is good practice to
match the gender of the enumerator with the gender of the respondent. This is especially true
of more sensitive aspects of WEE and in contexts where gender norms restrict the socially
acceptable interactions between women and men. At the same time, male researchers are
often required to engage with (largely male) community leaders when approval for community-
based research is sought. To build rapport and put the respondent at ease, it can also be
helpful if the enumerator is from the same region and ethnicity as the respondent, and should
also be able to speak any local languages. Because the identity of the enumerator can
potentially affect the answers given by respondents, ideally the same enumerators should be
used for the treatment and control group, and in baseline and repeat surveys.

21
Where to conduct interviews. When collecting potentially sensitive WEE data, whether the
respondent is alone or within earshot of other family members or members of the community
can affect the truthfulness of their answers. Ideally interviews should be conducted alone. In
some contexts, this may require the consent of a family member such as a husband (or parents
if interviewing young women). A consent script, explaining the purpose of the research and
guaranteeing the anonymity of the respondent, can help to alleviate any concerns. Interviews
can also be conducted within sight, but out of earshot, of the husband or parents. If it was not
possible to perform the interview in private, this should be recorded in the survey. Even if the
respondent is alone they may be reluctant to talk openly about certain issues. One solution is
to allow respondents to directly input their answers into a tablet or smart-phone. For
respondents with low levels of literacy, technology such as Audio Computer-Assisted Self-
Interview platforms can be used.

If respondents are required to travel, as can be the case for FGDs, there is also a need to
consider if this is safe for women and how to avoid introducing any sampling biases (for
example, by excluding women who are less mobile or have greater unpaid care burdens).
Other considerations include the provision of child-care for smaller children and whether the
participation of older children is appropriate.

When to conduct interviews. As noted above, the time of day when interviews are conducted
can potentially bias the survey results. For example, female retailers with childcare
responsibilities may open their shops later in the morning. Conducting interviews at a
marketplace early in the morning may therefore result in a sample that is bias towards women
without children or who do not have the same unpaid care burdens. The time of year can also
potentially affect survey responses. For example, women in rural households may have
different workloads at different times of year, depending on the agricultural cycle. When
conducting panel surveys, it is therefore important to consider whether each survey round
should be conducted during the same month each year.

Ethics and safeguarding. It is always important to maintain high standards of ethics when
conducting primary research. The following standards should be observed:

• Voluntary participation. Participation should be voluntary and free from external

pressure. Information should not be withheld from prospective participants that might
affect their willingness to provide data. All participants should have a right to withdraw
and withdraw any data already provided at any point without fear of penalty.
• Informed consent. Data enumerators should provide a clear statement of intent to
inform participants how information and data obtained will be used, processed, shared,
and disposed of, prior to obtaining consent.
• Child protection. Informed consent should be obtained from parents or caregivers and
from children themselves. Children should not be required to participate even if their
parents provide consent.
• Treatment of participants. Data enumerators must be aware of differences in culture,
local customs, religious beliefs and practices, personal interaction and gender roles,
disability, age and ethnicity, and should be mindful of the potential implications of
these differences when collecting data.
• Confidentiality. Data enumerators must respect people’s right to provide information in
confidence, and must ensure that sensitive information cannot be traced to its source.
They should also inform participants about the scope and limits of confidentiality.
• Data security. Data collectors should guard confidential material and personal
information by the proper use of passwords and other security measures. There is a
duty to state how data will be stored, shared, archived and (if necessary) disposed.

22
Given the sometimes sensitive nature of WEE data collection, it is even more imperative that
high ethical standards are maintained. In some cases this may require the development of
additional ethical and safeguarding guidelines. This is especially true if collecting data on
Gender Based Violence. This includes carefully wording questions to avoid unnecessary
distress, checking the national laws for the mandatory reporting of violence, and considering
what safeguarding and referral processes to put in place.18

3.5 Step 5: analyse and use WEE data

The final step is to analyse and use the WEE data collected.

Interpreting and analysing data is a vital step in ensuring that the programme generates useful
learning and insights around WEE which supports evidence-based decision-making, allowing
project staff to respond in a dynamic way to what is and is not working on the ground. Data on
its own carries no meaning: for data to become knowledge and learning it must be analysed,
interpreted, and put into context. Data analysis includes:

• Calculation of summary statistics – such as totals and averages,

• Outlier analysis – to identify interesting positive or negative outcomes,
• Trend analysis – to examine changes over time, and
• Correlation analysis – to examine relationships between variables.

The data from different tools (especially quantitative and qualitative tools) should be
triangulated and synthesised to build up a more complete picture of the change process.
Beyond the planned data collection activities from the Measurement Plan, additional field
observations and tacit information collected by project teams and consultants can also be
included in the data analysis process.

If any project-level learning questions were identified at the start of the project (see Step 1), the
data analysis process should aim to answer these questions. Data analysis can also be used to
answer the general learning questions: what works and why, for whom, in what circumstances.
Building on the idea of intersectionality, this recognises that women are not a homogenous
group: what works for some women might not work well for other women with different
characteristics or in different circumstances. For example, analysis of the access data may
reveal lower usage of a financial service among rural versus urban women. There may be clues
in the existing data, particularly qualitative data, to explain why this is the case. It may be, for
example, that rural women have to travel further to access the service, or that they tend to have
lower financial literacy and are therefore are less confident at using the service. Sometimes
further data collection may be required to differentiate between competing hypotheses
(represented by the dotted arrow in Figure 2).

By measuring change in each box in the empowerment pathway it should be possible to

validate the assumptions underlying the pathway, monitor progress in achieving WEE
outcomes, and identify any WEE barriers or unintended negative outcomes. By plotting data
and findings against the empowerment pathway it should be possible to identify whether the
project or intervention is on track to deliver the desired WEE outcomes and, if not, to pinpoint
where the causal logic is breaking down. It may be that certain assumptions are not holding in
practice, or that women face additional barriers that were not previously considered. As noted
above, further data may be required to narrow down the range of possible explanations. Data
collection is therefore seen as part of a dynamic problem-solving process, not a rigid exercise

For more in-depth guidance on measuring domestic violence and intimate partner violence, see
18

Annex 1 of the WHO Multi-country Study on Women’s Health.

23
designed only to record programme successes. Being curious is an important aspect of
analysing and using WEE data.

The process of generating lessons and insights from data can be a collaborative process
involving project partners. For example, towards the end of the pilot phase of an innovation
project, project staff may facilitate an insights and lesson learning workshop – involving the
project team, partner counterparts, and external consultants – to analyse the pilot data, share
observations, identify any issues, generate hypotheses, plan additional data collection (if
required), and brainstorm solutions. In some cases a partner financial service provider may
have tested multiple different versions or calibrations of a new model or innovation. Data
should be analysed to understand which version is most effective (which may vary across
contexts and client segments). The process of collecting and discussing gender data can in
itself challenge prejudice or gender stereotypes and facilitate stronger partner engagement
with WEE objectives.

Project teams are also encouraged to hold internal learning workshops where they can share
initial data and insights from their project(s), and brainstorm hypotheses and potential solutions
with colleagues. More formal review processes can also be used to promote intra- and inter-
project learning. For example, the learning from one innovation project might well be relevant
for other innovation projects also aiming to improve WEE outcomes. Creating diverse project
teams can increase the effectiveness of these sessions – different team members will bring
different ideas, perspectives, knowledge, and experiences, and can help to challenge each
other’s assumptions and biases.

To complete the cycle, project staff should use the insights and lessons generated to improve
the intervention, modifying or dropping what does not work, and scaling-up what does. This
might involve, for example, working with project partners to modify the design of a new
financial service so that it is more accessible to women (which may in turn require modifications
to the support package provided by the programme). In some cases it might be necessary to
pause or exit an intervention if, for example, serious negative WEE outcomes are identified.
The empowerment pathway should also be regularly reviewed and updated to reflect the latest
learning and insights (thereby completing the cycle in Figure 2).

24
4 Resources
This section lists useful external resources for conceptualising and measuring WEE.

WEE concepts and frameworks

Kabeer, N. (1999). “Resources, Agency, Achievements: Reflections on the Measurement of

Women’s Empowerment”. Development and Change, 30 (3): 435-464.
https://onlinelibrary.wiley.com/doi/abs/10.1111/1467-7660.00125

Golla, A. M. et al (2018). “Understanding & Measuring Women’s Economic Empowerment:

Definition, Framework & Indicators”. International Center for Research on Women.
www.icrw.org/publications/understanding-and-measuring-womens-economic-empowerment/

Measuring WEE

Anand, M. et al, (2014). “Practical Tools and Frameworks for Measuring Agency in Women’s
Economic Empowerment”. The SEEP Network. https://seepnetwork.org/files/galleries/2019-
WEE-MeasuringWomensAgency-_EN-DIGITAL.pdf

Badiee, S. et al (2022). “Transforming the Data Landscape: Solutions to Close Gender Data
Gaps”. Data 2X. https://data2x.org/resource-center/transforming-the-data-landscape-
solutions-to-close-gender-data-gaps/

Buvinic, M., and Furst-Nichols, R. (2013). “Measuring Women’s Economic Empowerment:

Companion to A Roadmap for Promoting Women’s Economic Empowerment.” United Nations
Foundation and Exxon Mobil Foundation.
http://womeneconroadmap.org/sites/default/files/Measuring%20Womens%20Econ%20Emp_
FINAL_06_09_15.pdf

Calder, R. et al (2020). “Measurement of Women’s Economic Empowerment: WOW Helpdesk

Guidance Note No. 2”. Work and Opportunities for Women.
https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_dat
a/file/895765/Guidance-Measurement-Womans-Economic-Empowerment2.pdf

Donald, A. et el (2020). “Measuring Women’s Agency”. Feminist Economics 26(3):200-26.

https://openknowledge.worldbank.org/handle/10986/35282

Glennerster, R. et al. “A practical guide to measuring women’s and girls’ empowerment in

impact evaluations”. J-PAL.

Glennerster, R., and Walsh, C. (2017). “Is It Time to Re-Think How We Measure Women’s
Household Decision-Making Power?” J-PAL. www.povertyactionlab.org/blog/9-6-17/it-time-
rethink-how-wemeasure-women’s-household-decision-making-power-impact

Lombardini, S. et al. (2017). “A ‘How To’ Guide to Measuring Women’s Empowerment”.

Oxfam GB. https://policy-practice.oxfam.org/resources/a-how-to-guide-to-measuring-
womens-empowerment-sharing-experience-from-oxfams-i-620271/

25
Qualitative data collection (general guidance)

Ager, A, et al (2010). “Participative Ranking Methodology: A Brief Guide: Version 1.1.” New
York, NY: Program on Forced Migration & Health, Mailman School of Public Health, Columbia
University. https://doi.org/10.13140/RG.2.2.34356.45448

Mack, N. et al (2005). “Qualitative Research Methods: A Data Collector’s Field Guide.”

Research Triangle Park: Family Health International.
www.fhi360.org/sites/default/files/media/documents/Qualitative%20Research%20Methods%2
0-%20A%20Data%20Collector%27s%20Field%20Guide.pdf

Raworth, K. et al. (2012). “Conducting Semi-Structured Interviews.” Oxfam. https://policy-

practice.oxfam.org.uk/publications/conducting-semi-structuredinterviews-252993

Researching Gender Based Violence

García-Moreno, C. et al (2005). “WHO Multi-Country Study on Women’s Health and Domestic

Violence against Women: Summary Report of Initial Results on Prevalence, Health Outcomes
and Women’s Responses”. Geneva: World Health Organization, 101.
www.who.int/reproductivehealth/publications/violence/24159358X/en/

Innovations for Poverty Action (2018). “The Safe and Ethical Conduct of Violence Research:
Guidance for IPA Staff and Researchers”. Poverty Action. www.poverty-
action.org/publication/ipv-ethical-guidance

26
ANNEX: example survey questions

Annan, J. et al (2016). “The Returns to Microenterprise Support among the Ultrapoor: A Field
Experiment in Postwar Uganda.“ American Economic Journal: Applied Economics 8(2): 35-64.

When you have small amounts of money, such as 500 or 2,000 shillings, can you decide how to
spend it on your own? (Yes / No)

Follow up with:

• When an expensive item like a bicycle or a cow is purchased by the household, is your
opinion listened to in the decision of what to buy?
• If you have some money you have earned, can you use it to purchase clothing for
yourself or children without asking the permission of anyone else?
• Are you allowed to buy and sell things in the market without asking the permission of
your partner?

If money is available, who in your household decides whether to pay school fees for a relative
from your side of the family? (You primarily / You with someone else / Someone else without
you)

Follow up with:

• If money is available, who in your household decides whether to purchase items like a
radio or a paraffin lamp?

If you have money that you have earned, can you refuse to give some to your partner if he/she
wishes to purchase alcohol? (Often / Sometimes / Rarely / Never)

Do you agree that a wife has a right to buy and sell things in the market without asking the
permission of her husband? (Yes / No / Don’t know)

If a wife has earned some money, does she have the right to buy clothing for herself or her
children without asking the permission of her husband? (Yes / No / Don’t know)

Almas, I. et al (2015). “Measuring and Changing Control: Women’s Empowerment and

Targeted Transfers.“ National Bureau of Economic Research Working Paper No. 21717.

In the last 2 weeks, did you and your spouse argue about managing money? (Yes / No / Not
applicable / Don't know)

Who in the household usually decides how much money to be spent on food? (Wife /
Husband / Together / Not applicable / Don't know)

Follow up with:

• Who in the household usually decides about the financial administration?

Imagine the following household composed of a wife, a husband and three children. The wife
is 40 years old and her husband is 43 years old. The three children are aged 5, 10 and 14.
Both wife and husband have been unemployed in the last 2 years and have been receiving
Financial Assistance. Today, the wife receives X Macedonian Denars (MKD) from her parents to

27
help the family. Who do you think should decide what to do with that amount? (Wife /
Husband / Together / Not applicable / Don't know)

In the following questions you will be facing different scenarios in which you will have to choose
between two alternatives, A or B. You cannot choose both. If you choose A it means you prefer
alternative A to alternative B. Which of these two alternative options do you prefer?

• 550 MKD paid to you (A) or 600 MKD paid to your partner (B)?
• 500 MKD paid to you (A) or 600 MKD paid to your partner (B)?
• 400 MKD paid to you (A) or 600 MKD paid to your partner (B)?
• 300 MKD paid to you (A) or 600 MKD paid to your partner (B)?

Björkman Nyqvist, M. et al (2017). “Mothers Care More, but Fathers Decide: Educating Parents
about Child Health in Uganda“. American Economic Review, 107(5): 496-500

To what degree do you agree with these statements? (Strongly agree / Agree / Disagree /
Strongly disagree / Not applicable)

• When women get rights they are taking rights away from men
• Gender equality, meaning that women and men are equal, has come far enough
already
• A wife should obey her husband, even if she disagrees.
• It is important for a man to show his wife/partner who is the boss.
• It is the job of men to be leaders, not women
• A woman should be able to choose her own friends, even if her husband disapproves
• A man should decide how to spend his free time on his own
• A woman should decide how to spend her free time on her own
• If a woman has power in the household, it means she is taking power away from her
husband
• A husband and wife can share power
• Women’s opinions are valuable and should always be considered when household
decisions are made

Green, D. et al (2018). “Silence Begets Violence: A Mass Media Experiment to Prevent

Violence against Women in Rural Uganda“. Working Paper.

In some of the other villages we have visited, (some) people think that a man has good reason
to hit his wife if she disobeys him, while (other) people in those communities do not think this is
a good reason to hit one’s wife. In your community, do people think a man has a good reason
to hit his wife if she disobeys him? (Yes / No / Don’t know / Refuse to answer)

Thinking of the opinions held by people in your community, do most people think that a man
has a good reason to hit his wife if she does not complete her household work to his
satisfaction? (Yes / No / Don’t know / Refuse to answer)

Karlan, D. (2017). “Impact of Savings Groups on the Lives of the Poor“. Proceedings of the
National Academy of Sciences

How would you rank your ability to do these activities on a scale of 1 to 10?

28
 • Run your own business
• Identify business opportunities to start up a new business
• Obtain credit to start up a new business or expand an existing business
• Save in order to invest in future business opportunities
• Make sure that your employees get the work done properly
• Manage financial accounts
• Bargain to obtain cheap prices when you are buying anything for the business
• Bargain to obtain high prices when you are selling
• Protect your business assets from harm by others
• Collecting the money someone owes you

How true is the following statement, on a scale of 1-10?

• While doing any task, it is important for me to do it better than others

• If I start working in a task, I definitely see the end of it no matter how difficult it is
• If I have the chance, I would make a good leader
• I want to be a respectful person in my village
• I do not care what others think about my success or failure
• I am in control of what happens in my life
• I often make plans for the future
• I believe that my future is determined by luck no matter how hard I work

Lucia, R. et al (2020). “Measuring And Understanding Unpaid Care And Domestic Work:
Household Care Survey”. Oxfam.

https://policy-practice.oxfam.org/resources/measuring-and-understanding-unpaid-care-and-
domestic-work-household-care-survey-621082/

29
Website
www.tandem-development.com

Contacts
[email protected]

Address
Tandem Development International Ltd,
56 Leman Street,
London,
United Kingdom,
E1 8EU