This presentation was given by Elena Martinez (International Food Policy Research Institute), as part of the Annual Scientific Conference hosted by the University of Canberra and co-sponsored by the University of Canberra, the Australian Centre for International Agricultural Research (ACIAR) and CGIAR Collaborative Platform for Gender Research. The event took place on April 2-4, 2019 in Canberra, Australia.
Read more: https://www.canberra.edu.au/research/faculty-research-centres/aisc/seeds-of-change and https://gender.cgiar.org/annual-conference-2019/
New Metrics for Sustainable Prosperity: Options for GDP+3 (preliminary study)
Who is empowered? Predictors of empowerment in Bangladesh and Ghana
1. Who is empowered?
Predictors of empowerment in Bangladesh
and Ghana
Elena Martinez and Greg Seymour
International Food Policy Research Institute
CGIAR Research Program on Agriculture for Nutrition and Health
Seeds of Change 2019
Canberra, Australia
April 4, 2019
3. Data: Population-based Feed the
Future baseline surveys
Ghana (2012),
representative for the FTF
Zone of Influence (ZOI)
Collected by the Institute of
Statistical, Social and
Economic Research and
USAID
Bangladesh (2011-12),
nationally representative for
rural
Collected by DATA, IFPRI,
and USAID
Bangladesh Integrated
Household Survey (BIHS)
GHANA
ZOI
BANGLADESH
ZOI
RURAL
BANGLADESH
4. Women’s Empowerment in Agriculture Index
(WEAI)
Measures empowerment and
inclusion of women in the
agricultural sector
Launched in 2012 by USAID, IFPRI &
OPHI
Survey-based index, constructed
using interviews of the primary
male and primary female adults in
the same household
An aggregate index in two parts:
Five Domains of Empowerment
(5DE)
Gender Parity Index (GPI)
5. Methods
Estimated predictors of women’s and men’s empowerment score using
fractional logistic regression
Empowerment score = (# indicators individual is adequate in)/10
Fractional variable, range=[0,1]
Estimated predictors of adequacy in each indicator of empowerment using
Poisson or OLS regression
Resources: individual, household, and community characteristics
Individual: literacy, marital status, age
Household: ethnic group, wealth/asset quintile, household hunger scale,
dependency ratio, household size, household type
Community: rural/urban, region
8. Empowerment score: Ghana
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=2,784) MEN (N=3,046)
9. Being literate is associated with higher
empowerment score, but only for women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=2,784) MEN (N=3,046)
Able to
read and
write
10. Being older is associated with higher
empowerment score, but only for women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=2,784) MEN (N=3,046)
Ref=18-29
11. Being married is associated with lower
empowerment score, but only for women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=2,784) MEN (N=3,046)
Currently
married
12. Being wealthy is associated with higher
empowerment score for men and women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=2,784) MEN (N=3,046)
Ref=
Poorest
13. Empowerment score: Bangladesh
WOMEN (N=5,400) MEN (N=4,530)
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
14. Being literate is associated with empowerment
score for both men and women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=5,400) MEN (N=4,530)
Able to
read and
write
15. Being married is not associated with
empowerment score
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=5,400) MEN (N=4,530)
Currently
married
16. Being older is associated with empowerment
score for both men and women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=5,400) MEN (N=4,530)
Ref=18-29
17. Being wealthy is associated with higher empowerment score for
both men and women
Empowerment score = (# indicators individual is adequate in)/10
Fractional logistic regression models, 95% CIs
WOMEN (N=5,400) MEN (N=4,530)
Ref=
Poorest
18. What did we learn?
Wealth, literacy, marital status, and age are associated with empowerment
score, though these associations vary by country.
Some of these resources matter more for women’s agency than for men’s
agency.
Examples:
Being married was associated with lower empowerment for women in Ghana but
not in Bangladesh.
In Ghana, age mattered for women’s empowerment score; in Bangladesh, age
mattered for both men and women.
20. Input in productive decisions
WEAI indicator:
Adequate if individual participates in decisions about at least one agricultural
activity
Regression outcome:
Number of agricultural activities the individual makes decisions about
Count variable, range=[0,9]
Model:
Poisson regression
21. Resources Input in productive decisions
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- -- ++
Household size ++ ++ ++
Dual-adult household
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
22. Resources Input in productive decisions
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- -- ++
Household size ++ ++ ++
Dual-adult household
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
23. Resources Input in productive decisions
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- -- ++
Household size ++ ++ ++
Dual-adult household
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
24. Resources Input in productive decisions
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- -- ++
Household size ++ ++ ++
Dual-adult household
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
25. Control over use of income
WEAI indicator:
Adequate if individual participates in decisions about how to use income from at
least one activity
Regression outcome:
Number of activities for which the individual makes income decisions
Count variable, range=[0,9]
Model:
Poisson regression
26. Resources Control over use of income
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- ++ -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- --
Household size ++
Dual-adult household ++
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
27. Resources Control over use of income
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- ++ -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- --
Household size ++
Dual-adult household ++
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
28. Resources Control over use of income
Direction of associations based on IRRs
Bangladesh Ghana
Predictor Women Men Women Men
Literacy ++ ++
Marriage -- ++ -- ++
Age ++ ++ ++
Wealth -- -- ++
Dependency ratio --
Household hunger -- --
Household size ++
Dual-adult household ++
++ Positive association 0.05 level of confidence
-- Negative association
Blank No association
29. What did we learn?
The associations between resources and agency vary for different
dimensions of empowerment.
The predictors of achievements in each dimension of empowerment vary by
country.
Use these results to think about how to target interventions.
E.g., marital status and control over income
30. Next steps
Data from other FTF countries
Mozambique, Nepal, Tanzania
Other WEAI indicators
Autonomy in production, leisure, speaking in public, ownership of assets, rights over
assets, access to and decisions on credit
Intrahousehold parity
WEAI also allows us to compare women’s and men’s agency in the same household.
Do men and women benefit equally from a household’s resources? What is the
association between household resources and the difference between men’s and
women’s agency?
Qualitative
How do these results compare with qualitative findings? Compare to qualitative results
from the WEAI pilots.
32. gender.cgiar.org
We would like to acknowledge all CGIAR Research Programs
and Centers for supporting the participation of their gender
scientists to the Seeds of Change conference.
Photo: Neil Palmer/IWMI
33. References
Alkire, S., Meinzen-Dick, R., Peterman, A., Quisumbing, A., Seymour, G., &
Vaz, A. (2013). The Women’s Empowerment in Agriculture Index. World
Development, 52, 71-91.
Kabeer, N. (1999). Resources, agency, achievements: Reflections on the
measurement of women’s empowerment. Development and Change, 30(3),
435-464.
FTF Bangladesh Baseline Report
FTF Ghana Baseline Report
Editor's Notes
There are many ways to define empowerment. In this analysis, we use Naila Kabeer’s definition (1999). She defines empowerment as an interrelated process made up of three dimensions – resources, agency, and achievements.
Many studies look at how increasing women’s resources or agency can improve achievements such as income, nutrition, and health. In this analysis, we look at the associations between resources and agency.
What resources and characteristics allow men and women to have more agency? In other words, what resources are needed to be able to make strategic decisions about one’s own life?
This study was based on analysis of data collected through USAID’s Feed the Future program in Ghana and Bangladesh. Feed the Future is a program launched in 2010 to address hunger and global food security, with a heavy focus on improving women’s and children’s nutrition. In its first phase, FTF had 19 focus countries in Africa, Asia, and Latin America. In this analysis, I will focus on the baseline surveys from two of those focus countries: Ghana and Bangladesh. The FTF baseline survey in Ghana was collected in 2012 by the Institute for Statistical, Social and Economic Research. It is representative for the FTF Zone of Influence (the area where FTF programming was implemented, which mainly covers northern Ghana). The FTF baseline in Bangladesh was collected in 2011-12 as part of the Bangladesh Integrated Household Survey. The sample used for this analysis is nationally representative for rural Bangladesh, and the data were collected by DATA and IFPRI in partnership with USAID.
We used the WEAI to measure women’s empowerment. The WEAI is… In this study, we used each individual’s empowerment score – or the proportion of the 10 indicators they are adequate in – as a measure of empowerment. We also looked more in detail at each of the 10 indicators.
We estimated predictors of women’s and men’s empowerment score using fractional logistic regression model. The outcome – empowerment score – is the proportion of indicators in which an individual was adequate. We also looked at predictors of achievements in each indicator using Poisson and OLS regression, depending on the type of indicator.
The predictors were “resources” – including human, social, and economic resources. In these surveys, we looked at individual characteristics – literacy, marital status, and age – household characteristics – ethnic or religious group, wealth quintile, household hunger scale, dependency ratio, household size, and household type – and community characteristics – geographic region and rural/urban.
These graph show the results of fractional logistic regression models where the outcome is empowerment score of women and men in Ghana. The y-axis shows the explanatory variables and the x-axis shows the marginal association of each predictor with empowerment score, including 95% confidence intervals.
Some of the predictors of empowerment vary between men and women. For example, being literate was associated with higher empowerment for women but not for men.
Similarly, being older was associated with higher empowerment for women, but age did not matter for men’s empowerment.
In contrast, being married was associated with lower empowerment for women. Again, there was no association for men.
Living in a wealthier household was associated with higher empowerment for both men and women.
Next, I ran a similar analysis looking at predictors of empowerment score in Bangladesh.
In Bangladesh, literacy was associated with higher empowerment for both women and men.
Unlike in Ghana, being married was not associated with empowerment score for either men or women.
Being older was associated with higher empowerment score for both men and women, except for women aged 60 or older.
Living in a wealthier household was associated with higher empowerment for both men and women, though wealth mattered more for women than for men.
The empowerment score is an aggregate across all 10 indicators of WEAI. However, the associations between resources and agency may vary for different areas of agency. Next, we looked at the associations between resources and each indicator of empowerment. In this presentation, I’ll highlight two of the indicators.
First, I’ll look at the indicator input in productive decisions. In WEAI, an individual is considered adequate in this indicator if they participate in decisions about at least one agricultural activity, where the activities are things like crop production, livestock production, etc. To bring out more variability in a regression model, the outcome that we used was the number of agricultural activities the individual makes decisions about. We modeled predictors of input in productive decisions using Poisson regression.
Photo credit: https://goo.gl/images/4tKWHi
Here, I am showing you the direction of significant associations with input in productive decisions for women and men in Bangladesh and Ghana. We see some interesting differences between the countries. For example, literacy was associated with higher input in productive decisions for both women and men in Bangladesh, but not in Ghana. Age was associated with higher input in productive decisions for both women and men in Bangladesh, but only mattered for women in Ghana. Some results were more difficult to interpret. Living in a wealthier household was associated with lower input in productive decisions in Bangladesh, however. This could be explained by less diversification in wealthy households. We will need to do more analysis to unpack some of these results.
Here, I am showing you the direction of significant associations with input in productive decisions for women and men in Bangladesh and Ghana. We see some interesting differences between the countries. For example, literacy was associated with higher input in productive decisions for both women and men in Bangladesh, but not in Ghana. Age was associated with higher input in productive decisions for both women and men in Bangladesh, but only mattered for women in Ghana. Some results were more difficult to interpret. Living in a wealthier household was associated with lower input in productive decisions in Bangladesh, however. This could be explained by less diversification in wealthy households. We will need to do more analysis to unpack some of these results.
Here, I am showing you the direction of significant associations with input in productive decisions for women and men in Bangladesh and Ghana. We see some interesting differences between the countries. For example, literacy was associated with higher input in productive decisions for both women and men in Bangladesh, but not in Ghana. Age was associated with higher input in productive decisions for both women and men in Bangladesh, but only mattered for women in Ghana. Some results were more difficult to interpret. Living in a wealthier household was associated with lower input in productive decisions in Bangladesh, however. This could be explained by less diversification in wealthy households. We will need to do more analysis to unpack some of these results.
Here, I am showing you the direction of significant associations with input in productive decisions for women and men in Bangladesh and Ghana. We see some interesting differences between the countries. For example, literacy was associated with higher input in productive decisions for both women and men in Bangladesh, but not in Ghana. Age was associated with higher input in productive decisions for both women and men in Bangladesh, but only mattered for women in Ghana. Some results were more difficult to interpret. Living in a wealthier household was associated with lower input in productive decisions in Bangladesh, however. This could be explained by less diversification in wealthy households. We will need to do more analysis to unpack some of these results.
Next, we’ll look at predictors of control over use of income. In WEAI, an individual is adequate in this indicator if they participate in decisions about how to use income from at least one activity. Like the last indicator, we use a count variable in the regression to bring out more variability. So, the outcome is the number of activities for which the individual makes income decisions, modeled using Poisson regression.
Photo credit: https://goo.gl/images/wkhygU
Again, we see some interesting differences between the countries. For example, literacy was associated with higher control over use of income in Ghana, but not in Bangladesh. Some results were consistent between countries, though. In both countries, being married was associated with lower control over income for women but higher control over income for men.
Again, we see some interesting differences between the countries. For example, literacy was associated with higher control over use of income in Ghana, but not in Bangladesh. Some results were consistent between countries, though. In both countries, being married was associated with lower control over income for women but higher control over income for men.
Again, we see some interesting differences between the countries. For example, literacy was associated with higher control over use of income in Ghana, but not in Bangladesh. Some results were consistent between countries, though. In both countries, being married was associated with lower control over income for women but higher control over income for men.
Photo credit: ILRI Ethiopia
The predictors of women’s and men’s agency vary by dimension of empowerment and country.
What can we do with these results? Some of these results can help us think about how to target interventions. For example, we found that being married was associated with lower control over income for women. Of course, this does not mean that women should not get married. Rather, interventions aimed at increasing women’s control over income or changing gender norms around how income is managed should focus on married women and consider how income and spending decisions change when someone gets married.
Some of this is already done, but I could not fit it into the presentation. So here is a preview of other elements of this work.
Based on regression coefficients
Based on incidence rate ratios (IRR)
*Higher for lower age group difference