We talk a lot about empowerment, but how do we measure it?


This page in:

It’s well-worn development wisdom that transfer programs specifically targeting women result in better child outcomes. Presumably this effect works through the empowerment of women in the household, where the shift in relative earnings gives greater weight to the preferences of the woman and less to those of her husband. Concepts of empowerment and autonomy often play a central role in development discussions – my colleague Markus has recently serially blogged about female empowerment and a few weeks back I included autonomy as a key feature in overall well-being. However empowerment and autonomy are fairly theoretical concepts and so the natural challenge for empiricists is to translate them into tangible constructs that can be measured through survey. I’ve reviewed several recent papers and notice that the same standard survey measures appear over and over in disparate settings. Are these measures fully up to the task?

Two recent thoughtful papers present good examples of the standard survey approach to individual empowerment:

-          Anderson and Eswaran define autonomy as women’s ability to make household decisions relative to their husbands’ and so ask the woman about her relative power in the decision process. The decision making measures distinguish a wide range of actions such as food purchases, durable goods purchases, and whether the wife can work outside the home. Many measures specifically involve choices regarding children, including the number of children to have, what to do when a child falls sick, and who the child will be permitted to marry. This focus on decision making was pioneered in the seminal 1960 book “Husbands and Wives: The Dynamics of Married Living” by the scholars (aptly named?) Blood and Wolfe.


-          Rahman and Rao look across a range of factors including decision making authority as above but also measure the relative mobility of the woman and whether husband permission is needed to travel to various destinations such as the health center, the local market, or the homes of relatives.

These two classes of measures appear in numerous studies that cover all major regions of the developing world. However these measures also present (at least) four challenges to the empiricist:

The measures of autonomy/empowerment ask about subjective perceptions. The reported perceptual assessments of family life may only partially and imperfectly reflect the true relative decision making power as it exists in the household. Ghuman, Lee, and Smith investigate the utility of survey measures of autonomy by identifying five surveys (from India, Pakistan, Malaysia, Thailand, and the Philippines) that contain both husbands’ and wives’ assessments of the wife’s autonomy. They find many within-marriage disagreements over the level of power in household decision making and the wife’s mobility. In one example from Pakistan, only slightly more than half of married couples give the same answer to whether the wife has to ask permission to go outside the home. In general, husbands report more egalitarian assessments of their wives autonomy than do the wives themselves, especially in South Asia.

Perceptions of autonomy are endogenous to previous life events. Interestingly, Ghuman and co-authors relate their autonomy measures to outcomes such as child mortality that have been linked to female autonomy by numerous previous studies. In India, when the wife reports that she has decision making authority over child health decisions, she is 23% less likely to have experienced a child death. On the other hand, if the husband ascribes decision making power to the wife then the family is 40% more likely to have suffered a child death.  Similarly divergent views are observed in Pakistani and Thai households. As the authors point out, it’s quite likely that the husband’s response is conditioned on past mortality experience where he rationalizes the negative event as the responsibility of the mother. What’s left unsaid by the authors, but also likely, is that the wife engages as well in ex-post rationalization.

The expression of autonomy differs by sphere of action. Autonomy in household decision making is one important domain, but not the only one. Recently two trials in South Asia have found that women’s health groups significantly reduce neo-natal mortality presumably in part by empowering women to act more confidently when interacting with health providers. This manifestation of empowerment may be completely unrelated to the relative power of decision making within the household and, if so, would not show up in the standard measures.

The expression of autonomy is shaped by community norms of acceptable social roles. It is possible that women exert influence indirectly through their relationships with male members of the household (and hence this influence may not be captured by the standard survey questions). This observation elides with an anthropological study on individual autonomy in Bangladesh. Devine, Camfield, and Gough argue that individual autonomy must necessarily be achieved within a particular institutional context which in turn varies dramatically across settings. Thus the particular expression of autonomy will also vary. In their fieldwork the authors note a clear gender difference in expressions of autonomy. Women describe autonomy in inter-relational terms often linked to the quality of interpersonal relations. For example, older women express a desire to maintain physical health and mobility so they wouldn’t burden their children. In contrast, men’s autonomy statements are more self-focused. Male respondents believe that an autonomous man “effectively manages the household”, while women associate autonomy with participation in household decision making and being consulted by the husband.

Noting how expressions of empowerment can vary across cultures, and across the lifecycle within a culture, Ghuman et al. conclude with the advice “rather than replicate the types of survey items (analyzed here) without question, future data collection efforts should involve careful consideration of whether they represent contexts faithfully and are adequate for capturing women’s empowerment (as conceptualized and expressed in that setting)”. Of course the off-the-shelf autonomy measures have some merit in a wide variety of contexts – after all they often yield seemingly sensible correlations to the outcomes that we care about. If you are in a rush to field a survey, perhaps the stock household decision making power questions will do in a pinch. But a little bit of local investigation and exploratory research should pay off with more robust and calibrated measures of autonomy and empowerment.



Jed Friedman

Senior Economist, Development Research Group, World Bank

Join the Conversation

osvaldo feinstein
August 17, 2011


Several answers to this question were provided at a 2003 WB workshop:



Jed Friedman
August 17, 2011

... (or lack thereof in some cases) over the past 8 years...

August 17, 2011

It's great that someone is trying to define empowerment operationally. We often read laudatory that literacy somehow produces empowerment but nobody knows what that means.

Helen Markelova
August 18, 2011

I really appreciate this post, especially the point on the subjectivity of the definition of empowerment and the endogeneity issues that I had not thought of before. I think that with concepts such as empowerment, there may not be a uniform and generalizable way to measure them (same with the actual definition of empowerment) precisely because of this inherent subjectivity due to social norms, personal perceptions, etc. Perhaps instead we should ask what it is we want to find out by measuring empowerment among women (as Markus' previous post showed, it is not always the case that investing in women leads to positive results in other outcomes). Can we instead measure something more "measurable" and less subjective if the goal is to ensure correct targeting of development programs? Or is women's empowerment a desired outcome in itself and thus should be measured as an outcome variable? (I don't have the answer to any of these, just wondering). I agree with your last paragraph---perhaps doing more exploratory research, including qualitative data collection and analysis, can uncover some unexpected things to include in the quantitative survey instrument that would help collect the "right" data. There are gender researchers that have combined qualitative with quantitative methods (Agnes Quisumbing, for example) to help guide the empirical analysis.

Jed Friedman
August 18, 2011

and Helen, your thoughtful reflections lead to some very pertinent questions! I think the subjective aspects of "empowerment" will always be present - but preliminary work in the study population should yield well targeted survey measures that still convey meaning and (hopefully) inform policy. Thanks!