Refining the Measurement of Women’s Autonomy: An International Application of a Multi-dimensional Construct

Women’s autonomy has long been a central concern for researchers examining the social position of women in developing countries. However, little emphasis has been placed on the measurement of autonomy, despite its importance for assessing the validity of comparative research. In this research, we use confirmatory factor analyses to determine (1) whether items thought to measure autonomy in fact form a reliable measure of autonomy, (2) whether the relationship between multiple dimensions of autonomy are strong enough to justify a discussion of autonomy as a single underlying construct, and (3) whether comparative research on autonomy is possible between two countries (India and Pakistan). We find that our indicators capture four distinct dimensions of autonomy that are moderately related, and that, while the model structures replicate fairly well across the two countries we study, there are measurement differences that make comparative research challenging. The concept of women’s autonomy has been an important one in sociology and social demography for more than two decades (e.g., Connell 1987; Cubbins 1991; Ferree and Hall 1996; Kane and Sanchez 1994; Mason 1986). Early literature defined autonomy as “the degree of access to and control over material and social resources within the family, in the community and in the society at large.” (Dixon-Mueller 1978) More recently, the definition of autonomy has been broadened to include “the ability to influence and control one’s personal environment” (Safilios-Rothschild 1982) or “the capacity to obtain information and make decisions about one’s private concerns and those of one’s intimates.” (Dyson and Moore 1983). These definitions assert a single construct that captures the multifaceted ability to gain control over the circumstances of one’s life. Among women, attaining such control is viewed as a key to improving their living conditions. An in-depth examination of this construct, therefore, can bolster our efforts to alleviate gender inequalities. In this paper, we investigate the empirical utility of the “autonomy” construct by examining (1) whether items thought – and commonly used – to measure several aspects of autonomy in fact form a reliable, unified scale and (2) assuming they do, whether the scale can be used successfully in comparative research. We use the term autonomy to reflect “the extent to which women exert control over their own lives within the families in which they live, at a given point in time.” (Jejeebhoy 2000: 205) In other words, we define autonomy in its static, individual form, rather than as a group process. As Jejeebhoy (2000) argues, various terms, including “status,” “autonomy” and “empowerment,” have been used over time to capture some element of gender equality in the household and community. “Status” was often We wish to thank Alaka Basu and five anonymous readers for their extremely valuable comments on earlier drafts. Direct correspondence to Rina Agarwala, Department of Sociology and Office of Population Research, 284 Wallace Hall, Princeton University, Princeton, NJ 08544. E-mail: [email protected]. © The University of North Carolina Press Social Forces, Volume 84, Number 4, June 2006 confused with “prestige” or “esteem” in the eyes of men, while “empowerment” referred to a more dynamic process of challenging existing power relations and gaining greater control over sources of power. Unlike autonomy, the term “empowerment” emphasizes the collective aspect of power, in addition to the individual one. While the collective aspect of power is no doubt paramount to gender equality, especially in certain cultural contexts, it is the individual aspect of power (captured through the term “autonomy”) that appears more frequently in the quantitative literature on gender and power. This paper, therefore, fits within the context of the existing literature on autonomy. Background Autonomy has been equated with mechanisms used to alter rigid gender stratification, a central concept in sociological discourse (Connell 1987; Ferree and Hall 1996). Throughout the 1970s, modernists attempted to narrow sexual inequalities by increasing women’s choices in education, employment and reproduction. In response to persisting gender inequalities, despite such attempts to assimilate women into modern life, a new generation of scholars pointed to the vital role that women’s power or autonomy plays in accessing and maximizing choices in the first place (Blumberg 1984; Fernandez -Kelly 1994; Nussbaum 2000). In the mid-1980s, scholars began to analyze the implications of autonomy through indepth empirical research in developing countries. Employing quantitative analyses of fieldlevel data, they showed that increased female autonomy was correlated with reduced fertility, improved child nutrition and education, and improved standards of living for women (Balk 1994; Basu 1992; Basu and Basu 1991; Dyson and Moore 1983). Based on such findings, the United Nations asserted in 1995 that women’s autonomy is essential to human dignity and must be considered a basic human right (UNDP 1995). Since then, several new studies have further examined the important determinants and effects of women’s autonomy in different contexts. (Balk 1997; Hashemi, Schuler and Riley 1996; Hehui 1995; Jejeebhoy and Sathar 2001; Kritz and Makinwa-Adebusoye 1999; Morgan and Niraula 1995). To date, however, there has been little evidence to show if and how these existing findings can be compared across the different contexts. Moreover, few have seriously considered how autonomy should be measured, within and across context. This is surprising given both the wealth of literature underscoring why autonomy is an important concept to study and the rich data available to analyze women’s autonomy in different countries. Precise measures of theoretical constructs, such as autonomy, can help provide empirical evidence for their existence and are essential to ensuring robust work on the roots and consequences of the constructs. In turn, such work can improve our understanding of the dynamics of gender stratification as well as our policy prescriptions to narrow gender gaps throughout the world. Indirect Proxy Measures of Autonomy Since Mason (1986) first highlighted the conceptual and methodological issues that arose in the early studies on women’s autonomy, a consensus on how we cannot measure autonomy has emerged. Autonomy cannot be measured, as it often was, using a single observable characteristic, such as women’s education or labor force participation rates, as an approximate indicator of autonomy. These one-dimensional proxies, or indirect measures, are highly imperfect and have grave policy implications, especially when used to analyze the predictors and effects of autonomy (Balk 1994; Jejeebhoy 1991; Vlassoff 1994). 2070 • Social Forces Volume 84, Number 4 • June 2006