2.2 CFA, self-compassion based empowerment and psychometrics

This study utilized a confirmatory factor analysis (CFA, henceforth) approach to examine the measurement of self-compassionate empowerment. CFA is useful since it “places a priori structure on the data and allows the explicit testing of competing hypotheses regarding the measurement properties of indicators thought to reflect the theoretical construct” [20]. Additionally, CFA allows for refining the empowerment measure based on the modification indices. Modification indices indicate how the model might be adjusted to improve its fit. The empowerment measure in this paper was developed based on a systematic review of empirical scholarship with a focus on measurement issues [8,20,25,27]. Initially, the theoretical model included a total of 18 items thought to capture empowerment that privileges self-compassion. Indicators for self-compassionate empowerment that directly or indirectly focus on the self were identified under four dimensions (1) bodily integrity (3 items); (2) economic (3 items); (3) decision making (6 items) and (4) interaction with non-family groups and public visibility (6 items). The identification process was guided by theoretical scholarship as well as gaps in empirical studies. For example, under the dimension of bodily integrity, indicators such as self-reported health, contraceptive practice and fertility choice have been explored. The attempt has been to identify indicators that demonstrate her perception and ability to do things for herself. For example, the empirical scholarship on self-compassion and its unique relationship to the body has been examined in sports studies [28] where women reported an active interest in understanding and taking ownership of their own bodies. Hence, negotiations around fertility and contraceptive choices have been used as body-related constructs that enhance women’s capacities. Again, women’s tolerance of domestic violence is an important indicator to assess her own self-esteem and bodily integrity. However, in the current data, the violence questions were worded in the context of the overall community (in your community, is it usual for husbands to beat their wives in the following situations) and hence could not be used. Similarly, under the dimensions of economic and decision making, variables were identified that indirectly reflect the woman’s value of self-worth within the household. For example, through questions on “who has the most say” on economic choices (e.g. buying an expensive item, land/property, expenditure on social functions) or her economic freedom (a marker of self-worth) achieved through the routes of financial acknowledgement (e.g. name on ownership papers; bank account). Again, under the dimension of “interaction with non-family groups”, the attempt has been to identify indicators that go beyond domestic duties and demands of kinship ties and represent some degree of physical freedom that the woman might enjoy. It is conceivable that variables such as her participation in the local self-help group, village/ward level administrative meetings or visiting friends on her own represent an enhanced capacity to negotiate the household environment while paying attention to her own social needs, self-worth and wellbeing.

Significantly, literature on self-compassion emphasizes on the theoretical as well as empirical relevance of autonomy, social interactions/support and subjective wellbeing. Hope and colleagues [29] in their study on University freshmen in Canada found a positive relationship between self-compassion and autonomy which in turn affected their goal pursuits and overall subjective wellbeing. Similarly, Toplu-Demirtas et al. [30] in their research on lesbian, gay and bisexual (LGB) in Turkey show how self-compassion mediate s relationship between social interaction and well-being. Although, it has not been empirically established and perhaps not immediately obvious how women’s physical mobility and her participation in local governance bodies are directly associated with self-compassion, it is reasonable to assume that a self-compassionate woman is more aware of her personal environment (through extra-familial social connections/ties and freedom of movement) to make empowering choices for herself; in fact in Neff’s original formulation [15], social connectedness and autonomy were important dimensions of self-compassion. Having said that, it must be noted that the variables chosen for this study are not completely unambiguous in their conceptual rendering of woman’s self-compassion and hence may not readily translate into their wellbeing. This is primarily because most of the scholarship on self-compassion is linked to a range of psychological health indices which are unavailable in the current dataset and perhaps theoretically unattainable in a non-phenomenological research design. Nonetheless, this conceptual ambiguity has been partially addressed by utilizing CFA to examine the associations between the observable items chosen for the overall empowerment measure.

Table 1 presents the list of all items in the original theoretical model under the above four dimensions.

Source: India Human Development Survey (IHDS-2, 2011–12)

The original empowerment measure (a composite index) included 18 variables covering the above four dimensions; a confirmatory factor analysis (CFA) was performed and it was found that 11 variables have higher factor loadings than the rest and hence were retained. The final empowerment measure (Cronbach’s alpha = 0.67; mean = 2.61; s.d = 1.82) was an index that took the value from 0 to 11; 0 = indicating either “no” or “no say” in any of the identified variables and a value of 11 = indicating “yes” or a “most say” in all questions. To demonstrate that CFA offers a more “parsimonious understanding of covariation among a set of indicators” [24], an alpha value for all the original 18 items was calculated. It resulted in a score of alpha = 0.52 (as opposed to an alpha = 0.67 with the 11 items following the CFA). Later, to ensure that the CFA-based measure offers an improvement over summative empowerment index, an OLS model estimating the effects of predictors on empowerment (summative index of all the 18 items) was carried out (results not reported here). Results show that a dampening effect on the positive association between empowerment and antenatal care (i.e. unstandardized beta coefficients, b = 0.016* as opposed to b = 0.129*** (CFA adjusted measure).

Additionally, construct validity of this measure was assessed by looking at the distribution of empowerment values (0–11) across the Indian states. The goal was to examine if this measure is consistent with the scholarship on the association between women’s agency and demographic behavior. The differential ranking of the states in terms of gender inequality and subsequently a woman’s agency (or social position) has been shown to be governed by sociological factors including marriage practices (endogamy/exogamy), cultural tolerance of violence, sexual autonomy, freedom of movement and finally, social institutions (e.g. legal/inheritance, lineage and descent systems) that either constrict or facilitate women’s decisions about her own education, paid work, health and family (See [11,31,32]). For example, as observed in previous scholarship the empowerment values in the current study were higher in the north eastern (such as Sikkim (3.26), Manipur 4.24 or Meghalaya 4.68) and the southern states (Kerala 3.77 or Tamil Nadu 3.21) of India when compared to the northwestern states-of Punjab (2.33) or Haryana (2.24) or Uttar Pradesh (2.19) or Gujarat (2.57)-that have values lower than the average value of empowerment of the country as a whole (2.61). Significantly, these state-level differences can be largely attributed to the practice of patriarchy with the northwestern and a few eastern states (e.g. Jharkhand, Bihar) known to be more deeply entrenched in a culture of patriarchal hierarchies of gender and generation resulting in substantially lower levels of education, employment and mobility among women.

Table 2 summarizes the post-CFA retained indicators of empowerment and Table 3 presents the factor loadings in indicators of empowerment. Factors were determined on the basis of loadings generated by the CFA model. Typically, factors with higher loadings (those close to 1 and with higher Eigen values) were selected since it is generally agreed that they fit the data better [8]. To evaluate the model fit, chi2 (χ 2) and comparative fit index (CFI) were used as proposed by [33]. The p-value associated with χ 2>0.05 and the CFI value >0.95 suggesting that the model meets the cut-off for good fit.

Source: IHDS (2011–12)

Source: IHDS (2011–12).