An Examination of the Validity of the Traditional Egalitarian Sex Role (TESR) Scale: Confirmatory Factor Analyses

Background: Attitudes toward sex roles are related to various psychological and behavioral consequences, including self-esteem (Sapp et al., 1996), depression (Krause, 1983), and domestic violence (Crossman et al., 1990). The 20-item Traditional Egalitarian Sex Role (TESR) scale was developed and validated in 1988 (Larsen & Long, 1988), and subsequent studies indirectly evaluated its psychometric properties by having it measure sex role attitudes. Literature search, however, failed to find one study where its psychometric evaluation was the primary purpose. This study aims to validate the unidimensionality of the TESR scale. Three different models regarding item composition are evaluated with considerations of 4 items that behaved differently from the rest of the 16 items in its development study. Model1 contains all 20 items, Model2 includes18 items, and Model3 only 16 items. Furthermore, the fact that no racial/ethnic information of the participants was disclosed in either its development or subsequent studies necessitates its validation among diverse racial/ethnic groups. This study validates the unidimensionality of the TESR scale among Korean married women in the U.S.

Methods: 187 Korean married women across the nation participated in the web-based survey in 2007. The survey invitation was posted on two online communities for Korean married women in North America and 42 Korean students associations. The mean age was 34.6 (SD=7.2). 45.7% (n=85) were either a U.S. citizen or a permanent resident, and 54.3% (n=101) were staying in the U.S. with a visa. Confirmatory factor analysis with maximum likelihood estimation was performed using EQS 6.1. Due to the small sample size and the multivariate non-normality, model assessment was based on Yuan-Bentler residual-based statistic (YB), its related F statistic (YB-F), RMSEA (<~.06), CFI (~>.90), and NNFI (~>.90).

Results: Results showed that Model 1 to 3 did not fit well to the data. Modifications followed with a few pairs of error covariances based on not empirical but substantive reasons. Modified models had a good fit to the data and were significantly different from the original models (p<.001). Model1: SB-X2 (165, N=179)=230.44, p<.001, CFI=.92, NNFI=.91, RMSEA=.05; Model2: YB-F (131, N=180)=1.48, p=.06, YB=143.50, p=.22, CFI=.92, NNFI=.91, RMSEA=.05; Model3: YB-F (101, N=181)=1.20, p=.20, YB=108.71, p=.28, CFI=.92, NNFI=.91, RMSEA=.05. Consequently, the unidimensionality of the TESR scale was validated among Korean married women.

Implications: Three different models seemed to fit the data equally well. There are possibilities, however, that a shorter form may perform better while improving the validity of the scale, it being considered that 4 items did not behave as solid as the other 16 items originally. In such a case, a shorter form would have double advantages of (1) simpler administration and (2) improved validity. Future studies can elucidate this matter. Given that since its introduction only a few studies have used the TESR scale and not a single study has solely aimed to investigate its psychometric values, this study provides an important basis on which psychometric evaluation studies of the TESR scale may increase so that it can be used more appropriately with the assurance of its reliability and validity.