Reverse-Coded Items Do Not Work in Spanish: Data From Four Samples Using Established Measures

Abstract

The potential for suboptimal psychometric performance of reverse-coded items may be particularly pronounced when scales are translated and administered in Spanish with these problems exacerbated in youth respondents. This is a significant concern, given the rapid rise in Hispanic-American and Spanish-speaking individuals in the US and their rightful, growing representation in psychological research and clinical care. The aim of this study was to examine the psychometric performance of reverse-coded items across four Spanish-speaking samples spanning developmental stages including youth, college students, and parents (N = 1,084; Adolescents n = 107; M = 19.79; SD = 2.09; 41.1% female; Caregivers n = 58; M = 40.79; SD = 7.94; 60.3% female; Spanish-speaking adults in the US n = 157; M = 33.4; SD = 9.5; 68.8% female; and College students living in Latin America n = 783; M = 21.04; SD = 3.13; 69.2% female) and four scales (Big Five Inventory; Strengths and Difficulties Questionnaire; Difficulties in Emotion Regulation Scale; Beck Hopelessness Scale); we expected reverse-coded items would demonstrate inadequate item-total correlations and their inclusion would compromise scale internal consistency. Hypotheses were supported with evidence of poor psychometric performance for at least two reverse-coded items on each instrument, such that un-reversing the items improved their item-total correlations. Across every instrument, alpha was either improved by excluding reverse-coded items or by including them in an un-reversed fashion and, overall, there was a moderate, negative effect of reverse-coded items on scale alphas. In growing consensus with previous authors, we recommend that reverse-coded items not be included in Spanish scales.

Keywords: Hispanic; Latino/a/x; reliability; reverse coding; reverse scored; translation.