Bias in testing; a study of bank selection tests

Abstract

Multiple choice objective type test is a popular method of selection. However, research indicates that these tests are biased towards specific socio-economic groups. Research in India on these issues is scanty and has been inconclusive. In this study, the problem of bias in the reasoning test for officers’ selection to a large public sector bank is examined. Test responses of all the candidates who took the test for the 1986 and 1987 all India selection rounds of the bank, along with their bibliographic/socio-economic information was used for the study. Using the automatic interaction detection algorithm (Sonquist, 1970) the low scoring and the high scoring groups are identified. The explanatory capability of the variables is examined. Family income and medium of instruction of the candidate at school were identified as the variables explaining a major part of the variance in performance. This supports the socio-economic bias hypothesis. Discipline of the candidate at degree level affects the performance significantly in the nonverbal part of the test. Other variables discussed in literature as sources of bias viz. region to which the candidate belongs, the urban-rural location of school/college and the sex of the candidate were not found to be significant. Bilingual test forms i.e. the test booklet in both English and Hindi have been introduced in 1987. The performance analysis of candidates who have benefited out of the introduction of bilingual test forms was done using two reference states with differing vernaculars. The introduction of bilingual test forms has reduced the explanatory powers of the socio-economic predictor variables. It was found that tests in the same language as the medium of instruction at school improves the test performance. However usage of bilingual test forms had been unintended consequences. Managerial and research implications of the findings are discussed.