I have translated a questionnaire for research purposes and in spss, the internal consistency reliability is coming low. The original questionnaire has good reliability. what could be the reason and what can be done to resolve this?
Cronbach Alpha is quite sensitive against sampling errors. So how many data do you have? If the data set is small it can happen that you observe a small Alpha just by bad luck. Alpha is based on the mean correlation of the items in a scale and correlations get stable only with higher ammount of data.
What type of questionnaire is it? Second reason that happens often with UX questionnaires is that the questionnaire shows a good Alpha only for certain types of products. If applied to other types of products the value can be much lower (since items in a UX questionnaire are always interpreted in the context of the actual product and thus for different products the correlations between the items of a scale can differ).
Low internal consistency on a pre-existing reliable scale could be due to many reasons. Double-check you reverse-scored relevant items. I believe in most versions of SPSS, it does not automatically reverse score items when you calculate alpha, unlike a factor analysis, where the item would simply have a negative loading. Then ask SPSS to report alpha if an item is deleted or run the factor analysis to help you identify if it's one or so troublesome items. Since you translated the survey, could it be that those items are understood differently in the different languages and/or cultures? Beyond that, it could just be small numbers or a scale that is not reliable with your participants or context (as Martin points out). Hope this helps some, Anna. ~ Kevin