Hello Viola,

Very likely the first order of business would be to verify that all items/variables you believe to be associated with a specific dimension do, in fact, demonstrate adequate association with that construct. Confirmatory factor analysis would be one place to start.

If CFA shows that your allocation of indicators to factors/dimensions was on target, then here are some possible explanations for low score consistency (however quantified, whether via Cronbach's alpha or some other statistic):

1. Not all indicators contribute substantively to identification of the latent variable;

2. Possible instances of mis-scaled or reverse-polarity indicators;

3. A too-homogeneous sample (insufficient variation in scores...measures of internal consistency reliability require score spread in order to yield "high" values);

4. Extraneous noise in the data collection process;

5. A speeded administration (insufficient time);

6. The dimension isn't uni-dimensional (despite the fact that CFA would appear to suggest that the simpler, one-dimension model adequately accounts for the observed relationships);

Do note that, for purposes of making decisions about groups, you can tolerate lower estimates of score reliability than would be the case when your intent was to make decisions about individuals.

Good luck with your work.

David Morse Thanks a lot for your suggestion!! I will run a CFA to start.

Wish you all the best :)