Dear Angel,

Did you check the matrix correlation between the items? With this, you can see if there is a item deviant. After you have find the item, you can delete it and calculate a new Cronbach's alpha.

You can also conduct a factor analysis in order to to identify the dimensions of your questionnaire. 

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4205511/

Dear Angel,

In research, you can take either established scale from exiting literature or develop a set of items for measuring the target construct, If you take established scale with known  cronbach's alpha (>0.7),  model fit would not be an issue. But, when you develop a list items to measure a construct , you need do a factor analysis and develop a scale with cronbach's alpha value >0.7. Thanks 

Checking the correlation matrix is a good idea, especially if there are any "reversed" items that generate negative correlations. Depending of what statistical program you are using to compute alpha, you may have have an option to look at each variable with a rating of "alpha if item deleted" to see any specific variable is reducing the overall alpha.

There is another way... you can look into studies similar to yours. See how much alpha value is used to consider acceptable by them. Sometimes, when working with certain variables the alpha value is hard to reach 0.7 benchmark.

Cronbach's alpha is sensitive to the number of items in the scale and generally tends to underestimate the internal consistency reliability. So go for composite reliability. Composite reliability values of 0.60 to 0. 70 are acceptable in exploratory research, while in more advanced stages of research, values between 0. 70 and 0.90 can be regarded as satisfactory (Nunally & Bernstein, 1994 ).

You could also perform a Principal Component Analysis to see whether the specific item loades onto other items in your questionnaire, in which case you could take those items together to analyse as composed component/item.

Cronbach's alpha is a "lower bound" for reliability (it assumes that all the items load on one "factor" and that all the loadings are equal). Running factor analysis will relax these assumptions.

Rather than principal components (which makes some unrealistic assumptions), I would recommend maximum likelihood factor analysis (which is the basis for Structural Equation Modeling).

Hello, Angel - in addition to all of the very good responses thus far, I would like to add that Cronbach's alpha is also affected by sample size...

Albeit Cronbach‟s Alpha is widely used as an estimator for reliability tests, it has been criticized for its lower bound value which underestimates the true reliability (Peterson, R.A. and Y. Kim, 2013). Composite Reliability can be used as an alternative as its composite reliability value is slightly higher than Cronbach‟s Alpha whereby the difference is relatively inconsequential (Peterson, R.A. and Y. Kim, 2013).

Peterson, R. A., & Kim, Y. (2013). On the Relationship between Coefficient Alpha and Composite Reliability. Journal of Applied Psychology,98(1), 194-8. DOI:10.1037/a0030767

What to do with low cronbach's alpha (0.6)?

If you try all other methods to increase your Cronbach Alpha as mentioned above by other scholars, you can refer to some references to justify lower Cronbach Alpha according to different research disciplines - see RG link below.  Wishing you all the best.

https://www.researchgate.net/post/Any_Citation_for_Cronbach_alpha_more_than_05_is_acceptable

Did you already checked for inverse items (negative connotations opposite to the rest of the items in that part; these will show up with a negative loading after running a factor analysis)? In case of more binair scoring, alpha is not the best indicator (lower abound as already mentioned). One is recommended to use the KR-20 or KR-8 formula. Besides, the number of items involved really matter in how to interpret a low alpha coefficient (see Cortina, JM 1993 in the Journal of Applied Psychology).   

In addition to Alexander Minnaert submission, which is very valid, you may need to conduct item analysis for you to identify redundant items. if found such would have to be removed or restructured for relevance. there is also the need to ensure content validity of the items and be sure that all that should be included is actually captured.

Will  your major study have a larger sample?  sometimes just making the sample larger will increase the reliability value.  The responders above have already suggested the more complex solutions.

thanks everyone. i was able to proceed with my study

Angel Andrada : What was you justification to proceed with low cronbach's alpha?

There are a few ways to deal with it:

1. Since Cronbach is based on the number of items, you can add more statements/items that constitute to your construct (or factor). If there are more items, then there is option of 'Scale if item removed'.

2. If you cannot add item or remove item that causes trouble, you have another way out provided that you have managed to collect more responses that minimum sample size. Check for 'Variance' among the items. You can do that by creating a new variable by going to transform, using Variance under the option 'Statistical' and chossing variance. You can then check in the data view and remove those responses with high value of variances. e.g. some response might have variance of 0.35, 0.67, 1.25 only while some responses might have variance of 5.33, 6.2 or something like that. So, remove them and your Cronbach Alpha will rise up. (Note: you should have more responses than min sample size so as to refine the data).

3. If still you cannot do 1 & 2, follow the book of Hinton et al. 2014, named SPSS explained. Cite this book and also put an expert's positive opinion to back up your action. This should put you on a safe side.

Cronbach’s alpha results should give you a number from 0 to 1, the closer Cronbach’s alpha coefficient is to 1.0 the greater the internal consistency of the items in the scale. George and Mallery (2003) provide the following rules of thumb: “_ > .9 – Excellent, _ > .8 – Good, _ > .7 – Acceptable, _ > .6 – Questionable, _ > .5 – Poor, and _ < .5 – Unacceptable”