From what I have read in the literature, the value of reverse coding items seems to be related to reducing or eliminating acquiescence and boredom. Is it necessary to reverse code? Why, or why not?
Ruth, there are of course pro and cons. But despite the potential benefits it is well-known that several problems result from reversed coding, and that is why I wouldn't use reverse coded items.
1. With reverse coded items you will never know whether a test person understood the question correctly or whether the person missed the reversing of the scale and just used the scale as before.
2. It is easy for respondents to misinterpret phrases that include negation, e.g., being not unhappy does not mean that one is happy.
3. Reverse-coded items tend to load on a separate factor, a method factor.
Here you will find an overview and some references under the heading "Don't Use Reverse-coded Items in Scales": http://scaleresearch.siu.edu/petpeeve5.html
Quite recently, an article was published in Psychological Methods which may be of interest for you:
Weijters, B., Baumgartner, H. & Schillewaet, N. (2013). Reversed Item Bias: An Integrative Model. Psychological Methods, 18(3), 320-334.
Gosh, Karin - what great information you have provided me! That you so very much!
The research site that you so kindly provided indicated that "the tide has turned" on reverse coding, and it is not advised. I really appreciate this input!
Karin's feedback is spot on! This article provides an excellent example of the problems that come from using reverse coded items. The reverse coded items were so different that they actually measured a different item.
We consistently find problems with reverse scoring when we carry out Rasch analyses to examine the internal construct validity of outcome measures. Respondents also feed back that they appear to measure something different and that it is hard to switch between positively and negatively worded or scored questions. Regards, Paula
I agree with the points raised above. However, it does rather leave one with the difficulty of dealing with response sets (i.e., mindlessly ticking "yes" to everything) which reverse scoring was intended to address. A partial solution to this can be found in the physical layout of the scale, not lining up response categories across items. However, it would be interesting to hear of other solutions.
In classical authoritarianism studies, reverse items were introduced because it was found that the tendency of some respondents to say "yes" and the tendency of other respondents to say "no" was meddling with the measurement of the actual attitudes (see Couch & Keniston, 1960).
I personally write laboriously as many negative items as positive ones, in every scale.
Nevertheless, it is not unusual researchers to write surprisingly poor negative items, such as "I don't feel vertigo often", which do have a confounding variable in common: they are badly worded and unnatural. This is bad practice, and one has to take the pains of writing items that could be in fact uttered by respondents that hold negative attitudes towards the object, and not simply restating the existing items in negative.
Regarding the loading of negative items to a separate factor, I could argue that the method captures their unnaturalness as a unique component. Moreover one can empirically determine whether the second factor reflects a true structural property of the construct under investigation or a wording side-effect. This is accomplished by comparing the correlations of the two factor with other variables. If all the correlations are in the same direction, then it is not plausible to think that the second factor reflects some real structural property, rather than wording options. Carmines and Zeller (1987) demonstrate an example on Rosenberg's self-esteem scale.
I also think that the weird factor structures that the critics refer to may stem from use of PCA. PCA in itself does not reflect the true structure of a construct, rather than an efficient summary of the correlations between items. So if the funny structures resulted from PCA, this would not be a good argument against negative items, especially badly worded ones, as the second component would only summarize the higher correlation between all the badly worded items rather than between all the negatively stated items.
Couch, A., & Keniston, K. (1960). Yeasayers and naysayers: agreeing response set as a personality variable. Journal of Abnormal and Social Psychology, 60, 151–174.
Carmines, E. G., & Zeller, R. A. (1987). Reliability and Validity Assessment. SAGE.
Excellent feedbacks. I would like to add that if you want to use some reverse-coded items anyway, you can specify a "method factor" for them in a CFA to account for the reverse-coding influence.
I also think reverse coding is a disaster in cross-cultural research esp. with participants who have not done a lot of studies or if the research is conducted in a second language--at least this has been my personal experience in East Asia.
Colleagues, these discussions have been very instructive. While the majority opinion seems to be against reverse-coding, does anyone know if there are disciplines in which the practice might actually be a benefit? Thanks in advance...
Mary I. Bresnahan You have mentioned, "My experience is that reverse coded items often have to be dropped from the analysis even after they are reflected."
Could you please elaborate more on this? For what reasons did you drop reverse-coded items? Is too many people wrongly marking reverse-coded items one of the reasons? What are other potential reasons for dropping them from the analysis? How to justify their drop in the research paper? Awaiting eagerly for your answer.