One free measures of the Big 5 is the IPIP. I have noticed that IPIP items seem to have a positive bias towards items loading on the extraversion (EX), emotional stability (ES), conscientiousness (C), agreeableness (A), and openness/intellectance (O) pole of the scale.

For example, for agreeableness, you have items "I am interested in people" (positive) and "I insult people" (negative), or for conscientiousness you have items "I pay attention to details" (positive) or "I make a mess of things" (negative).

However, at the level of a construct there are positive aspects to all the other poles of the Big 5. Being disagreeable could mean that you are comfortable confronting people when you believe they have done something wrong or that you are comfortable pointing out social injustice. Being lower on conscientiousness may mean that you don't feel the need to do everything perfectly in life or that you believe that sometimes it's better to be a little messy so that you can focus on more important things.


Thus, I'm interested in the positive bias in measures of the Big 5 towards conscientiousness, agreeableness, extraversion, emotional stability, and openness.

  • Is there any research that quantifies the degree to which different measures of the Big 5 display this positive bias?
  • Are there any Big 5 measures that specifically try to avoid this positive bias?
  • To what extent is this positive bias inherent to the concepts of the Big 5?
  • 2
    $\begingroup$ BTW: Do people with high agreeableness agree with more statements (positive or negative, or at least: positive) than ones with low agreeableness? $\endgroup$ Aug 31, 2012 at 11:50

3 Answers 3


Backstrom et al (2009) have a paper devoted to the question of whether the intercorrelations between the Big 5 scales is due to social desirability bias or substantive factors. In Study 2 they re-wrote each item on a 100 item IPIP scale so that each item had reduced positive bias. They then administered the original and neutrally worded items to a sample of participants. They found that when IPIP items were worded in a more evaluatively neutral fashion, the correlations between the Big 5 were substantially reduced. Specifically, they state that:

The first component accounted for 19.3% of the variance in the IPIP items as compared with the variance of the first component from the analysis of the neutral items of 10.5%. The mean correlation between the first principal component and the items was .40 for the IPIP items and .12 for the neutral items.

Their first paragraph in the discussion captures several interesting conclusions:

Taken together, the two studies presented here support a model that still appears to be controversial. The model can be described as follows: when we use self-ratings for the measurement of personality traits, some items tend to activate social desirability concerns. In the scores, the participants’ tendency to respond in a socially desirable way is mixed with the descriptive aspect of the items, which is supposed to measure the FFM. The result is a common factor related to all of the other factors in the inventory (Edwards and Edwards, 1991 and Peabody and Goldberg, 1989). It has been shown several times that this factor can be reduced, and that when this is done the items’ descriptive quality stands out more clearly. In previous studies the social desirability of the items has been reduced by simply selecting items that correlate weakly with social desirability inventories (e.g. Jackson et al., 1996). In the present study, the focus was on the semantic content of the items. By systematically making the items from the IPIP-100 less attractive, without changing their meaning in any crucial way, we could get rid of almost all common variation between the factors, i.e. variation related to social desirability. These new items appeared to measure the same thing as the original items in the IPIP-100, except social desirability. This shows that it is relatively easy to create personality inventory items that measure FFM traits with less influence of social desirability. We therefore encourage inventory constructors to henceforth strive for making inventories where descriptive scales are less confounded with social desirability.


  • Bäckström, M., Björklund, F., & Larsson, M. R. (2009). Five-factor inventories have a major general factor related to social desirability which can be reduced by framing items neutrally. Journal of Research in Personality, 43(3), 335-344. PDF

In a previous post on the global personality factor, you provided examples to describe how self-report items indicating higher trait levels are phrased in more socially desirable terms, hence the positive bias.

I think it's important to distinguish between two (or three) kinds of biases here: one is related to the wording/content of the items (the positive bias your refer to) and the other stems from trait desirability, i.e. the extent to which individuals find a trait desirable for personal reasons (e.g. their own standing on the trait) and for socio-cultural reasons (e.g. the extent to which a trait is valued by society or his/her culture).

To use your own examples, one could perhaps mitigate the positive bias by including items that make the lower ends of each trait seem less undesirable (.e.g., "I enjoy being in my own company most of the time" for low extraversion) and the higher ends less desirable (e.g. "I like to dominate in social situations" for high extraversion), but you would still have the effects of the trait desirability biases. It's not to say that the task of improving the wording or content of the items to reflect less bias isn't important, and to what extent there is bias is itself arguable.

Has the purported positive bias been documented or described in any articles? Perhaps that should be the main question at this point.

I'm only aware of approaches to dealing with the desirability biases mentioned. One is to obtain reports from multiple informants in addition to self-reports. Another is to use ipsative (i.e. forced-choice) measures of the Big Five (e.g., Hirsh & Peterson, 2008).


Hirsh, J. B., & Peterson, J. B. (2008). Predicting creativity and academic success with a “fake-proof” measure of the Big Five. Journal of Research in Personality, 42(5), 1323-1333.

Ludeke, S. G., Weisberg, Y. J., & Deyoung, C. G. (2013). Idiographically Desirable Responding: Individual Differences in Perceived Trait Desirability Predict Overclaiming. European Journal of Personality.


Here is a work about IPIP 50 in New Zeeland but there are also some references to USA findings: http://www.psychology.org.nz/cms_show_download.php?id=617

  • 1
    $\begingroup$ Thanks. The article seems to provide standard psychometric information: i.e., the factor structure of the test, gender differences, whether test scores predict relevant criteria including job satisfaction and counter productive work behaviour; it doesn't directly seem to go into whether items have a positivity bias. $\endgroup$ May 31, 2013 at 0:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.