Wellbeing Measurement:
Social Desirable Responding

May 11, 2008

Introduction

A common concern about self-report measures in general is the possibility of social desirable responding. That is, a tendency to report overly positive levels of well-being.
Although most theories predict positive biases, it is also possible that ratings are negatively biased. Thus, the empirical question is to examine whether ratings are biased either in a positive or negative direction.

The main problem in examining this question is the lack of a clear validation criterion to establish biases. However, a number of empirical findings can be used to assess the problem of social desirable responding.

The most direct approach is to compare self-ratings of well-being with informant ratings of well-being. The advantage of informant ratings is that both self-ratings and informant ratings are made on the same scale. Thus absolute scores are directly comparable. In contrast, it is impossible to compare absolute levels of self-ratings to absolute levels of other well-being indicators (e.g., income, GDP).

Below I present the evidence on the absolute level of well-being measures in studies that included self-ratings and informant ratings using the same measurement instruments.

Meta-Analysis

Few studies report means for self-ratings and informant ratings. Thus, the empirical evidence is limited. However, the existing evidence shows very similar levels of self-ratings and informant ratings (DATA). The average effect size of the difference in means is d = .02. A value of .00 would indicate identical levels of wellbeing. A value of 1 would indicate a difference of 1 standard deviation (e.g., the actual IQ is 100, but the average self-rated IQ is 115).

Conclusion

There is no evidence to support the claim that self-ratings of well-being are inflated by social desirable responding.

Limitations

The main limitation is that informant ratings themselves could be biased to the same extent, which would mask inflated self-ratings.
Another limitation is that all studies included in the meta-analysis used anonymous ratings of wellbeing. Thus, it is possible that other method of assessment produce different results.

Do you know additional studies that are important and should be included in this meta-analysis?
Do you have additional comments about these studies or my comments?
Please email me (uli.schimmack at utoronto.ca).