Social Desirability, Environmental Attitudes, and General Ecological Behaviour in Children

Abstract

Socially desirable responses have been widely discussed as potentially biasing self-reported measures of environmental attitude and behaviour assessment. The direct and moderating effect of social desirability on children has not been analysed before. By applying a Lie scale together with a two-factor environmental attitude set measure and a scale of self-reported General Ecological Behaviour (GEB) to 198 pupils, we found a moderate impact of Lie scores on only one of both attitude measures and a small impact on GEB. In a multiple regression analysis general behaviour was predicted by attitude, social desirability, and the interaction of both. Social desirability had no moderating effect on the relationship between environmental attitudes and behaviour. Implications of these outcomes for research on environmental issues with children are discussed.

Keywords:

• Social desirability
• Ecological behaviour
• Environmental attitudes

Acknowledgements

The study was supported by the University of Bayreuth. First, we appreciate the cooperation of all the students and teachers who participated in this study. Second, we are very grateful to Florian Kaiser and Michael Wiseman for reading and commenting the manuscript.

Notes

Values were corrected for errors due to sampling and due to differences in reliabilities of the measurement instruments.

As fit statistic, the averaged mean square (MS) statistics of 0.90 corresponds to a 10% lack of variation and a MS of 1.10 to a 10% excess of variation (underfit) in the model prediction compared to what is in the data.

The amount of Preservation and GEB variance that is explained both by Lie score and age, gender, and stratification level is consistent with the difference between the explained variance calculated by means of correlations and the one in Model 2.

To avoid multicollinearity in regression analysis with interaction terms (see Tabachnick & Fidell, 2007), we created centred independent variables.

We did not include background variables to avoid that common variance with age, gender, or stratification level is not attributed to Lie scores.