Association for Public Policy Analysis & Management

The traditional absence of easy-to-access systematic performance data may have buffered leaders of public organizations from accountability, but also protected them from biases that accompany data use. In this study we use experimental data with actual elected officials to demonstrate such biases, as well as the effectiveness of stakeholder arguments about the quality of data.

The data come from survey experiments of local elected officials in Denmark, who are asked to assess the degree of school leadership control over school performance. The respondents are provided with three treatments.

The control group receives no performance data. A second group receives data showing if their school performance is ranked in the top, middle, or lowest third of peers. We find that the provision of performance data enhances a willingness to believe that school leaders control outcomes, but only among those with low performance. This is consistent with leadership attribution bias – altering ones belief about the degree of control leaders have over outcomes in response to cases of unusual performance. Leadership attribution bias is widely observed in private sector studies, but usually takes a curvilinear pattern. Our findings suggest that in a public setting, it may take effect largely on a negative valence, suggesting a mix of leadership attribution bias and negativity bias.

The credibility of performance data is also shaped by political framings of the data itself. We examine if comments by advocates about performance data can influence how it is used to assess public sector leaders, and whether this influence depends upon sharing the same political ideology. A third group received performance data and comments from teachers unions that cast doubt on the validity of performance data. We find that the comments of these advocates are effective in reducing the use of performance data, but only among elected officials who are members of liberal parties.