Association for Public Policy Analysis & Management

Theoretically, two perspectives may explain why employee assessments offer a viable alternative to managers’ self-assessments. First, using employee assessments can be preferable in labor-intensive organizations, which characterizes most public service organizations, because management practices can be expected to have the greatest impact on employee behaviors, when they are indeed perceived by employees (Wright & Nishii 2013). We therefore expect that employee ratings of management provide better predictions of organizational performance compared with managers’ self-ratings. Second, employee assessments allow researchers to draw on multiple raters (i.e., the number of employees in the organization), and this can greatly alleviate challenges from low reliability and measurement error linked to single assessments of management practices. In line with the ‘wisdom of the crowd’ argument (Surowiecki 2004), idiosyncratic noise linked to single assessments is expected to cancel out with an increasing number of raters. Our second expectation is therefore that a higher number of employee ratings yield better predictions of organizational performance.

Empirically, the expectations are tested on a data set from high schools containing identical measures of transformational and transactional leadership as rated by principals and teachers, and objective indicators on educational performance (e.g., school value added to student grades). Including high schools with more than 30 teacher responses, we investigate how manager and employee rated leadership are associated with organizational performance, and how well employee-rated leadership predicts organizational performance, when we rely on an increasing number of employee responses (randomly drawn from the pool of valid responses within organizations). Empirical results are expected to demonstrate that employee-rated leadership is correlated with organizational performance more strongly than manager-rated leadership, and that the strength of the former correlation increases, when more employee responses are included to reduce the random errors associated with single observations.

In sum, the paper potentially contributes to the debate on how best to measure management practices, how much researchers gain by investigating the management-performance nexus using employee ratings of management, and whether investing in obtaining more employee responses within single organizations is characterized by diminishing returns.