Positive-Findings Bias in Quality of Work Life Research: Public-Private Comparisons

Abstract

Consistent but modest support exists for the positive-findings bias in business quality of work life applications, but the data provide no support for the hypothesis in the public sector. Two features of the common wisdom preoccupy this analysis. First, observers have speculated that the generally positive results of planned change-as in the literatures on organization development and quality of work life (QWL)-are artifactual. In one critical version, good results merely reflect bad methodology and research design (Terpstra, 1981, 1982). Such a positive-findings bias means that the attractiveness of outcomes decreases as the rigor of evaluative studies is enhanced. This analysis will refer to the positive-findings bias as PFB or PFB hypothesis. Second, observers have tended to emphasize the differences between public and private organizations. Thus, some critics trace essential differences to the absence of market forces in public agencies (Hayek, 1944). More subtle versions of this perspective focus on differences in publicness (Bozeman, 1987). Using a large panel of QWL evaluative studies, this analysis tests these two prominent features of the common wisdom. Distinguishing between public and business subbatches of the QWL panel, it assesses each evaluative study in terms of three measures of methodological rigor and two measures of outcomes. These basic data will test whether a PFB applies to public or business subbatches of evaluative studies of QWL applications. This analysis does not begin from square one, as it were. In fact, it rests on two significant demonstrations. First, as Table 1 indicates, the outcomes of the QWL panel are known and attractive-both overall and separately for public and business sectors (Golembiewski and Sun, in press). No great differences exist between sectors, despite the clear advantage that the business sector seems to enjoy in category I outcomes, and the overall pattern is consistent with the results of over two dozen survey studies relying on populations that average between fifty and sixty cases