Sector Bias in Public Programs: US Nonprofit Hospitals

This study investigates whether the public perceives nonprofit organizations as different from private for-profit and public organizations and whether introducing new performance management systems would provide positive credits to the organization. Using two randomized survey experiments on US hospitals (one with an adult sample and the other with a student sample), we replicate the study of Hvidman &amp; Andersen (2016) in Denmark with an extension of adding a nonprofit organization cue. The results show no sectoral differences among the hospitals and no positive feedback for adopting a new performance management system.

The public's ability to evaluate government programs is an essential part of the feedback process that links individual citizens to elected officials in the democratic process. Of major concern, however, is that the perceptual biases of individuals that are well documented in the behavioral public administration literature preclude objective evaluation (Belle et al., 2018;James et al., 2016;Marvel, 2015;Olsen, 2017). Among the findings in the literature are that there is a perceived bias against public organizations, that is, public organizations are evaluated more negatively than private organizations even when the objective outcomes are identical (Marvel, 2015;Hvidman & Andersen, 2016). Such findings are not without challenge. Meier et al. (2019), replicating a Danish study (Hvidman & Andersen 2016) of hospitals that found bias, showed no sector bias in two replications in the US. Experimental subjects in the replication study rated US public hospitals no different from private hospitals.
Differences in findings of the two studies can result from differences in context, differences in subjects, or differences in the structure of the experiments (Walker, James and Brewer 2017). Given the similar design of both experiments and given the US study used two different subject pools (one MTurk adults and one students; the Danish study used students), differences in context are the likely explanation. Among the differences in delivery of public programs via hospitals in Denmark and the US is that the US hospital sector is dominated by nonprofit hospitals (62%); public hospitals total 20% and for-profit hospitals 18%. In Denmark, 97% of hospital beds are publicly owned. To keep the replication of Hvidman and Andersen as parallel as possible, Meier et al. (2019) only reported on the public versus private differences in the US setting.
Studies examining sector bias in public programs need to include nonprofit organizations for both empirical and theoretical reasons. In the United States as well as in many developing countries, public programs are often implemented by nonprofit Electronic copy available at: https://ssrn.com/abstract=3556967 organizations rather than by government agencies or private sector organizations. In addition to hospital care, nonprofits play a major role in in implementing programs in education (charter schools and private universities), substance abuse treatment, mental health services, housing, family planning and other social programs. Omitting such organizations could lead to misleading results. Theoretically, nonprofit organizations rely heavily on normative incentives and altruistic motivation (Donahue and Zechhauser 2012), and this orientation might create positive halo effects that favor nonprofit programs over those delivered by government agencies. This might be especially the case in assessing performance criteria such as benevolence.
This study contributes to the literature in terms of empirical generalization and extension by replicating Hvidman and Andersen (2016); in other words, we report a nonprofit organization cue that was included in the Meier et al. (2019) study. Our results, which are based on all three sectors (public, private, and nonprofit), reveal no differences in evaluations based on the sector of the hospital. Given that the results also show no differences across the three sectors, the difference in results between the US and Denmark studies cannot be attributed to the sector differences in implementing health policy and is likely to come from other contextual factors that differentiate the two countries.

The Experiment
Hvidman and Andersen presented a basic 2×2 factorial design using university undergraduates. The subjects were exposed to vignettes about a hospital that was designated as either public or private; a second experimental manipulation randomly provided half the respondents with a performance evaluation cue, that is, the hospital contracted with an outside consultant to evaluate the hospital, and this evaluation produced positive results.
Our US replication was a 3×2 design with respondents randomly assigned to the hospital being public, private, or nonprofit and then randomly assigned to the same performance evaluation cue. 1 Two replications were conducted. 2 The first replication used 554 adult subjects selected from Amazon MTurk; the second replication used 638 undergraduate students from a large public university. 3 The full documentation of the experiment can be found in the Appendix.

Findings
Tables 1 to 4 include the results from the two randomized survey experiments using both adult and student samples; each table presents the results of adult and student sample above and below, respectively. Note. PM indicates the performance management cue group; ideology is on a five-point scale from 1 (very liberal) to 5 (very conservative); religious service attendance is on a four-point scale from 1 (never) to 4 (weekly); for gender and White, 1 indicates female and white, respectively (otherwise, zero); age is a continuous variable.   Note. * p<0.05; robust standard errors in parentheses; two-tailed tests of significance; nonprofit organizations are the reference group across models.
Both experiments followed the lead of Hvidman and Andersen (2016) using an experimental manipulation of the hospitals' use of performance information; the vignette stated that the hospital hired an outside consultant to assess the hospital's performance and that assessment reported positive results (for more details, see Appendix). The hypothesis was that public hospitals would not get credit for good performance, but private hospitals would. The performance manipulation had no impact on the evaluation of the Danish hospitals whether public or private, and the US study showed similar insignificant results with the exception of a positive change in terms of benevolence in the adult sample (but not the student sample) before the experimental condition was assessed by sector. Note. * p<0.05; robust standard errors in parentheses; two-tailed tests of significance; nonprofit organizations are the reference group across models. Table 3 reports the impact of the performance appraisal cue controlling for the sector of the hospital, and Table 4 interacts the performance appraisal cue by sector. With the exception of the benevolence impact (direct impact only) for the adult population (Table   3), none of the results are statistically different from zero; the direct impact of performance management on benevolence even disappears when introducing interaction terms (Table 4).
The basic conclusion is that US hospitals get no credit for engaging in an external performance appraisal that shows positive results, and this lack of results is consistent across the public, private and nonprofit sectors. Overall the experiments found no sector biases in the US hospital sector, either as a direct effect or in interaction with a performance appraisal.
Nonprofit hospitals were not different from either public or private hospitals.

Conclusion
Our replication of a Danish sector bias study involving hospitals failed to show any sector biases; public, private, and nonprofit hospitals with similar performance were rated no different from each other in two samples of subjects. Given that generally the perception is that the US has a pro-private sector orientation, additional research is clearly merited. One possibility for the absence of perceived differences, in fact, might be that at least one observational study of US hospitals found that on average they did not differ in terms of a wide range of performance indicators that included both the criteria for good medical treatment, hospital readmission rates, mortality rates, and public evaluations of the hospitals (Cheon et al., 2019). The similar performance ratings across US hospitals might have diluted the sectoral biases of the public since the public organizations are not necessarily performing poorly compared to other types of organizations. In other words, the similar performance ratings of hospitals across sector in the US may contribute to our null findings. Other possibilities that might reveal biases should be considered such as the lack of negative information (the only cue was the positive performance management cue), the credibility of the performance evaluation information (the current cues were not especially precise), or the inclusion of other evaluation criteria that are more salient to the general public. We encourage future scholars to examine above mentioned possibilities replicating and extending the sector bias experiment.