![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
Conventional wisdom of ‘sector matters’ suggests that those working in the government are more risk averse than those employed by business enterprises. However, whether public sector workers tend to be more risk averse than non-profit sector workers is unknown. Our paper examines whether the levels of organizational risk aversion as perceived by managers differ between public and non-profit organizations and explore reasons leading to this potential difference. Statistical results show that organizational risk aversion is more pervasive in the public sector than in the non-profit sector. Mediation tests further indicate that managerial trust and an organization's formalized rule constraints in rewarding good performers and removing poor performers are decisive to this difference. The findings imply that the top management's feeling of insecurity and structural reasons embedded in the merit system are the keys to organizational risk aversion in the public sector.
Notes
Non-US readers not familiar with this aspect of US tax code or this definition of non-profit organizations are referred to: IRS, Publication 557 ‘Tax-Exempt Status For Your Organization’, pp. 65–6 (Rev. June 2008), Cat. No. 46573C.
We have no population data from which to compare our sample data. However we did perform missing variable analysis, important itself as well as suggestive for response bias. Including all for the variables used in the present study, we found an average of seventeen missing responses (0.014 of for all responses). More directly relevant to response bias potential, we have analysed correlations of missing data with a variety of demographic variables including gender, job type, age, race and state. None of these yielded correlations significant at the 0.10 level. While these results are encouraging they are not sufficient to rule out bias.
Indeed, using two single items to measure organizational risk aversion can be deemed a limitation as single items may fail to truly reflect the scenario of organizational risk aversion. Many business management scholars prefer measuring risk-related concepts such as risk perception and risk propensity by using multiple items (Simon et al., Citation1999; Sitkin and Weingart, Citation1995). However, the use of single items is not uncommon and unprecedented (Keil et al., Citation2000). Given the research of organizational risk aversion is burgeoning in the field of public management, and the only related study by Bozeman and Kingsley (Citation1998) uses items identical to what we use in the present study, we believe that measuring organizational risk aversion with the current two single items may not be impeccable but fairly acceptable.
We recognize that a Cronbach α of this magnitude is close to borderline in terms of conventional views about thresholds. We also note, however, that α measurement is unstable when applied to only two variables and, thus, is a less useful test coefficient than is the case with five or more variables (Miller, Citation1995).
The NASP-III data collectors focused on the population of individuals in ‘management’ positions, so they selected job titles such as ‘director’, ‘coordinator’ and ‘manager’ from the name lists. However, not all of those having a title of manager or director are responsible for management. Some of them mainly perform either technical or professional tasks. As a result, there is a question in the NASP-III dataset asking respondents to choose among ‘managerial’, ‘professional’, ‘technical’ and ‘others’ in terms of their main responsibility of the current job. Regarding whether the current position was obtained through promotion, another question in the NASP-III dataset asks respondents: ‘Is the current job a promotion to a higher position from the same organization, an upwards move from a different organization, a lateral move within the same organization, or a lateral move from a different organization?’ We created a dummy variable of ‘promotion’ by summing up the first two choices. In this way, we can compare perceptional differences between those whose current job was obtained through promotion and those whose current job was obtained through a lateral move.
Pearson's r is used for two continuous variables; Spearman's ρ is used for two ordinal variables; Spearman's ρ is used for an ordinal variable and a continuous variable; point-biserial correlation is used for a continuous and a dichotomous variable; Rank-biserial correlation is used for an ordinal variable and a dichotomous variable; φ (fourfold point correlation) is used for two dichotomous variables.
The standard Sobel mediation test formula is z-value = z-value = a × b/SQRT(b 2 × s a 2 + a 2 × s b 2) and Goodman mediation test formula is z-value = a × b/SQRT(b 2 × s a 2 + a 2 × s b 2 − s a 2 × s b 2), in which ‘a’ and ‘b’ denote coefficients and ‘s’ denotes standard error.
The authors employ clustered standard errors for all regression models in the current study because intraclass correlation may exist. Intraclass correlation denotes that observations in the same cluster are likely to show similar attributes and behaviours. In the current study, public and non-profit organizations belong to two different clusters. We thank one of the anonymous reviewers to suggest the use of clustered (robust) standard errors.
For readers interested in the interpretation of ordinal logit regression coefficients, y-standardized coefficients can be interpreted as, for example, ‘employee risk aversion in public organizations is 0.55 standard deviations higher than that in non-profit organizations, holding other variables constant’. We also report marginal effects in Appendix 2 for readers interested in more detailed information.
