Job Satisfaction – An International Comparison of Public and Private Sector Employees

ABSTRACT

Benefits of job satisfaction include increased productivity, performance, creativity, innovation, motivation, and involvement. The current study examines a global sample from 37 countries to examine the effects of work-life balance, intrinsic and extrinsic rewards, and work relations on job satisfaction for public and private sector employees, using data from the International Social Survey Program. Descriptive results show a significant difference between these workers in the study’s main variables, with several areas higher for public workers. Additionally, regression analyses identify significant differences, mostly favoring public workers. A key contribution of the study is the finding that a one-size-fits-all model of job satisfaction does not work equally across the globe (context matters). However, the findings do lead to specific, actionable items for managers. Future research should examine more detailed country-specific variations and the corresponding causes and explore private/public sector determinants on a global basis.

KEYWORDS:

• Job satisfaction
• public sector
• private sector
• global comparisons

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. ISSP Researchers collected the data via self-administered questionnaires, personal interviews, and mail-back questionnaires, depending on the country. For a full summary and description of this research, see https://www.gesis.org/issp/modules/issp-modules-by-topic/work-orientations/2015/

2. Countries include, in alphabetical order: Australia, Austria, Belgium, Chile, China, Taiwan, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Georgia, Germany, Hungary, Iceland, India, Israel, Japan, Latvia, Lithuania, Mexico, New Zealand, Norway, Philippines, Poland, Russia, Slovak Republic, Slovenia, South Africa, Spain, Suriname, Sweden, Switzerland, United Kingdom, United States, Venezuela

3. One of the primary limitations of the available attitudinal data is that each question represents a subjective single item indicator. As Souza-Poza and Souza-Poza aptly point out “[Subjective Well Being] scores depend on the type of scale used, the ordering of the items, the time-frame of the questions, the current mood at the time of measurement, and other situational factors” (Sousa-Pouza & Sousa-Pouza, 2000, p. 5; see also Diener et al., 1995; Diener, 1984; Schwarz & Strack, 1991). They further point out that as the ISSP data set only measures job satisfaction as a single-item indicator, variance due to the wording of the item cannot be averaged out and the single item further makes the evaluation of internal consistency problematic.

4. One limitation of the use of ISSP data in this analysis is the possibility of common source bias/common method bias, while theoretically and statistically speaking we don’t anticipate this being large problem (see Favero & Bullock, 2014).

5. Categories for this variable include: (1) Male, (2) Female.

6. Continuous variable.

7. Continuous variable.

8. Response categories for this variable include: (1) married, (2) civil partnership, (3) separated from spouse/civil partner(s), (4) divorced from spouse/ legally separated, (5) widowed/ civil partner died, (6) never married/ never in a civil partner.

9. Continuous variable.

10. Continuous variable.

11. Categories for this variable include: (1) Managers, (2) Professionals, (3) Technicians and Associate Professionals, (4) Clerical Support Workers, (5) Services and Sales Workers, (6) Skilled Agricultural, Forestry and Fishery Workers, (7) Craft and Related Trades Workers, (8) Plant and Machine Operators and Assemblers, (9) Elementary Occupations, and (10) Armed Forces Occupations

12. Categories for supervising others: (1) Yes, (2) No. 13 No.

13. Categories for this variable include: (1) Employee, (2) self-employed without employees, (3) self-employed with employees, and (4) working for own family’s business.

14. Categories for type of organization: (1) Public, (2) Private.

15. Despite these data and methodological problems, the use of data standardization, as an adjustment of raw scores in cross-cultural research to correct for such response tendencies, is used to reduce or eliminate unwanted cross-cultural differences that are not due to variables of interest, but rather response sets and methodological artifacts (see Hofstede, 1980; Van de Vijver & Leung, 1997). Detecting potential response bias requires researchers to identify different response patterns based on particular methods used and eliminate them. Furthermore, researchers need to detect and control for this bias or error variance in cross-cultural research, and assuming that different patterns are some form of bias, researchers need to standardize their data to reduce this error variance (Fischer, 2004). In the ISSP data, the original researchers have already taken appropriate methodological precautions against response bias, and additionally I have adjusted the raw scores through data standardization and reporting beta coefficients, thus “remov[ing] variation that is substantial and related to culture” (Fischer 2004, p. 264).

16. All correlations, cross-tabulations, t-tests, ANOVA, ANCOVA, post-hoc tests, and full descriptive statistics have not been included here due to space limitations, but are available upon request. Additionally, appropriate tests for multicollinearity were conducted. There are no issues with multicollinearity of variables in the OLS model. Additionally, all outliers were Winsorized in the initial data cleaning stages, prior to final models and analysis.

17. Due to the ordinal nature of the dependent variable, it is most appropriate to use an ordered probit regression to look at the effect of different job characteristics on one’s overall job satisfaction. However, many researchers have argued that using OLS regression is appropriate when looking at satisfaction variables on a Likert scale, where most respondents understand that the difference between responses of 1 and 2 is the same as the difference between responses of 2 and 3, and so on. Additionally, using OLS regression results allows us to report an r-squared and adjusted r-squared value for the model and compare coefficients across models, which comparison is not appropriate in a probit model. Therefore, all regression results reported herein are OLS regression result. It is important to note that when the same OLS models where run in an ordered probit regression, the same significant results appeared for each of the independent and control variables across countries and waves (full ordered probit model results, are available upon request).

18. Descriptive results by country are available upon request.