http://orcid.org/0000-0002-5387-5455
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
Previous studies on Canadian nurse wages were limited to individual factors and did not take into account contextual factors such as hospital market share, labour market size or unionization. Based on market share, some refer to the nursing labour market as a monopsony, which depresses wages and might explain the shortage. However, this has not yet been tested empirically in the Canadian Registered Nurse (RN) labour market. This article aims to fill this gap by using the microdata files of the Labour Force Survey for the years 2010–2012 and the multilevel analysis to shed light on this issue. The contribution of this work is that it takes into account both individual and contextual variables to try to explain nurses’ hourly wage. In accordance with the monopsony model, we hypothesize a negative correlation between hourly wage and level of market share; i.e. monopsony employers would pay a lower wage rate. The results do not support the monopsony model to explain nursing labour shortage: there is no statistically significant relation between RN wages and market share; no relation was found for market size either. This suggests that an explanation for RN labour shortage must be investigated elsewhere.
Acknowledgements
The authors would like to thank Professors Anyck Dauphin and Samir Amine from UQO; Professor Patrick González from Université Laval; Professor Lori Curtis from University of Waterloo; and Naisu Zu, PhD, for their helpful inputs and comments on this work. They are also grateful to participants at the 50th Canadian Economics Association Meetings for additional feedback. This research was supported by funds to the Canadian Research Data Centre Network (CRDCN) from the Social Science and Humanities Research Council (SSHRC), the Canadian Institute for Health Research (CIHR), the Canadian Foundation for Innovation (CFI) and Statistics Canada. Although the research and analysis are based on data from Statistics Canada, the opinions expressed do not represent the views of Statistics Canada or the CRDCN.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. Matsudaira (Citation2014) also found no evidence of monopsony: facilities in California did not have to raise their wage offers relative to their competitors in order to hire more nurses, particularly nurse aides, which is more in line with the perfect competition model.
2. Note that in a competitive market, the wage equals the marginal cost. Any difference between the two is interpreted as a measure of the degree of monopsony, as in the Lerner index.
3. Some could argue that one determinant of Canadian RN wages is the wages paid to RNs in neighbouring US states. This has not been controlled for. Canadian-educated nurses working in the US account for about 5% of the Canadian nursing supply in 2008 (19,699 Canadian-educated nurses working in the US over a supply of 365,686 regulated nurses (HRSA Citation2010; CIHI (Canadian Institute for Health Information) Citation2011). Close to half move to the US to obtain full-time employment and only 8% for higher salary and benefits (Hall et al. Citation2013). Still, if US hospitals are a fallback option for Canadian nurses living on the US border, negotiated union wages in Canadian provinces might reflect this.
4. Statistics Canada uses the concept of ‘financing’ to classify jobs. Thus, all hospital employees are classified in the public sector.
5. In 2012, about 3.5% of RNs in Canada were nurse supervisors and 7.2% of RNs were males. For the purpose of the multilevel modelling, it would be difficult to have enough of them in each cluster. For these same reasons, we have not included Licenced Practical Nurses or Orderlies either.
6. Personal communication with Labour Statistics Division staff, December 2013.
7. These two wage dimensions are found in the LFS database.
8. We chose this lower and upper bound to account for outliers. The ceiling of $150 was chosen because we take into account the fact that an employee with a bachelor’s degree and who is not part of the management team generally earns less than this amount.
9. The breakdown of the different geographic regions is only in the RMF. The Public Use Microdata File (PUMF) contains only the three largest CMAs (Montreal, Toronto and Vancouver) with the others grouped as a fourth category. Neither the RMF nor the PUMF contains the postal code.
10. For example, according to the structure of the CMA/UA/CA, no specific region exists for North-West Quebec. Municipalities such as Témiscamingue, Val-d’Or, Rouyn-Noranda would be paired to either Gatineau or Saguenay, which is not too straightforward. However, the EI structure includes a region for North-West Quebec.
11. Look up EI Economic Region by Postal Code. Accessed in October 2013: http://srv129.services.gc.ca/regions_ae/fra/codepostal_recherche.aspx.
12. These authors identified eight market sizes. We have modified their size structure based on the Canadian context. On top of the HI, they also used market size to find monopsony effects.
13. A given EIR may include a dozen or even several hundred census subdivisions (municipalities). For example, the EIR of Gatineau includes 10 municipalities, while Central Quebec includes over 500 municipalities.
14. Statistics Canada now uses the term ‘population centre’ to replace the term ‘urban area’. Population centres are classified into three groups, depending on the size of their population: (1) small population centres, with a population between 1000 and 29,999; (2) medium population centres, with a population between 30,000 and 99,999; and (3) large urban population centres, with a population of 100,000 or more (Statistics Canada Citation2011) .
15. One could argue that the notion of market size that matters for finding monopsony effects is related to employment opportunities in the local labour market. However, working age people tend to concentrate where the jobs are located. Therefore, the number of working age people in a region is a good proxy for employment opportunities.
16. Where market share data are not available, some use market size as a proxy.
17. For example, employees of a given region can be similar and very qualified because of the selection process and the availability of resources.
18. Most multilevel regression software assumes that the residual variance is the same for all regions.
19. Procedures such as generalized least squares (GLS) produce unbiased estimates standard errors as they can correct the loss of degrees of freedom due to the intra-class correlation (Kreft Citation1996). However, they do not permit variance breakdown.
20. Wage can be a nonlinear function of the number of years of experience; hence, we have also included the square of years of experience.
21. In principle, all variables that vary between regions and labour markets are excluded in the first step. However, we cannot add too many variables in the second step. It is recommended to have a minimum of 10 groups for each second-level variable (Bressoux Citation2008, 325). Some suggest getting a minimum of 20 groups (U. of Bristol, Center for Multilevel Modelling: http://www.bristol.ac.uk/cmm). So there is a trade-off. With less than 60 groups, we cannot reasonably accommodate more than two second-level variables in the model. Given this constraint and the fact that wage harmonization by the unions in Canada limits regional differences in wages, we have chosen to include union as a first-level variable in the model.
22. Random sampling with 50 replications based on 55 clusters was performed for M2, M3 and M4.
23. The REML (restricted multilevel) models produce about the same results as the FML. However, the deviance is lower in the FML. Results from the REML are available upon request.
24. It should be noted that in the random part, there is no difference between the asymptotic estimate for the second-level variable and the final bias-corrected (bootstrap) estimate (0.0108). It is as an indication that there is no bias in the second-level variance estimate.
25. dlnS/dExp = 0.01509–2(0.0003)Exp. First-order condition: max = 0 = 0.01509–2(0.0003)Exp. So, 0.0164 = 2(0.0003)Exp and Exp = 0.01509/0.0006 = 25.2 years. Roughly similar results are obtained when age and square of age are used.
26. Details of the empirical analysis are available upon request.
27. Even though tenure was not included to be a proxy for mobility, one can argue that longer tenure could be associated with lower worker mobility, thus with higher monopsony power.
28. It should be noted that the coefficient for Alberta was statistically significant in M2 (level 1 and 2 predictors, without random coefficient).
29. Basic household costs in Manitoba were the second or third lowest, depending on the family structure.
30. Moreover, comparing RNs with a control group (such as women teaching in elementary school) is challenging with the MLM because other professional categories cannot be assigned to a hospital and have an HI. Therefore, it is difficult to perform simultaneously the multilevel estimation and comparison with a control group.