Monopsony Power and Relative Wages in the Labor Market for Nurses

Appeared in Journal of Health Economics, Vol. 14, No. 4, October 1995, pp. 443-76.






Monopsony Power and Relative Wages in the Labor Market for Nurses



Barry Hirsch
Florida State University
Tallahassee, Florida, USA

and

Edward J. Schumacher
East Carolina University
Greenville, North Carolina, USA





Abstract

This paper examines the thesis that monopsony power is an important determinant of wages in nursing
labor markets. Using data from the 1985-93 Current Population Surveys, measures of relative nurse/non-
nurse wage rates for 252 labor markets are constructed. Contrary to predictions from the monopsony
model, no positive relationship exists between relative nursing wages and hospital density or market size.
Nor is support found for the presence of monopsony power based on evidence on union wage premiums,
slopes of experience profiles, or the mix of RN to total hospital employment.*

Key words: Monopsony, Nurses, Wages, Labor Markets

JEL classifications: J42, J44, I11





*Correspondence to Barry Hirsch, Department of Economics, Florida State University, Tallahassee,
Florida 32306-2045. Fax: 904-644-4535.


* The authors appreciate helpful suggestions from Carson Bays, Thomas Buchmueller, Marie Cowart, Mike
DuMond, David Macpherson, Lee Mobley, William J. Moore, Tim Sass, and two anonymous referees. The CPS
ORG files, from which the primary data in the paper are drawn, were developed by Barry Hirsch and David
Macpherson.

---

I. Introduction
The labor market for registered nurses has received considerable attention from researchers
owing, at least in part, to the reported shortages that have appeared periodically in nursing markets.1 One
explanation for nursing shortages, popular in the nursing literature and economics textbooks, is that
hospitals face an upward sloping labor supply curve and thus possess monopsony (or oligopsony) power
[one of the earliest statements is Yett (1970)]. The upward sloping supply curve results in a lower wage
and employment level for nurses than would occur if the market were competitive. Monopsony would
help explain reported shortages, since hospitals will list vacancies and desire to hire additional workers at
the monopsonistic wage, but would decrease their profitability were they to raise wages to attract more
applications.
Although the monopsony model has considerable theoretical appeal, empirical evidence for
monopsony power in nursing labor markets is mixed. Previous studies have focused either on the
potential for monopsony power, based on estimates of labor supply curve elasticities facing hospitals
[e.g., Sloan and Richupan (1975), Link and Settle (1979, 1981), Sullivan (1989), and Hansen (1991)], or
examined the relationship between hospital wages, employment, and market structure [e.g., Hurd (1973),
Link and Landon (1975), Sloan and Elnicki (1978), Feldman and Scheffler (1982), and Bruggink et al.
(1985)].
In perhaps the most careful and detailed study of monopsony to date, Sullivan (1989) utilizes
data for 1979-85 from the American Hospital Association’s Annual Surveys of Hospitals in order to
estimate the inverse elasticities of labor supply facing hospitals. Following research by Bresnahan
(1981) and Baker and Bresnahan (1988), he takes the equilibrium market structure as given and then
estimates the supply elasticities under three alternative assumptions about the nature of the market
equilibrium (employment setting, wage setting, and consistent conjectural variations). These estimates
translate into a one year labor supply elasticity of about 1.25, and a three year elasticity of about 4.0.
Contrary to expectations from the monopsony model, Sullivan’s results do not differ substantially

1 See, for example, Aiken (1987), Buerhaus (1987), McKibbin (1990), and Hassanein (1991). More recently,

1

---

between metropolitan and non-metropolitan hospitals. Nor does he compare these results to those
performed on any alternative “non-monopsonistic” occupation.2 We argue below that an important
limitation of this approach is that the presence of an upward sloping labor supply curve is necessary but
not sufficient evidence of a monopsonistic outcome. Rather, wage and employment outcomes predicted
by the monopsony model must be directly tested.
Previous studies examining the relationship between market structure and nursing wages
generally find the positive relationship between wages and the degree of competition predicted by
monopsony theory. Because larger (more competitive) markets tend to have both higher skill workers
and higher wages in nursing and alternative non-nursing occupations (due in part to cost-of-living
differences), these studies do not provide a convincing test of monopsony power. An exception is a
careful study by Adamache and Sloan (1982), who examine the real wages of RNs, LPNs, kitchen
workers and secretaries employed in hospitals. In contrast to the above studies, they find no effect of
concentration on entry level compensation for union or nonunion workers, after controlling for cost-of-
living and population density.
Adamache and Sloan (1982) also analyze the relationship between unions, tenure, and
monopsony. They argue that a negative correlation between tenure and turnover suggests that
monopsony power should depress wages relatively most for workers at the top of the wage scale; that is,
flatten the wage-experience profile. In regressions with the variation in bottom-to-top pay for kitchen
workers, RNs, and LPNs as dependent variables, they find no significant effect of concentration on wage
dispersion among RNs and LPNs, contrary to the prediction of the monopsony model of less dispersion
in more concentrated markets. They also argue that unions should have countervailing power that offsets
the effects of monopsony. They again find no evidence for this proposition.3

reports of nursing shortages have declined (Brider, 1993).
2 Hansen (1991) provides an extension of the Sullivan study. She sets up a general nonlinear supply- demand model
that includes both competition and monopsony as subcases. In her model, an exogenous shock in an input market
that pivots rather than shifts the supply curve allows the competitive case to distinguished from the monopsonistic
case. Empirically, Hansen can not reject the null hypothesis of competition in the market for nurses. When she runs
the test separately for rural and urban nurses, she again finds no evidence for monopsony.
3 Robinson (1988) addresses the monopsony question by looking at differences across markets in employment and
occupational mix. Under the assumption that non-nurse labor markets are competitive, the marginal factor cost of

2

---

Our study extends the work of Adamache and Sloan (1982) and others by analyzing how wage
and employment outcomes differ across markets more and less likely to be monopsonistic. We test the
prediction that relative nursing wage rates for registered nurses (and perhaps other nursing personnel)
will be lowest in relatively small labor markets with a limited number of employers. An important
contribution of the study is the use of a large dataset on individual workers constructed from the monthly
Current Population Survey Outgoing Rotation Group (CPS ORG) files for the period October 1985
through December 1993. These data allow us to examine the relative wage rates of hospital and non-
hospital registered nurses (RNs), licensed practical nurses (LPNs), and nursing aides, as compared to
alternative control groups of non-health related workers, across 202 metropolitan areas and 50 non-urban
state groups. The 1985-93 period is particularly appropriate for study, since most of these years have
been described as characterized by widespread and sustained nursing shortages [McKibbin (1990)].
Contrary to the predictions of the monopsony model, we find no evidence that the relative wages of
nursing personnel are positively related to either labor market size or hospital density.
The scope of the paper is as follows. In Section II, we examine the theory and testable
implications of monopsony models of nursing labor markets. The data set is described in Section III,
followed in Section IV by a presentation of descriptive evidence on the relative wage rates for nursing
personnel during the 1985-93 period. Our estimation approach is outlined and empirical results are
presented in Sections V and VI. A brief conclusion follows.
II. Monopsony in Nursing Labor Markets: Theory and Implications
Monopsony here refers to a labor market where there is a limited number of employers (e.g.,
hospitals). Each firm faces an upward sloping labor supply curve of nursing personnel when making its
hiring and salary decisions, with the marginal labor cost (MLC) exceeding labor's opportunity cost at
each level of employment. Figure 1 illustrates the standard single buyer monopsony model. Profit

nurses is more expensive relative to non-nurses in monopsonistic than in competitive labor markets. Both
employment and the proportion of all hospital jobs filled by nurses, therefore, should be higher in competitive than in
highly concentrated nursing labor markets. Consistent with the monopsony model, he finds that in a cross section of
hospitals, total employment (controlling for measures of output) and the ratio of RNs to LPNs initially increases as
market concentration decreases. The relationship is nonlinear, however, with a reversal among markets with

3

---

maximization by the hospital would lead to employment at Em, where the hospital's marginal revenue
product (MRP) equals MLC, and a wage Wm that will just attract employment Em. Both employment and
wages would be lower than would exist in a competitive labor market, where employment and wages
would tend toward Ec and Wc. At the profit maximizing wage for the monopsonist, there exists a nursing
“shortage” in the sense that the hospital would prefer to hire more nursing personnel at wage Wm, but is
unable to do so. An increase in employment beyond Em would require the hospital to raise wages, but
this action would in turn lower its profits.4
In the more likely case where there exists several employers, wage-employment outcomes are
theoretically indeterminate. If hospitals and other large employers act in collusion (e.g., implicitly or
explicitly through information sharing arrangements) to “cooperate” and limit wages and employment to
the level that maximizes joint profits, then something similar to the monopsonistic outcome can be
achieved. It is widely recognized that this is not a stable equilibrium. Wage-employment outcomes
similar to competitive outcomes can obtain, even when the number of employers is small and labor
mobility is weak, if employers behave in a noncooperative and rivalrous fashion. That is, it will be in the
interest of individual employers to raise wages above the prevailing monopsonistic level in order to
attract large supplies of nurses away from their competitors. If enough employers behave in this fashion,
maximizing individual rather than joint profits leads to the competitive outcome.
An alternative to the collusion model are noncooperative leader-follower models with a dominant
employer and a competitive fringe. For example, a large hospital may set its wage, based in part on the
expected reaction (i.e., employment) by smaller wage-taking employers. The precise wage-employment
outcome in such models varies with the specific assumptions. But these models typically lead to a
prediction of lower wages and employment than with a competitive outcome [see Sullivan (1989) for a
presentation of alternative models].
An additional possibility is that of a monopsonist that can wage (i.e., price) discriminate. Perfect

extremely low concentration.
4 If monopsony is accompanied by monopoly in the product market, then output, employment, and wages will be
even lower than with monopsony alone.

4

---

wage discrimination would imply that a monopsonist pays individual nurses exactly their reservation
wages; that is, bids up the labor supply curve. In this (admittedly unlikely) case, monopsony would
imply, as before, lower wages (except for workers at the margin), but would not imply a decrease in
employment or a “shortage” of nurses. Wage discrimination by employers with monopsonistic power
might be evinced by a larger residual variance of wages in monopsonistic than in competitive markets.
That is, there would be greater wage dispersion among nurses with given characteristics (i.e.,
productivity), corresponding to the dispersion in reservation wages, whereas employers in competitive
markets must pay the same market wage to all workers of equal productivity.
Monopsonistic power may also take the form of a flatter wage-experience profile in smaller, less
competitive markets. If experienced (often married) nurses are less mobile than younger entry level (and
less often married) nurses, wages should not rise as quickly with respect to experience in less competitive
markets.5 Reputation and implicit contracts play an important role here. On the one hand, if wages are
determined competitively at entry and contracts are short run, entry-level nurses will demand even higher
initial wages since they do not expect employers to increase wages fully with respect to future
productivity. This reinforces our prediction about relatively flatter slopes of earnings profiles in smaller
markets.6
The monopsony outcome requires that there be limited long-run mobility of nursing personnel
across and within labor markets. Existence of an upward sloping market supply curve (i.e., a positive
employment-wage relationship) does not imply non-competitive outcomes. A necessary but not
sufficient condition for the monopsonistic wage-employment outcome is that there be an upward sloping
labor supply curve facing individual hospitals (employers). One may find empirical evidence of upward
sloping short-run labor supply curves facing hospitals [Sullivan (1989)] because of long-run implicit
contracts between nurses and hospitals. Such contracts require that hospitals are able to structure

5 This is essentially the argument made by Adamache and Sloan (1982).
6 On the other hand, hospitals may develop a reputation as non-opportunistic employers, in which case long-run
implicit contracts should lead to a compensation profile that maximizes the joint surplus of hospitals and nurses. In
this latter case, the present value of lifetime earnings is at least as high as in markets where wages equal workers’
spot marginal products. Absent further assumptions about differences in implicit contracts, incentive effects, and pay

5

---

compensation in ways that increase the attachment of nurses to their current employer, and that hospitals
develop a reputation for non-opportunistic behavior. Firm-specific training, pay sequencing that back-
end loads compensation (wages or fringes), and various forms of incentive contracts act to lower worker
mobility and break down the equality between wages and workers' spot marginal products [see Lazear
(1991) and Hutchens (1989)]. A decrease in wages owing to a hospital-specific demand shock, for
example, would not result in the loss of all nurses to other employers. What appears to be evidence for
monopsony (i.e., a less than perfectly elastic labor supply schedule) may in fact reflect implicit contracts
that maximize the joint surplus of hospitals and nurses.7
For these reasons, testing whether hospitals face upward sloping labor supply curves is not a
sufficient test of the monopsony model. Rather, one must examine directly whether the wage-
employment outcomes predicted by the monopsony model, as compared to outcomes from the
competitive model, in fact occur. The approach taken in this paper, therefore, is to examine directly
testable implications that follow under most variants of the monopsony model. Most important is the
prediction that wages will be lower in monopsonistic labor markets than in otherwise similar competitive
markets. And, except in the unlikely event that an employer can perfectly price discriminate,
employment and the ratio of employment to other factors will be lower.
Because monopsony requires limited mobility of labor and a relatively small number of
employers, we should find that the monopsony outcome is more prevalent the smaller the labor market
and the fewer the number of employers. Rural and small town markets are likely to have relatively few
hospitals, nursing homes, and doctors’ offices over large geographic areas, while nurses in such markets
(particularly those who are married) may have limited mobility. If monopsonistic power is a major factor
in the nursing labor market, it is these markets where its effects should be observed. By contrast, highly

sequencing in competitive and monopsonistic labor markets, few testable implications arise.
7 The arguments in this paragraph apply not only to nursing labor markets. In fact, the prevalence in nursing of
transferable general training, rather than nontransferable firm specific training, implies that implicit long-term
contracts are even more important in many non-nursing labor markets. If we are correct, then studies such as the one
by Sullivan (1989) should find evidence of upward sloping firm-level labor supply curves in a wide variety of
occupations. The presence of implicit contracts may help account for the "surprising" finding by Sullivan that the
inverse elasticities of nursing labor supply curves did not differ between hospitals in large metropolitan markets and
those in smaller markets.

6

---

populated urban areas, with numerous hospitals and many alternative nurse and non-nurse employment
opportunities, are least likely to show the effects of buyer power in the labor market.
Our empirical test of monopsony therefore requires that we determine whether wages in a labor
market are low relative to a counterfactual competitive outcome, and to measure market characteristics
such as size and/or number of employers. Such information allows us to examine whether wage rates
diverge from those observed in competitive markets, and whether the pattern of divergence is consistent
with that predicted by the monopsony model. Because wages vary across areas, in part because of
unmeasured site-specific amenities and locational attributes [Roback (1982)], our wage measure will rely
on a comparison of nursing wages to those of a control group of non-nursing workers with similar
characteristics in the same market. We will also examine alternative implications of the monopsony
model, including its potential effects on wage-experience profiles, distortions of the employment mix,
and the role of labor unions as a countervailing force.
III. Data
The primary data for this study are drawn from the monthly Current Population Survey (CPS)
Outgoing Rotation Group (ORG) files for the period October 1985 through December 1993 (99 surveys).
The CPS ORG files have not been used previously in the nursing literature. The CPS is conducted by
the Bureau of the Census and includes large representative samples of U.S. households. In each month's
survey a quarter of the sample (the outgoing rotation groups) are asked not only the standard
demographic and employment questions, but also questions from an earnings supplement that includes
responses on weekly earnings, hours worked per week, and union status.8 The annual CPS earnings files,
containing data for all 12 monthly surveys in each year, are not public use tapes, but are made available
through the research and data services staff at the Bureau of Labor Statistics. Beginning in October
1985, the CPS identified each individual's location as either in or out of one of 202 Metropolitan
Statistical Areas or Consolidated Metropolitan Statistical Areas (MSA/CMSA) with populations of

8 The sample design of the CPS is that a household is included for 4 months, followed by 8 months out, followed by
4 months in. Only the outgoing rotation groups (rotations 4 and 8) are asked the earnings questions. Hence, the
ORG or earnings files include all individuals surveyed in the CPS, but only once in a year.

7

---

100,000 (in July 1983).9 Those not in a designated MSA/CMSA either reside in a small MSA or a non-
metropolitan area. Thus we have representative national samples with all workers assigned to one of 252
market areas (202 MSA/CMSAs and 50 non-urban state groups).
Our nursing sample includes all hospital and non-hospital registered nurses (RNs), licensed
practical nurses (LPNs), and nursing aides who are employed wage and salary workers ages 18 and over
with positive weekly earnings and hours. Workers whose primary activity is school, whose implied real
wage rate is less than a dollar, or who have had their occupation allocated by the Census are excluded
from the sample. Sample sizes of these groups for the October 1985 through December 1993 period are
as follows: RNs - 24,345, RNs employed in hospitals - 17,296, LPNs - 6,119, and aides - 20,166.
In addition, we utilize three control groups to construct measures of relative wages by labor
market. We construct a large initial control sample that includes all female workers (meeting the same
criteria as above, except the occupation non-allocation requirement) in non-health related occupations
within the following broad occupational groups: executive, administrative and managerial; professional
specialty; technicians and related support; sales; administrative support and clerical; and service (except
protective and household services). This large control sample is then divided into three distinct control
groups along educational lines. Those with at least a college degree comprise the control group with
whom the wages of RNs are compared (n=127,831); those with a high school degree or some college
make up the control group for LPNs (n=341,365); and those with less than a high school degree are used
as the control group for aides (n=38,689). We should emphasize at the outset that our basic results and
conclusions are not sensitive to the choice of control groups.10
The CPS ORG files have several advantages relative to other data sources. In contrast to data

9 There are 181 MSAs and 21 CMSAs. The latter (with the exception of St. Louis) contain two or more primary
metropolitan statistical areas (PMSAs). Prior to October 1985, there were identifiers in the CPS for only 44 large
Standard Metropolitan Statistical Areas (SMSAs).
10 Because monopsony power might in principle affect the wages of health professionals outside nursing, we chose to
exclude all workers in health-related occupations from the control groups. By way of summary, excluded from the
control group samples are males; female workers in health-related occupations within the selected broad
occupational categories; and workers in the following non-selected occupational groups: private household services;
protective services; farming, forestry, and fishing; precision production, craft, and repair; machine operators,
assemblers, and inspectors; transportation and material moving; and handlers, equipment cleaners, helpers, and
laborers.

8

---

from the AHA surveys, the CPS provides information on individual worker (but not hospital)
characteristics.11 CPS data also are available in a timely fashion on an annual basis and information is
available on current earnings, hours, union status, and occupation, as opposed to the previous year's
earnings and occupation on the longest job held last year as in the Census of Population (union status is
not reported in the Census). The CPS also includes information on nursing personnel employed not only
in hospitals, but also outside of hospitals within the same labor market. Previous literature has focused
primarily on hospital employees. Most important for our study, the CPS provides large representative
samples of potential groups with whom nurses can be compared, with information taken from the same
surveys and with detailed area (i.e., labor market) identifiers. The primary disadvantage of the CPS is
that it lacks information on occupational grade and responsibilities (apart from the designation of RN,
LPN, or aide) or on employer characteristics, apart from industry (hospital, nursing home, etc.) and
sector (private, federal, state, or local).
The comparison of nursing and non-nursing wages is a critical component of this study. By
measuring nursing wages relative to control groups within the same labor markets, we are able to control
not only for differences in measurable worker characteristics, but also for cost of living differences, area
amenities or disamenities, area-specific unmeasured labor quality, differences in working conditions, and
other market-specific wage determinants that otherwise are not easily measured. The measurement of
relative wages for nurses within areas is made possible by comparing nursing wages (RNS, LPNs, and
aides) within each labor market to the wages of their respective control group in that market, after
controlling for measurable worker and market characteristics. The implication of the monopsony model
is that nursing wages, relative to their control groups, will be lower in labor markets that are small and
with a limited number of employers. Our method should be reliable unless unmeasured area-specific
determinants of relative nursing to control group wages are systematically related to market size and
number of employers.
A second source of data is Hospital Statistics, 1989-90 edition, a summary of information

11 Sullivan (1989) provides a good discussion of the AHA data.

9

---