RES E A R C H Open Access
Global Health Workforce Labor Market
Projections for 2030
Jenny X. Liu
1*
, Yevgeniy Goryakin
2
, Akiko Maeda
2
, Tim Bruckner
3
and Richard Scheffler
4
Abstract
Background: In low- and middle-income countries, scaling essential health interventions to achieve health
development targets is constrained by the lack of skilled health professionals to deliver services.
Methods: We take a labor market approach to project future health workforce demand based on an economic model
based on projected economic growth, demographics, and health coverage, and using health workforce data (1990–2013)
for 165 countries from the WHO Global Health Observatory. The demand projections are compared with the projected
growth in health worker supply and the health worker “needs” as estimated by WHO to achieve essential health coverage.
Results: The model predicts that, by 2030, global demand for health workers will rise to 80 million workers, double the
current (2013) stock of health workers, while the supply of health workers is expected to reach 65 million over the same
period, resulting in a worldwide net shortage of 15 million health workers. Growth in the demand for health workers will be
highest among upper middle-income countries, driven by economic and population growth and aging. This results in the
largest predicted shortages which may fuel global competition for skilled health workers. Middle-income countries will face
workforce shortages because their demand will exceed supply. By contrast, low-income countries will face low growth in
both demand and supply, which are estimated to be far below what will be needed to achieve adequate coverage of
essential health services.
Conclusions: In many low-income countries, demand may stay below projected supply, leading to the paradoxical
phenomenon of unemployed (“surplus”) health workers in those countries facing acute “needs-based” shortages.
Opportunities exist to bend the trajectory of the number and types of health workers that are available to meet public
health goals and the growing demand for health workers.
Keywords: Health workforce, Labor market projections, Global health
JEL Classifications: J210, J230, J440
Background
The Sustainable Development Goals (SDGs) for health and
well-being lay out ambitious targets for disease reduction
and health equity for 2030, including universal health cover -
age (UHC) [1]. Health systems are highly labor intensive,
and health workers play a key role in performing or mediat-
ing most of the health system functions. Thus, an effective
health care delivery system depends on having both the right
number and the appropriate mix of health worker s, and on
ensuring that they have the required means and motivation
to perform their assigned functions well [2].
In many low- and middle-income countries, efforts to
scale-up health services to achieve UHC and health devel-
opment goals are confronted by acute shortages and in-
equitable distribution of skilled health workers that present
a binding constraint to delivering essential health services
[3, 4]. These countries face a “crisis in human resources for
health” that can be described in terms of (1) availabil ity,
which relates to the supply of qualified health workers; (2)
distribution, which relates to the recruitment and retentio n
of health workers where they are n eeded most; and (3) per-
formance, which relates to health worker productivity and
the quality of the care they provide [5]. Multiple conditions
* Correspondence: Jenny.Liu2@ucsf.edu
1
Institute for Health and Aging, Department of Social and Behavioral
Sciences, University of California, San Francisco, 3333 California Street, Suite
340, San Francisco, CA 94118, United States of America
Full list of author information is available at the end of the article
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Liu et al. Human Resources for Health (2017) 15:11
DOI 10.1186/s12960-017-0187-2
contribute to this problem, including inadequate education
and training capacity, negative work environments, weak
human resources regulatory and management systems, and
inadequate financial and non-financial incentives [6, 7].
National policymakers, researchers, and international
agencies have called attention to this global shortage and
maldistribution of the health workforce, and for govern-
ments to make concerted efforts to address these chal-
lenges in order to achieve UHC [4, 8].
Given the criticality of the health workforce in the health
system, and substantial time and resources invested to edu-
cate and develop skilled health workers, it is crucial to
understand the factors that affect the size of the future
health workforce and plan appropriately today. Traditional
approaches to addressing human resource constraints in the
health sector have focused on “needs-based” workforce
planning, which estimates health workforce requirements
based on a country’s disease burden profile and commensur-
ate scale-up of education and training capacities to increase
the supply of health workers to provide those services [9,
10]. In this approach, health workforce density has been
found to be associated with decreases in maternal and infant
mortality rates [2, 11], as well as in the total burden of dis-
ease as measured in disability-adjusted life years (DALYs)
[12]. Using this approach, the World Health Organization
(WHO) estimates that a health workforce density of around
4.45 health workers per 1000 population corresponds to the
median level of health workforce density among countries
that have achieved, or have come close to achieving, UHC
[13]. Policy makers could then identify the production cap-
acity and associated financing necessary to increase the
stock of health workers to meet these health service require-
ments [4, 13].
However, this needs-based approach neglects other im-
portant factors that influence the size of the health work -
force, notably labor market dynamics that are defined by
demand and supply interactions [5, 14]. It should not be
assumed that labor markets always “clear,” in other words
thatthesupplyanddemandforworkersperfectlymatch.
There are a number of reasons for an imbalance between
the demand and supply for workers. For example, prices
may not adjust easily due t o fixed wage rates established by
legislative or bureaucratic processes, or may be tied to civil
service schedules that make them relatively insensitive to
the numbers of health workers employers seek to hire or
who are willing to be employed. Other institutional rigidi-
ties, such as regulatory guidelines and trade unions, can also
restrict the extent to which the number of workers
demanded or suppl ied responds to price signals. These situ-
ations can lead to either a shortage (i.e., quantity demanded
exceeds the quantity supplied) or surplus (i.e., quantity
demanded falls behind the quantity supplied) of health
workers. Further, the number of health workers estimated to
be “needed” to achieve the national health goal of UHC may
not necessarily coincide with the demand for health workers
due to economic capacity and other market conditions in
the health system. Countries may also face unemployment
among health workers when the supply of health workers
exceeds demand generated by the country’s underlying eco-
nomic capacity to employ them. A labor market analysis will
help to identify such mismatch of labor supply and demand,
and lead to more effective policy design to address these is-
sues [15].
This study estimates the demand for health workers in
2030 (the target achievement year of the SDGs) as a func-
tion of economic, demographic, and health coverage factors
basedonaneconomicmodel.Themodelassumesno
change in technology or organization of health services and
thus projects the demand for health care as if the current
system of healthcare and technology remains the same in
2030. We then compare this demand projection with the
supply and the “needs” projections (based on the WHO
SDG threshold density of 4.45 health workers per 1000
population [13]), and discuss the potential policy implica-
tions of the findings.
Methods
Theoretical framework
The demand for health workers reflects the willingness to
pay of the purchasers of healthcare (e.g., government, pri-
vate sector firms), which in turn drives the demand for
employing health workers in clinics, hospitals, public health
centers, and other parts of the health system. The demand
for health workers is influenced by factors including house-
hold income (i.e., the ability of consumers to purchase
health services), the fiscal capacity of the government to
support the health system and employ public sector
workers, demographic and epidemiologic conditions of the
population (e.g., aging and burden of disease that determine
the relative types of health services consumers want), and
the level of health coverage in terms of risk pooling and fi-
nancial protection available to enable consumers to access
and utilize healthcare at times of need.
The supply of health workers can be defined as the total
number of health professionals with the appropriate skills
and qualifications who are willing to enter into job market
in the health sector and find acceptable jobs. Labor eco-
nomics predicts that, as the level of compensation offered
increases, more qualified workers should be willing to
become employed as a health professional [16, 17]. In turn,
higher wages encourages more students to apply for health
professional education, and increases the demand for med-
ical training and eventually the number of skilled profes-
sionals a vailable. In a global labor market where workers’
skills may be transferable across country boundaries, migra-
tion flows also play an important role in determining the
supply of health workers within a country. In particular,
outflows of health professionals from low- and middle-
Liu et al. Human Resources for Health (2017) 15:11 Page 2 of 12
income countries to other, more attractive markets offering
better compensation have been identified as one of the
biggest challenges facing health systems [6, 18].
Traditional labor economic analyses assume that, in well-
functioning labor markets, disequilibrium (i.e., imbalances
between demand and supply) is short-lived. A core assump-
tion is that the wage rate is flexible and freely adjusts the
incentives to both employers and health workers, influen-
cing their employment behaviors and preferences such that
equilibrium is restored. Figure 1 depicts a static health
worker labor market in which employers’ demand (D) to
employ health workers interacts with the health workers
available to supply (S
1
) their labor to determine the market
wage rate (W*) and the number of workers (H*) that will be
employed at that rate. Countries face a binding constraint
on the amount of financing available to employ more health
workers, and resource constraints are more severe in lower-
income countries. A shortage of workers results, for ex-
ample, when a wage rate (W
L
) is lower than the market
optimum (W*), or when the number of workers supplied
(H
S
) falls below the number demanded (H
D
). All else being
equal, shortages in this market could be alleviated through
(1) additional compensation to increase wages to W* and
attract more workers into the market; and/or (2) increasing
the production of workers or import of workers from exter-
nal markets to shift the supply curve outward (S
2
) while
keeping wages at W
L
.
In reality, markets can fail to “clear” because prices are
either not flexible, or demand and/or supply does not readily
adjust to price signals. Both types of rigidities are common
in the health labor market [14]. First, price of labor in the
health sector is often not flexible because wages in the pub-
lic health sector (often the largest employer in many coun-
tries) are usually set by legislative processes and tied to civil
servant pay scales. Second, health professional associations
(especially for physicians) use their bargaining power to re-
strict labor supply and negotiate set wage rates. Third, the
regulation of health services, such as licensing by profes-
sional bodies and governmental jurisdictions, to ensure
quality standards and monitoring results in additional rigidi-
ties in the ability of workers to become employed wherever
positions may be available.
Projecting demand
The demand model builds on a previous economic model
for projecting physician numbers [19]. We apply a similar
theoretical approach but incorporate factors in addition to
economic growth that are expected to influence demand for
health workers, including population demographic structure
and health coverage. We also use more recent and robust
health workforce data, which are the result of concerted
efforts by the WHO to gather cross-national data on work-
force numbers since 1990. Because the demand model
requires rich historical data on health worker densities, sep-
arate models for nurses/midwives and all other health pro-
fessionals could not be estimated; data for these cadres were
insufficient to produce demand projections. Rather, we first
predicted the number of physicians from the demand
model, and then applied constant ratios of physicians to
other cadres to obtain estimates of nurses/midwives and all
other health workers (AOWs). The total projected number
of health workers therefore reflects the sum of the estimates
for physicians, nurses/midwives, and AOWs.
In most health systems, spending on health workforce
wages and benefits represents a significant share of total
health expenditures [20]. Previous studies indicate that
overall economic growth, as measured by national income,
is the best predictor of health expenditures from which the
demand for health workers is derived [21, 22]. In other
words, spending on healthcare tends to increase as overall
income increases, which in turn suggests that more workers
can be employed to deliver health services [19, 21].
To our knowledge, few have previously projected future
health workforce labor market demand. Owing to data
requirements, early efforts largely focused on specific devel-
oped countries for which data on health workers are more
readily available [23]. Leveraging efforts to obtain cross-
national and longitudinal data on health workers, Scheffler
et al. [19] were the first to forecast the demand, need, and
supply of physicians for 158 countries with suitable data.
While notable in the scope of global coverage, their result-
ing model relied on only one model parameter input—gross
national income—to generate projections.
Our demand model expands on previous methods
developed by Scheffler et al. [19]. In addition to income (i.e.,
Fig. 1 Health worker static labor market theoretical framework. Legend:
Demand (D) and supply (S) interact to determine the number of workers
(H*) that will be employed at a market wage rate (W*). At a wage rate
(W
L
) that is lower than the market optimum (W*), a shortage of workers
results, and the number of workers demanded (H
D
) exceeds the number
supplied (H
S
). To alleviate shortages in this market, either (1) additional
compensation could be given to increase wages to W* and attract more
workers into the market, or (2) theproductionofworkerscouldbe
increased such that supply shifts outward (S
2
)andthequantitydemand
(H
D
) is achieved while keeping wages at W
L
Liu et al. Human Resources for Health (2017) 15:11 Page 3 of 12
gross domestic product (GDP) per capita), we included
measures for demographic and health coverage patterns
that also drive demand for he alth workers [21, 22, 24]. The
size of the population aged 65 or over was used as an indi-
cator of aging, which is known to increase the demand for
health services [24]. We also included private per capita
household out -of-pocket (OOP) spending on medical care
as a proxy indicator for the extent of social protection for
healthcare spending. While overall healthcare spending may
trend upward with national income, the portion spe nt OOP
is largely determined by the level of health coverage by
health insurance, government subsidies, and other forms of
risk pooling and financial protection [25]. Less generous
health coverage leaves individuals to pay more OOP, which
is expected to lower the demand for and use of health ser-
vices. Thus, we expect higher OOP health spending to be
correlated with lower demand for health workers. We ex-
clude additional structural factors affecting the labor mar -
ket, such as attrition, training capacity, labor regulations,
and migration, as these data are largely unav ailable across
countries or over time.
The following section provides a summary description of
the estimation model. A detailed and technical description
of the demand projection methodology, model specification
choice, and imputation for missing data is provided in
Additional file 1: Annex A. Due to the missing data prob-
lems for nurses/midwives and AOWs, the economic model
was first used to predict the demand for physicians. System-
atic ratios were then applied to predicted physician densities
to estimate the number of nurses/midwives and AOWs.
The economic model specifies physician density
(dependent variable) as a function of GDP per capita, OOP
per capita, and the size of the population over 65 years.
Given data availability constraints, we included a vector of
dummy variables by country (i.e., fixed effects) to account
for constant differences in characteristics across countries.
All independent variables were lagged up to 5 years to
ensure that historical levels of these measures predict future
worker density numbers, and which allows time for such
factors to work through the economy and affect the labor
market. A stepwise approach was used to select the
combination of year lags of the different predictors that
maximized the predictive power of each variable. The
resulting optimal model is as follows:
lnðphysicians per 1000 population
it
Þ
¼ β
0
þ β
1
lnðGDP per capita
it−1
Þ
þβ
2
lnðGDP per capita
it−4
Þ
þβ
3
lnðGDP per capita
it−5
Þ
þβ
4
lnðOOP per capita
it−2
Þ
þβ
5
lnðPop65
it−3
Þþμ
i
þ ξ
it
ð1Þ
where μ
i
represents the vector of country dummy
variables, ξ
it
is the disturbance term, and β coefficients
are unknown parameters to be estimated from the
model. To allow for a more flexible functional form,
quadratic terms for income and health spending indica-
tors were investigated but ultimately excluded because
they yielded little additional predictive value.
Equation 1 was fit through a generalized linear
model (GLM). Estimated coefficients from the regression
model were then applied to the future values of each
predictor variable to compute the future predicted phys-
ician density.
Alternative model specifications were explored, which in-
cluded different ways in which input parameters could be
calculated (e.g., percentage of the population aged 65+,
OOP health spending as a percentage of total health expen-
ditures). To select the appropriate model, the data were
split into an “initializing” dataset (data years 1995–2004)
and an “attestin g” dataset (data years 2005–2013) with
which to assess the precision of predicted values resulting
from different specifications [26]. The model in Eq. 1
yielded predictions with the lowest mean errors.
We conducted two additional sensitivity analyses of the
projections of physician demand resulting from alternative
input parameters for the optimal demand model chosen.
First, we assessed the stability of the predictions to alterna-
tive estimated future values of GDP per capita (US$2010)
obtained from the Economic Research Service International
MacroeconomicDataSetpublishedbytheUnitedStates
Department of Agriculture (USDA) [27]. There was a rela-
tively small (9%) difference in the total estimated shortages
in 2030 based on the two methods (15.6 million with the
main method we used, and 17.0 million using USDA
numbers).
Second, we examined the possible upper and lower
bounds of the predictions resulting from high and low
future estimates of the size of the population over 65 years
generated by the United Nations Population Division [28].
Because population is the largest driver of demand in our
model, we can obtain predicted total health worker deficits
that may result from population growth among people
older than 65 that is higher and lower than expected, com-
pared to the median estimate that is presented in the main
results. These alternative low and high estimates (shown in
Additional file 1: Appendix Figure A1) indicate a tight band
for the resulting predicted values.
Projecting supply
The supply of health workers was projected to 2030 using
historical data to predict the changes in health worker
densities (per 1000 population) for each country. We as-
sumed that the historical growth rate of health worker
densities for each country would continue to 2030 at the
same rate each year. This assumes that supply growth only
trends with time, which may be plausible if the health labor
Liu et al. Human Resources for Health (2017) 15:11 Page 4 of 12
market is relatively rigid, for example, due to the strong in-
fluence of professional associations and trade unions in the
sector. Given that data were only available beginning from
1990 and that many countries only have a few data points,
thegrowthratemethodalsoenablesustoobtainprojec-
tions using minimal empirical data inputs.
We separately estimate densities for physicians and
nurses/midwives for each country from time t = {1990, …
2013} using the following equations:
Physicians per 1000 population
t
¼ α
0
þ α
1
year
t
þ ε
t
ð2Þ
Nurses=midwives per 1000 pop ulation
t
¼ β
0
þ β
1
year
t
þ ε
t
ð3Þ
where ε
t
is the random disturbance term and α
0
, β
0
, α
1
,and
β
1
are unknown parameters; the last two parameters repre-
sent the linear growth rates estimated from the model.
For countries where more than two historical dat a points
were available for physician and nurse/midwife densities,
the future projections of worker densities were predicted
based on the model parameter estimates. This occurred in
136 and 81 countries, respectively, out of the total 165
countries in our sample.
For countries with fewer than two historical data points
or where the linear regression predictions yielded implaus-
ible values, several alternative approaches were used to esti-
mate future supply (see Additional file 1: Annex A for
details). Briefly, for 72 and 118 countries for physicians and
nurses/midwives, respectively, the median density growth
rate for the country’s region and income group was applied
to the most recent data year to obtain future year predicted
values. In a number of countries where no empirical data
for nurses/midwives was available but information on physi-
cians was available, a global ratio of 2.517 nurse/midwives
to physicians was applied to obtain the estimate for nurse/
midwife density. The constant ratio for AOWs was then
similarly applied to obtain an estimate for all other health
workers. Supply estimates for all three types of health
workers were summed to obtain the total supply of health
workers per country (see Additional file 1: Annex A for
more details).
We explored alternatively specifying Eqs. 2 and 3 as log-
linear models, which assumes an exponential functional
form. T he resulting projections from the exponential speci-
fication yielded estimates that were magnified compared to
those from the linear specification, exaggerating both posi-
tive and negative trends. Coupled with the sparse number
of data points for many lower-income countries, resulting
predicted values appeared to be less stable. Within-sample
specification tests were also not possible given the lack of
sufficient time trend data for each country. We therefore
adopted the more conservative linear specification.
Data
Country-level data were collated from multiple sources.
Historical data on physician and nurse/midwife densities
(1990–2013) were obtained from the WHO Global Obser-
vatory Health database [29]. Historical and projected total
population and population aged 65 and over were ob-
tained from the United Nations Population Division [28].
We used the World Bank Development Indicator data-
base to retrieve historical (1995–2013) GDP per capita (in
constant 2011 dollars, purchasing power parity (PPP)
values), total health expenditures per capita (in constant
2011 international dollars, PPP values), and the share of
health spending OOP [30]. Projected real GDP per capita
through 2030 were obtained from the analysis carried out
by Patrick Huang-Vu Eozenou (see Additional file 1: Annex
B for details). Historical data on OOP expenditures per
capita (1995–2013) were calculated from health expenditure
and OOP percentage data. Estimates f or fu ture OOP per
capita were obtained by projecting the OOP percentage of
total health spending from 2014 to 2030 and selecting the
models giving the smallest predicted error for each country
(see Additional file 1: Annex A for details). Projections of
future workforce demand were made for 165 countries for
which both demand and supply input data were available.
To obtain the estimates for nurses and midwives and all
other workers, we use the approach taken by WHO to fill
missing data by multiplying the projected physician density
for each country by a constant 2.517 ratio of nurses and
midwives to physicians based on the WHO Global Obser-
vatory data [29]. This estimation assumes that the skills mix
for healthcare workers will stay constant and is the same for
all countries. To obtain estimates of AOWs, we similarly
applied constant multipliers, but did so stratified according
to World Bank income groups (high income = 0.373; upper
middle income = 0.406; lower middle income = 0.549; low
income = 0.595). These ratios were derived by WHO from
the ratios of nurses/midwives to physicians in 2013 (the
most recent data year) and multipliers used for AOWs [31].
Results
Table 1 summarizes the estimated coefficients from the
demand model. The income elasticity for physic ians per
capita is 0.23, which indicates that a 10% increase in per
capita GDP (lagged 1 year) is correlated with a 2.3% in-
crease in physician density. The elasticity of physician
density with respect to OOP health expenditures is
negative (consistent with our hypothesis) and significant,
indicating that a 10% increase in OOP expenditures
(lagged 2 years) is associated with a 1.0% decrease in
physician density. In addition, a 10% increase in the size
of the population aged 65 or older (lagged 3 years) in-
creases physician density by 5.2%.
Table 2 presents projected health worker demand, supply,
and net differe nces (supply-demand) by region and income
Liu et al. Human Resources for Health (2017) 15:11 Page 5 of 12