Variations in labor supply between female and male hospital physicians: Results from a modern welfare state

Health Policy 107 (2012) 74–82

Contents lists available at SciVerse ScienceDirect

Health Policy journal homepage: www.elsevier.com/locate/healthpol

Variations in labor supply between female and male hospital physicians: Results from a modern welfare state夽 Karl-Arne Johannessen a , Terje P. Hagen b,∗ a b

The Employers Organization Spekter, Norway Department of Health Management and Health Economics, University of Oslo, Norway

a r t i c l e

i n f o

Article history: Received 1 October 2011 Received in revised form 20 May 2012 Accepted 25 May 2012 Keywords: Labor supply Physicians Gender differences Working hours

a b s t r a c t In industrialized countries, female physicians have up to 10 h lower labor supply a week than male physicians. At the same time, the number of female physicians is increasing. The question analyzed in this article is whether these differences in labor supply for female and male hospital physicians persist in a modern welfare society, such as Norway, where comprehensive welfare reforms aim to reduce gender inequality are implemented. Information on weekly working hours from all hospital physicians in Norway during the period 2001–2007 was merged with economic variables (wages, income from other sources, net personal dept), demographic variables (age, sex, marital status, children born in the year, number of children), managerial positions and variables describing the hospital, specialty and time (year). The estimation method employed both random and fixed-effects models. Labor supply for women was 10–11 percent or 4–4.5 h per week lower than among men. The effects of children diverged strongly between the sexes. For instance, childbirth in a given year reduced the supply of working hours by women by approximately 80% but had no effects for men. After controlling for children and other factors, female physicians worked some 3–4% or 1–1.5 fewer hours than comparable male physicians. Although significant, variation in labor supply between female and male physicians is much lower in Norway then in other advanced industrialized countries. © 2012 Elsevier Ireland Ltd. All rights reserved.

1. Introduction The proportion of female hospital physicians has increased considerably across OECD countries in the past 30–40 years and is now close to 50 percent in

夽 This research is supported by Spekter and the University of Oslo. The authors would like to thank seminar participants at the University of Oslo, at the HERO/HEB workshop, Geilo 21st and 22nd March 2011 and two anonymous reviewers for useful comments on an earlier draft. ∗ Corresponding author at: Department of Health Management and Health Economics, University of Oslo, P.O. Box 1098 Blindern, 0317 Oslo, Norway. Tel.: +47 975 64 771; fax: +47 22 84 53 01. E-mail address: [email protected] (T.P. Hagen). 0168-8510/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.healthpol.2012.05.009

countries such as the Czech Republic, Finland and Poland [1]. In Norway, more than 60 percent of medical students in 2009 were female [2]. This increase in female participation in the hospital workforce may have important consequences for the supply and distribution of physician services. For example, evidence from Canada and Australia indicates that female physicians favor certain specialties, are less likely to work in rural areas, prefer working outside their childbearing age and are more likely to retire early than comparable male physicians [1,3–5]. Evidence from a number of European countries also suggest that female physicians work fewer hours than male physicians – in Northern Europe 10–12 h fewer hours a week while differences are smaller in southern European countries [1].

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

A common perception is that the Scandinavian countries are gender-equality pioneers given the implementation of many-sided and comprehensive equality policies. They may therefore exhibit a differing pattern of female physician labor supply than other countries in the OECD-area. Both women’s representation in politics [6] and female labor participation increased during the 70s and 80s and induced a political drive for equality policy measures in all Nordic countries. In Norway, enterprises, NGOs and public institutions today exercise at least 40% of both genders in all their executive boards. Already in 1978 an Equal Pay Act was implemented that made within-job wage discrimination illegal [7]. A few years later, Esping-Andersen [8] concluded that Norway had an ‘exceptionally aggressive drive for equality’ and provided evidence showing considerably less income inequality within occupational groups than Denmark or Sweden. Female labor participation has increased further. During the 1990s the share of children placed in kindergarten in Norway increased due to subsidies (earmarked grants) from the central state to the local governments. Furthermore, since 2009, all children were guaranteed a place in kindergarten from the age of one, up to a maximum cost to the family of approximately D210/US$250 per month. Also during the 90s, a system of parental leave was introduced and today parents with a newborn can choose whether they want to receive 100 percent of benefits (wages) for 46 weeks or 80 percent of benefits for 56 weeks. From 2003, and as the first country in the world, Norway chose to set aside parts of the parental leave as a ‘daddy quota’. The question analyzed in this article is whether the differences in labor supply between female and male hospital physicians persist in a modern welfare society were strong measures for equality has been in place for many years. We implement the analysis using a general economic model of physician labor supply that takes into account wage elasticities, income from sources outside the hospital, individual characteristics such as age, sex, childbirth, the number of children, employment in a managerial position, and medical specialty along with hospital-specific characteristics. The literature has so far rarely compared the effects of nonlabor variables, such as childcare, between the sexes. Baltagi et al. [23], for example, report a nonstable negative effect of children aged less than three years on the labor supply of male physicians. A recent and rather more sociologically flavored Swiss study based on a questionnaire survey reports that female physicians, especially those with children, have both lower rates of employment and place lower values on career success and career support experience than their male counterparts [9]. In this register based analysis, we concentrate on how differences in labor supply varies between the genders and on how the different genders react to decisive life events, namely marriage/cohabitation, childbirth and the number of children. We begin the analysis with a description of the Norwegian healthcare sector, including some special features concerning wage negotiations and the physician labor market. We then describe our analytical approach before turning to an overview of data sources and the definitions of variables. This is followed by the presentation of the results and a final discussion.

75

2. Institutions and wage setting In 2007, Norway ranked second among OECD countries in total health expenditures per capita ($4763 after adjusting for purchasing power parity). This is significantly less than the level of expenditure in the US ($7290) but higher than in neighboring Nordic countries ($3512 in Denmark, $3323 in Sweden and $2840 in Finland) [10]. The relatively high level of health expenditure in Norway is mainly a reflection of the general price and wage levels, partly explained by the country’s booming petroleum revenues. Part of this variation may also be related to differences in registration practices across OECD countries [11]. The Norwegian hospital sector is predominantly public with only a few non-profit private hospitals and some for-profit hospitals specializing in elective surgery. No private for-profit hospitals are included in this study. From 2002, all public hospitals in Norway came under central government ownership and were organized as health enterprises (HEs) within five regional health authorities (RHAs) [12,13].1 The RHAs overlap the former health regions, which were voluntary cooperation organizations between counties as the former hospital owners. During the period analyzed, the counties (and later the RHAs) were financed by a combination of risk-adjusted capitation and activity-based financing (ABF). In this system, the number and composition of hospital treatments, as measured by diagnosis-related groups (DRGs), quantified activity. When ABF was introduced in 1997, the central government refunded 30 percent of the DRG-based cost of a treatment through ABF. This increased to 60 percent after the 2002 hospital reforms but was later scaled back to 40 percent [14]. Within each of the health regions, the organization of the hospitals is hierarchical according to its functions and specialties, with a university hospital topping the specialty hierarchy in each of the regions. Responsibility for primary health and care services is at the municipal level. Following the 2002 hospital reforms, the responsibility for wage negotiations was transferred from KS – the Norwegian Association of Local and Regional Authorities – to Spekter, an employers’ organization which has most of the previously public organizations that have been converted into enterprises, as members. The existing mechanisms for the compensation of the differing sources of labor and their duties were simplified and the wage structure changed. This technical change had no intention to increase the salaries per se, but aimed at a structural change within the limits of the ordinary wage negotiations for that year. The wage system for Norwegian hospital physicians today defines a standard for basic working hours per week between 35.5 and 40.0 h. A minimum salary for this basic work is guaranteed by central regulations. The hospitals may add individual or collective salary to this basis according to local conditions. Due to insufficient physician work force, the hospitals are depending on extended working hours from the physicians, particularly in the specialties with high rate of emergency services that drain physician

1 The number of RHAs was reduced to four from 1 July 2007. However, this has no effect on the analysis in this paper.

76

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

working hours from daytime to night time and weekends, such as anesthesiology, surgery, internal medicine, orthopedics and gynecology. Physicians may voluntarily negotiate such planned extended working hours on an individual basis, typically varying between 0 and 10 h per week, which are paid with approximately double hourly salary as the basic payment. Despite its voluntary character, the local conditions on a given unit usually implicate that all doctors must participate to some degree. We evaluate the wage elasticities in the period while a full analysis of the effects of the package of structural wage reforms is outside the scope of this article. 3. Analytical approach 3.1. Statistical model The voluminous labor supply literature has been surveyed several times [15–17], most recently by Blundell and MaCurdy [18] and Koebel et al. [19]. Former economic analyses of the labor supply of hospital physicians include Sloan [20], Noether [21], Rizzo and Blumenthal [22] and Baltagi et al. [23], the last based on Norwegian data on male physicians during a five-year period in the 1990s. These analyses primarily address wage elasticities and effects on local labor markets on the physicians supply of labor and are usually formulated within lifecycle models where the actors maximize the lifetime sum of discounted utility derived from consumption, leisure and some individual attributes [18,19]. Our estimated model was derived from the lifecycle model and has the form: log Hit = a0 +a1 log Wit +a2 log Yit +a3 Xit +uhe +yt +spit +eit where H is total weekly hours at work, W is the wage rate, Y is a vector describing non-hospital income and X are individual attributes. In addition to individual variables, we include dummies that are specific for each health enterprise (uhe ), year (yt ) and specialty (spit ). This provides us with a fixed-effects analysis that utilizes the variation between individuals within each health enterprise, year and specialty. We assume all remaining errors are white noise. 3.2. Data and variables The main data sources are the wage and personnel register administered by Spekter and data from Statistics Norway on other individual characteristics. Data from the two sources were linked by the National Person Number (comparable to national insurance numbers in other countries) by Statistics Norway. Spekter collects employer reported data on wages and hours of work from all Norwegian hospitals in October each year. The former problem of discrepancy between reported and actual working hours was removed during the 1990s and does not create any problem to our analysis. However, the collection of data once a year induces some bias related to the reporting of overtime work as some of the physicians appear to accumulate overtime hours over a long period and report them only erratically. We have taken steps to reduce the problems

this creates. All 25 public health enterprises are included in the analysis.2 We excluded data from residents and physicians without the Norwegian Person Number (usually short-term workers). We also excluded from the database individuals aged less than 25 years or more than 67 years. The dataset was balanced (only physicians employed during the entire period were included), providing a final sample of 3559 physicians, comprising 2448 men and 1111 women (31.2 percent). As stated in Section 1, the wage system for Norwegian hospital physicians defines a standard for basic working hours per week between 35.5 and 40.0 h and additional voluntary regular extended working hours negotiated individually, typically 0–10 h per week. We calculate total planned weekly working hours (Heo ) as the sum of these hours. For individuals working part time, the basic weekly working hours is the fraction of basic working hours corresponding to their fractional position. The total weekly working hours (Hio ) includes the total planned working hours plus working hours from casual overtime. The registration of working time for physicians on parental leave was challenging, as the actual working hours deviated from the registered total planned working hours because of reimbursement rules under the National Insurance System (the reimbursement of hospitals uses planned working hours). We corrected for this by reducing the number of registered total planned working hours using information on the date of the baby’s birth (where we have exact information) and the average length of parental leave for individuals with higher education in public healthcare organizations. We assume that the period of the mother’s parental leave is 36 weeks from the date of birth while the period of the father’s leave is eight weeks from the conclusion of the mother’s leave. The data on the average length of parental leaves for different groups are based on information from NIS [24]. This approach will give us planned working hours corrected for parental leave. However, we are not able to capture individual variation in how much of the parental leaves that are actually taken. Turning to the explanatory variables, the hourly wage (W) is calculated as total wages (per month) before tax divided by total working hours (per month). Ideally, the after tax wage rate should have been used but since tax rates varies with wage level, deductibles and wages coming from public and private sector we have chosen this simplified approach. Our approach is in line with former analyses [23]. This yields an average hourly wage for each individual that varies with position, experience and education, for example Ph.D. We describe other income with two variables, both based on data from individual tax records made available by Statistics Norway: (i) non-hospital income (NonHospInc) is total income minus labor income from the hospital; (ii) net capital income (NetCapInc) is capital income minus interest payments. Individual characteristics are described by three variables, age, age squared and

2 Baltagi et al. examined 40 acute hospitals representing approximately 80 percent of the entire population of hospitals in Norway. We sample only 25 health enterprises (HE) in this study as each of the new HEs includes from one to four of the former acute hospitals.

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

gender, and we include marital status, children born during the year, children born during the previous year, the number of children in the household and managerial position to describe additional individual characteristics. To analyze the effects of gender further, we included interaction terms between gender and children and gender and marital status. We instrumented all endogenous variables (Wage, NonHospInc and NetCapInc) by their previous values (Wt−1 , NonHospInct−1 and NetCapInct−1 ). We could do this because the prior values are predetermined in the actual year of the analysis and consequently comprise valid instruments. Table 1 details the variable definitions. 4. Results

Table 1 Variable definitions and data sources. Variables

Description

Data source

Log Heo

Log of total planned weekly working hours (not including overtime) Log of total weekly working hours (including overtime) Log of hourly wage (in Norwegian kroner, NOK) Nonhospital income = total yearly income minus yearly income from the hospital (in 1000 NOK) Net capital income = yearly capital income minus yearly interest payments (in 1000 NOK) Birth year minus actual year 1 = female, 0 = male 1 if married or cohabiting, 0 else 1 = if one or more children were born during the calendar year, 0 = else Number of children in the household 1 = if in a managerial position (leader of department or section), 0 = else

Spekter

Log Hio

Log Wage NonHospInc

NetCapInc

4.1. Descriptive statistics Age

Female physicians have a smaller labor supply than male physicians before any corrections for childbirth and other factors over the entire lifespan (Fig. 1). The difference is greatest for physicians aged 30–35 years (5.8 h on average, including overtime) and least for physicians aged 43–60 years (2.0 h on average, including overtime). Weekly working hours both exclusive and inclusive of overtime increased during the first part of the period and fell slightly later (Table 2). The peak in 2003 reflects a year with tremendous activity growth (7–8 percent). During the period, overtime work fell, reflecting the changes in the wage system described above. Increased staffing throughout the period may also have contributed. The transformation of the wage system induced a favorable wage increase from 2002 to 2004 of approximately 19 percent, followed by successive years with wage increases below the Norwegian average. In this dataset, we observe a slight wage decrease in the final 2 years. This reflects a combination of low wage increases in real terms and a changing mix of planned working and overtime hours. The employment share of female physicians is stable and the employees’ average age increased by one year each calendar year. This reflects that the dataset is balanced. On average, 3–5 percent of physicians had a child each year.

Gender Married ChildThisYear

#Children Manager

Spekter

Spekter SN and Spekter

SN

SN SN SN SN

SN Spekter

SN = Statistics Norway.

The average number of children per physician increased from 2.07 to 2.31. This is slightly above the national average. Approximately 70–75 percent of the sample is married or formally cohabiting. Finally, the proportion of physicians in a managerial position increased from 18 to 21 percent during the period, which is as expected as the sample of physicians becomes gradually more experienced.

50 48

Working hours per week

77

46 44 42 40 38 36 34 32 30 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 Men (incl. overme work)

Men (excl. overme work)

Women (incl. overme work)

Women (excl. overme work)

Fig. 1. Working hours for men and women by age (average 2001–2007).

3557 3557 3557 3557 3557 3557 3557 N

All economic variables (Wage, NonHospInc and NetCapInc) deflated by the consumer price index (2007 = 100).

43.30 (8.47) 45.04 (9.14) 394.45 (113.73) 266.05 (364.58) −30.04 (274.11) 46.31 (8.69) 0.31 (0.46) 0.73 (0.44) 0.05 (0.21) 2.17 (1.26) 0.19 (0.39) 41.28 (7.79) 43.56 (9.34) 349.46 (119.60) 208.81 (288.64) −46.09 (188.35) 45.31 (8.69) 0.31 (0.46) 0.73 (0.45) 0.05 (0.22) 2.12 (1.28) 0.19 (0.39) 41.39 (7.24) 43.17 (8.54) 344.08 (87.26) 214.69 (312.58) −42.37 (200.62) 44.31 (8.69) 0.31 (0.46) 0.72 (0.45) 0.06 (0.25) 2.07 (1.29) 0.18 (0.38)

a

42.44 (8.27) 42.76 (8.50) 400.94 (118.56) 180.24 (305.81) −16.84 (207.81) 50.31 (8.69) 0.31 (0.46) 0.75 (0.44) 0.03 (0.16) 2.31 (1.23) 0.21 (0.41) 42.96 (8.35) 43.31 (8.68) 415.60 (144.29) 187.85 (318.64) −17.55 (216.51) 49.31 (8.69) 0.31 (0.46) 0.74 (0.43) 0.03 (0.18) 2.28 (1.23) 0.20 (0.40)

2006 2005 2003 2002 2001

Heo Hio Wage (NOK) NonHospInc (1000 NOK) NetCapInc (NOK) Age Gender Married ChildThisYear #Children Manager

2004

4.2. Regression analyses

Variables

Table 2 Descriptive statistics (means with standard deviations in parentheses).a

43.17 (8.319) 43.54 (8.57) 428.14 (144.29) 195.33 (497.67) −3.89 (358.51) 48.31 (8.69) 0.31 (0.46) 0.74 (0.44) 0.03 (0.18) 2.25 (1.24) 0.18 (0.38)

2007

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

42.93 (8.22) 43.45 (8.61) 416.57 (145.86) 182.59 (445.87) −6.33 (381.71) 47.31 (8.69) 0.31 (0.46) 0.73 (0.44) 0.05 (0.21) 2.22 (1.25) 0.19 (0.39)

78

We rejected an initial random coefficients model based on a Hausman specification test. Table 3 presents the results from the static multilevel models. Model 1 includes fixed effects for year and health enterprise, and model 3 additionally includes fixed effects for specialty. Models 1–3 analyze weekly hours excluding overtime work and model 4 analyzes weekly hours including overtime. Models 2–4 present results from the analyses of the interaction variables, for example between gender and child birth. All of the economic variables included provide reasonable estimates. The wage coefficients, defined as elasticities, are low and not stable across the different specifications indicating that wage increases during the period had only small effects on the labor supply. In accordance with earlier findings, non-hospital income reduced labor supply while the effects of capital income increased labor supply. The effects of age and age2 are close to zero. Holding a managerial position has a significant and positive effect of 3.2–3.8 percent on the labor supply (approximately 1.4 h a week). Women worked 10–11 percent fewer hours than men (model 1). In terms of the number of hours, the difference is approximately 4 h per week. The effects of children diverged strongly between the sexes (models 2–4). In general, a child born in a certain year had no effect on the supply of working hours among male physicians and reduced the supply of working hours among female physicians by over 80 percent. Interestingly, the lagged effects (a child born in the previous year) reduced the number of weekly hours including overtime among men slightly and had no significant effect on the supply of labor by women. This may in part be due to the fact that fathers usually take their child birth leave after the mothers. The labor supply also increased with the number of children for men, whereas for women there was no effect. For illustrative purposes, we depict this process further in Fig. 2, which shows the relationships between gender and the number of children. There was no difference in working hours between men and women among those without children, whereas there was a marked drop in working hours among women with children. In Fig. 2, the effect captures both short term effect of child birth and long-term effects of having children. The fact that this difference declined with increasing number of children probably reflects that having more children also means having children with higher age. We find that the effect of being married or cohabiting is negative for women (−4 percent) and non-significant among men. The effects of specialty on labor supply are interesting as they reduced the gender differences when the fixed effects for specialty were included (model 3 compared with model 2). The effects are small but we observe that introduction of dummies for specialty reduces the effects of both Gender, Married*Gender, ChildThisYear*Gender and #Children. This indicates that women partly chose a lower number of working hours by their choice of specialty. We further illustrate this in Fig. 3, where we depict working hours and the share of women for each specialty.

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

79

Table 3 Regression results. Log of weekly hours (estimates with standard errors in parentheses). Variables

Excluding overtime work, Heo 1

LogWage NonHospInc NetCapInc Age Age*Age Gender Married Married*Gender ChildThisYear ChildThisYear*Gender ChildThisYear t−1 ChildThisYear t−1 *Gender #Children #Children*Gender Manager Intercept Time periods N per period Fixed effects −2 Res Log likelihood * ** ***

Including overtime work, Hio

2 0.025*** (0.008) −0.004*** (0.000) 0.003*** (0.000) 0.005** (0.002) −0.000*** (0.000) −0.109*** (0.004) −0.025*** (0.005) –

3 0.032*** (0.008) −0.004*** (0.000) 0.003*** (0.000) −0.031*** (0.000) −0.000*** (0.000) −0.029*** (0.005) −0.003 (0.005) −0.046*** (0.009) −0.005 (0.012) −0.842*** (0.018) −0.017 (0.011) −0.021 (0.018) 0.007*** (0.002) −0.003 (0.003) 0.035*** (0.005) 3.675*** (0.067)

−0.345*** (0.010) – −0.017* (0.009) – 0.006*** (0.002) – 0.032*** (0.005) 3.662*** (0.069) 6 3535 Years, health enterprises 4872.7

4 0.016* (0.008) −0.004*** (0.000) 0.003*** (0.000) −0.003 (0.002) 0.000 (0.000) −0.031*** (0.008) −0.007 (0.005) −0.041*** (0.009) −0.018 (0.012) −0.847*** (0.018) −0.020* (0.011) −0.012 (0.018) 0.006*** (0.002) −0.001 (0.003) 0.038*** (0.005) 3.730*** (0.088)

0.013 (0.008) −0.004*** (0.000) 0.003*** (0.000) −0.002*** (0.000) 0.000*** (0.000) −0.026*** (0.008) −0.005 (0.005) −0.042*** (0.009) −0.005 (0.012) −0.840*** (0.018) −0.016 (0.011) −0.009 (0.002) 0.006*** (0.002) 0.001 (0.003) 0.037*** (0.005) 3.690*** (0.005)

6 3535 Years, health enterprises 2856.4

6 3535 Years, health enterprises, specialty 2335.6

6 3535 Years, health enterprises, specialty 3058.6

Significant at 10% level. Significant at 5% level. Significant at 1% level.

44.5 43.5 42.5 41.5 40.5 39.5 38.5 37.5 36.5 35.5 0

1

2 Men

3

4

5

Women

Fig. 2. Working hours for men and women by number of children (average 2001–2007).

80

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

250.0 %

47.5

45.5

200.0 %

43.5 150.0 % 41.5 100.0 % 39.5 50.0 %

37.5

35.5

0.0 %

W %/M %

Average planned working hours (Heo)

Fig. 3. Female physicians’ specialty choice and working hours by specialty.

As shown, the shares of women are lower in the specialties with highest number of working hours per week (in particular in the four specialties surgery, anesthesiology, internal medicine and orthopedics). Only gynecology stands out for having both a high share of female physicians and a high level of planned working hours. That females are overrepresented in this specialty is in line with results reported by Gjerberg [25]. The other specialties with high percent of females have close to normalized working hours and little on call duties, which mainly is due to a low fraction of emergency services in these categories. 5. Discussion and conclusions This study described and analyzed variation in labor supply for a balanced panel of hospital physicians in Norway during the period 2001–2007. We found differences between male and female hospital physicians of 10.9 percent or approximately 4 h a week. Although still significant, our main conclusion is that these differences are minor compared to the variations found in countries like Netherlands, Germany and the US where differences of 10 h or more are discovered [1,26]. The other important finding is that the differences between men and women in working hours in Norway largely can be contributed to differences between men and women in the effects of having a child and the belonging differences in maternity leave between the two genders. Controlling for children (and marriage) reduced the difference from 10.9 to 2.9 and further to 2.6 percent with the inclusion of specialty (3.1 percent if overtime is included) which correspond to 1–1.5 h a week. The analyses do not allow us to draw conclusions of causal mechanisms. Although generalizations must wait for true comparative analyses, it is not unreasonable to speculate that the

welfare reforms implemented during the last 10–20 years and presented in the opening of the article may be one of the reasons behind the lower differences in Norway than in other countries. Our hypothesis is that both the generous parental leave with the specific ‘daddy quota’ and the expansion of the supply of kindergartens, which has caused a substantial increase in female workers combining child care with early return to the workforce in most sectors, are important measures in our case. However, that full gender equality is still not achieved may be illustrated by the fact that the effect of being married or cohabiting on labor supply is negative and significant for women and non-significant among men. This is observed also in other professions of the Norwegian workforce. Despite the political focus on equity, and all the steps to stimulate women to increase their labor supply, a majority of female workers in most public sectors in Norway still take more domestic and family duties than men. In general, female workers with children and a husband in general still have a stronger tendency to adapt her carrier than a man with children and a wife. In the total cohort, the wage elasticities were close to zero, indicating that wage increases from a relatively high level had no or only small effects on the labor supply. This contrasts with findings by Baltagi et al. [23], who concluded positive wage effects in their study of 1303 male Norwegian hospital physicians. The difference in findings between the studies may arise for several reasons. To start with, the two studies took place almost a decade apart, and during this period, physician salaries as well as other important factors may have changed significantly. This may also apply when comparing our study to some of the previous studies from other countries. Furthermore, the sample period in Baltagi et al. included 1996, when physician salaries included an extraordinary but permanent increase in wages intended

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

to stimulate both extended working hours and overtime work. An alternative specification (data not shown) indicated that wage elasticities were slightly higher among women (0.03) than among men (0.004 and insignificant). Projecting future physician supply and requirements is not an easy task, and the methods used to predict such longterm consequences may vary [27]. However, most studies seem to conclude that several specialties will face major challenges in the long run if no action is taken. Based on the results in our study, the combination of an increasing share of women, their somewhat shorter working hours and, particularly, their differing choice of specialty strongly indicates that some specialties may run into recruitment challenges. The female choice seemed in particular to avoid the specialties with the highest workload of emergency and night work, as surgery, anesthesiology, internal medicine and orthopedics. These specialties consume more physician working hours to unpredictable services than other specialties, and therefore have a higher need for extended working hours per week to cover the total demand of daytime workforce. Our study therefore may indicate that female choice of specialty in part is a mechanism to avoid the burden of unpredictable shift work and long working hours. If current developments in Norway continue, women will constitute approximately 65 percent of Norwegian hospital physicians by 2025. All other factors being equal, simple calculations show that the 3 h shorter working time for women today, currently constituting a hypothetical 3 percent “loss” of the total labor years, will increase to 6 percent by 2025 with this demographic change. This, however, appears to be a relatively minor problem, as the same gender shift could result in a 20–25 percent reduction in recruitment to core specialties such as surgery and anesthesiology, which may have even more serious consequences. Such a perspective has been described by others also [28–30] and appears to remain a global medical challenge in that doctors will continue to favor part-time employment if they do not obtain sufficient flexibility in their full-time working conditions. In addition to individual labor supply and working hours, the number of hospital physicians needed in the future will depend on several factors. First, the lower working hours observed for women is not restricted to hospital care [29]. Therefore, long-term policies should consider hospital and general practice recruitment, as well as considering any central versus rural dimensions, if we are to address these challenges in the next 15–20 years. Moreover, future planning should not be restricted to increasing enrollment alone, but should also focus on a better retention of existing physicians, and on encouraging more physicians to postpone their retirement. Accordingly, future actions in this area should therefore be a balance between regulatory matters regarding working time, retirement plans and flexible working hours. In addition, the medical professions themselves must take ownership of a process aimed at developing flexible ways of delivering health services. The time needed to train physicians, as well as the time usually needed to change the requisite training infrastructure, indicates that we need at least a decade to generate major shifts in physician supply.

81

Four limitations with the article should be mentioned. First, the comparison of weekly hours of work across countries can be flawed by the facts that data are from different years. Although available data from other European countries are only 5–6 years older than our Norwegian study, more recent data also from these countries may very well show less gender variations in weekly hours. Second, due to a very complicated tax system we have chosen to use wage rate before tax in the analyses. Although this is not uncommon in our setting [23], using net wage rate will give more precise wage elasticities. Third, we have decided not to model the changes in the wage structure that was implemented in 2002–2003. Although re-estimation of the models on data from 2004 to 2007 did not give different conclusions from what we have presented, a more careful modeling of the changes will give information on how the changes affected labor supply. Fourth, next generation models in this area need to handle the choice of specialty in a way that reflects the endogeneity problem that is revealed by this analysis.

References [1] Simoens S, Hurst J. The supply of physician services in OECD countries. OECD health working papers. Paris: OECD Publishing; 2006. [2] Taraldset A. Fleste kvinnelige medisinstudenter i Oslo og Tromsø. Tidsskrift for Den Norske Legeforening 2009;129(16):1685. [3] Cohen ED, Korper SP. Women in medicine: a survey of professional activities, career interruptions, and conflict resolutions. Trends in medical education and specialization. Connecticut Medicine 1976;40(2):103–10. [4] Woodward CA, Hurley J. Comparison of activity level and service intensity of male and female physicians in five fields of medicine in Ontario. CMAJ Canadian Medical Association Journal 1995;153(8):1097–106 [see comment]. [5] McMurray JE, Cohen M, Angus G, Harding J, Gavel P, et al. Women in medicine: a four-nation comparison. Journal of the American Medical Womens Association 2002;57(4):185–90. [6] Karvonen L, Selle P. Women in Nordic politics: closing the gap. Aldershot: Darthmouth; 1995. [7] Petersen T, Snartland V, Becken LE, Olsen KM. Within-job wage discrimination and the gender wage gap: the case of Norway. European Sociological Review 1997;13(2):199–213. [8] Esping-Andersen G. Politics against markets: the social democratic road to power. Princeton, N.J.: Princeton University Press; 1985. [9] Buddeberg-Fischer B, Stamm M, Buddeberg C, Bauer G, Haeemmig O, et al. The impact of gender and parenthood on physicians’ careers – professional and personal situation seven years after graduation. BMC Health Services Research 2010;10(1):40. [10] OECD. Health at a glance. Paris: OECD; 2009. [11] Søgaard J. International reliability of SHA total expenditure on health. Working paper. Copenhagen: Danish Institute of Hospital Research; 2009. [12] Hagen TP, Kaarboe OM. The Norwegian hospital reform of 2002: Central Government takes over ownership of public hospitals. Health Policy 2006;76(3):320–33. [13] Hagen TP, Vrangbæk K. The changing political governace structures of Nordic health care systems. In: Magnussen J, Vrangbæk K, Saltman RB, editors. Nordic health care systems. Recent reforms and current policy changes. Berkshire, UK: Open University Press, McGraw-Hill Education; 2009. [14] Biorn E, Hagen TP, Iversen T, Magnussen J. The effect of activitybased financing on hospital efficiency: a panel data analysis of DEA efficiency scores 1992–2000. Health Care Management Science 2003;6(4):271–83. [15] Heckman JJ, MaCurdy TE. Labor Econometrics. In: Griliches Z, Intriligator MD, editors. Handbook of econometrics. Amstedam: North-Holland; 1986. [16] Killingsworth MR, Heckman JJ. Female labor supply: a survey. In: Ashenfelter O, Layard R, editors. Handbook of labor supply. Amsterdam: North Holland; 1986.

82

K.-A. Johannessen, T.P. Hagen / Health Policy 107 (2012) 74–82

[17] Pencavel J. Labor Supply of men. In: Ashenfelter O, Layard R, editors. Handbook of labor supply. Amsterdam: North-Holland; 1986. [18] Blundell R, MaCurdy T. Labor supply: a review of alternative approaches. In: Handbook of labor economics. Elsevier; 1999. [19] Koebel B, Laisney F, Pohlmeier W, Staat M. Life cycle labor supply and panel data: a survey. In: Mátyás L, Sevestre P, editors. The econometrics of panel data. Berlin, Heidelberg: Springer-verlag; 2008. [20] Sloan FA. Physician supply behavior in short run. Industrial and Labor Relations Review 1975;28(4):549–69. [21] Noether M. The growing supply of physicians – has the market become more competitive. Journal of Labor Economics 1986;4(4):503–37. [22] Rizzo JA, Blumenthal D. Physician labor supply – do income effects matter. Journal of Health Economics 1994;13(4):433–53. [23] Baltagi BH, Bratberg E, Holmaas TH. A panel data study of physicians’ labor supply: the case of Norway. Health Economics 2005;14(10):1035–45. [24] Grambo A-C, Myklebø S. Moderne familier – tradisjonelle valg [Modern families – traditional choices]. NAV-rapport Oslo, NAV (The Norwegian Labour and Welfare Administration), vol. 2009/2; 2009.

[25] Gjerberg E. Women doctors in Norway: the challenging balance between career and family life. Social Science and Medicine 2003;57:1327–41. [26] Staiger DO, Auerback DI, Buerhaus PI. Trends in the work hours of physicians in the United States. JAMA – Journal of the American Medical Association 2010;303(8):747–53. [27] Staiger DO, Auerbach DI, Buerhaus PI. Comparison of physician workforce estimates and supply projections. JAMA – Journal of the American Medical Association 2009;302(15):1674–80. [28] Bland KI, Isaacs G. Contemporary trends in student selection of medical specialties – the potential impact on general surgery. Archives of Surgery 2002;137(3):259–67. [29] McKinstry B, Colthart I, Elliot K, Hunter C. The feminization of the medical work force, implications for Scottish primary care: a survey of Scottish general practitioners. BMC Health Services Research 2006;6. [30] Williams TE, Sun B, Ross P, Thomas AM. A formidable task: population analysis predicts a deficit of 2000 cardiothoracic surgeons by 2030. Journal of Thoracic and Cardiovascular Surgery 2010;139(4): 835–40.