In France, as in many OECD countries, there has been a recent and strong feminisation of medicine, particularly for general practitioners (GPs) 1 2 . In 2009, 29% of GPs were women vs. 13% in 1983 (source: Eco-Santé), and more than two thirds of medical students who chose general practice were women. Numerous studies showed that women GPs typically work less than men (both number of days per week and number of weeks per year) 3 4 , which may result in short-term potential problems in terms of both access to ambulatory care and general practice regulation. It is therefore essential to understand the determinants of GPs provision in terms of quantities and services provided and thus the determinants of the income gap of GPs.

There have been many studies on the measurement and analysis of the income gap between men and women in the field of labour economics 5 6 , and these studies always show an income gap in favour of men regardless of country 7 8 , business sector 8 9 , status (self-employed or salaried) 10 and executive or non-executive status 11 . In the context of health, the first studies that analysed the income gap between male and female physicians were conducted in the United States during the 1970s 12 13 . Kehrer and Langwell (1976, 1982) found differentials of approximately 30% and 22%, respectively, in hourly income in favour of men. Less than one third of each of these differentials is explained by individual doctors’ characteristics (28% and 13%), such as experience, status (salaried or private practitioner) and medical specialty. More recently, Baker (1996) 14 also highlighted an income gap of 41% in favour of men and an hourly wage gap of 14%. After controlling for speciality and practice setting, Baker was the first to show that female GPs had a 13% higher hourly income. Using the same physician dataset, Bashaw and Heywood (2001) explained these results by the differences in working time productivities: women typically worked fewer hours than men, and because of the “law” of diminishing marginal productivity, women had higher hourly earnings than men. Thus, the decrease in working time productivity of physicians is according to the authors in favour of women because they worked fewer hours 15 .

Only two similar studies were recently conducted in Europe. Theurl and Winner (2011) obtained similar results to those obtained for U.S. doctors: a 32% income gap and a 20% hourly income gap that both were in favour of Austrian male doctors 16 . Gravelle et al. (2010) measured the income and hourly income gaps between male and female GPs in the United Kingdom of approximately the same magnitude (30% for income and 11% for hourly income). Again, the gaps are in favour of male GPs. The differences in doctors’ characteristics were found to explain between 35% and 66% of the income gap, and interestingly, the marginal return of one working hour is greater for women than that of men 17 .

To date, no comparable study has been conducted in France. Dormont and Samson (2008) measured the income gap between male and female French GPs at 34% in favour of men, accounting for experience, type of practice, duration of work, region of practice and the density of doctors and specialists but failing to adequately analyse the determinants of the gap 18 . In France, a fee-for-service payment prevails to compensate private medical practices. This system remains tightly regulated by health insurance: there are annual negotiations between national health insurance and physicians’ unions, where a convention fixes the costs of services (e.g., the cost of a consultation, visit, or specialised procedure). Only doctors belonging to “Sector 2” have the option of charging extra fees (representing only 12% of French GPs in 2009, source: Eco-Santé). GPs may also perform varying activities as a doctor in a private practice and being an employee (hospital, rest home). In this context, the income from a private practice activity should be linked to the number and type of services performed. This mechanism, in theory, should reduce the income gap between male and female GPs, or at least address the gap associated with the choice of activity.

In this study, we measured the income gap between male and female GPs working in the southwest region of France, and we thoroughly analysed its determinants to better identify potential differences and consequences in choices of medical practice and care delivery between men and women doctors.

This paper is the first to examine the nature of the income gap between male and female GPs in France. Women have incomes that are 26% less than those of their male colleagues, which is comparable to the wage gap measured for salaried workers in France, which was 25.3% in favour of men 26 . Our results were also similar to European studies with respect to income for physicians 16 17 .

As shown in the literature, the descriptive analysis confirms that the female GPs’ practice differs from male GPs’ practice, notably with respect to workload (e.g., number of hours worked per week, number of working days per week and number of weeks off) 3 and to the number of services provided (consultations and visits) with a 33% differential.

With a fee-for-service payment, we could expect that the number of services provided and the income were strongly associated such that the production of services should explain a large part of the income gap. Using the Oaxaca-Ransom decomposition (1994), we showed that less than three quarters (73%) of the income gap was actually explained by the average differences in the characteristics between male and female GPs and that less than two thirds (61%) of the income gap can be explained by the difference in the number of services provided. The impact of the differences in medical services provision between male and female GPs (33%) is thus far from fully explaining (less than two thirds, 61%) of the gender income gap (26%).

Considering the unexplained part of the income gap, the decomposition also allowed us to highlight the differences in the marginal returns to characteristics of GPs. Similarly to Kehrer, Langwell and Gravelle et al. (1976, 1982, 2010), we showed that female GPs have higher marginal returns that decrease the income gap between men and women by partially compensating for the differences in workload 12 13 17 . In particular, women have a higher marginal return relative to the average of the sample (which is correspondingly lower for men) when they perform an additional service. In the context of the FFS system, this result may be due to the different types of consultations and/or to the different patients’ profiles, implying the provision of specialised procedures paid for in addition to the fee. According to our data, female GPs reported to perform more often than male GPs paediatric follow-ups and gynaecological follow-ups, which imply additional payments in the French system (e.g., five Euros for follow-ups with infants less than two years old corresponding to an extra fee of nearly 25%). These different types of consultations could explain at least partially the difference in marginal returns of a service provided between male and female GPs. Another additional explanation may be derived from the induced demand theory: Delattre and Dormont (2008) showed that facing an increase in medical density and thus being subject to rationing, doctors tend to compensate by increasing the volume of treatments performed per consultation or number of visits 27 . Our results may also be interpreted by following a similar intuition: female GPs who appear to choose to work less or undergo a professional time constraint compensate for this by increasing their workload either by intensifying the type of their consultations and/or by favouring patients and consultations that incur additional payments. Nevertheless, it is rather well-known that female doctors generally see more women and children 28 , which can also result from patient demand and preferences.

It is important to note that by controlling for the number of medical services provided, the GP’s gross income can be interpreted as the productivity of the GP (income for a given production). For example, all things being equal, providing an additional medical service for a female GP generates a higher additional income, suggesting a higher productivity. Another important dimension of GP productivity is the length of medical services provided: several studies agreed that female practitioners have longer consultations/visits than male practitioners 24 25 26 27 28 29 . By considering the length of a consultation/visit as a proxy of quality 30 31 , a higher marginal return for women may be viewed as a compensation for quality. Following this line, we could have studied the hourly income gap between male and female GPs; however, in France, physicians are not paid hourly, and thus studying the hourly income gap was not appropriate 15 .

The differential between the estimates of constants in the two models finally explains a large part of the income gap. This differential may be derived from unobservable characteristics (preferences, risk aversion, etc.) in two possible ways. First, it could be the result of either the differences in unobservable characteristics. For example, Rizzo (2007) showed that male and female physicians appeared to reveal different ‘target incomes’ 32 . Second, it could be the result of differences in the estimated coefficients of these unobservable characteristics, such as discrimination effects (i.e., patients in rural areas or older patients that are less likely to see a female GP).

We used declarative data for this reason, and therefore, there is a self-reporting bias. However, there is no reason to believe that this bias is different for the two subgroups. Considering the selecting variables, we have chosen to select the significant variables that all were in accordance with the literature. In addition, we did not consider the simultaneity of GPs’ decisions between workload (number of services performed) and income: this could generate an endogeneity bias. We tried to correct this bias with instrumental variables, but we did not find valid instruments in our data. However, Fortin et al. (2010) showed that because the bias is the same for male and female GPs, such decomposition results remain valid 33 .

The findings of this study are helpful for understanding the determinants of income gap between male and female GPs. As is well-known, female GPs work less than male GPs; however, this difference appears to only explain 61% of the total income gap. Even if workload is clearly an essential determinant of income and thus, a determinant of the gender incomes gap, our results contradict our assumption because FFS does not reduce the gender incomes gap, where there is an imperfect relationship between the provision of medical services and income. More interesting in the context of feminisation, female GPs receive a lower income but demonstrate a higher marginal return when performing an additional consultation. To consolidate our results, it would be interesting to use a larger sample and to take into account the life partner’s income. Moreover, a complementary approach could be used to perform a similar analysis of other medical specialties and to consider the simultaneity of the two key decisions: workload and income.

Oaxaca-Ransom decompositionWe ran three models:Pooled model for all GPs:

l n Y i = β 0 + ∑ k = 1 K X i , k β k + u i

Two-group models (male, female):

l n Y i g = β 0 g + ∑ k = 1 K X i , k g β k g + u i g g = M , F

In these equations, i indexes the GPs, Y is the GP’s income, X_k represents the matrix of observable characteristics used to explain Y (GP’s characteristics, workload, type of practice) and the superscript g = {M, F} denotes gender.The expression could be written as

D = 1 n Y M ¯ − 1 n Y F ¯ = ∑ k = 1 n X k M ¯ β ^ k M + β ^ 0 M − ∑ k = 1 n X k F ¯ β ^ k F − β ^ 0 F

Oaxaca and Ransom 26 proposed a decomposition based on a non-discriminate norm, which decomposes the average income gap as follows:

D = 1 n Y M ¯ − 1 n Y F ¯ = β ^ 0 M − β ^ 0 F ︸ 1 + ∑ k = 1 n X k M β ^ k M − β ^ k Ref 2 + ∑ k = 1 n X k F β ^ k Ref − β ^ k F 3 ︸ Differences of return to characteristic + ∑ k = 1 n β ^ k Ref X k M ¯ − X k F ¯ 4 ︸ Differences in characteristic, Explained part

1. This component measures the part of the income differential that is associated with the gap between constants. Gravelle et al. (2010) highlighted that the difference between the regression constants for males and females can arise from unobserved differences in returns, differences in the means of unobserved variables and differences in the returns on these omitted variables 17 .

2. This term is the difference between the estimated coefficients of men relative to the reference (male advantage if >0).

3. This is the difference between the estimated coefficients of women relative to the reference (female disadvantage if >0).

4. This term represents the differences in characteristics between the two groups, which can be interpreted as the part of the income gap that is associated with the differences in observed variables (e.g., personal characteristics, workload, and the doctor’s practice). The term is one of the income differentials explained by the average differences in characteristics.

URML: Union régionale des médecins libéraux; GPPMP: Gross profit of private medical practice; GI: Gross income.

The authors declare that they have no competing interest.

MD participated in the analysis and in writing the manuscript. MLV helped with the analysis and interpretation of the findings. CF was involved in the project conception, design, participated in the analysis, and drafted the manuscript. MD and CF contributed towards preparing and revising the manuscript for the article. All authors read and approved the final manuscript.