EFFECTS OF THE ACA MEDICAID EXPANSION ON THE COMPENSATION OF NEW PRIMARY CARE PHYSICIANS

Abstract

Background:

It is well-documented that the Affordable Care Act Medicaid expansion increased health care utilization by low-income Americans. Emerging studies also found that the expansion changed the geographical distribution of new physicians. However, the effect of the expansion on physician compensation has not been studied.

Objectives:

We aimed to assess how the Medicaid expansion affected the compensation of new primary care physicians (PCPs) and whether the effect differed by specialty, gender, and geography.

Research Design:

We used a quasi-experimental difference-in-differences design to assess changes in compensation for new PCPs from before to after the Medicaid expansion in states that expanded Medicaid compared to states that did not expand.

Subjects:

Our study included 2,003 new PCPs who responded to the Survey of Residents Completing Training in New York between 2009 and 2018.

Measures:

Our primary outcome was respondents’ self-reported starting salary for their first year of practice. Our secondary outcomes were respondents’ self-reported additional anticipated income and incentives they received for accepting the job offer.

Results:

We found that starting salaries for new PCPs, especially new general internists and family physicians, grew faster in expansion states than in nonexpansion states. In addition, we found that the expansion was associated with a statistically significant increase in receiving additional anticipated income as part of the compensation package for new PCPs practicing in rural areas.

Introduction

The Affordable Care Act (ACA) expanded Medicaid eligibility to nonelderly adults with income up to 138 percent of the federal poverty level. As of June 2021, 39 states and the District of Columbia have adopted the ACA Medicaid expansion,¹ leading to significant increases in insurance coverage: 14.8 million newly-eligible adults have enrolled in Medicaid in 2014–2020 as a result of the ACA Medicaid expansion;² an additional 3.9 million previously-eligible adults enrolled in Medicaid owing to enhanced outreach efforts and streamlined enrollment process.²

Since the ACA Medicaid expansion began in 2014, numerous studies have noted that it improved health care access, utilization, and health outcomes for low-income Americans.^3,4 Studies of the effect of the Medicaid expansion on primary care and behavioral health workforce have also emerged in recent years.^5,6,7 In terms of primary care workforce, a recent study showed that new general internists increasingly chose to practice in expansion states after the expansion was implemented.⁷ A qualitative study found that health care organizations in expansion states tried to recruit more primary care providers to meet the higher demand for health care.⁸

Yet to our knowledge, there is a lack of empirical evidence regarding the Medicaid expansion’s effect on physician compensation. This study aims to fill the gap by quantifying the expansion’s impact on the compensation of new primary care physicians (PCPs). Our primary hypothesis is that starting salaries and incentives for new PCPs grew faster in expansion states as compared to nonexpansion states following the Medicaid expansion. We also hypothesize that such effects may differ across primary care specialties, gender, and geographic areas, as prior studies have consistently identified differences in physician income across these factors.^9,10 We tested our hypotheses using a unique dataset, as described below.

Methods

A. Study Design

We used a quasi-experimental difference-in-differences design with two-way fixed effects to assess changes in compensation for new PCPs from before to after the Medicaid expansion in states that expanded Medicaid compared to states that did not expand. Our study period, from 2009 to 2018, includes five years before the Medicaid expansion started in 2014 and five years afterwards. States that expanded their Medicaid programs during the study period were deemed expansion states, and the post-expansion years for each expansion state was defined based on the implementation date of Medicaid expansion. During our study period, 25 states and the District of Columbia expanded in 2014, 3 states expanded in 2015, 2 states expanded in 2016, and 1 state expanded in 2017 (Online Appendix 1).

B. Data Sources

New York Resident Exit Survey:

We used the 2009–2018 Survey of Residents Completing Training in New York (New York Resident Exit Survey). Each year, the Center for Health Workforce Studies of the University at Albany, State University of New York, surveys all new physicians completing a residency or fellowship training program in New York to inform the medical education community on the outcomes of training and the demand for new physicians.¹¹ The average response rate for the survey over the study period was 60 percent.

The survey collects detailed compensation information for the first non-training position accepted by new physicians, including both the “base salary/income” (starting salary) and the “additional anticipated income” received by new physicians for their first year of practice. The survey also asks whether the physician received other incentives for accepting his/her position, such as “relocation allowances” and “J-1 visa waiver”.

The survey serves the purpose of our study for two reasons. First, to our knowledge, it is the largest publicly available survey that consistently collects new physicians’ compensation information across specialties both before and after the ACA Medicaid expansion. While several other physician compensation surveys have become available in recent years, they do not date back to the years before the 2014 ACA Medicaid expansion.¹² Second, the survey captures compensation information for new physicians who completed their residency or fellowship training in New York and accepted their first non-training positions across all 50 states and District of Columbia during the study period. This is because New York trains more residents and fellows than any other US state, as pointed out in previous studies using this survey.^9,13 During our study period, new PCPs trained in New York constituted approximately 14% of all primary care residents in the U.S. In 2018, there were 6,465 primary care residents in training in New York, exceeding California—the state with the second largest number of primary care residents in training—by over 50%.^14,15,16

Other Data Sources:

We obtained state-level information for (1) the Medicaid expansion from the Kaiser Family Foundation (KFF) website,¹⁷ (2) annual population and number of PCPs from the Area Health Resources Files, (3)annual share of nonelderly population with non-Medicaid insurance coverage from the KFF website,¹⁸ and (4)annual median household income from Census Bureau’s Small Area Income and Poverty Estimates program.¹⁹

To control for concurrent time-varying state-level policies that may impact new physicians’ compensation during the study period, we drew on several data sources to collect information on whether a state had continued making increased Medicaid payments for primary care after the ACA national Medicaid “fee bump” expired on December 31, 2014,²⁰ as well as whether a state expanded scope of practice for nurse practitioners.²¹

C. Measures

Outcomes:

Our primary outcome of interest was new PCPs’ self-reported starting salary for their first year of practice. The data was recorded in bracketed values in the survey. Following prior studies using the same survey,^9,13,22 we converted the bracketed values to continuous dollar values by replacing the raw categories with the midpoint of each category to facilitate a direct interpretation of our results. The continuous dollar value was adjusted for inflation using the Consumer Price Index and was reported in real 2018 dollars. (Online Appendix 3.I)

Our secondary outcomes of interest included respondents’ self-reported additional anticipated income and incentives they received for accepting the job offer. To examine the additional anticipated income, we constructed a binary variable indicating whether a respondent received any additional anticipated income as part of the compensation package, as well as a continuous measure quantifying the dollar amount amongst those who received additional anticipated income. To examine the incentives, we constructed a binary indicator to document whether a respondent received any of the following: “sign-on bonus”, “income guarantees”, “on-call payments”, “relocation allowances”, or “educational loan repayment”, “H-1 visa sponsorship”, “J-1 visa waiver”, “spouse/partner job transition assistance”, “support for maintenance of certification and continuing medical education”, “career development opportunities”, or “other”. (Online Appendix 3.II)

Policy Variable:

We constructed a binary indicator for state implementation of Medicaid expansion. The policy indicator was assigned a value of 1 for years after the expansion in an expansion state, and a value of 0 for the years prior to the expansion in that state as well as for comparison states. (Online Appendix 1)

Covariates:

Our study accounted for physician- and state-level factors that may affect the compensation of new PCPs. First, we included physician characteristics including age, gender, citizenship, race/ethnicity, location of medical school education (US, Canada, or other countries), type of medical school education (allopathic or osteopathic), primary care specialty, and obligation to practice in Health Professional Shortage Areas. We also included the number of job offers a physician received to measure the physician’s perceived “quality” and bargaining power. Similarly, we included indicators for the level of educational debt that a physician had to capture the physician’s potential motivation to seek positions offering particularly generous compensation.

Second, we included characteristics of the accepted clinical position for each physician including practice setting (solo practice/partnership, group practice, hospital, or other), whether the practice is located in small city or rural area, and weekly hours anticipated for direct patient care. Third, we controlled for state-level characteristics associated with potential changes in health care demand unrelated to the Medicaid expansion including number of PCPs per capita, share of nonelderly population with non-Medicaid insurance coverage, and median household income.

Finally, we controlled for time-varying state-level policies that may affect compensation of new PCPs during the study period, including whether a state continued the Medicaid “fee bump” for primary care after 2014,²³ and whether a state allows nurse practitioners to have full practice authority.^24,25 (Online Appendix 4.I).

D. Study Samples

Our study samples included new PCPs who reported having specialized in family medicine, general internal medicine, or general pediatrics during their residency training. For purpose of our analyses, we subset the samples to those who had accepted a job by the time of the survey and whose primary activity was patient care in a non-training position. After excluding observations with missing values for the outcomes or covariates, we had a sample of 2,003 PCPs over the ten-year period for starting salary, 1,260 for additional anticipated income, and 1,841 for incentives (Online Appendix 2 and Appendix 6.I).

E. Statistical Analysis

Implementing the Difference-in-Differences Design in Empirical Models:

For continuous outcomes (e.g., starting salary), we used log transformed dollar values as dependent variables in ordinary least square (OLS) models to quantify the growth rate²⁶ (approximate percentage change) associated with the Medicaid expansion, adjusting for covariates mentioned above as well as state and year fixed effects; for binary outcomes (e.g., whether received any incentives for accepting a job), we estimated linear probability models to facilitate a direct interpretation of the effect of Medicaid expansion. (Online Appendix 4.II).

Heterogeneous Effect of Medicaid Expansion:

We explored whether the effect of ACA Medicaid expansion varied across subgroups of the study population by extending the above difference-in-differences design in three ways. First, to assess how the expansion’s impact varied over the distribution of the starting salary or additional anticipated income for new PCPs, we applied the difference-in-differences design to a quantile regression framework and analyzed the deciles of new PCPs’ starting salary and additional anticipated income respectively. Second, to examine how the impact of Medicaid expansion varied by primary care specialty, we analyzed each specialty separately. Third, to study whether the expansion’s impact differed by gender or geography, we extended the above difference-in-differences design by including interaction terms between the Medicaid expansion and gender as well as rural-area indicators. (Online Appendix 4.III).

Validity of Difference-in-Differences Design:

To verify the parallel trend assumption required for the difference-in-differences analysis, we conducted an event study and found no evidence of differential trends in the outcome variables between expansion and nonexpansion states prior to the Medicaid expansion (2009–2013). We also conducted placebo tests using 2010, 2011, 2012, 2013 as the hypothetical Medicaid expansion year respectively. Each of these tests showed that there was no evidence of differential trends in compensation for new PCPs from before to after the hypothetical expansion year between expansion and nonexpansion states. In addition, to verify the differences in physician characteristics between expansion states and nonexpansion states did not change from before to after the Medicaid expansion, we regressed each physician-level covariate on the Medicaid expansion indicator²⁷ and showed there was no statistically significant association between Medicaid expansion and changes in any physician-level covariates (Online Appendix 5).

Sensitivity Analyses:

We conducted numerous sensitivity analyses with alternative model specifications, variable definitions, and sample definitions, and yielded results similar to our main findings (Online Appendix 6).

All analyses were conducted using R, version 4.0.3, with robust standard errors clustered at state level.

Results

A. Descriptive Statistics

Table 1 summarizes the new PCPs’ characteristics. Both before and after the Medicaid expansion, the characteristics of the newly-trained PCPs who accepted clinical positions in expansion states were different from those in nonexpansion states. The former are more likely to be female, white, below the age of 30 years, and to have attended medical schools within the US, while they are less likely to be Hispanic, specialize in general internal medicine, practice in rural area, and have received multiple offers during their job search. As discussed above, even though there were differences between new PCPs practicing in expansion and nonexpansion states, such differences remain largely unchanged before and after the Medicaid expansion (Online Appendix 5.III).

Table 1.

Characteristics of the Study Sample by State Medicaid Expansion Status, 2009–2018

	Pre-expansion			Post-expansion
	Expansion States	Nonexpansion States	p value	Expansion States	Nonexpansion States	p value
Gender
Male	50.5%	68.5%	0.000	47.1%	54.8%	0.041
Female	49.5%	31.5%		52.9%	45.2%
Age
Less than 30	25.8%	14.6%	0.003	32.3%	20.7%	0.000
30–40	65.0%	75.3%		61.3%	66.4%
40 and above	9.2%	10.1%		6.3%	12.9%
Citizenship
US citizen or permanent resident	69.3%	59.0%	0.009	82.8%	72.8%	0.000
H-1, H-2, H-3 Temporary worker	23.4%	25.3%		9.4%	9.7%
J-1, J-2 Exchange visitor	7.3%	15.7%		7.9%	17.5%
Medical School Attended
US	40.3%	18.5%	0.000	49.2%	22.6%	0.000
Canada	0.3%	0.6%		0.0%	0.0%
Other country	59.4%	80.9%		50.8%	77.4%
Race / Ethnicity
White	33.4%	14.6%	0.000	36.7%	24.4%	0.000
Hispanic	8.4%	18.5%		8.7%	19.4%
Black/African American	6.8%	15.7%		5.7%	13.8%
Asian or Pacific Islander	40.1%	40.4%		39.2%	35.0%
Other	11.1%	10.7%		9.7%	7.4%
Medical Education Type
M.D.	88.5%	94.4%	0.007	83.6%	90.3%	0.004
D.O.	11.5%	5.6%		16.4%	9.7%
Specialty
Family Medicine	16.6%	12.4%	0.000	23.9%	18.4%	0.002
General Internal Medicine	59.2%	77.5%		54.5%	67.3%
General Pediatrics	24.2%	10.1%		21.6%	14.3%
Level of Education Debt
None	44.9%	57.9%	0.007	32.9%	48.4%	0.101
Less than $250,000	43.9%	32.0%		37.3%	27.2%
$250,000 and above	11.1%	10.1%		29.8%	24.4%
HPSA Obligation
Yes	14.8%	24.2%	0.009	11.4%	20.3%	0.003
No	85.2%	75.8%		88.6%	79.7%
Practice Type
Solo practice / Partnership	4.5%	6.7%	0.008	2.7%	2.3%	0.849
Group practice	26.4%	29.8%		30.1%	29.0%
Hospital	61.8%	61.2%		60.5%	63.1%
Other	7.3%	2.2%		6.7%	5.5%
Practice Location
Major city / Suburban	75.0%	61.2%	0.001	83.2%	72.4%	0.001
Small city / Rural	25.0%	38.8%		16.8%	27.6%
Hours for Direct Patient Care
Less than 60	88.9%	83.7%	0.092	83.6%	75.6%	0.012
60 or above	11.1%	16.3%		16.4%	24.4%
Number of Offers
0 or 1	18.9%	6.7%	0.000	17.1%	7.8%	0.000
2–5	61.1%	62.9%		62.9%	59.0%
6 and above	19.9%	30.3%		20.0%	33.2%

SOURCE: Author’s analysis of 2009–2018 Survey of Residents Completing Training in New York.

NOTES:

The sample includes residents and fellows completing primary care training in New York. Residents and fellows are only included if they accepted non-training clinical positions and did not have missing information for analysis. The pre-expansion period is 2009 through the year prior to the Medicaid expansion for expansion states, and 2009 through 2013 for nonexpansion states. P-values are results of chi-square tests for differences between physicians practicing in expansion and nonexpansion states.

B. Primary Outcome: Starting Salary

Difference-in-Differences Estimates:

The top panel of Table 2 summarizes the unadjusted average starting salary for new PCPs. In expansion states, the average starting salary increased by 13.8%, or $26.0K, from $174.7K prior to Medicaid expansion to $200.7K post Medicaid expansion. In contrast, the average starting salary experienced a smaller increase in nonexpansion states, with a growth rate of 6.1%, or $16.3K, from $214.1K prior to Medicaid expansion to $230.3K post Medicaid expansion. The average unadjusted starting salary grew significantly faster in expansion states after the Medicaid expansion, both in terms of the increased dollar amount ($9.8K, p=0.0203) and the growth rate (7.8%, p=0.0021). While the average unadjusted starting salary in expansion states had been lower than that in nonexpansion states both before and after Medicaid expansion, the gap between expansion and nonexpansion states shrank from $39.4K ($174.7K vs. $214.1K) before the expansion to $29.6K ($200.7K vs. 230.3K) after the expansion.

Table 2.

Difference-in-Differences Estimates for Starting Salary, by State Medicaid Expansion Status, 2009–2018

	Salary (in $1,000)		Change		Dif-in-Dif Estimate
	Pre-Expansion	Post-Expansion	$ Amount	Percent	$ Amount	Percent
Unadjusted Model
Expansion States	174.7	200.7	26.0	13.8%
	(11.1)	(9.5)	(2.9)	(2.2%)	9.8**	7.8%***
Nonexpansion States	214.1	230.3	16.3	6.1%	(4.2)	(2.5%)
	(3.8)	(5.0)	(3.1)	(1.3%)
Adjusted Model
Expansion States	178.0	202.7	24.7	13.0%
	(1.6)	(1.5)	(2.0)	(1.1%)	11.4**	7.3%***
Nonexpansion States	205.9	219.1	13.2	6.2%	(5.2)	(2.6%)
	(4.2)	(3.3)	(6.1)	(2.9%)

SOURCE: Author’s analysis of 2009–2018 Survey of Residents Completing Training in New York.

NOTES:

^a.There are 2,003 observations.

^b.Unadjusted estimates present the average observed starting salary and its changes from pre- to post-expansion period without controlling for any covariates. The pre-expansion period is 2009 through the year prior to Medicaid expansion for expansion states, and 2009 through 2013 for nonexpansion states. Dollar changes are estimated using starting salary as dependent variable, and percent changes are estimated using logarithm of starting salary as dependent variable.

^c.Adjusted starting salary represents the average predicted starting salary by study period (i.e., pre- and post-expansion) and state Medicaid expansion status (i.e., expansion and nonexpansion) after accounting for state-level and physician-level covariates, state fixed effects, and year fixed effects. For each study period and Medicaid expansion status, adjusted starting salary was calculated for each of the 2,003 survey respondents using the OLS coefficient estimates assuming the state where he/she accepted the job offer had/had not expanded Medicaid coverage in that time period. Dollar changes are estimated using starting salary as the dependent variable, and percent changes are estimated using logarithm of starting salary as the dependent variable.

^d.Standard errors are clustered at state level.

^*p<0.1,

^**p<0.05,

^***p<0.01.

The bottom panel of Table 2 presents the adjusted difference-in-differences estimates for starting salary. Consistent with the unadjusted estimates, the adjusted estimates show that the Medicaid expansion was associated with a 7.3% (p= 0.0053) higher growth rate for starting salary in expansion states (13.0%) relative to nonexpansion states (6.2%).

Heterogeneous Effect of the Medicaid Expansion Across Starting Salary Distribution:

Consistent with the average starting salary summarized in Table 2, the top panel of Figure 1 shows that the unadjusted starting salary for new PCPs in expansion states was lower than that in nonexpansion states throughout the starting salary distribution both before and after Medicaid expansion. A comparison between the plot on the top left and top right panel indicates that the gap between expansion and nonexpansion states narrowed substantially post the Medicaid expansion, again revealing that starting salary for new PCPs grew faster in expansion states than nonexpansion states.

FIGURE 1.

Distribution of Starting Salary and Quantile Regression Estimates, 2009–2018

Source: Author’s analysis of 2009–2018 Survey of Residents Completing Training in New York.

Notes:

The top panel presents the observed empirical cumulative distribution function of starting salary by state Medicaid expansion status before and after Medicaid expansion, respectively. The bottom panel presents the quantile regression difference-in-difference coefficient estimates (dotted black line), adjusting for physician-level and state-level covariates, as well as state and year fixed effects. The 95% confidence interval for starting salary at each decile of the starting salary distribution is shown in grey shade. If the grey shade is entirely above or below zero, then it indicates statistically significant results. The OLS difference-in-difference coefficient estimates (solid red line) with 95% confidence intervals (dashed red line) are also presented in the same figure for reference. The dotted black line almost always falls between the two dashed red lines, suggesting the estimates yielded by the OLS and the quantile regressions are not statistically different.

The bottom panel of Figure 1 presents the estimated effect of Medicaid expansion on the deciles of starting salary distribution. We found that the expansion was associated with a statistically significant increase in the growth rate of starting salary between the 10^th and the 70^th percentile in expansion states. In addition, the effect estimated at each decile using the quantile regression was almost always within the 95% confidence interval of the OLS estimate, suggesting consistent results between the OLS and the quantile regressions.

Heterogeneous Effect of the Medicaid Expansion by Specialty, Gender and Geography:

The left panel of Figure 2 shows that the Medicaid expansion was associated with statistically significant increases in starting salary for general internists and family physicians, but not for pediatricians. The middle and right panels of Figure 2 show that we did not find any statistically significant differences by gender or geography in the effect of Medicaid expansion on PCPs’ starting salary.

FIGURE 2.

Heterogeneous Effect of the Medicaid Expansion by Specialty, Gender and Geography, 2009–2018

Source: Author’s analysis of 2009–2018 Survey of Residents Completing Training in New York.

Notes:

The left panel presents the difference-in-differences coefficient estimate with 95% confidence interval(error bar) for starting salary for each specialty. The middle and right panels present the OLS coefficient estimates with 95% confidence intervals (error bars) for each outcome measure for (1) the interaction between Medicaid expansion and rural indicator and (2) the interaction between Medicaid expansion and female indicator, respectively. Each color represents a separate OLS regression. (Online Appendix 4.III).

C. Secondary Outcome: Additional Anticipated Income and Incentives

We did not detect any statistically significant association between the Medicaid expansion and the secondary outcome measures overall, or across different gender or specialty. However, we did find that the effect of Medicaid expansion differed significantly between rural and non-rural areas regarding the likelihood of having additional anticipated income. As shown in the middle panel of Figure 2, new PCPs practicing in rural areas of expansion states were 10.1% (p=0.0147) more likely to have additional anticipated income in their compensation packages relative to their peers in non-rural area after the Medicaid expansion. (Online Appendix 7)

Discussion

We found that the ACA Medicaid expansion was associated with a significant increase in the growth rate of starting salaries for new PCPs in expansion states relative to nonexpansion states. We found significant effects on both average starting salaries and on salaries along most of the starting salary distribution. Despite the faster growth after the Medicaid expansion, the average starting salary for new PCPs was still lower in expansion states relative to nonexpansion states. In addition, we found that the Medicaid expansion was associated with statistically significant increase in the chance of having additional anticipated income as part of the compensation package accepted by new PCPs to practice in rural areas relative to non-rural areas in expansion states.

Our main finding——starting salaries for new PCPs growing faster in expansion states after the Medicaid expansion——reflects the fact that the demand for new PCPs has grown faster than the supply of new PCPs in expansion states as compared to nonexpansion states after the expansion, forcing health care employers in expansion states to offer higher compensation for new PCPs. Before the Medicaid expansion, the U.S. had faced a shortage of PCPs.²⁸ The Medicaid expansion further exacerbated such shortage because newly eligible enrollees use more primary care services after gaining Medicaid coverage.³ To combat the shortage of PCPs due to the expanded insurance coverage, health care organizations in expansion states were willing to further enhance their compensation packages to attract new PCPs.^8,29

As a consequence of the faster growth of starting salary, more new PCPs may choose to practice in expansion states. This, in turn, can reduce the supply of new PCPs in nonexpansion states as noted by prior research.⁷ Indeed, a recent study published in Medical Care found that the Medicaid expansion was associated with more new general internists choosing to practice in expansion states instead of nonexpansion states.⁷ It will be interesting for future studies to examine whether the increased starting pay together with other ACA provisions directly targeting primary care(such as investment in continued development of primary care workforce)³⁰ enhanced the job satisfaction of PCPs in expansion states, and whether the change in new PCPs’ distribution, especially new general internists, ultimately leads to better access to primary care and better health outcomes for people in expansion states.

It is not surprising that our study did not find any significant difference in the growth rate of general pediatricians’ starting salary between expansion and non-expansion states. The ACA Medicaid expansion targeted low-income adults and may have a limited effect on children’s health insurance coverage. Consistent with our results, a previous study found no significant association between the Medicaid expansion and the supply of pediatricians.⁵

Our finding that the Medicaid expansion was associated with an increased chance of having additional anticipated income for rural-practicing PCPs in expansion states relative to nonexpansion states is consistent with existing evidence showing that bonuses and incentives constitute the main difference between rural and non-rural physician compensation in recent years.³¹ Our finding also reflects the fact that the Medicaid expansion has boosted health insurance coverage to a greater degree in rural areas relative to urban areas.³² Since including additional anticipated income in the compensation package has the potential to attract more new PCPs to practice in rural areas of expansion states, it remains an important topic for future research to determine whether the Medicaid expansion may alleviate the issue of persistent uneven geographic distribution of PCPs.^33,34

It is not surprising that we did not find any significant effects of the Medicaid expansion on the gender gap in new PCPs’ compensation. This is because the Medicaid expansion was not designed to address the factors identified in the literature that potentially lead to differential starting salary for male and female physicians (such as difference in career goals or workplace biases).⁹

This study had several limitations. First, the New York Resident Exit Survey data were self-reported and therefore subject to recall bias. Second, the survey did not ask about respondents’ preference for work-life balance until 2014, a potentially important factor influencing the starting salary for new PCPs. Third, the compensation information from the survey may not be representative of the compensation for new PCPs across the US. Fourth, the limited sample size made it hard to consistently observe significant results over time in our event study. Fifth, the continuous measures for starting salary and additional anticipated income may understate the Medicaid expansion effect on new PCPs’ compensation. To the extent new PCPs’ starting salary or additional anticipated income increased yet stayed within the same income bracket after the Medicaid expansion, such increases would not be captured by our measures. Sixth, the effect captured in this study is driven by a combination of demand and supply factors. While nonexpansion states may need to raise compensation for new PCPs to compete with expansion states to attract new PCPs, our study did not attempt to quantify the effect of Medicaid expansion on nonexpansion states.

Conclusion

Our study provides new evidence regarding how the ACA Medicaid expansion has shaped the compensation of new PCPs since 2014. We find the expansion was associated with faster growth in starting salary for new PCPs and higher likelihood of having additional anticipated income as part of the compensation package for those who accepted patient care positions in rural areas relative to non-rural areas in expansion states. Although the results are encouraging, the starting salary for new PCPs in expansion states still fell below that in nonexpansion states after the Medicaid expansion. It remains an important topic for future studies to investigate whether the combination of market force and the ACA provisions directly targeting primary care services can further enhance compensation of new PCPs in expansion states.