. Author manuscript; available in PMC: 2023 Aug 4.

Published in final edited form as: Am Econ J Econ Policy. 2023 Aug;15(3):184–214. doi: 10.1257/pol.20200841

Table 3:

The effect of organizational concentration on utilization, identified from PCP exits

Second stage	(1)	(2)	(3)	(4)
Second stage	OrgConc_it	Log(total utilization)_it
$O r g C o n c_{i t}^{P C P}$	0.583 (0.027)	−2.120 (0.152)	−1.625 (0.287)	−2.385 (0.440)
First stage	(5)	(6)	(7)	(8)
First stage	$O r g C o n c_{i t}^{P C P}$
$O r g C o n c_{i, o r i g}^{P C P} \times p o s t_{i t}$	−0.349 (0.005)	−0.349 (0.005)	−0.249 (0.007)	−0.203 (0.007)
F-statistic	33,199	33,199	8951	4648
PCP provider concentration			X	X
PCP characteristics, org. size				X

Notes: Each column represents an instrumental variables regression. The instrument is the exiting PCP’s organizational concentration (jackknifed) multiplied by a post-exit indicator. Specification 1’s outcome variable is the individual patient’s realized organizational concentration. Specifications 2-4’s outcome variable is the patient’s log total utilization. All regressions control for calendar year fixed effects, relative year fixed effects, and patient fixed effects. Specifications (3) and (4) include PCP provider concentration as an additional endogenous variable, instrumented by the original PCP’s provider concentration multiplied by a post indicator. Specification (4) controls for PCP characteristics: gender, experience quartile indicators, residency training indicators (internal medicine vs. family practice), and the PCP’s organization size (log total number of claims billed to the PCP’s TIN, and the log number of unique providers billing to the PCP’s TIN). Standard errors in parentheses are clustered at the PCP and patient levels. Cragg-Donald Wald F-test reported for first-stage. The PCP Exit Sample has 304,954 patient-year observations.