Table 10.

Result: pooled instrumental variable model

	Dependent variable
	Quality
	Bladder tumor	Cataract	Adenoid and tonsil
	(1)	(2)	(3)
Constant	2.094	−2.380	0.617
	(1.349)	(1.560)	(0.889)
ms	−1.829***	−0.753**	−0.603
	(0.707)	(0.294)	(0.545)
frac_female	−1.134	2.412	−0.317
	(1.485)	(1.922)	(1.433)
frac_65	−0.159	1.432	−18.521
	(1.691)	(1.311)	(26.590)
Com	0.016	0.046	0.193
	(0.205)	(0.229)	(0.202)
HHI_ins	−0.993	0.293	−0.285
	(0.889)	(0.681)	(0.612)
Lowvolume	−0.175	0.055	−0.235
	(0.261)	(0.120)	(0.229)
Acad	−0.137	−0.243	−0.384
	(0.271)	(0.325)	(0.290)
Observations	199	286	191
$R^2$	0.088	0.071	0.055
Adjusted $R^2$	0.084	0.069	0.053

* $p<0.1$; ** $p<0.05$; *** $p<0.01$. For each diagnosis group, we report the results from the pooled model, where we use the simulated weighted market share as an instrument for the weighted market share. The simulated weighted market share is based on a multinomial logit model. We report the MacKinnon and White (1985) Heteroskedasticity-Consistent standard errors (in parentheses under coefficients). We used data from 2008 to 2011