Table 5.

IV Estimates and specification tests for the effect of caregiving

Outcome	Obs.	OLS or Probit		IV-2SLS or IV-Probit		Marginal effects of IV-Probit		IV specification tests
	(1)	(2) Coeff.	(3) (S.E.)†	(4) Coeff.	(5) (S.E.)†	(6) %points	(7) (S.E.)†	(8) Overid	(9) Exo.
Table 5A: Daughter-in-law sample
Prob(Pain affecting daily activities)	2528	0.41**	(0.18)	0.62***	(0.23)	13.1***	(4.9)	NS	NS
Prob(Self-reported health fair to poor)	2528	0.23	(0.17)	0.46**	(0.23)	15.7**	(7.9)	NS	NS
Prob(Any outpatient care use)	2525	−0.09	(0.18)	−0.18	(0.23)	−6.6	(8.6)	NS	NS
Log(Out-of-pocket costs) if any use	1548	0.19	(0.21)	0.47*	(0.27)	-		NS	**
Prob(Any regular prescription drug use)	2528	0.21	(0.17)	0.27	(0.23)	8.2	(7.2)	NS	NS
Log(Out-of-pocket costs) if any use	721	0.29	(0.29)	0.30	(0.31)	-		NS	NS
Table 5B: Daughter sample
Prob(Pain affecting daily activities)	4108	0.31*	(0.17)	1.73***	(0.40)	41.3***	(9.4)	NS	***
Prob(Self-reported health fair to poor)	4108	0.22	(0.15)	1.62***	(0.37)	56.1***	(12.5)	NS	***
Prob(Any outpatient care use)	4105	0.10	(0.17)	0.82**	(0.38)	29.8**	(13.9)	NS	**
Log(Out-of-pocket costs) if any use	2567	0.46**	(0.21)	0.75*	(0.39)	-		NS	NS
Prob(Any regular prescription drug use)	4108	0.22	(0.18)	1.07***	(0.39)	34.0***	(12.2)	NS	***
Log(Out-of-pocket costs) if any use	1268	0.41	(0.31)	0.57	(0.54)	-		NS	NS

Notes: NS denotes ‘not significant’ at the 10% level. F-statistics for first-stage regressions for the four conditional models were over 50 in the daughter-in-law sample, and 22 and 15 in the daughter sample. All models control for other explanatory variables presented in Table 4, namely age, education, home ownership, health insurance status, and whether both parents(-in-law) are living together, type of residential area, and survey wave.

^†Robust standard errors in parentheses.

^***p<0.01,

^**p<0.05,

^*p<0.1.