Table 5:

Sub Sample Summary Coefficients of Lagged Height-for-age Z-scores on Current Z-score

	CHNS : China 1989–91		CHNS : China 1989–93		KIDS : South Africa		LSMS : Nicaragua
	OLS	IV	OLS	IV	OLS	IV	OLS	IV
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Z-score<-1 at baseline
Coefficient	0.501	0.290	0.433	0.226	0.256	0.279	0.254	0.245
(Standard error)	(0.050)	(0.100)	(0.053)	(0.104)	(0.152)	(0.264)	(0.109)	(0.196)
t-statistic	10.02	2.90	8.17	2.17	1.68	1.06	2.33	1.25
R-squared value	0.437	0.306	0.479	0.345	0.279	0.301	0.428	0.409
F(instruments in first stage)		11.55		11.55		1021.10		1.10
Observations	705		705		258		149
Age<=24 months at baseline
Coefficient	0.363	0.336	0.308	0.318	0.281	0.223	0.427	0.351
(Standard error)	(0.053)	(0.084)	(0.053)	(0.069)	(0.070)	(0.182)	(0.059)	(0.141)
t-statistic	6.85	4.00	5.81	4.61	4.01	1.23	7.24	2.49
R-squared value	0.619	0.306	0.536	0.483	0.416	0.358	0.479	0.315
F(instruments in first stage)		16.09		16.09		1148.00		1.02
Observations	331		331		152		475
Age >24 months at baseline
Coefficient	0.596	0.520	0.512	0.520	0.216	0.258
(Standard error)	(0.035)	(0.12)	(0.043)	(0.128)	(0.053)	(0.119)
t-statistic	16.87	4.30	11.95	4.06	4.08	2.17
R-squared value	0.62	0.395	0.510	0.395	0.342	0.306
F(instruments in first stage)1		8.08		8.08		9700.00
Observations	790		790		362

1/ Instruments are lagged prices, lagged prices interacted with mother’s education dummy, lagged wealth and shocks. Standard errors for the IV models are bootstrapped with 1000 repetitions. Coefficients significant at 5% are in bold. The Nicaragua sample only ranges from 0–36 months at baseline so estimates for the older sub-sample are not possible.