The most significant variable is the cross-sectional standard deviation which enters with a very large positive coefficient. This positive relation holds up for most countries when country-specific coefficients are allowed. This results may seem surprising given that the relation between dividend yields and cross-sectional standard deviation was significantly negative, a potential effect of increased stock market development reducing the cost of capital. One interpretation that explains these results is that the cross-sectional standard deviation is an imperfect proxy for risk. One would then anticipate expected returns to be positively related to the predictable component in risk and their innovations to be negatively correlated (increased uncertainty reduces returns). The regression results suggest that the correlation between the predictable components somewhat surprisingly dominates the unconditional correlation. One would expect this relation to weaken with the liberalization since local volatility should no longer be the relevant risk measure, which is indeed what structural regressions of both dividend yields and returns reveal. Since dividend yields and returns are contemporaneously negatively correlated, the opposite sign obtains for the dividend yield regression.

Table
The Impact of Liberalizations on Excess Returns
A. Intro=ADRs
l	NUMC	CONCR	STDL2	INFL	FXV	XMGDP	CCR	Intro
0.9	-0.003	0.030	0.591	-0.013	-0.188	0.002	-0.046	0.0005
announc.	-0.63	1.03	21.90	-3.04	-2.77	0.10	-4.20	0.36
0.9	-0.003	0.030	0.591	-0.013	-0.189	0.001	-0.046	0.000
effective	-0.66	1.05	21.90	-3.05	-2.79	0.04	-4.19	0.42
0.5	-0.004	0.030	0.590	-0.012	-0.191	0.001	-0.047	0.0021
	-0.74	1.06	21.91	-2.99	-2.81	0.06	-4.18	0.55
0.1	-0.005	0.030	0.590	-0.012	-0.195	0.001	-0.049	0.0059
	-1.04	1.10	21.91	-2.91	-2.86	0.04	-4.33	1.05
B. Intro=Country funds
l	NUMC	CONCR	STDL2	INFL	FXV	XMGDP	CCR	Intro
0.9	-0.003	0.031	0.590	-0.013	-0.189	0.001	-0.046	0.0003
	-0.49	1.10	21.80	-3.12	-2.79	0.06	-3.90	0.22
0.5	-0.005	0.031	0.588	-0.012	-0.192	0.001	-0.048	0.0038
	-0.97	1.13	21.72	-2.97	-2.83	0.06	-4.31	0.96
0.1	-0.005	0.035	0.588	-0.012	-0.193	0.003	-0.047	0.0066
	-1.00	1.26	21.70	-2.96	-2.84	0.18	-4.33	1.16
C. Intro=ADRs and Country Funds
l	NUMC	CONCR	STDL2	INFL	FXV	XMGDP	CCR	Intro
0.9	-0.003	0.031	0.590	-0.013	-0.189	0.001	-0.046	0.0002
	-0.48	1.08	21.88	-3.06	-2.78	0.07	-3.93	0.20
0.5	-0.004	0.032	0.589	-0.013	-0.190	0.001	-0.047	0.0022
	-0.77	1.16	21.79	-3.03	-2.80	0.08	-4.21	0.60
0.1	-0.004	0.034	0.588	-0.012	-0.191	0.003	-0.047	0.0053
	-0.91	1.24	21.75	-3.00	-2.81	0.17	-4.30	0.96
D. With Regulatory Liberalization Indicators
	NUMC	CONCR	STDL2	INFL	FXV	XMGDP	CCR	PRE	DURING	POST	AFTER	WaldTest
	-0.012	0.036	0.583	-0.011	-0.204	0.009	-0.047	0.011	0.011	0.024	0.017	3.700
	-1.96	1.31	21.62	-2.64	-3.00	0.52	-4.05	1.74	1.17	2.86	1.66	0.054
Weighted	-0.008	0.036	0.584	-0.012	-0.205	0.007	-0.045	0.014	0.010	0.035	0.016	3.010
	-1.41	1.30	21.61	-2.77	-3.00	0.41	-3.89	1.30	0.65	2.43	0.91	0.083
E. With ADR, Country Fund and Regulatory Liberalization Indicators
	NUMC	CONCR	STDL2	INFL	FXV	XMGDP	CCR	PRE	DURING	POST	AFTER	WaldTest
	-0.007	0.040	0.583	-0.010	-0.206	0.005	-0.044	0.008	0.006	0.019	0.017	2.510
	-1.38	1.46	21.51	-2.49	-3.03	0.31	-3.83	1.19	0.68	2.46	1.97	0.113
Weighted	-0.006	0.038	0.583	-0.011	-0.206	0.005	-0.043	0.011	0.008	0.030	0.024	2.500
	-1.14	1.37	21.51	-2.58	-3.02	0.32	-3.79	1.00	0.51	2.28	1.56	0.114
F. With Cumulative Net Capital Flow Break Points
	NUMC	CONCR	STDL2	INFL	FXV	XMGDP	CCR	PRE	DURING	POST	AFTER	WaldTest
	-0.006	0.049	0.592	-0.011	-0.204	-0.004	-0.054	-0.003	0.015	0.001	0.022	0.430
	-1.25	1.73	21.96	-2.70	-3.00	-0.22	-4.78	-0.52	1.87	0.20	2.32	0.513
Weighted	-0.005	0.063	0.651	-0.013	-0.251	-0.011	-0.064	-0.009	0.016	-0.004	0.044	0.130
	-0.82	1.78	21.43	-2.56	-3.26	-0.48	-3.86	-0.73	0.99	-0.25	1.93	0.714
Group-wise heteroskedasticity and autocorrelation-consistent t-statistics are reported below the coefficients. In panels A-C, we estimate a time-series cross-sectional estimation with the U.S. dollar returns in excess of a U.S. T-bill as the dependent variable. l represents how fast the additional impact of further liberalizations declines. With low ls, additional issues generate almost no additional effect. The Intro variable is defined in the panel title. In panels D-F, we estimate a model with dummy variables around the liberalization definition. In the regressions labeled 'weighted', we weight the dummy variables by the correlation with the world market return before the liberalization. The Wald test is whether the average returns declines from Pre to Post liberalization.