doi:10.1016/j.jhealeco.2006.12.002 
Copyright © 2006 Elsevier B.V. All rights reserved.
Explaining China's regional health expenditures using LM-type unit root tests
Win Lin Chou
, a, 
aDepartment of Economics, Chinese University of Hong Kong, Shatin, Hong Kong
Received 12 October 2004;
revised 30 November 2006;
accepted 6 December 2006.
Available online 12 January 2007.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
This paper investigates the relationship between health care expenditure, income, and other factors that are not related to income for China with pooled cross-section and time series data. To study the stationarity property of these variables, we use panel Lagrange Multiplier (LM) unit root tests that allow for structural changes. To perform the LM unit root tests, we employ finite-sample critical values derived through the bootstrap method, instead of relying on the critical values from the asymptotic normal distribution. An important finding based on the estimated panel cointegrated regressions is that the government budget deficits have a significant long-run impact on China's health care expenditure. This provides supportive evidence on the differences between rich and poor areas in China's health care financing policy, and the substantial disparities in health service coverage in China.
Keywords: Lagrange multiplier test; Panel unit root tests; Bootstrapped critical values; Panel cointegration tests
JEL classification codes: C12; C23; I10
Table 1.
LM unit root tests on per capita health expenditure by region, 1978–2004
Notes: (–): Refers to the case of no breaks. (k): lag order. Bootstrapped critical values are from Table A1 of Appendix A.
a MinLM = Minimum LM test of
(Lee and Strazicich, 2003) and
(Lee and Strazicich, 2004).
b Max|
tb| = test statistics of
Nunes (2004).
* Denotes significance at the 10% level.
Table 2.
LM unit root tests on per capita income by region, 1978–2004
Notes: (–): Refers to the case of no breaks. (k): lag order. Bootstrapped critical values are from Table A1 of Appendix A.
a MinLM = Minimum LM test of
(Lee and Strazicich, 2003) and
(Lee and Strazicich, 2004).
b Max|
tb| = test statistics of
Nunes (2004).
* Denotes significance at the 10% level.
** Denotes significance at the 5% level.
Table 3.
Panel LM unit root tests on HE, GDP and four non-income series, 1978–2004
Notes: HE = real per capita health expenditure; GDP = real per capita income; NI1 = the dependency ratio of old aged population (NI1); NI2 = the proportion of population aged 65 and over; NI3 = the ratio of the health expenditure funded by public sector; and NI4 = real government budget deficits.
** Denotes significance at the 5% level.
*** Denotes significance at the 1% level.
Table 4.
Results from Pedroni panel cointegration tests
Note: N = 28, T = 27. The 5% and 10% critical values are −1.65 and −1.28, respectively. All Pedroni statistics use the left tail of the normal distribution to reject the null, except the panel ν test which variance ratio statistic and takes positive values.
** Denotes significance at the 5% level.
Table 5.
Parameter estimates of cointegrated panel regressions using OLS, DOLS and FMOLS methods (dependent variable: HE)
Notes: Equations are based on the pooled data from 1978 to 2004 for 28 provinces. Figures in parentheses are t-values.
*** Denotes significance at the 1% level.
Table A1.
Bootstrapped critical values, empirical sizes and frequency of true break for HE, 1978–2004
a Empirical size is computed by allowing one break under the null but the break is neglected.
b True break dates are assumed to be the same as those estimated in
Table 1.
Table A2.
Bootstrapped critical values, empirical sizes and frequency of true break for GDP
a Empirical size is computed by allowing one break under the null but the break is neglected.
b True break dates are assumed to be the same as those estimated in
Table 2.
Abstract
Purchase PDF (187 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (0)
Purchase PDF (51 K)
