Abstract
This paper employs cointegration and error‐correction modelling to test the causal relationship between real income, real investment and tertiary education using data for the People’s Republic of China over the period 1952–1999. To proxy tertiary education we use higher education enrolments and higher education graduates in alternative empirical specifications. One of the paper’s main findings is that real income, real investment and tertiary education are cointegrated when real investment is the dependent variable, but are not cointegrated when either tertiary education or real income is the dependent variable. We also extend the in‐sample analysis to examine the decomposition of variance and impulse response functions.
Acknowledgements
The authors thank two referees for helpful comments on an earlier version of this paper.
Notes
1. The tests are available from the authors on request.
2. Note that the Zivot and Andrews (Citation1992) test implies that real investment is I(0) and the other variables are I(1). Tsay and Chung (Citation2000) show that, regardless of whether the variables used for Granger causality are I(0) or I(1), as long as their fractional orders of integration sum up to a value greater than 0.5, then it is sufficient for analysis. This requirement is met in this study. See also He and Maekawa (Citation2001), who perform a simulation study to test for spurious results from the Granger causality test. One of the three cases for model specification they consider is where one variable is I(0).