The role of R&D and economic policy uncertainty in Sri Lanka’s economic growth

Financial Innovation

Table 2 Cointegration tests

The table shows the cointegration test results. In Panel A, we report the Gregory-Hansen test results. This test has a null hypothesis of no cointegration, which is tested against the alternative of cointegration with regime shifts. The optimal lag for this test is based on the Schwarz information criterion. We find that the regime shift occurred either in 1999 or 2002. In Panel B, we report the nonlinear autoregressive distributed lag (NARDL) bounds test results. The null hypothesis under this test is that there is no cointegration. The main intuition underlying this test is that negative and positive changes in a predictor has nonlinear impact on the predictand. k, I(0), and I(1) denote, respectively, the number of predictors in the equation, and the critical values for the lower and upper bounds. In both tests, we included a maximum of 2 lags in the test regressions. To test the semi-endogenous growth theory, we estimate \(\nu_{t} = lnX_{t} + \left( {\left( {\phi - 1} \right)/\sigma } \right)lnA_{t}\), where \(A\) is TFP and \(X\) is R&D expenditure. Similarly, to test the Schumpeterian growth theory, we estimate \(\varsigma_{t} = lnX_{t} - lnQ_{t}\), where \(Q\) denotes product quality (i.e. \(Y\) or \(AL\)), and \(ln\) denotes the natural logarithm
*, **, and ***Statistical significance at 10%, 5%, and 1% levels, respectively.