Revisiting Financial Development and Economic Growth Nexus After Incorporating Structural Breaks: Evidence from India

The paper empirically investigates the co-integrating relationship between India’s financial development and its economic growth. Whereas economic growth has been taken to be natural log of India’s GDP; natural log of domestic credit to private sector has been considered as the proxy for financial development variable. The methodology employed has been autoregressive distributed lag (ARDL) bounds testing c-integration approach, and we have analysed log transformed yearly data for the 58-year period, i.e., 1960–2017. First the two time series were checked for unit ischolar_main both at level and 1st difference and the test employed was Perron (1997) Unit ischolar_main test with a single break. Then, the ARDL test was carried out with the two variables; Log GDP and Log Domestic Credit to Private Sector. Further model optimality model was tested using both AIC and SC criteria and model stability using CUSUM Plots. Also, ACF and PACF Plots along with ‘Q’ statistics results were used in the study for detection of serial correlation. The results of the study revealed positive co-integrating relation between India’s economic growth and financial development when GDP was taken as dependent variable under ARDL Breakpoint Model. The ‘F’ statistics of this model was 39.02885 which was significantly higher than the upper bound limit as given in both Pesaran, M. H., Shin, Y., and Smith, R. J. (2001) and also Narayan (2004) tables. The ECM term was negative and statistically significant and time to achieve long-run equilibrium was approximately 17 ½ years. On the other hand, when dependent variable was domestic credit to private sector, no co-integration was detected in the results as the computed ‘F’ value of 2.917512 was less than the lower bound. Further, no serial correlation was noticed even at lag 10, when the same was checked by ACF and PACF Plots supported by ‘Q’ statistics. The model was found to be stable using CUSUM Stability plot and the long-term results of the ARDL model for both the variables were found to be quite robust.