Modeling nonlinearities and asymmetries in quarterly revenues of the US telecommunications industry

Abstract

Using the modeling cycle methodology proposed by Teräsvirta [Journal of the American Statistical Association 89 (1994) 208–218], we find strong evidence of nonlinearities in the growth rates of ten quarterly real revenues data of the US telecommunications industry. Estimation of LSTAR models shows interesting dynamics of adjustment and show that non-linear models improve simple linear models in terms of the estimated residual variance and other criteria. All estimated models pass the corresponding diagnostic statistics for nonautocorrelation, constancy of parameters and non-remaining nonlinearity. Performance of out-sample forecasts is also evaluated. Our results suggest that dynamics in adjustment of the enterprises is associated to internal regulation decisions more than innovation and competitiveness structural changes. In this respect, the Telecommunications Act of 1996 produced dramatic impact in the dynamics of the growth rate of real revenues and the structure of the enterprises analyzed.