Empirical applications of novel Time Series Econometric techniques

Resumen

In this Doctoral Thesis I make use of several novel Time Series Econometric techniques in order to address several topics of interest in a variety of fields of interest. More specifically, in Chapter I, I make use of Chebyshev polynominals in time in order to study energy related variables potentially presenting non-linear behavior. In Chapters II and III I make use of fractional integration and cointegration methods in order to study optimal currency areas in Africa and Latinamerica. Finally, I analyze the Kenyan energy sector employing the recently introduced fractionally cointegrated VAR (FCVAR) methodology. Studying variations of natural gas prices in relation to consumer prices may give us better indicators for the analysis of economic activity. The first chapter deals with the analysis of natural gas spot prices using fractional integration techniques in the context of non-linear deterministic trends. I find nonstationarity with mean reverting coefficients (i.e., orders of integration in the range (0.5, 1)) in the daily and monthly series, as well as in their logarithmic transformations. Evidence of non-linearities are only obtained in the monthly series which may be a consequence of the higher degree of volatility associated to this frequency. The second and third chapters deals with the feasibility of monetary and currency unions in two different regions, namely Eastern Africa (Chapter II) and Central America (Chapter III). In Chapter 2 I make use of some other novel time series techniques to support the Generalized Purchasing Power Parity (G-PPP) and the Optimal Currency Area (OCA) theory in the East African Community (EAC). I provide reasons for the feasibility of a new common currency in the region and explain why new econometric techniques such as fractional cointegration can be a useful tool for policy makers when studying macroeconomic variables. Though not working towards an imminent transition to a monetary or currency union, the Central American Monetary Council (or CMCA, from Spanish Consejo Monetario Centroamericano) serves as an institution promoting economic and financial stability among five Central American countries (Costa Rica, El Salvador, Guatemala, Honduras and Nicaragua) and the Dominican Republic. Econometric studies conducted by researchers from CMCA have mostly focused on studying inflation levels of these countries, making use of econometric tools such as VECM and cointegration. In Chapter 3 I expand the study of inflation stability in the member countries of the CMCA by adopting a long memory and fractionally integrated approach and implementing cointegration methods that have not yet been used in the study of the Central American Monetary Council. The results first show that all the series of prices are nonstationary, with orders of integration equal to or higher than 1 in all cases. Looking at long run equilibrium relationships among the countries, I only found strong evidence of a cointegration relationship in the case of Honduras with El Salvador. In the final chapter, I conduct a fractional integration and cointegration study of several Kenyan electricity price series in order to determine whether signs of persistence or mean reversion can eventually be discovered. Such features can be considered as relevant when considering the possibilities of shocks affecting the energy market of Kenya, which has recently been subjected to major debate. Among the factors affecting electricity prices, I find oil prices and interest rates have significant positive effects on electricity, and based on the fact that all the series are I(1), long run relationships are examined by means of fractional cointegration. The recently introduced FCVAR model is implemented, with the results showing that the series under study are fractionally cointegrated, with oil price shocks affecting electricity prices.