Global Hemispheric Temperature Trends and Co–Shifting: A Shifting Mean Vector Autoregressive Analysis

This paper examines trends in annual temperature data for the northern and southern hemisphere (1850-2010) by using variants of the shifting-mean autoregressive (SM-AR) model of González and Teräsvirta (2008). Univariate models are first fitted to each series by using the so called QuickShift methodology. Full information maximum likelihood (FIML) estimates of a bivariate system of temperature equations are then obtained. The system is then used to perform formal tests of co-system in the hemispheric series. The results show there is evidence of co-shifting in the temperature data, most notably since the early 1980s.