Does temperature contain a stochastic trend? Evaluating conflicting statistical results

: Kaufmann RK, Kauppi H, Stock JH

Publisher: SPRINGER

: 2010

: Climatic Change

: CLIMATIC CHANGE

: 3-4

: 101

: 3-4

: 395

: 405

: 11

: 0165-0009

DOI: https://doi.org/10.1007/s10584-009-9711-2

We evaluate the claim by Gay et al. (Clim Change 94:333-349, 2009) that "surface temperature can be better described as a trend stationary process with a one-time permanent shock" than efforts by Kaufmann et al. (Clim Change 77:249-278, 2006) to model surface temperature as a time series that contains a stochastic trend that is imparted by the time series for radiative forcing. We test this claim by comparing the in-sample forecast generated by the trend stationary model with a one-time permanent shock to the in-sample forecast generated by a cointegration/error correction model that is assumed to be stable over the 1870-2000 sample period. Results indicate that the in-sample forecast generated by the cointegration/error correction model is more accurate than the in-sample forecast generated by the trend stationary model with a one-time permanent shock. Furthermore, Monte Carlo simulations of the cointegration/error correction model generate time series for temperature that are consistent with the trend-stationary-with-a-break result generated by Gay et al. (Clim Change 94:333-349, 2009), while the time series for radiative forcing cannot be modeled as trend stationary with a one-time shock. Based on these results, we argue that modeling surface temperature as a time series that shares a stochastic trend with radiative forcing offers the possibility of greater insights regarding the potential causes of climate change and efforts to slow its progression.