DP18114 Heterogeneous Predictive Association of CO2 with Global Warming
Global warming is a non-uniform process across space and time. This opens the door to a heterogeneous relationship between CO2 and temperature that needs to be analyzed going beyond the standard analysis based on mean temperature found in the literature. We revisit this topic through the lenses of a new class of factor models for high-dimensional panel data, labeled Quantile Factor Models (QFM). This technique extracts quantile-dependent
factors from the distributions of temperature across a wide range of stable weather stations in the Northern and Southern Hemispheres over 1959-2018. In particular, we test whether
the (detrended) growth rate of CO2 concentrations help predict the underlying factors of the different quantiles of the distribution of (detrended) temperature in the time dimension. We document that predictive association is greater at the lower and medium quantiles than at the upper quantiles and provide some conjectures about what could be behind this nonuniformity. These findings complement recent results in the literature documenting steeper
trends in lower temperature levels than in other parts of the spatial distribution
