Does sunspot numbers cause global temperatures? A reconsideration using non-parametric causality tests

• A non-parametric Singular Spectrum Analysis based causality test is proposed.
• SSA-based causality test outperformed time and frequency domain causality tests.
• The new non-parametric technique can capture the possibly existing nonlinearities.
• Predictive ability is detected from sunspot numbers on global temperatures.

In a recent paper, Gupta et al., (2015), analyzed whether sunspot numbers cause global temperatures based on monthly data covering the period 1880:1–2013:9. The authors find that standard time domain Granger causality test fails to reject the null hypothesis that sunspot numbers do not cause global temperatures for both full and sub-samples, namely 1880:1–1936:2, ​1936:3–1986:11 and 1986:12–2013:9 (identified based on tests of structural breaks). However, frequency domain causality test detects predictability for the full-sample at short (2–2.6 months) cycle lengths, but not the sub-samples. But since, full-sample causality cannot be relied upon due to structural breaks, Gupta et al., (2015) conclude that the evidence of causality running from sunspot numbers to global temperatures is weak and inconclusive. Given the importance of the issue of global warming, our current paper aims to revisit this issue of whether sunspot numbers cause global temperatures, using the same data set and sub-samples used by Gupta et al., (2015), based on an nonparametric Singular Spectrum Analysis (SSA)-based causality test. Based on this test, we however, show that sunspot numbers have predictive ability for global temperatures for the three sub-samples, over and above the full-sample. Thus, generally speaking, our non-parametric SSA-based causality test outperformed both time domain and frequency domain causality tests and highlighted that sunspot numbers have always been important in predicting global temperatures.