Time Series, Nonsense Correlations, and the Principle of the Common Cause

Time series, that is, time-ordered sets of observations on a random variable or random variables, are of fundamental importance for empirical inferences in sciences as diverse as neurophysiology, climatology, epidemiology, astro- and geophysics, and many of the social sciences. In this chapter I shall argue that a number of particularities of time series pose serious difficulties for one of the most prominent kinds of account of causal inference: probabilistic theories. A core assumption of probabilistic theories is the principle of the common cause, according to which a correlation between two variables is indicative of a causal connection between these variables. ‘Nonsense correlations’—i.e., correlations that are artifacts of the statistical properties of the variables or that obtain for other non-causal reasons—pose an obvious problem for probabilistic theories.