Causality Over Time and Economic Dynamics

6.1 Time and Causality

Economic phenomena unfold in time. While much of causal inference focuses on static comparisons for example, comparing outcomes between treated and untreated units at a point in time, real economic systems are dynamic: the effect of an intervention today may evolve, decay, amplify, or interact with subsequent events. Distinguishing static causal effects from dynamic causal effects is therefore central to understanding temporal causality. The difference between intervention and control in a single point is captured by the static effects, whereas the dynamic effects take into consideration that prior states and shocks affect future outcomes.

The temporal causal reasoning focuses on time sequence: an antecedent has to come before an effect. This ordering in dynamic systems demonstrates the propagation, accumulation or decays of shocks. As an example, an export-led economy can produce an upward-escalating cumulative and compounding growth in output and employment in the long run, which is consequently known as dynamic cumulative causation in the industrial growth literature (Dávila-Fernandez and Oreiro, 2023) [107]. On the same note, structural change and growth can also have the feedback of productivity improvement and investment factor reallocation in medium-term horizons (Cyrek, 2024) [108].

At a methodological level, capturing dynamic causal effects demands causal tools that explicitly incorporate time whether through time‑series models, lag structures, or longitudinal panel frameworks, which are taken up in the next sections.

6.2 Time‑Series Causal Models

Time‑series data are ubiquitous in macroeconomics, finance, and policy analysis. Classical regression approaches ignore temporal structure, but causality in time entails not just association but predictive precedence and temporal sequencing.

Granger Causality

A workhorse of temporal causal analysis is Granger causality, which operationalizes causality in a predictive sense: a variable X Granger‑causes a variable Y if past values of X contain information that helps predict future values of Y, beyond past values of Y itself. According to Shojaie and Fox (2022), Granger causality and recent developments in methodology are thoroughly reviewed, which also explains its use in revealing directional effect in multivariate time series and its use in economics and social sciences [103].

Granger causality extensions deal with heterogeneity and cross-sectional dependence between units. Nazlioglu and Karul (2024) create tests permitting to test Granger causal relations in panels in which cross-sectional consequences are not negligible - vital when examining macro panels of nations or industries [105]. There are practical examples: Spectral Granger causality is applied by Alola, Adebayo, and Onifade (2022) to investigate ecological footprint dynamics in China and it is evident that the direction of causality in different frequency domains and the magnitude of disturbances differs [109]. Xu and Zhang (2023) use linear and nonlinear causality tests of the information flows of house prices across Chinese cities and show that causality may differ depending on city pairs and over time [110]; similar studies investigate contemporaneous and dynamic relationships between residential housing markets [112]. Financial markets are not an exception: Elsayed, Gozgor, and Lau (2022) study the dynamic spillovers and causality between cryptocurrencies and currency markets and demonstrate the interaction between financial innovations over time [113]. We can represent it as below