Production versus circulation: An empirical investigation of causal relationships in the U.S. economy

This paper examines competing theoretical frameworks in economics regarding causal relationships between production and circulation variables. Using three complementary methodological approaches—Error Correction Models enhanced with multivariate adaptive regression splines (ECM-MARS), Bayesian generalized linear models with AR(1) structure (BGLM-AR1), and Bayesian Structural Time Series (BSTS)—we analyze U.S. monthly economic data from January 1971 to December 2019 to determine whether production variables (capital stock, investment per worker, labor productivity) fundamentally determine circulation variables (exchange rates, interest rates, inflation) or vice versa. Our findings reveal limited but consistent evidence supporting the Marxist classical perspective: while the stringent temporal stability requirements eliminate most relationships, those that persist show production variables predicting circulation variables with greater robustness than the reverse. The ECM-MARS approach identified one stable relationship (Real Social Consumption per Worker → Federal Discount Rate, 90% support), while BGLM-AR1 found Real Net Profit → Consumer Price Index with 80% support. BSTS models failed to meet the stability criteria entirely. These results suggest that policy approaches prioritizing circulation-oriented interventions while neglecting production fundamentals may have limited efficacy, though the scarcity of stable relationships underscores the complexity of macroeconomic causal inference.