Latent" Correlations of flows in the economy: spline analysis

The article proposes a new approach to the analysis of relationships that allows detecting "latent" correlations of flows. The methods of classical econometrics show some limitations that make it difficult to find correlations between decelerations or accelerations of growth. The disadvantages of the existing methods include the procedures for smoothing the empirical dynamics, which distort the information about the real acceleration or deceleration of growth at the "nodal" points. The paper also points to the absence of a time sequence in the models of communication, which makes it difficult to find possible regression switches. As an effective alternative, it is proposed to model the dynamics of the compared processes using cubic splines. The spline approximation preserves the empirical dynamics with zero error in the models under construction. By differentiating spline models, the transition to continuous flow velocity trajectories is performed. By comparing the flow velocity trajectories, correlations that are" latent " for classical econometrics tools are identified, and regression switching points are determined. The analysis is carried out on the example of the dynamics of oil export flows.