Vectorization and econometric model simulation

Linear analysis of nonlinear models, stochastic simulation, evaluation of forecast errors and optimal control are examples of econometric analysis that require many datasets to be solved for the same model. In this paper, we present an algorithm which demonstrates significant gains in solution speed-up and cost reduction, using vectorization, a technique implemented on supercomputers. A loop over datasets is made around each model equation. Here, it is very important to allow for computer memory constraints in the design. The Gauss-Seidel solution procedure requires little addition effort. But at least two options are available for vectorization of the Newton-Raphson algorithm. Measurements indicate that a performance improvement of 7–8 times can be achieved from vectorization