The paper qualitatively approximates the conceivably asymmetric dynamic relationships among GDP growth, inventory growth and three types of aggregate demand growth (internal public demand, internal private demand and external export demand) during the Peruvian experience of market-based growth (1993-2010). In order to capture the potentially important asymmetries in the conditional mean vector, a particular flexible dynamic model (neural VAR) is proposed. The model's perturbation vector is distributed multivariate Student-t distribution with conditional scale matrix following an autoregressive conditionally heteroskedastic (ARCH) process. The parameters of these conditional objects are then robust to the presence of outliers, which reduces the misspecification arising from spurious asymmetries in the conditional means. After covering the required computational cost, the approximated parameter estimates could reveal incentives to maintain inventories which are additional to the traditional production smoothing incentive. A statistically significant parameter estimate inside the contemporaneous structural matrix points out that a positive shock in private demand growth will be absorbed mainly by a more than proportional increase in the production growth on impact. This amplifier impact from demand shocks to production evolution is consistent with the numbers advanced on the average incidence of inventory investment over production growth during the four most recent recessions. This average incidence can then be explained by the aggregate inventory cycle (not necessarily of final goods).