Correlações em séries temporais de preços de frango, soja e milho

Abstract

With the evolution of the agricultural market, the process of production, export and consumption of food commodities has changed. Given this scenario, food prices can be affected by several factors, such as the energy market through strategies that even divert food cultures for the production of biofuels, causing interest in a better understanding of these relationships. In recent years, many studies were developed on the relationship between the food market and other markets, seeking to explain the link between the prices of different commodities with the prices of agricultural commodities (raw materials). However, Brazil still needs more attention in its food market. Here, the objective was to investigate intrinsic long-term correlations between Brazilian food markets, using Econophysics techniques. The daily series of price returns for chicken, soybeans and corn were analysed for the period from 02/02/2004 to 06/16/2017, obtained by the Center for Advanced Studies in Applied Economics / Escola Superior de Agricultura Luiz de Queiroz / Universidade of São Paulo - CEPEA / ESALQ / USP. Chicken prices depend mainly on the cost of the feed, which includes corn and soy as a source of energy and protein, respectively. The correlations were analysed using methods the Detrended Cross Correlation Analysis (DCCA) and the correlation coefficient associated with it and the recently proposed Detrended Partial Cross Correlation Analysis (DPCCA) useful to quantify the intrinsic cross correlations between two non-stationary time series. The results point to the absence of cross correlations for temporal scales up to 30 days. The intrinsic correlations presented by the DPCCA between chicken and corn price returns are stronger than the correlations between chicken and soybeans, especially from 250-day scales, signalling that the interactions between the markets for these commodities are greater in the long run. Furthermore, it was observed that after the 2008 crisis, the correlations decreased for temporal scales up to 200 days.