Seasonality as an unobservable component in South African agricultural market data

Abstract

The shortcoming of most of the tests for seasonal patterns is that the problem under investigation is formulated in a stringent manner, leading to a test of the null hypothesis of no seasonality against the alternative of deterministic seasonality. These tests would typically involve using models that incorporate deterministic dummy variables that are then used to capture seasonal effects in the data by means of a regression. This implies that the possibility of stochastic seasonality, which is manifested by changing seasonal factors over the sample period (deterministic seasonality implies constant seasonal factors), is ignored. A possible more comprehensive/rigorous approach would thus be to test for the presence of stochastic seasonality versus deterministic seasonality. The aim of this paper is to investigate the possible presence of unobserved seasonal components in South African agricultural market data, with the emphasis to demonstrate the additional insight such an approach can provide to researchers and especially for pairs traders