The main objective of this study is to investigate the nature of price relationships existing amongst geographically separated agricultural commodity markets in Malawi. Market liberalisation provides the overarching framework on which this investigation is based. The liberalisation of commodity markets is assumed to facilitate the integration of geographically separated markets, increase gains from trade and the general welfare of the cOuntry. The investigation is performed using modern cointegration techniques. The cOintegration techniques applied in this thesis are based on the concept of spatial price equilibrium (SPE) initially proposed by Samuelson in 1952 and threshold cointegration techniques. The two categories of cointegration techniques, linear and non-linear, have been ) applied to the analysis of dry beans and maize markets in Malawi. For linear cointegration techniques, the Johansens' (1988) model and the Engle-Granger (1987) model have been applied (see chapters 6 to 8); while for non-linear model techniques, two types of threshold cOintegration (TAR and M-TAR) models have been applied (see chapter 8). Both linear and non-linear models investigate different aspects of price behaviour in the dry beans and maize markets. The analysis uses monthly price averages for the January 2000 and January 2012 period. The dry beans and maize markets have been selected for analysis because of their distinct characteristics. The most prominent distinguishing characteristic between them is that dry beans markets are fully liberalised while maize markets are partially liberalised. The dry beans market fully satisfies the key assumptions of the market liberalisation concept. Therefore, the a priori expectation is that the dry beans market is perfectly integrated. The principal maize market is not expected to be perfectly integrated because it is not fully liberalised. These distinctions between the two markets provide some sort of a 'natural experiment' on which the assumption that market liberalisation facilitates market integration can be tested.