Spatial variation of the food price index in Canada's Northwest Territories: a ridge regression analysis

Abstract

This thesis investigates the spatial variation in the 1985 Food Price Index (FPI) at destinations in the Northwest Territories (NWT). It follows a conceptual model that assumes the FPI varies directly with the accessibility of a given northern destination to the metropolitan South. Accessibility has been defined in many ways in the geographical literature, and other studies of the region have produced results that suggest a diversity of influences are operating on consumer prices. For this reason, a multivariate analysis is employed to test a wide range of variables. The analysis demonstrates how sensitive is the response of statistical methods, in this case multiple regression analysis, to relatively minor changes in the definitions of some variables from a previous study. It also suggests that the wide range of interpretations sometimes observed in the literature pertaining to high prices in the NWT, may well be the result of a combination of this effect and that of inherent multicollinearity among the variables. For this reason it is felt that at the present stage of the investigation of northern food prices, priority must be given to constructing clear, unambiguous statistical models before further interpretive analysis can be meaningful. Procedures to deal with multicollinearity exist, and the choice of the present study is the ridge regression technique. The linear model of northern accessibility that it provides is effectively free of multicollinearity. It posits the cost of transportation, the availability of a highway link with the South, and the 'willingness' at the destination to interact with the South as the predictors of the FPI in descending order of importance. Its predictions account for almost 89 percent of the variation that is observed in the 1985 FPI. Its interpretation suggests that in 1985 the accessibility of NWT destinations had already been substantially optimized.