Modelling Changing Rural Land Use in New Zealand 1997 to 2008 Using a Multinomial Logit Approach

Abstract

Rural land use in New Zealand is an important driver of economic activity and has clear implications for the environment, including for biodiversity, climate change emissions and water quality. The spatial distribution and aggregate shares of rural land use is always changing, but change occurs slowly. To better understand the drivers of rural land use change, this paper addresses three questions using the popular multinomial logit modelling approach. First, do recent commodity prices have any predictive power on land use conversions? Second, is recently sold land more likely to change use? Third, does land which is marginal between uses have identifiable characteristics? The data used consists of the New Zealand Landcover Database version 3 (LCDB3), with observations in 1997, 2002 and 2008; 6 year average profitability data for dairy, sheep and beef and forestry; QVNZ land sales data; and I control for land quality and Maori tenure. In answering the first and third questions, I evaluate the predictive power of a spatially explicit land use share multinomial logit model, estimated from 2002 cross-sectional variation. To supplement the land use share multinomial logit for questions one and three, and to address question two, I use a land use transition multinomial logit, estimating the likelihood of transition from a single starting land use between 1997 and 2002, similar to Lubowski et al. (2008). Finally, I compare the two modelling approaches.