1 Introduction
Consumption expenditure is the largest component of total spending in the New Zealand economy, accounting for around three-fifths of expenditure based gross domestic product (GDP) since the late 1980s. Despite a growing international literature on consumption modelling, past research on New Zealand consumption models is scarce. The main New Zealand research is McDermott (1990), Corfield (1992) and Rae (1997). One shortfall of all three studies is the inadequate measurement for the wealth variable. McDermott (1990) and Corfield (1992) used housing value, M3 money supply and equity prices as proxies for gross wealth, but no adjustments were made to account for household debt. Rae (1997) derived a net wealth measure based on data from various sources. This is not a criticism of the earlier studies. Comprehensive measures of household assets and liabilities were not available at that time, and the authors had to improvise with whatever data were available.
This paper presents two models of consumption, building upon earlier work done by Downing (2001). The primary purpose of this consumption modelling is for forecasting. While it is possible to set up models to test various hypotheses, such as the impact on consumption from rising house or equity prices, that is not the main intended purpose.
An improvement over earlier New Zealand consumption models is the use of the household wealth information from the WestpacTrust Household Savings Indicators (HSI). In addition, new variables are introduced into the modelling to capture short-run variations in consumption growth specific to New Zealand, aimed at addressing some of the criticisms cited by Hendry et al. (1990), who argue that the basic formulation for most consumption functions used for forecasting omits five potentially important influences on consumer behaviour. These are income uncertainty, credit constraints, demographic changes, liquidity, and dynamic adjustment.
The unemployment rate, common in the consumption literature, is used to capture income uncertainty. New Zealand households hold a disproportionately large proportion of their assets in housing compared to other industrialised countries. Households are increasingly borrowing against the value of their housing assets for consumption or other purposes. Introducing a mortgage equity withdrawal variable is aimed at capturing the credit constraints and liquidity elements. Due to New Zealand’s small population, net migration has a larger impact on population growth than other industrialised countries. Modelling consumption on a per capita basis only addresses part of the net migration impact. A net migrant transfers variable, which measures the amount of funds brought into the country, is introduced to capture the consumption impact from net migration.
Data availability restricts the analysis to the 1989:4 to 2002:1 period. This restriction does qualify any conclusions from the modelling. However, one advantage of the restricted time period is that the data series are unlikely to suffer from any structural breaks, as most of the economic reforms and financial deregulations in New Zealand occurred in the mid-1980s. The tests performed on the models support the absence of structural breaks.
This paper is structured as follows. Section 2 discusses the modelling approach and model specifications, including data choice and properties. In Section 3, the empirical results of the models are presented. Out-of-sample forecast comparisons from the models are also presented in Section 3, along with a historical decomposition of New Zealand consumption growth over the 1990s. Concluding remarks and possible extensions are contained in Section 4.
