SpendInfo
SpendInfo2001 is a revolutionary development in the modeling of household expenditure in Australia. No matter where you are in Australia people like to have a beer, go out to a restaurant or go on an overseas holiday, but knowing how much each region is likely to spend on these activities is what business really needs.
The amount that an average household spends in any area will be the result of many factors. Of course income will be one determinant, but other issues such as the type of households in the area will also be important. Imagine the difference in spending on baby goods between a retirement area such as Victor Harbor in SA and a new young family growth area like Cranbourne in Victoria. Geographic differences also play a part in the determination of spending patterns; in general you don’t need to spend as much on clothing in Darwin, or sunglasses in Hobart. Other important trends in spending patterns in Australia included lower levels of smoking amongst higher-educated households.
In fact, for as many different regions, for as many different household types and for as many different expenditure types, peculiar differences in spending patterns exist. The art of combining those trends is left to the cutting edge statistical tools developed especially for Spendinfo2001. Finally, the computing power caught up with the theory of estimating such as model, and it still took over 400hrs to estimate in computing time alone.
The raw data on which these estimates are based is found in an Australian Bureau of Statistics survey called the Household Expenditure Survey (HES). This survey released in early 2001 followed the spending characteristics of over 6800 households. Along with their spending behavior, a significant number of socio-demographic variables on each household were collected. The survey period stretched from a fortnight to a year. For everyday items such as groceries, the survey followed a fortnight’s expenditure whilst for items purchased less frequently, the amount spent were collected over longer periods.
Spendinfo2001 takes the information contained in this survey to model the likely spending behavior in smaller areas. The modelling process doesn’t rely on the very small number of people that are likely to have been surveyed in any given area, but rather exploits the similarities between households contained in the HES and those on the ground that we are interested in. For instance, assume that an area has many households with unmarried 25-year-old plumbers who rent, when modelling their spending patterns, not those households in the HES that are a perfect match are used. Instead we will take information from other plumbers, other unmarried people, other people that rent, other people that are around 25 years old, other people who earn similar amounts of money and other people that live in similar areas. From the amalgamation of these people an estimate is derived. Of course, determining how to select which people in the HES correspond and how much weight to place on each, is the difficult task. The process that does this is called micro-simulation, and in this case utilizes what is technically termed the modified simulated annealing estimation of a pseudo maximum likelihood function.
Spendinfo2001 estimates the spending characteristics for over 30,000 individual areas in Australia, called CCD’s or Census Collector Districts, and the results are aggregated into broader areas, such as postcodes. Within each of these 30,000 regions, the spending characteristics are modelled with respect to the approximately 200 households contained within each. Put simply, the micro-simulation process is one of matching; we need to match these 200 households as closely as possible with those contained in the HES. This match is based on over 70 characteristics of these households such as income, age distribution, and educational attainment along within information about the region in which the area is situated.
From a technical viewpoint, a number of additional processes are applied to the raw estimates from Spendinfo2001. National Economics believes that most accurate figures are obtained if they can be benchmarked to ABS National Accounts estimates, and to income relativities that can be obtained from Australian Taxation Office (ATO). As such, a significant amount of post processing is applied to the estimates so that practitioners can be confident that the National Economics Spendinfo2001 estimates match as closely as possible in aggregate with the nation or statewide estimates made by the ABS.
Although over 450 expenditure types ranging from nuts to hosiery are detailed in the HES and estimated in Spendinfo2001, broader aggregates or categories of spending types are presented in the general National Economics product. These broader categories are designed to reflect groupings relevant to the retail sector.
Reiterating, Spendinfo2001 answers the question; how much can you expect the households in the region to spend on various products and it does this buy consider who lives where.
Additional information available from National Economics in the Spendinfo2001 range includes the more detailed expenditure types and an exciting range of estimates varying by types of households.
For further information regarding SpendInfo please contact Peter Hylands at National Economics on (03) 9488 8444