| An advanced index calculates a value for different properties |
|
Markus Halonen, Manager, Statistics and Data Analysis, Grey-Hen Monitoring the price level of used cars can be challenging because the variety of the vehicles makes it difficult to define the prices. So-called hedonic indexes also note the properties that have an effect on the price of the vehicle. Statistics Finland monitors the price development of different commodities and calculates the consumer price index based on that data. This index also includes used cars. The current method for the price index of used cars was developed at the end of the 1990's, and yours truly was one of the members of the Eurostat team who produced it. 
Defining trade-in prices based only on categorisation is inaccurate. The so-called hedonic indexes lead to a better result.
Challenge lies in the variety of trade-ins With homogeneous products, i.e. products that have very little differences in quality, it’s rather easy to follow up on the price developments. For example, the price development of skimmed-milk can be followed by simply checking the price of the milk at certain stores at regular intervals. Milk is such a homogeneous product that it’s easy to form price relations between two separate months and then calculate the index based on these relations. Used cars are a very differentiated and non-homogeneous product; there can be various kinds and the prices vary. There are over 10,000 versions of used cars on the market at a time and the accessory levels are different. The age and mileage of the vehicle have an effect on the prices, and so does the area where the car is being sold. The prices of a particular vehicle can’t be monitored from one month to another, for the same vehicle is usually not on sale on consecutive months. Even if they were, the ageing of the car can lower the price, so the prices of an individual car do not describe the general price development of similar vehicles. If you count the average price of all vehicles on sale in a month, then the average price is highly influenced by the type of cars on sale. So the changes in average prices describe both the development of the quality of the cars on sale as well as the development of the price level. For a price index, we need to be able to differentiate the pure price level change from the change in quality, and so the index should give us the pure price development. Categorisation is not enough We get to measure price development more accurately, if we categorise a non-homogeneous product and do it so that as much as possible of the price dispersion is between the categories and as little as possible within them. So within a category, the product is homogeneous and the development of the average price is quite well representational of the pure price development of similar vehicles. A reasonable price index is reached from monitoring the price development of each category and calculating a weighted average based on those developments. However, there may be problems with this method as well, if the categorisation is not successful and there is dispersion within the groups so that they are not internally homogenous. For example, if the categorisation is done by make, there is still plenty of variation in prices within the make, due to age, mileage, and model and version properties. So the categorisation must be more accurate than this to give precise results. More accuracy from hedonic indexes An even better way to solve the problem of quality differences is to use statistical modelling, where the price of a commodity is explained through properties and time periods. We get an estimate of the effect of each factor on the price and we can evaluate the statistical significance. So we also get an estimate on how much the expected value of a price of a similar car has changed on average from the previous month. Such indexes that utilise statistical modelling in standardising differences in quality are called hedonic indexes. This means that the value of a differentiated product such as a car is though to be comprised of the properties of the car, such as engine power, make, fuel consumption, and size. This way we can model the connection between the properties of a car and the price. The price development of homes was calculated with categorisation still in the mid 1990’s. However, in my master’s thesis it was shown that the categorisation used was internally non-homogeneous and a more accurate result could be gained from hedonic price models. Statistics Finland then changed the method of housing price index to the hedonic index, the first part of which was developed as part of the thesis. Price forecasts make leasing easier An example of the end uses of these indexes is the leasing contracts of financing companies. Financing companies should be able to measure and predict the price developments of different types of cars. Let’s say that a leasing company sells a three-year-old Skoda Octavia 1.6 Ambiente HB at the end of the agreement with a mileage of 90.000 kilometres and it sells these kinds of vehicles each month. Right now, the expected value development or the price index of these vehicles tells the leasing company how much more or less it can expect when it sells that car that has a similar type, age, and mileage. So the prediction of the future prices is on the other hand a question of predicting the price index development of that particular model in that age group. The official price index for used cars is calculated at Grey-Hen with the method developed by the Eurostat group, and then delivered to Statistics Finland. It’s a chain index, so the level is always compared to the previous month. The monthly changes in the price index are linked to an index series. Grey-Hen has also developed group-specific sales price indexes as a variation of the previous model. Even with quite small vehicle groups it’s possible to make out quite accurate expected value developments for prices. This requires that the expected values for prices have been calculated for different types of cars. This work is continuous. By statistical modelling of price data we can form price estimates for vehicles, i.e. expected values for prices of the cars. More advanced cars with the same money By monitoring the change in expected value of a particular type of car on the market we receive the pure price change. We can calculate the change in expected value for an individual vehicle as well; for example by comparing the example Skoda Octavia’s expected values in February and March. By calculating the changes to a large group of individual vehicles on sale and the calculating the averages, we receive an estimate of a price development of that particular group. The development of average prices can in the long run be quite different from the price development of a particular car. The markets constantly receive better and better cars. That is why the development of the average price is usually above the price index curve and ascends more or descends less. The difference comes from the effect of the quality change. The same phenomenon has been clear in the price development of new cars, measured by Statistics Finland. Due to the tax cuts in the beginning of 2003, you could get a new car approximately 7 % cheaper than at the end of 2002, according to the index from Statistics Finland. However, the average price of new vehicles did not change much from 2002 to 2003. The car buyers took advantage of the lower price level, and chose better and often bigger cars, so the average price did not change significantly. The same kind of development can be expected after the latest taxation changes, as the sale of more expensive diesel vehicles has grown compared to petrol models. In 2008, the average price of new cars should not be significantly lower than in 2007, though the same kind of new car is clearly cheaper than it was.  The sales price development of cars aged 3-6 years, according to body (12/2002=100). The price development of vans has been significantly different from passenger cars for several years now.
There have been great differences in standardised price development between models in the past three years with cars aged 3-6 years.
 This article was published in Suomen Autolehti magazine. |
|
