| The intricate world of used car prices |
|
Managing Director Mikael Teerilahti, Grey-Hen How many car sale examples are enough to ensure reliable market price calculations? How come the Internet doesn't give the correct answer?  Price observations made online concentrate strongly on the most common models of the most common makes. However, half of sales come from vehicles that are not part of this group.
To be able to calculate prices as arithmetic averages to a significant part of models, the size of the database should be millions of cars. Even then, the effect of age and mileage on prices would not be clear for most of the vehicles. Internet gives us the wrong impression We are used to getting a general impression of the price level by checking car prices on the Internet. Before, the source was magazine adverts. Another essential source of information is the own experience from actual purchase and selling prices. The number of own sales is much too small to help form a view of the general prices. It can sound surprising, but also the amount of data online is quite limited in relation to the number of different vehicles on the market. Even the largest Internet search service has only enough data to calculate the prices for the most common cars, and even then only the asking prices. The prices should be calculated or evaluated by model and market area. In addition, it’s necessary to have a clear picture of the effect of mileage, as well as age up to the accuracy of months. After all, a car from December is over ten percent more expensive than one from January. The issue can be illustrated. The following chart includes all BMW 316 vehicles from years 2003-2005 on sale at the same time in the Autotalli web service. By observing the list, we get a typical idea of the price level of that particular model. The models from 2005 cost between 28,000–27,000 euros and vehicles from 2004 between 27,000–23,000 euros (most expensive 28,900 and the cheapest 19,900). The models from 2003 cost between 23,000-21,000 euros. Next we should evaluate the effects of mileage. For example, prices for cars from 2004 seem to have no logic for the effect of mileage. With human reasoning we get a result that cars with a low mileage cost 27,000 and with a high mileage 23,000. This result is thus very superficial. The aim is to clarify the reason behind the asking prices of different vehicles in the chart. This can’t be done by calculating the group average and comparing each car to that, because the cars have very different properties. Therefore, the column PI has each car’s average price noting the age and mileage, calculated with statistical methods. The column PE shows the difference between the asking price and the statistic average price per vehicle. This calculation, or the reasoning, has not been able to note from which area the cars are from or which month they were registered. The effect of these factors may have been over 1,500 euros to either direction, each separately. The major flaw in Internet data is that it lacks the cars that are sold quickly, and these cars tend to be cheaper than the average. It is also common that cars more expensive than average are in the end sold with a much lower price than what was originally given. In other words, cars that are more expensive than the average are overrepresented online. The column AV includes the prices from a calculation where the overpricing based on selling times has been eliminated. The difference between the online asking prices and the corrected prices in this example group is 3.7 percent and 857 euros. It might not seem like such a great difference, but it must be remembered that average profit is only a fraction of those numbers. General price level data is not enough Below is an example of all BMW 3 series models from 2000-2005 in Autotalli.com search service. The size of the database was then around 38,000 cars. All 3 series BMWs are placed in the graph so that the vertical column includes different models and horizontal column includes the years. The graph has 73 lines for models and 6 year columns so that makes a total of 432 cells. There were 280 examples in total. The cars are not divided into market areas or age as we don't have that data. So each model line and age cell should be thus divided into 20 areas and at least ten year and mileage groups. Then, the examples should be divided to these small cells, so that the accurate effects of age and mileage could be calculated as averages. Most of the examples are from the common models. Still over half of cars are such where the same model has been sold only sporadically. Half of all sales are made up from the less common models, so it's not enough just to know the prices of common cars. The large national trade-in search services in Finland nowadays include around 30,000-40,000 vehicles. To the calculations it makes no real difference whether that number is 30,000 or 50,000. Most models still remain without a reliable result. The Finnish car collection contains around 20.000 different models and the examples are distributed unevenly between them. They are focused on the most common models of the most common makes. The number of examples for the most common models enables the producing of national annual average prices for average mileages (depending on the model 17,000-25,000 km / year). However, the area or month specific prices (both have an effect of over 10%) are not available even to the most common models without methods from statistical mathematics. The inaccuracies are too large for the results to be usable in pricing of cars or especially in pricing management and monitoring. The same inaccuracy is reached through simple reasoning, without any kind of calculations. 
All BMW 316 cars from years 2003-2005 on sale at the same time in the Autotalli.com search service. The difference between the online asking prices and the corrected prices is 3.7 percent and 857 euros in this example group.
All examples of BMW 3 series vehicles years 2000-2005 from Autotalli.com search service. The examples do not provide enough information for pricing of cars.
 This article was published in Suomen Autolehti magazine. |
|
