Are used vehicle prices falling or not?

According to our March 2017 Guidelines publication, prices fell significantly in February and have fallen for 8 straight months. This index looks at the marketplace in a particular way. It is based on vehicles that are 8 years old or younger (selling wholesale) and is designed to mimic the price of a fixed pool of vehicles through time, adjusted for seasonality and depreciation. Other publications and other methods of looking at the marketplace have different results; the key being methods that look at average dollar values in general. This is critical, because average prices have been relatively stable.

The stakes are very high. In February, for example, our AuctionNet database—which includes most of the wholesale transactions in the United States—recorded $6.7 billion in sales. And this is only a tiny fraction of the value of used vehicles on the books of financial institutions across the country.

So are prices falling, or are they stable?

The “strange but true” answer to our question is this: Both things are happening. Prices for individual vehicles (e.g. the 2014 Honda Civic EX) are falling, but average prices of all vehicles are relatively stable because the composition of the vehicle pool is changing. Companies and individuals holding vehicles as assets or collateral, and tracking their value, are experiencing declines, while companies and individuals watching the flow through the marketplace—dealers, auction houses, potential car and truck buyers—are seeing stability. While the value of individual vehicles is falling, the pool of vehicles is becoming richer.

Answers lay in the details.

Our price index looks at the marketplace in total, but it is difficult to really understand what is going on without more detail. The index value fell by 3.6%[1] in February. Specifically, the prices for all 8-year-old or younger vehicles fell by an average (weighted by volume) of 1.4% in February. After removing 1.2% of normal depreciation (the impact of aging and normal mileage accumulation) equals a 0.02% decline. Based on our statistical analysis, the normal seasonal price change in February is 3.4%. Because we expected prices to increase after adjusting for depreciation by 3.4%—and they actually declined by 0.02%—the seasonally adjusted change is the difference between these two, or 3.6%. Figure 1 shows this seasonal component in more detail. Over the last 20 years, the typical movement in prices (after removing normal depreciation) is in the +1% to +5% range, and averages approximately +3%. Note this February is significantly out of line with history.


At the same time, a weighted average of prices at auction seems to give a different story. Figure 2 gives the weighted average of prices, seasonally adjusted, for vehicles in the same age range as our index (vehicles up to 8 years old). The series has fluctuated up and down within 4% bounds since plateauing at the end of 2014. While by this measure prices were still significantly down in February, the stability that preceded February isolates it as an anomaly.


The critical change to the composition of the used vehicle pool in this context is not the mix of cars versus trucks, or the increased volume of utility vehicles. All of these fluctuations are occurring, but most, if not all methods of measuring prices, control for them. The mix of ages, however, is the major, often unaddressed change to the used vehicle marketplace that is supporting average price measurements as the prices of individual vehicles are dropping. Figure 3 gives the average age of vehicles at auction in years. Please note that as the quality and durability of vehicles has increased, the mix of vehicles has moved older over time. This accelerated after the financial crisis as a direct result of the crash in new vehicle sales in 2008 and 2009. The situation is now reversing and strong new vehicle sales are leading to a significant decline in the average age at auction.


Figure 4 helps bring this into perspective a little better by showing auction volume ratios of the 1- to 5-year-old age group over the 6- to 10-year-old age group.  In January 2007, the ratio was right at the historical average of 2.4. This means the auctions sold 2.4 vehicles in the 1- to 5-year-old range to each vehicle in the 6- to 10-year-old range.  Following the financial crisis this plummeted to almost even, but has since reversed and moved back toward 2 to 1.


To understand how this impacts the way indices and other aggregations of the market are calculated, consider an example.  Suppose the market consists of only two vehicles and we are measuring price movements over 6 months.  In the first three months, vehicle A had a volume of 1,000 units, while vehicle B had a volume of 50.  In the second three months, vehicle A’s volume shifted down to 500, while vehicle B’s volume shifted up to match (also 500).  Now suppose that vehicle A is a 2-year-old vehicle starting at a price of around $30,000, while vehicle B is a 6-year-old vehicle starting at a price of $10,000.  Suppose the price of both is declining by 1% each month.  You can see in the first three months the weighted average of the two would be very close to $30,000 and in the last three months the average would shift down to around $20,000.  This shift is simply because of the change in volume, and had little to do with the minor price declines of both.  To create an index, analysts use different methods to adjust away the influence of this shift in volume.

A little more about our index.

Because our index is specifically designed for those who are tracking the value of a portfolio, or any other relatively fixed set of vehicles, it is based on the price series of individual Model Year/Model/Style combinations (e.g. the 2015 Honda Civic LX), which are indexed before aggregation. With this design, the influence of the mix of vehicles in terms of age or segments is significantly reduced. Using the same example, we would create two series. One is for vehicle A and one is for vehicle B. They are both indexed to the current price of each. As such, we would have two series with a current value of 100 and both declining by 1% each month. The two identical series are then aggregated. When the mix shifts, the influence is felt on the rates of change rather than on the aggregated dollar prices. This limits the bias considerably, and we have observed that it limits the bias more than mix adjustment methods.

The shift in the mix of ages has a huge influence on aggregated dollar value measures since younger vehicles are more expensive. Approximately 80% of the difference in our index and the weighted average auction price can be explained by these movements in the average age at auction over the years. That stated, the largest shift in the index’s 20-year history has taken place over the last two years.

There is one reason for caution and one silver lining hidden behind the February numbers. The reason for caution is simply that it is only one month. While the previous seven months also declined, February stands out and contained about half of the total decline over the last eight months. A significant part of this finding depends on the specific seasonal adjustment used, and the seasonal pattern has been very difficult to pin down. Initial results for March do not show a correction to February, but are actually down again (albeit by a tiny amount with one full week of auction results to go). Last year, the seasonal swing bulged out into May and June. If this happens again in 2017, the February results will look less severe in retrospect.

The silver lining is buried within the prices for different ages. While the volumes have increased significantly for vehicles in the 1- to 3-year-old age range, in terms of price these vehicles have actually out-performed their elders. Based on a different index calculation we maintain that looks at ages individually, 1- to 3-year-old vehicles have declined in price over the last year by about 5% compared to a nearly 10% decline for 6- to 8-year-old vehicle. See figure 5 for more detail.


Our closing thoughts.

As we mentioned before, the J.D. Power index is designed with a very specific purpose in mind, which is to identify the price movements felt by companies that have a portfolio of vehicles held as assets, and in many cases sitting on inventory lots. To accomplish this, removing the influence of vehicle mix is critical. J.D. Power produces price indices and price series that look at the market place in many alternative ways using our comprehensive wholesale and retail transaction datasets. Our message to investors, remarketers, rental companies, and other companies is this:

The price index is a good overall barometer of the movement in prices for a fixed pool of vehicles, but it is always better to look closer. A subprime lender, for example, would find more specific information by looking at an index or aggregated price series weighted to match their portfolio of loans on older vehicles, while a rental company would need to focus more on 1- and 2-year-old vehicles. Look at the index to get the overview, but take action on measurements tailored to your situation.

[1] The March 2017 Guidelines publication quoted a 3.8% change. The 3.6% change reflects the most current result, and any change is due to delayed sales records entering the database.