Through a combination of family generosity, stinginess, and luck, I managed to go 35 years without doing business with a professional car salesman. So when the hour finally arrived for me to put on my big-boy pants and buy a car, I took the task seriously. I wanted to bring robust econometric analysis to bear on the car market. I wanted to become an automotive Bill James.
Buying used was a given—I’m not the kind of sucker who absorbs the depreciation that hits the moment a new car leaves the lot. Instead, my first important decision was to avoid sellers who negotiate price. The received wisdom is that outfits with “no-haggle” prices charge more than the best price you can get from a dealer who bargains. My assumption, however, was that I would be no match for a car salesman. I’m an amateur, and car salesmen do this for a living. If I went up against a pro, I would almost certainly pay more than the built-in mark-up at a no-haggle store. So I chose a no-haggle vendor called Carmax.
Carmax is a national chain of used car dealerships. You can search their entire inventory—tens of thousands of vehicles—online. The vehicles’ specifications, mileage, history, and price are all readily available to consumers. Since the prices you see are essentially final-sale prices, I treated their site as a data mine.
The first bit of gold I found was a market failure. I had initially targeted the Ford Fusion as my prospective car. It’s a sensible family sedan with excellent reliability and acceptable performance. Most important, it’s an obvious value-play compared with the other cars at the top of its class, such as the pricey Toyota Camry and Honda Accord.
I started tracking Fusions in early 2009. In July, something odd happened: The prices on used Fusions started going up, an after-effect of the Cash for Clunkers program. This bit of government interference greatly benefited new car buyers but drove pre-owned prices sky-high. (Thanks, President Obama!) I was waiting for Fusion prices to drift back to earth when I noticed the Mercury Milan.
The Milan and the Fusion are corporate twins. The cars are made by the same parent company on the same design platform in the same manufacturing plant using the same components. The only minor difference is that the Milan has a slightly upgraded interior, which makes it a tad more expensive new.
But on the used market, it was a different story. Used Milans were anywhere from 10 percent to 20 percent cheaper than equivalent Fusions. Why? The only explanation was that Ford spends lots of advertising dollars on the Fusion, but none on the Milan. The market’s ignorance was my bliss.
With the Milan as my new target, I concentrated on finding the optimal model year/mileage value point. I panned Carmax’s website for every variation of the Milan and created a scatter graph for each model, plotting price against mileage. Running a best-fit curve for each of these graphs, I examined the slope of the curves, looking for the moment where the dx/dt—the rate of depreciation—began to flatten out.
To optimize value you want a car that sits at the precise moment where the rate of its value loss most quickly decelerates. I found it, with a large enough data set to give me confidence in my numbers.
For good measure, I repeated the exercise with nearby model years—looking at 2007, 2008, and 2009 Milans—and then plotted those curves on top of one another to calculate how much of a price premium each successive year commanded. (The answer is that the model-year premium is not constant. It varies depending on where you are on the mileage curve. Assuming, of course, that you’re within the same design cycle.)
Eventually, I found my dream car—which is to say, the car I was convinced represented the best value. And the deal Carmax was offering was better than it should have been. Much better.
On the scatter graph, the car was a gigantic outlier, with a price significantly lower than my calculations said it should be. As I examined the data point, I saw that the car was loaded, too. It had a leather interior, a moonroof, and all-wheel drive.
In fact, it was such an outlier that its superior value would have leapt out at anyone even without my statistical analysis or asset depreciation calculus. What I couldn’t understand, staring at that mountain of data, was why the price was so good.
It was only when I finally called up a picture of the car that I saw in a flash: There was one factor I hadn’t accounted for—a factor so powerful that it delivered more value than all of my sabremetrics and market failures put together.
My new car is purple.
Jonathan V. Last
