The single biggest way to save money on a used Nissan Leaf? Don’t buy from Carvana.
That’s what the data say. Other results are in the RESULTS section below.
Hedonic price function
I didn’t make up the term. A hedonic price function is the method economists use to parse out the factors that contribute to the market value of an object.
Basically, you take what everybody knows — newer cars cost more, high-mileage cars cost less — and you quantify it. You calculate a data-based estimate of just how much those factors affect the price of the object in question.
Probably the single most important use of this approach is in the U.S. Consumer Price Index. That index is no longer a simple average of observed prices for various goods. That’s because the qualities of the underlying goods keep changing. Take computers for example. A new PC is typically far more capable than a ten-year-old PC. If you simply lump those two together as “PCs”, you ignore the increase in the value of what the consumer is buying. So, the Bureau of Labor Statistics uses hedonic price estimates to determine how much “better” newer models are, compared to older models, and adjusts the prices of newer models (down) accordingly. The other way to look at that is that this is how Uncle Sam systematically understates the true rate of inflation. And, accordingly, saves money on all of its payments linked to inflation, such as Social Security payments.
Practically speaking, I “ran a regression”. Which is, at root, just an average. Instead of taking a simple average price, it’s an adjusted average. It adjusts the average asking price for the factors that you have identified (e.g., age, mileage).
The nice thing about it is that it gives you the independent effect of each factor. So, newer used cars also tend to be lower-mileage used cars. To what extent is their higher price for late-model cars due to low age, versus low mileage? In theory, a regression parses that out for you.
Regressions, much like the Magic 8 Ball from which they were derived, sometimes give you a non-committal answer. By that I mean that there’s so little systematic impact of some factor that you can’t really say whether or not you’re just picking up random associations. The ones above that are egregiously so, I’ve put in faint gray type. You should ignore any estimates that don’t pass a standard hurdle for “statistical significance” this way.
(I hope it goes without saying that regressions may sometimes give a wrong answer, and may sometimes give a right answer. Again, much like the Magic 8 Ball.)
Results
By far, the most important factor for saving money on a used Leaf is not to buy it from Carvana. Averaged across all the cars in the dataset, buying from Carvana added $3K to the price of the car. I don’t think that’s news, as I vaguely recall that they had a reputation for being a convenient-but-relatively-costly source for used cars.
But just as interesting are the things that don’t matter.
The trim level or options level had no (statistically significant) effect on the price. This, I think, also matches conventional wisdom. As stated by my mother-in-law: Buy a loaded car because you want to drive a loaded car. Don’t buy it for the additional resale value, because there isn’t any. (But keep in mind, the effect of difference in range is accounted for elsewhere, and so differences in battery capacity across trim lines should not, in theory, contribute to the pure effect of the trim level.)
Curiously, history of having been in an accident also did not much matter. I have no explanation for that, except that maybe most of those aren’t major accidents?
Now we get to the big three: Year, Mileage, and Range. To interpret these, you need to know that the average asking price in this dataset was about $15,000.
All other things equal, each year that a used Leaf ages, as a used car, it loses about $900 in value, or about 6% of its value. See the note below on comparing this to the depreciation of a brand-new Leaf. It’s lower, but not vastly lower, than the depreciation of a new car. I think this matches our received folklore about car values. The fastest reduction in value that you will experience is in the moment you drive your new car off the dealer’s lot.
Mileage hardly matters. Every additional 10K miles on the odometer only drops the value of the car by about $200, or just over 1%. I think that’s quite different from what you see with gas cars, and speaks to the robustness of the motors and drive train. It appears that the typical buyer isn’t concerned with expensive repairs for a fairly high-mileage used Leaf.
Alternatively, these cars may simply have such low average miles, that mileage had not yet begun to matter. Median mileage was just 36K, median age was about five years. The low average mileage is a function of both low car age and few miles per year, consistent with a car in which long-distance travel was likely awkward for most.)
Slice that either way you like. The fact is, odometer miles didn’t much matter.
Oddly, battery range only matters a bit. This analysis was restricted to models that had an estimate of the actual remaining range, via some remote monitoring service that tests battery life. Every 10 miles of additional battery range was only worth about $430, or maybe 3%. I’m guessing that people interested in a used Leaf are probably coming from the same place that I am, in that they plan to use it as a “city car”. So extra range is nice, but not a deal-breaker, as long as it’ll get you around town for the next few years.
I’m not sure the same would hold true for the price of a new EV, where the typical buyer may want to be able to take long trips in the car. There, I’d expect to see a higher value on longer range.
Final takeaways.
First, I’m not going to buy from Carvana. As if I ever were.
Second, on paper, as new cars, Leafs appear to have suffered from shockingly steep depreciation. So much so that it gave me pause. How could a car that lost value that quickly be a good buy as a used car?
But I think that’s largely an artifact of the (then) $7500 tax credit. A person buying a $32500 Leaf actually only paid $25,000 for it, net of the tax credit. If I factor the tax credit in, depreciation on a new Leaf that appeared to be in excess of 12%/year falls to a far more reasonable 8% per year. Or just modestly higher than the rate of depreciation shown among the sample of used Leafs.
Third, at the price range I’m considering (~$10K), it probably won’t cost me much to buy the model with the bigger battery/higher range (SV) instead of the base model (S, SL). The additional 30 miles of range (100 vs 70) ought to cost me just over an additional $1200. And that additional range should be cheap insurance against the future degradation of the lithium-ion battery.
Finally, as long as the battery is in good shape, I probably shouldn’t worry too much about the mileage. Or, at least, that’s what the market is telling me. Buyers seemed to be almost indifferent to the miles on the odometer. To me, that crowd-sources the conclusion that, battery aside, these cars don’t tend to wear out within the range of mileages typically found for used Leafs.
Epilogue
A wise philospher once said, “You cannot step into the same used-car query twice.” I thought about adding more information (e.g., Carmax as seller). But I “keypunched” the cars in the order they showed when I ran my first query. To get back to those Edmunds.com search results, I’d have to re-run that query. That could easily give me a slightly different set of cars, depending. Which means the existing dataset might not synch up with the new data, as punched. Not without extraordinary measures.
In this case, the juice is not worth the squeeze. I’ll let this be the end of it.
