Alexander Micek (tumbledry.org) has posted a refutation of a recent “paper” published by SquareTrade — a vendor of extended warranties, which in turn got coverage on a lot of popular tech sites.
I’m not sure the results are quite as worthless as he argues. By way of comparison, consumerreports.org in reporting their laptop reliability survey results states that “differences of fewer than three points aren’t significant” (albeit deep in the prose below the chart) and in their FAQ that their standard for statistical significance is p = 0.01 (vs. 0.05 for most statistical reporting). When a survey is based on a really big sample (in SquareTrade’s case n = 30,000) you’d think it probably has some kind of usefulness. I doubt most people know what to do with this stuff anyway — does it really matter if SquareTrade doesn’t give you its confidence interval or Consumer Reports buries the necessary information deep in verbiage? In any event, in my experience methodology (i.e. how the data are acquired) is both more likely to distort statistics and less likely to be reported in any comprehensible way than confidence intervals.
Since the organization reporting the results sells extended warranties, it seems highly likely that their figures come from people with extended warranty coverage. Indeed, I’m guessing that the reason they don’t have three years of accidental damage data is that they’ve only been offering accidental damage for two years. This isn’t a reporting omission, of the kind Alexander Micek is complaining about, but a sampling error, which is much worse. (It’s like the difference between trying to randomly sample probable voters using some reasonably scrupulous methodology (e.g. calling people randomly selected from electoral rolls) and — say — just asking the people you find in a particular shopping mall.)
It’s worth noting that SquareTrade’s 30,000 laptops found Acer and Toshiba to be the most reliable, while Consumer Reports’ 65,000 respondents found Toshiba and Sony to be the most reliable. In both cases, the ballpark failure rate is around 20% (although the Consumer Reports figures cover laptops that are anything up to four years old — which (on the assumption that the average age would be around two years or less*) would seem to imply a higher failure rate than SquareTrade’s figures).
In short, even though it seems as if SquareTrade’s figures ought to have a sampling bias upwards, their figures are below Consumer Reports’s sample (which we would suppose to be less biased), and the general agreement over brand reliability makes their figures seem reasonable**. My suspicion is that Consumer Reports’ survey has a softer criterion for “major problem” than SquareTrade has for “failure” (going back to methodology: this criterion is much more valuable than “confidence interval” and is not documented in either case).
* Laptop ownership is increasing, and some people will sell or give away or discard an older laptop (especially if it breaks down), so I’d expect a population of laptops to skew towards the younger end of a capped age range.
** I’d put in the charts side-by-side but this would violate Consumer Reports’s usage rights.
The Simple Cost Equation
The basic question is this: is my extended warranty a good deal? Ignoring brand names (because the information provided doesn’t break down the sample sizes per brand, making that information useless), if the raw data indicate that ~30% of laptops fail over three years — 20% of that due to reliability and 10% to accidents — with an n of 30,000 then the chances are that this is pretty damn accurate. I’ve had three PC laptops (TI, Dell, and HP) and two experienced a significant failure in their first three years, and seven Apple laptops of which three failed in the first three years — so I’d have to say based on my own experience these figures are optimistic.
So let’s suppose that a typical laptop comes with a 12 month warranty, so that your extended warranty justifies itself by paying for repairs in the second and third years. Based on SquareTrade’s figures that’s a roughly 15% chance of failing during the extended warranty period. So the ballpark value of an extended warranty is 15% of average repair cost — and repairs for laptops are frequently expensive or simply impractical (i.e. the laptop has to be replaced). So let’s be generous and call it 15% of 50% of replacement cost — i.e. around 7.5% of replacement cost.
Well, AppleCare costs $349 for a $1999 MacBook Pro. Dell’s extended warranty coverage costs (at minimum) $120 for a $720 Dell Latitude E5500. (I tried to get a quote on a more comparable Dell, but Dell’s Pro laptops include three years of on-site coverage.) Neither Dell nor Apple cover you against accidental damage, SquareTrade essentially doubles the cost of coverage if you want accidental damage cover. SquareTrade just quoted me $200 to cover a $2000 laptop for three years — $200 extra to cover accidental damage. That’s a better deal than Apple or Dell, but probably much less convenient. (My guess is that while accidental damage is less common, it tends to be more catastrophic. A big dent, broken screen, or major spill is often effectively a write-off.)
Incidentally, AppleCare is hugely profitable for Apple.
I bought my wife a top-of-the-line iPod for Christmas in 2004, and I got the extended warranty (which included free replacement for any reason — completely bizarre, but then CompUSA is out of business now). So when it became obsolete — seriously, what were they thinking? — we just returned it and got the new 60GB iPod Photo as a replacement. (Nothing was wrong with the old iPod.) One day, the new iPod then wouldn’t start up, so I called Apple and they had a courier pick it up and sent a new replacement. The new replacement soon had the same problem, and we solved it the same way. The next iPod didn’t seem to have the problem but some time later — after the warranty expired — it recurred. I spent a few minutes Googling the problem, learned how to reset an iPod, and solved it. So that iPod “failed” three times — entirely because it had an extended warranty. Actual number of failures: zero.
Not only will insured users get non-defective products fixed, but the insured do not care how much a repair costs, while the uninsured care deeply. The laptops I’ve had that have failed have either been impossible to repair (e.g. two laptops’ physically fell apart) or very cheap (e.g. in two cases it was a hard disk failure — replacing hard disks is easy and cheap). In one case the unit was effectively DOA (it died after a day’s use), which was covered and thus doesn’t count (electronic devices are most likely to fail when brand new). So even though I’ve had five of my ten laptops fail (50%), the total cost of repair/replacement is less than the cost of extended warranties would have been — and of course mechanical failures are far more likely in older equipment that you’re happier to replace.
And of course there’s fraud. I once had a friend and business partner who — I discovered to my cost — was not the most honest person in the world, and he would always buy maxed out insurance on any piece of electronics (cell phone, camera, computer) he bought, and a remarkable number would prove “defective”, or be stolen, or accidentally destroyed before the coverage expired. In just the four years of our association he had one of his cars stolen (he’d had a car “stolen” the year before we went into business together, but I hadn’t observed a pattern), a laptop was left too close to a window during a rain storm, a cell phone was dropped in a toilet, another cell phone died some other way, and all of our video gear (but none of our office computers except for one laptop which happened not to be working) was stolen. And that’s just the stuff I know about. For the honest consumer, this means that when you pay for extended coverage, you’re paying for people like my friend. For the dishonest consumer, this is a wonderful opportunity for — effectively — perfect crimes. Oh dear, my 30 month old laptop just happened to fall out the window.
This class of problem is known as “Moral Hazard” and has been understood since at least the 16th century. It’s why free support lines keep you on hold for fifteen minutes (even when they have no other calls), why our healthcare costs are skyrocketing, and why figures on laptop reliability from a company that sells extended warranties are likely to be highly suspect no matter how the data are analyzed or reported.
The interesting thing here is that we don’t seem to see a lot of Moral Hazard at work in these figures. Consumer Reports seems to have higher failure rates for its respondents (many of whom do not have extended warranties) than SquareTrade (all of whose presumably do). If SquareTrade wanted to frighten people into buying warranty coverage, they’d have been better served by quoting Consumer Reports’ figures than doing their own study! The really interesting question may be whether SquareTrade is going to see an uptick in “accidental” damage to laptops as their coverage draws to an end.
Peace of Mind
Extended warranties have other benefits beyond the raw figures. To begin with, laptops often develop problems that aren’t major enough that one would try to get them repaired on one’s own dime (loose hinges, say) but which are both easy and inexpensive to fix. Typically, if you have a laptop without extended coverage then you will just ignore such a problem, while if you do have extended coverage you will get it fixed. If you tried to get it fixed without extended coverage it would probably cost an arm and a leg, but often the cost to the vendor for repairing the problem is minuscule (the person doing it is on salary, and the parts cost a few cents).
Again, this is a lot like health insurance (and for the same reasons). Because most people seeking a service (laptop repairs, medical care) do not pay directly, the actual costs are hidden and the fees actually paid for them are negotiated in back rooms. For the few who pay directly the costs seem incredibly high because they can’t negotiate an “insider” rate. So, in essence, the problem with laptop repair costs is that most of the people who get laptops repaired don’t pay for it, so the people who do get horribly overcharged. It thus follows that if you’d rather not live with minor, niggling issues you may want to get an extended warranty because it’s better than just being irritated by minor issues all the time.
You Win When You Lose
On the other hand, not getting an extended warranty also has benefits. If your laptop fails after two years and you have an extended warranty, you can fix it for free. But if your laptop fails and you don’t have cover you can just buy a new one (which will probably be better and/or cheaper as well as being new). Based on the figures above, you’ll not only be ahead financially on average and — in an odd way — when you do “lose” you “win” (yay, I get a new laptop). In my experience, catastrophic failures in laptops tend to occur when they’re getting old and mainly serve as an excuse to get a replacement. Of my ten laptops and five failures, one was a simple DOA (instant new replacement), and the others were all old. (In fact, I think two were more than three years old, and thus wouldn’t benefit from a typical extended warranty anyway.)
The one real justification for an extended warranty, in my opinion, is if you’re the kind of person who will pay for “peace of mind” or the kind of person who will fume over niggling minor things — the kind of thing you can get fixed for free — at least in theory. It’s not really good value for money (because you’re essentially paying through the nose for stuff that costs the vendor almost nothing to do and subsidizing criminals) but it’s cheaper than stressing out over shit. In my experience, a laptop that survives a year of solid use is only going to die if you physically damage it or the hard disk dies. It’s really expensive to cover the first, and it’s dirt cheap to fix the second.
Of course, if your plan is to “accidentally” back over your obsolete laptop on its 30 month birthday, then get all the coverage you can get.