Some years ago, I heard it told that the American car companies had done the market research and found that the typical car-buyer in the US bought a car and kept it for 3 years, on average. And so, armed with this info, the cars were designed to last 3 years on average. This is also why the 3-year lease is a thing.

Now, I don't know how true this is, or what definition of "average" was being used in those statements (mean, median, or mode), but let's roll with it. Assuming a normal distribution, that would mean half the cars built last less than 3 years, some significantly less so, and half the cars built last longer than 3 years, again, some significantly more so. Depending on how wide the distribution is, that could mean that there's simultaneously a lot of shitbox lemons and a lot of indestructible workhorses at the same time. And regardless of how wide that distribution actually is, that would explain the varying reports of various US-built cars I've personally heard: some glowingly praising reliability that rivals the Japanese stereotype, and some that would bring more value to the world as crushed metal cubes.

I want to argue against engineering the cars to last only as long as the typical new-car-buyer keeps one, but I have a hard time doing so. If most new cars aren't kept longer than that, why should you care how long the car lasts after the first owner has sold it? I think you could argue that the poor reputation of your used cars could harm new car sales in the long term, but I don't know how to quantify that. In the absence of numbers of how such reliability planning harms long-term sales, I could totally see decision-makers coming to favor designing to a 3-year reliability target, with the car being considered disposable thereafter.

In either case, if the average car buyer buys a car every N years, and thus, you treat that as reliability target, then targeting for a median or mean reliability of N years strikes me as the wrong way to to that. IMO, the ideal reliability distribution would be something of a skewed one, with a long right-hand tail. Keeping the assumption above of N = 3 years, then you'd want the mode on the reliability distribution to be 3 years, with the band below the mode as narrow as possible. If you couldn't reasonably make it any narrower, than push the target modal reliability up, such that the number of cars that fall below N are significantly less than the ones meet it. Yes, that means there will be more cars exceeding your reliability target, but having cars that lasted longer than the minimum necessary strikes me as a lesser evil than having ones that didn't.

Of course, for all I know, the US manufacturers do exactly as I described in the above paragraph, and the German and Japanese brands simply have a longer tail.

Edit: I should make it clear, I personally don't agree with this statement, "If most new cars aren't kept longer than that, why should you care how long the car lasts after the first owner has sold it?". I'm not trying to excuse or justify the low reliability ratings of domestic manufacturers, I'm trying to understand how they got there. I'm trying to model the decision-making of US car manufacturers, and it's a mindset I can see them running on, given the paragraphs above it.

Those down-voting, I'd love to hear your counterarguments. If I'm wrong, then I'm wrong, but it's hard to make less-stupid arguments without knowing how I'm wrong.