About two years ago, I took Statistics in college. The second most interesting part of the course was the industrial manufacturing focus, specifically quality control.
Even more specifically, reliable ways to determine how much quality you can take out of the product and still meet some specification with some good probability.
Basically, the whole course was a primer on using statistical techniques to make cheaper products by decreasing quality. All anecdotes the prof had to tell about industry experiences was about taking out quality. If this is typical, I think it explains a lot.
Now, this is actually a neutral tool, not the inherent evil the cynical parts of our mind would label it. Haphazard quality control techniques often resulting in ensuring quality by massively overspecifying certain components, with corresponding seriously inflated expenses. Sometimes that would also translate into improved reliability as parts can be stressed more before failing, but often not; one example that comes to mind is a cast-iron heating duct cover in an old house I saw that would probably survive an automobile impact, which is several hundred times better then the wall it was embedded in would have fared. Really, the strength and expense of cast-iron is major overkill in this application, at least looked at with modern eyes. Cutting those extra expenses in modern designs plays a large role in how we can manufacture many things more cheaply then ten years ago, even in domains where technology advance hasn't played much of a role.
On the other hand, these are the techniques used to reliably manufacture a device that will fail in 2 years, plus or minus 3 months, with 50% probability. This is the source of the flood of cheap garbage that has really only been gaining steam in the past four or five years; yes, in the 80s and 90s people were decrying "consumerism" but it's gotten several times worse as some of these techniques become refined and universally applied.
The upshot of all of this is yes, the quality of consumer electronics has been steadily declining for a decade or two now, along with everything else that comes off the factory floor, and the better the statistical techniques get, the closer to the consumer rejection threshold this stuff will get. The good news is I don't think it can get much worse... like almost everything else, the statistical manufacturing has diminishing returns and we should already be well past the part of the curve where a little effort gives big gains.
Oh, and the first most interesting part of the class? That with the exception of the distributions we covered beyond the Gaussian distribution, the entire 300-level college honors statistics class was stuff I had covered in high school. That was pretty disappointing, though it made for easy homework. Apparently, you are actively not expected to actually retain that stuff for any length of time after high school.