I've experienced the same frustration. There are certain technical circles in which I am well known, yet when I decided to get an 8-5 I had a heck of a time because I left college to run my business. I eventually took a job making half as much as someone with my experience would normally make.*
Not long ago, my boss ask me to look over some resumes for an opening we have. Most of the resumes looked pretty similar. It's one page, after all, and they are all applying for the same position. Those that were different had positives and negatives, so they all "scored" about the same. How to decide who to call first? It felt like forensic science, trying to find some clue of who might be better among the pile of nearly identical resumes. A typo MIGHT hint that the person isn't as careful with their work. That seems silly, but those were the kind of clues we had to go on.
We interviewed a couple of people and it was similarly "tied" - both seemed like they would be a decent fit. Without actually knowing either of the candidates, we had to choose one based on the tiny bits of information we had. Compared to the tiny clues we were looking for, a degree vs. no degree would have been huge. It's not a great indicator, but it sure is better than any of the other differences between two otherwise good resumes.
To look at it another way, suppose you have ten resumes that look okay, ten that made it past the "obviously no" filter. You have to find some way to narrow it down by eliminating nine of the ten candidates. Four years of study and carrying a four year project to successful completion obviously helps narrow it down.
It's now time for me to get off Slashdot and get back to my studies. With 17 years of experience I don't want to be narrowed down for lack of a degree, so I'm getting my degree from WGU. The "final exam" for most courses is an industry-recognized certification exam, so I'll end up with a degree and about a dozen certifications.
* The "low paying" job turned out to be a blessing, due to working with wonderful people.