I'm not sure it follows that a person you hire with a degree will be worse than the average person that you hire without a degree. After all, you said "if all else is equal. . . ." If all else is equal than the candidates are equally qualified and the presence or absence of the degree made no difference. I don't see how having a degree by itself (which is the scenario you envision) can ever count against you.

If you hire someone without a degree, the most you can conclude is that there was probably at least one person with a degree that was inferior to that individual. The problem is that you don't know how many people with degrees are inferior to this individual. Exact numbers matter in this case; I think that the reason that people prefer people with degrees is because having the degree is more often associated with the required skills than not having the degree.

This is actually true about any discriminated against group (which people without degrees are). On average the discriminated against group may be worse (for whatever historical reasons), but the ones who are good enough to become seriously considered despite that are actually better than the ones you would consider equally desirable. Therefore if you're on the fence about a decision, you should prefer the one who lacks the most obvious signals like degrees, etc.

This may sound like an abstract and weird hypothesis. But it is a testable one. For example see http://hbswk.hbs.edu/item/6498.html for evidence that being willing to hire women into management in a culture where women are discriminated against results in better financial returns for the company that is willing to do so.

As for your hypothesis that people with a degree are more likely to have the required skills, that depends heavily on the job. Certainly if you're going to hire someone to work on a compiler, you'd prefer people who know how compilers work. Which gives a big edge to CS grads. But for general software development I'd prefer someone who can quote Code Complete back at me than an average CS grad.

(Disclaimer, I have a masters in math, and almost finished my PhD. If you are looking to hire me, and you have an alternate candidate without my academic qualifications who you think is a tossup compared to me, then yes, I am saying that you should hire them instead.)

For example, suppose you have two positions to fill that are basically the same. You fill one position with person A that doesn't have a degree and you fill another position with person B that has a degree. Person A may have gotten the position because they beat people with and without degrees that were also applying to the position. The same applies to person B. On one hand, you're saying that person B will get some level of preferential treatment because he/is has a degree. But on the other hand, you're saying that person A is probably better than person B in a manner that isn't reflected in the basic competency tests that were used to hire them in the first place. However, at this point you're comparing apples to oranges because they weren't hired from the pool of candidates.

Moreover, I don't know what you mean by "better". You might say that someone that succeeds despite lacking a degree is probably more passionate than an equally hireable person that had a degree. He/she may have more job experience. He/she may have passed with an A+ from the school of hard knocks. It seems to me that the only way to quantify this and aggregate over individual differences would be to compare the long-term salaries of people with degrees versus those without. Pay may seem like a coarse measure of whether an employee is better or not (I should know because I left a job in which I was underpaid for years despite being productive) but it is at least a simple measure of how much a company actually values an employee.

However my personal, anecdotal experience is that peer coworkers that I've had who did not have college degrees have, on average, been better than ones who did. By "peer coworker" I mean "working with me, with a job title similar to or better than mine". By "better" I mean "impressed me more". My measure of being impressed is what I thought of the quality and quantity of work that I saw them doing.

I have no idea what their salaries were like. There were some that I know were making less than coworkers with degrees, even though I thought that their work was better. Most I never had a discussion about salaries with.

That said, on the whole I would wager that discriminated against workers get a worse salary even if their productivity is equivalent or better. Why? Because salary is the result of a negotiation, one of whose inputs is what your alternate options are. People are not paid what they are worth to the company. They are paid what the company thinks it needs to pay them to keep them happy, and the difference between that and their worth is kept by the company as profit. (Companies that do not act this way soon find that they are not able to make a profit and some time later find themselves out of business...) If some employees have a hard time being paid more elsewhere, then they will often be satisfied with less from you.

This phenomena is presumably why the paper that I pointed you to measured productivity on the basis of company growth and profitability, and not on paid salary.

Then again, there are people (occasionally) like Steve Blank and Ed Fredkin.