I still remember when I read the data in business school 40 years ago, that Unions outside monopolies and oligopolies did not increase wages, because I was just amazed, but then realized that is what economic theory dictates - it was at once to me elegant, and true, that that had to be the case.
This is not what the data show.
Here (pdf) is a paper by economists
David Blanchflower and
Alex Bryson using data from the Current Population Survey that finds a union wage premium across a variety of areas; I've given the chart below. As you can see, the largest difference was in construction. Their literature review at the beginning of the paper suggests that others have found the same, and that the main disagreements are over the extent of the difference. It also appears to have been larger 40 years ago than it is now.