Here is Wendell Cox's summary of 1920-2010 U.S. urbanization data. This stuff is indispensible because the Census' urbanized areas (UZAs) are periodically redefined to capture the boundaries of actual development. From 1950-2010, there are seven decennial observations. How do the top-ten population rankings change over this period? Without looking, everyone knows that the New York UZA was #1 in each year. So the number of rank changes for #1 was zero. In the #2 slot, Chicago was #2 in 1950, but LA has been #2 ever since. That means just one rank change for the title of second biggest UZA.
This is how the top ten rankings changed over the sixty years:
UZA rank # of rank changes
The greatest stability is at the top. (Doing the same exercise for central cities would make no sense because they involve political boundaries that may or may not budge, depending on local laws and conditions.)
There are costs and benefits of UZA size. The trick is to exploit the benefits of size without being overhwelmed by the costs? I have mentioned often that flexible enough land markets are the only explanation. Allow for the propitious location and re-location of land uses. The propitious spatial arrangements are the ones that allow profitable interactions. These are too complex to be knowable by policy makers. Experimentation by risk takers is the only option.
Here is the Chatterji, Glaeser, Kerr study of "Clusters of Entrepreneurship and Innovation". Looking across their various cases and examples, the authors report a mixed record of success as well as a mixed set of policies.