One approach to urban areas emphasizes the existence of certain immutable relationships, such as Zipf's or Gibrat's Law. An alternative view is that urban changereflects individual responses to changing tastes or technologies. This paper examinesalmost 200 years of regional change in the U.S. and finds that few, if any, growth relationships remain constant, including Gibrat's Law. Education does a reasonable jobof explaining urban resilience in recent decades, but does not seem to predict countygrowth a century ago. After reviewing this evidence, we present and estimate a simple model of regional change, where education increases the level of entrepreneurship.Human capital spillovers occur at the city level because skilled workers produce moreproduct varieties and thereby increase labor demand. We find that skills are associatedwith growth in productivity or entrepreneurship, not with growth in quality of life, atleast outside of the West. We also find that skills seem to have depressed housing supplygrowth in the West, but not in other regions, which supports the view that educatedresidents in that region have fought for tougher land-use controls. We also present evidence that skills have had a disproportionately large impact on unemployment duringthe current recession.