Jump to:Page Content
Among the many wonderful things in the United States Constitution, the third paragraph of Article 1, Section 2, is one of my favorites. That paragraph mandates an enumeration of each state’s population, the census, "within three years after the first meeting of the Congress of the United States, and within every subsequent term of ten years."
The census is indispensable, not only to determine the apportionment of representatives and presidential electors, but also to give us a picture of our changing country. The 220 years of census data give us a picture of a nation constantly in flux.
I’ve just released a paper – jointly written with Giacomo Ponzetto and Kristina Tobio – that looks at county-level changes in America over the very long run. I’ve also released a policy brief that looks at the county level over recent decades. Some trends persist, and others change dramatically over time.
For our work on the long run, we focused on the eastern United States – those counties bounded by the 90th meridian (west), which is approximately the location of Memphis, and by the 30th parallel and 43rd parallels, which exclude southern Florida and northern New England. These restrictions allow us to focus on areas that were populated 150 years ago, at the beginning of the Civil War.
If we include the entire United States, population shifts are dominated by the continuing march westward.
The first, quite striking fact about this part of the country is an enormous stability in population patterns over time. Those counties that were highly populated remain so, and those less populated remain relatively less populated.1
This long-run stability is associated with Gibrat’s law – which posits that there is no correlation between initial population and percentage growth in population. Over the 140 years between 1860 and 2000, Gibrat’s law holds — counties of all initial population levels had, on average, the same percentage population growth. But for individual decades or shorter time periods, the data suggests that the strength of Gibrat’s law over the entire period seems almost like an accident of history.
During every 19th-century decade, except for the war-torn 1860s, population growth was faster in lower-population countries. During the first six decades of the 20th century, population growth was faster in more populated counties.
First, we spread out as Americans populated the continent to take advantage of its vast natural resources. Then as we first left our farms and then our natural-resource-based manufacturing centers, we clustered in denser areas, taking advantage of the benefits of proximity.
Over the last decade, there was also a strong tendency of population to move into all but the densest counties, which is shown across the entire United States in the chart below. I’ve grouped counties into 10 bins based on density levels as of 2000.
The top line shows that population rises with initial density until we get to the densest counties. It isn’t hard to understand why Americans have moved to more dense places – those areas earn more income.
The bottom line shows the powerful connection between density and median family incomes. Declining transportation and communication costs seem to have made urban density more valuable. My book “Triumph of the City” argues that this connection reflects the fact that globalization and new technologies have increased the returns to being smart and that we get smart by being around other smart people in cities.
Whatever the reason, denser areas now pay higher wages. People follow the money and move to places with more people.
The changing relationship between initial population and subsequent population growth suggests that growth is best seen as a result of changes in technology, not immutable laws that govern urban change.
The knowledge of the historian and the theories of the economist are better guides to understanding our changing nation than approaches, borrowed from physics, that emphasize permanent statistical relationships.
There are other sharp differences between the 19th and 20th centuries. In the 19th century, even within our Eastern sample of counties, the population moved west. During most of the 20th century, again within this sample, the population moved back east.
During the last census, there was also a tendency – throughout the entire country – to move to old ports. These trends again reflect the 19th-century tendency to spread out to get access to natural wealth and the 20th-century tendency to cluster in the metropolitan areas that were anchored by old ports.
These facts actually contain a lesson for current economic policy. In the 19th century, it was incredibly valuable to build transportation infrastructure that crossed America and allowed access to our hinterland.
But as America increasingly clusters in dense counties around the edges of the country, it is far less important to invest in far-flung transportation infrastructure and far more important to invest in ways that make our metropolitan areas more productive, like education. I will return to these long-run trends next week.
1. In the language of economics, the logarithm of population in 1860 can explain 44 percent of the variation in the logarithm of population levels in 2000, across the 1,124 counties in this sample. If we restrict our sample even further, to the 80th meridian, near Erie, Pa., that effect is even more pronounced: population in 1860 can explain 57 percent of the population 10 years ago.