There are two ways for cities to grow when they’ve begun to run out of housing: They can expand outward, carving new neighborhoods from the countryside, or they can turn inward, replacing small buildings with larger ones, tucking new construction on unused lots, stretching higher instead of farther out.

The first option has long been the American way.

Since 1950, the vast, vast, vast majority of housing growth around U.S. cities has come from building where there were no buildings before. In every decade since then, nearly 90 percent of new housing construction in the U.S. has taken place on land that was either previously undeveloped or very low-density (with less than four homes per acre).

That’s according to a novel new analysis of American Community Survey data on the construction vintage of all U.S. housing by Issi Romem, the chief economist at the platform BuildZoom that tracks building permit and contractor data. His study uses building ages at the Census block-group level to trace where cities expanded — and what was already on the land when new housing was built.

“Densification just doesn’t account for the bulk of housing growth,” Romem says.

The above GIF of the Washington-Baltimore area shows how the two metros have expanded into each other since 1940. Nearly all of the region’s growth since then has come from the spread of new low-density neighborhoods (in light blue) into rural parts of Maryland and Virginia. Far fewer neighborhoods grew denser — darker blue — over this time.

Here is what the same timeline looks like in Atlanta:

And Phoenix:

This history — and why U.S. development has taken this shape — also suggests that it will be very difficult for cities to grow denser in the coming years, despite rising worries about the environmental costs of sprawl and the individual toll of commuting.

Romem’s data show that cities produce new housing in proportion to their rate of outward expansion. Metros that spread out the most add the most housing, and have kept their housing costs in check as a result. Metros that have resisted sprawl (like Portland, which has an urban growth boundary, or San Francisco, which is hemmed in by mountains and water) haven’t built much.
This chart showing all U.S. cities with a population of at least 250,000 illustrates this relationship:

That picture means that metros that didn’t spread out failed to compensate by turning inward — at least on the scale necessary. In the choice between more sprawl and greater density, they chose neither. And that’s driving the increasing divide in the United States between what Romem calls expansive and expensive cities.

In the national picture, these two charts tracing the U.S. housing supply since the 1950s show, at top, how the share of all housing at the densest levels has shrunk with time. Below, we can see where the share of new housing was added, with about a quarter of it consistently coming on undeveloped land:

One tiny detail there — the slight uptick in the 2000s in the share of new construction at the highest density — speaks to the new apartments now rising in cities like Washington.

“That’s huge,” Romem says of what might seem like a small numerical change. “That is the reflection of the urban renaissance, the shift back to denser suburbs and downtowns.”

The larger historical trend, however, shows that densification has slowed down since the postwar period. Part of this has to do with the fact that easily developable land disappears with time as cities grow. Fewer large, usable parcels remain. It becomes harder to cobble together small ones.

But the other obstacle blocking more density today comes from choices that we’ve made. By zoning large swaths of cities for single-family housing only, we effectively outlaw greater density. Limits on building heights have the same effect. Development flows where it’s easiest to build. And not only is it easy to build on the empty edges of town, but we have intentionally made the alternative that much harder.

That leaves us with a choice between ever more sprawl or ever higher housing costs (with growing cities falling into one camp or the other). Or, Romem suggests, we could do the really hard thing: fundamentally rewrite the rules for how land can be used. Curb the ability of cities to block new housing. Kill off single-family zoning (which is not the same as killing off single-family homes). “Such a change,” Romem adds in his paper, “would need to be coupled with a broader acceptance of multifamily housing as a legitimate place for raising children.”

He acknowledges that massive change like this sounds impossible today.

“But eight or 10 years ago, the thought of having mandatory health care in this country seemed probably not far from this in terms of how outlandish it was,” he says. “And yet that happened. There is hope.”

In the meantime, here are a few more GIFs to ponder: