Never calculate densities for MSAs

Metropolitan Statistical Areas (MSAs) are the areas most often used for reporting statistics for urban areas in the United States. They do a reasonable job of encompassing entire urban or metropolitan areas–suburban areas along with the principal cities. More data are reported for MSAs than for any alternative delineation of urban areas.

Densities–population, housing units, or anything else per unit of area–are important characteristics of urban areas. (Indeed, Louis Wirth, in “Urbanism as a Way of Life,” included density among the three characteristics defining what it means to be urban.) But it is never appropriate to calculate densities for MSAs.

Since density is a count of something divided by the area in which they are counted, density depends critically on the area being used. The problem with calculating densities for MSAs arises because MSAs are defined using counties as the basic building blocks. (This makes a lot of sense because many types of data are available at the county level.) An MSA starts with the county in which a large urban area is located. Counties are then added that have substantial commuting ties. But some of those counties can have large amounts of rural land that is in no meaningful sense part of the metropolitan area. For reporting counts such as the population of an MSA, this overbounding makes little difference because these rural areas have very small numbers of residents and their inclusion does not significantly affect the MSA totals. But for densities, which are equally dependent upon the area included, the inclusion of large amounts of rural territory greatly affects the values.

And this is not simply a problem that will produce some minor random error if MSAs are used to calculate densities. The magnitudes of the rural areas added can be extreme, and differences vary systematically with the region of the country. In the Northeast, Midwest, and South, most counties are reasonably small, so the county-based MSAs reasonably approximate the boundaries of the actual urban area. In parts of the West, however, counties can be extremely large, adding great expanses of rural territory to their MSAs.

The Riverside-San Bernardino-Ontario, California MSA is east of and adjacent to the Los Angeles MSA. It consists of two counties, San Bernardino and Riverside Counties, that extend from Los Angeles County east across the remainder of California, in some places up to around 200 miles, to the Colorado River and the states of Arizona and Nevada. San Bernardino County is the largest county in the lower 48 states with an area of 20,105 square miles. This is larger than 9 states. Nearly all of the population is concentrated in the southwestern corner of the county. A major portion of the county is the Mohave Desert. This area is not just rural, it is essentially empty. Driving across the county on Interstate 40, you encounter a warning sign before an interchange that informs you that it will be over 80 miles until the next gas station. Obviously including all of this empty land in any calculation of density gives an extremely artificially lowered value that does not reflect the character of the actual urban development in the MSA adjacent to Los Angeles.

Some readers of this who are familiar with how MSAs are defined may think that I’m getting too carried away, stating the obvious. Anyone at all familiar with MSAs and how they are defined would never calculate densities for these areas. I would have thought so, too. Until I encountered this publication issued by…the U.S. Bureau of the Census!

Patterns of Metropolitan and Micropolitan Population Change:
2000 to 2010
, 2010 Census Special Report.

This report presents a wide range of statistics for Metropolitan and Micropolitan Statistical Areas (the latter being the smaller counterparts of MSAs). Chapter 3 addresses the topic “Population Density.” It begins by discussing overall population densities for these areas, first presenting the population densities by size category. It goes on to list the areas with the highest and lowest population densities for both the Metropolitan and Micropolitan Statistical Areas. The authors observe that the areas with the lowest densities are in the West. This is the first time they hint at the problem, noting that counties in the West typically have larger land areas. Then it’s back to discussing the highest density micropolitan areas.

It’s only at this point that the authors actually acknowledge that densities “can be heavily affected by the size of the geographic units for which they are calculated.” They go on to say that the areas are delimited using counties, which can vary greatly in terms of area.

The chapter continues by switching to reporting on population-weighted densities, which are somewhat more reasonable for MSAs. But as far as I am concerned, the damage had already been done. The report had calculated and presented overall densities for MSAs and had discussed these as if they were meaningful. And they are not.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s