Gender and the Geography of Work

I've had a long-standing interest in the multi-decade slide in the fraction of men that work in the United States (eg here, here, here, and here).  Yesterday, I posted some maps of the fraction of men over age 16 working by county based on the 2000 census, including this overview map, which has white being 25% or less of that population employed, and fully saturated blue being 75% or more.

In this post, I start to look at the equivalent data for women and the combined geographical patterns of male and female work, which raise a variety of fascinating questions.

Here is the basic data: fraction of women over 16 working, by county in the 2000 census.  White is less than or equal to 25%, fully saturated red is greater than or equal to 75%.

The first and obvious things: women are a little less prone to work overall, so the overall color is less saturated.  But then, also, there is a little less variability: whether women are working is slightly less dependent on where they live, so the color is more uniform across the map.  To put the same thing in statistical terms, the interval from two standard deviations below the average county to two standard deviations above is 43.6% to 82.6% for men, but 34.8% to 67.5% for women.  You'd expect that range to include about 95% of counties in each case.

But the interesting questions to me are things like: to what extent does male employment and female employment vary together (men and women are both prone to work because the local economy is strong, or both not prone to work because this is a retirement area, for example) and to what extent do they differ (traditional male employment has declined and women are working more to take up the slack in the family budget).  It's a little hard to see that by staring at the two maps together, and so I wanted to combine them.

To start with, I rescaled both sets of employment-population ratios by their own average and standard deviation.  So, for example, a county with 34.8% of the adult female population working would get rescaled to -2.0 for women (two standard deviations below the average), whereas a county with 82.6% of men working would get a +2.0 for men.  So this basically measures how much men or women are working in a given county relative to the overall set of counties in the US.

If you just plot that data on a scattergram for the three thousand and some counties in the US, you get this:

Here each blue dot is a county, and the x-axis is the rescaled male E/P ratio, while the y-axis is the equivalent thing for women.  The straight black line is a linear fit to the data (which explains 56% of the variance).  In fact it looks to me like there's two things going on - there's a fairly strong tendency for male and female employment to vary together with only modest separate fluctuations (the region of the pink diagonal oval, and then there's a long tail of very low male employment counties (the green oval).

Next, I made some maps of these two variables together.  Basically, I combined the blue-color of male employment with the red color of female employment, via some complicated formula which I'll spare you the details of, and came up with maps like this:

Here, the darkness overall indicates the fraction of people overall working.  So in very pale regions like Eastern Kentucky/West Virginia, or the Four Corners region where Arizona, New Mexico, Utah, and Colorado intersect, neither gender is heavily involved in the formal economy (the former being primarily Appalachian mountain people, and the latter being a heavily Native American area).  Meanwhile, dark gray counties - for example Denver and the counties just to the north of it - have large fractions of both genders working, to an extent about equally above trend for their respective populations nationally.  The Blue/Orange scale then shows the degree to which men are working more than women (blue) or women working more than men (brown/orange) - again, not in absolute terms, but relative to what is typical for each gender in counties nationwide.

The map above immediately makes it clear that there are all kinds of interesting things going on geographically, with a "brown belt" that runs across the northern midwest, upstate New York, and New England, and various blue regions in the rural south, the western great plains, and Utah/Idaho.

To get a better feeling for this, I created a 3-D visualization of the data in which the height of each county corresponds to the population density (so the volume of each county's "block" corresponds to the population of that county), and the color scale is as above.  So herewith a quick tour of the country.  In each case, you can click on the picture for a larger version to scratch your head over.

To begin, we're looking eastward over the "brown belt" from hovering somewhere over about South Dakota.  In the lower left is the Minneapolis/St Paul region, the suburbs of which seem to be a particular center of brown work.  Then in the middle distance is Milwaukee, Wisconsin and Chicago, then the Michigan peninsula.  In the distance is the eastern seaboard, with New York the tallest pillar (which is cut off), and Boston being the smaller brown pillar to the left of it.

My guess would be that the reason this area trends brown is that it is the old industrial heartland of the US, the male working population has been depressed by decade-long trends of globalization and automation, and women are picking up the slack.

Next up, we hover over the North Atlantic and look down the eastern seaboard, with the brown tower of  Boston in the foreground, and then the slightly bluish grey of New York next.  Presumably, New York, with it's male-dominated financial industry, tends to have stronger male employment than Boston (though the suburban counties around the latter tend more gray, perhaps due to the technology industry there).

Moving further down the eastern seaboard, we have the very pale brown cities of Philadelphia and Baltimore - old industrial cities with generally weak employment, particularly in traditionally male pursuits (I assume).  As we move south, the urban counties of Virginia look darker (ie economically healthier), with a variety of different colors that I don't have the first clue the cause of:

Moving inland, here we look north-west up the Appalachians, which are a mountain chain in real life, but a valley in this artificial terrain, because the population density is low.  It's an amazingly large region in which an astonishingly small fraction of either gender works, at least in the sense measured by census forms.

Next we turn around and hover over Florida and look west across the northern Gulf of Mexico, with Atlanta in the lower right of the picture, and the dark gray masses of the main Texas cities in the upper left of the picture.  In the South, while there are exceptions, it seems that the rural areas are more apt to have higher male employment (bluish), while the cities tend to be dark gray - economically strong with both genders working:

Moving westward, we look north over Texas: Austin and Dallas/Fort Worth look like they have higher employment fractions than Houston and San Antonio (this is as of 2000, however - the tech crash and oil boom since could have changed this quite a bit).  Meanwhile, we can look up the "blue belt" up the western great plains, where apparently the "real men" reside who still provide for their families unaided.  Even here though, there are enormous variations county to county which I have no idea how to account for.  Presumably it has to do with the concentrations of particular types of industry and agricultural work that are more apt to employ either men or women.  Texas seems particularly variable.

Next up, we hover over Baja and check out California:  San Diego and Los Angeles are gray, while Silicon Valley trends a little bluish with, no doubt, all the male engineers, while San Francisco itself has stronger female employment.

Finally, we look back down the Pacific Coast with the dark gray/brown cities of Seattle and Portland in the foreground and the blue bastions of traditional maleness off in Idaho and Utah in the upper left of the picture.

In a lot of ways these maps raise more questions than they answer, but it will certainly be fascinating to see how the patterns have changed when the 2010 census data come out later this year.