On February 1, the Bureau of Labor Statistics released its employment projections through 2020. Despite President Obama’s vision in his State of the Union speech of “an economy built on manufacturing,” the BLS predicts U.S. manufacturing employment will decline over the course of the decade even as the overall economy grows.

More specifically, manufacturing’s share of jobs will drop from 8.1 percent in 2010 to 7.0 percent in 2020.

Professional services, health care, and education are the broad sectors predicted to grow the fastest, along with a rebounding construction industry. Within those sectors, the fastest growing will be home health care services, individual and family health care services, and management/scientific/technical consulting services. Apparel and leather manufacturing will shrink the most.

What does this mean for individual U.S. metro areas? Although the BLS makes predictions for the nation, industry growth shapes local economies. If manufacturing jobs continue their long decline, cities that depend on manufacturing will continue to suffer more than others.

Most metro areas’ economies are skewed toward some industries and away from others. Los Angeles has movies, New York has finance, and Detroit has cars. Let’s say that the BLS is spot-on and each narrow industry grows just as fast until 2020 as they predict. The metro areas that should grow fastest will then be those where fast-growing industries are concentrated. Among the largest 100 metro areas, the big winner would be the Washington, D.C. area, thanks to its large share of professional services jobs. (Note that this analysis considers only private-sector employment predictions.)

At the other extreme, Greensboro, North Carolina, and other manufacturing centers in the South and Midwest would grow slowest between 2010 and 2020.

Below then is the BLS’s list of the top 10 U.S. metro areas (out of the top 100 largest metros) for projected job growth:

1. Washington-Arlington-Alexandria, DC-VA-MD-WV
2. Bethesda-Rockville-Frederick, MD
3. Colorado Springs, CO
4. New York-White Plains-Wayne, NY-NJ
5. El Paso, TX
6. Springfield, MA
7. Baton Rouge, LA
8. Tacoma, WA
9. Baltimore-Towson, MD
10. San Antonio-New Braunfels, TX

And the bottom ten metros (the first metro listed is projected to have the slowest growth):

1. Greensboro-High Point, NC
2. Gary, IN
3. Los Angeles-Long Beach-Glendale, CA
4. Grand Rapids-Wyoming, MI
5. Columbia, SC
6. Detroit-Livonia-Dearborn, MI
7. Cincinnati-Middletown, OH-KY-IN
8. Milwaukee-Waukesha-West Allis, WI
9. Oxnard-Thousand Oaks-Ventura, CA
10. Salt Lake City, UT

What is Los Angeles doing on the bottom 10 list? It’s dragged down by the expected carnage in apparel manufacturing and a projected small decline in movie industry employment.

The map below shows the predicted job growth for the largest 200 metros. The dark blue metros include Washington, D.C. and others with the highest expected job growth based on the industry projections. The mid-Atlantic states, Massachusetts, Florida and parts of Texas and the West all have metros where faster-growing industries are concentrated. The Midwest and the South, where manufacturing is concentrated, would have the slowest growth.

The good news for the metros on the bottom end of the BLS projections is that industry mix is not necessarily destiny. Looking back at the last decade (2000-2009), industry mix is a good but far from perfect predictor of actual metro job growth. Some metros do better than industry mix alone predicts, and some do worse.

New Orleans, for instance, had much slower job growth last decade than its industry mix would predict, thanks to Hurricane Katrina’s destruction. But metro performance often follows more predictable patterns. Even after taking industry mix into account, metros grew faster last decade if they had a milder, drier climate; a more educated workforce; and lower population density. A mild, dry climate attracts workers, which in turn draws businesses; a more educated workforce raises local productivity; and lower population density generally reflects a lower cost-of-living, including cheaper housing. (Density turns out to be a better predictor of job growth, statistically speaking, than housing prices. As the world has painfully learned, housing prices can fluctuate wildly, so housing prices at any one time might not be a great indicator of local costs over the long run. Density, in contrast, is fairly constant over time.)

The effect of climate and education on local growth should be no surprise – lots of academic and popular research has examined these and other factors. And even though unique and unpredictable events are, um, unpredictable, it’s a safe bet that metros with a nice climate and educated workforce will continue to outperform.

The effect of density might be a surprise. We know that density raises productivity and wage growth, and policies that limit density can stifle growth. Nonetheless, high-density places have slower job growth because high density comes with high real estate prices, which push many businesses and households elsewhere. Keep in mind, too, that high-density places tend to have both higher educational attainment and a more favorable industry mix (more professional services, less manufacturing) – so high density often goes hand-in-hand with factors that boost job growth. Still, holding all other factors equal, the data clearly show that high density places have slower job growth.

Combining the industry employment predictions with climate, education, and density gives us the full view of where the jobs will be in 2020*. Good climate and low density gives the South and Southwest a boost and knocks some big Northeast cities down. The 10 fastest and slowest job growth metros this decade now look like this:

Note: Top and bottom 10 out of 100 largest metros. The first metro listed on the bottom-10 list is projected to have the slowest growth of all 100 largest metros.

Southwestern metros now dominate the fastest-growth list, led by Phoenix, Tucson, and El Paso. The slow-growth club is all Northeast and Midwest, plus Los Angeles. Mild climate and low density mean that southern metros like Greensboro, North Carolina and Columbia, South Carolina, should grow faster than their unfavorable industry mix suggests, so they’re off the bottom-10 list when all factors are included.

The map below shows that employment growth will be fastest in the Southwest, Texas and parts of Florida and California. The South looks more blue than before, while the Northeast and Midwest are pretty solidly light and dark red. Turns out, metros with good climate, higher education and lower density tend, on average, to have a more favorable industry mix to begin with. But some metros blessed with industries that are likely to grow – like New York and Boston – will be held back by harsher weather and a higher cost of living. Other metros – like Phoenix and Las Vegas – should grow fast despite having concentrations of industries projected to grow more slowly.

Could these predictions be wrong? Of course. They probably will be, in fact. The BLS experts are forecasters, not psychics: some industries will grow faster nationally than they predict, and others will fall short. Weather, education or density could matter more or less this decade than last decade.

And surely there will be unpredictable factors that affect individual metros: even events far less dramatic than Hurricane Katrina can throw a city off its growth path. To underscore the uncertainty, I’ve purposely not included numerical growth predictions – those depend not only on how metros do relative to each other but also how the national and global economy fare, and I’m not going anywhere near those big questions.

But people looking to move for work or pleasure need some sense of where the growth will be this decade, and industry employment projections and other factors together give us a reasonable view into where the growth is likely to be. The decades-long shift of American jobs and people from the Rustbelt toward the sun will continue.

*For the econometrically inclined: I took the coefficients from a 2000-2009 metro-level regression (weighted by 2000 metro employment) of actual growth on industry-mix-predicted growth, % with at least bachelor’s degree, several climate variables, and log tract-weighed population density and calculated fitted values for 2010-2020 growth by applying the coefficients from the 2000-2009 model to industry-mix-predicted growth for 2010-2020, % with at least bachelor’s degree in 2010, log tract-weighed population density in 2010, and the original climate variables. In the regression, all variables are statistically significant at the 5% level, and the R-squared is .39.

(Originally published on The Atlantic Cities blog)