No Room at the Inn for Amazon

Such an Enormous Investment Must be Transformative

Since Amazon announced that they’ll be launching a second headquarters with, over 10–15 years, 50,000 employees and gazillions of dollars of investment, every urbanist and regional economist out there has put forward their preferred set of cities, the places they think will make good candidates.

I want to take a different tact. I want to explain why every city is bad for Amazon and nobody can fulfill their RFP. And then, once I’ve explained why a “good city” for Amazon does not exist, I’ll explain what I think Amazon should be (and perhaps actually is) looking for.


There Are No Houses for Amazonians

Amazon is talking about 50,000 people over 10 years. Let’s assume at least 10,000 workers are relocated into the new city in the first 5 years. That means we need to find a city that can plausible create something like 10,000 housing units for Amazon’s workers over 5 years, and do so in excess of existing demand. That is, if the pre-Amazon economy of a city adds 5,000 households over 5 years we need to find a city that can add 15,000 housing units.

But let’s be charitable. Let’s assume Amazon’s entry will stimulate above-trend construction, and that there will be some displacement of current residents, and some hiring of existing households. Let’s say Amazon only needs net new housing for 5,000 worker-households.

Can any city provide a net increase in 5,000 housing units? Well, that’s a speculative question, but one thing we can do is ask if any city did produce 5,000 extra housing units beyond household formation over the last 5 years, using data from the American Community Survey. So basically, we want to look at change in the total number of housing units in a metro area minus the change in the total number of households. Has any city brought on excess supply in the last 5 years? And of course, given Amazon’s RFP, we must restrict to cities with at least 1 million residents.

Here’s a map of metro areas that have added at least 5,000 more housing units than they added households since 2010.

If they go to any other metro area, Amazon is making a bet that the metro they move to can substantially quicken its pace of housing supply expansion. That may be a reasonable bet. But the point is, the list of safe bets for sufficient housing supply is very short.

It’s also a really weird list. Charlotte is no big surprise. New York City kind of is: nobody thinks of NYC as having excess supply! But this may be catchup; NYC supply was very tight 2005–2013 when growth heated up, 2010–2015 new supply has come online, but 2013–2015 saw slower expansion in households. Plus, there’s a chicken-egg problem: slow housing supply growth can slow new household formation.

On the other hand, cities with very weak household growth make a strong showing: St. Louis, Baltimore, Philadelphia, Rochester, and Hartford all have extremely weak growth in the household population. They also have lackluster total population growth. But they have been adding new housing! Much of their housing supply may be dilapidated, but that might not be a big problem for Amazon.

Basically, this metric is a way to quantifiable demonstrating that population decline can give a metro area an advantage in absorbing a huge new investment. Now, Amazon may disagree… but Amazon’s workers may be very unhappy when the shock of 10,000 new employees cannot be absorbed by the housing market and prices shoot through the roof. And when Amazon finds that the salary differential they expected vs. Seattle vanishes within 5 years thanks to the need to compensate for the fact that their workers have to live in prefab housing tentcamps, they may be disappointed in their investment. I’m joking about Amazon tentcamps of course, but the point is that an investment in an area that is not keeping up with existing household formation may cause problems down the road.

But hold on. That chicken-egg problem about household formation seems significant. Is there some other way we can gauge housing supply?


The Houses for Amazonians Are Too Expensive

Amazon wants cheaper wage rates. Don’t believe me? Note that their RFP says:

It’s subtle, but note also the jobs listed are mostly back-office. In other words, Amazon is looking to relocate a lot of jobs that do not require the tech and creative agglomeration economies of Seattle, and where they can offer a lower prevailing wage and still recruit people. That is, if you’re an Amazon-quality coder, you’re mobile. You may prefer the Seattle wage-rate. You can command a high wage rate to boost your real income given some fixed nominal costs. But accountants? Accountants are geographically diffuse with less bargaining power with regards to geography. Amazon is hoping to get a break on wage rates for these folks. That means, they don’t want Seattle prices. That means they care about local prices.

So where are local prices low?

Let’s assume we need two things:

  1. Local prices at least 10% lower than Seattle
  2. Local prices exhibiting slower-than-metro-area-average inflation

Which cities with over 1 million people fit that bill?

We can use the BEA’s “Implicit regional price deflator” to assess these things. Here’s a map of qualifying cities:

Very cool map but hold on.

It has no overlap with the map of cities that actually have available housing.

Now, these cities’ low prices do suggest that they are able to robustly add housing, or at least have in the past. But at least some of that seems to be about comparatively weak demand, that is, yes, they’ve added lots of housing, they haven’t added lots of housing compared to household formation.

These cities all have a swell track record of keeping prices competitive. But we don’t really know that that means they could add the housing supply Amazon needs at the price that it wants. Though, sure, it seems plausible.

So on the housing supply front, there are cities that have produced excess housing units, and then there are cities that are priced competitively, and there is no overlap between them.

Let’s look at other factors.


There Is No Labor Pool for Amazon

Amazon wants to add 50,000 workers in the information technology, management, and administrative sectors over 10–15 years.

How big a shock would this be for the sectors in each metro area, and how does it compare to metro area population?

Well, forecasting is hard. So let’s do something charitable. Let’s look at growth over the last 15 years, and see what an extra 30,000 workers in these sectors would have done.

Let’s look for cities that meet two criteria:

  1. Cities that have above-average employment shares in these sectors
  2. Cities where this shock would not make up more than 25% of workers in these sectors
  3. Cities that already had above-average growth in these sectors

Basically, we’re looking for cities where this sector already shows some signs of life.

Okay, so now we have some new contenders out west, and some repeat-appearances. Atlanta, Nashville, Charlotte, Indianapolis, Phoenix, Columbus, and Orlando all showed up in the price map as contenders and also in this map. Now, none but Charlotte showed up as actually having a proven track record of adding excess supply. But the price chart may be more important.

But there’s a different way to think about the labor pool that may be more important: university systems.


There Are No Students for Amazon

We can use National Center for Educational Statistics to look at the pipeline of STEM students. What we will want to see is an area that has a large pipeline of STEM graduates, but also one where those graduates are a comparatively small share of pre-Amazon STEM-field employment. That is, a city with lots of STEM folks but even more demand for STEM folks may not be very good for Amazon, because they’ll be competing for those people. And while they may like there to be some ecosystem of STEM-graduates, one might presume Amazon likes being the big dog in town.

So, let’s say Amazon wants a metro area with at least 2,500 annual STEM degrees awarded and where that pipeline makes up at least 5% of IT and scientific jobs in the metro area.

This is yet another basically new map! Philadelphia and Rochester have shown up before on the first housing supply map. Minneapolis, Houston, Washington DC, and Raleigh-Durham are total newcomers. Detroit showed up on the price map. Chicago showed up on the previous labor force map.

San Jose also shows up. The problem there is that California metro areas are sort of small-ish here, so the strong demand from other Bay Area metros isn’t showing up. If we consolidated the Bay Area into one metro area of similar size to Washington, Minneapolis, or Chicago, it drops off this list.

So depending on how we specify the “available labor pool,” we can get a pretty different list of cities, although Chicago does show up on both.


No City Has Amazon’s Preferred Idiosyncracies

Amazon wants on-site transit access, direct flights to Seattle and DC, a robust international airport, and direct-to-customer fiber connectivity. This is trickier data to get your hands on. Let’s start by scoring up every city that’s made any of our criteria thus far, and then see how they do on these criteria. We’ll give 1 point for each list appearance, and an additional point if they manage to show up in a labor-pool list and a housing-supply list.

Let’s start with the key point of this whole blog post: no city has a perfect score (5), or even a just-shy-of-perfect score (4). The best cities score a 3, meaning they get one of the housing-related components and one of the labor-related components, and a bonus for having reasons to think both labor and pricing conditions may be good for Amazon. But no city manages to crush these rankings. They all have clear shortfalls. The only city that crushes even one side (pricing vs. labor) of the equation is Chicago, showing up in both labor pool specifications, but neither pricing specification.

Let’s now review each of these metro areas, and see if they meet other qualifications.

First off, let’s add a point if the metro area has the highest category of advertised download speed. Here’s a map:

Next, let’s add a point for a large airport. How can we define a large international airport? Well, I don’t want to get fancy with this, so I’ll just roll with the airports that the Bureau of Transportation refers to as “major airports.” I’m not going to bother with debating whether Atlanta should get more airport points than Detroit. If it’s a big hub airport with international flights then it counts. As a plus side, this encapsulates international flights and flights to Seattle/DC, since basically every major international airport in the US also has direct flights to DC and Seattle. Also I should note, even if a metro area has zero base points, I add it if it has a big metro area, and assign it the minimum of 0.5 base points, on the assumption that a big airport means I’m missing something.

Finally, we can look at transit systems. I’m going to use two measures of transit since Amazon specified both actual transit access and roadway access and distance from city center. The first measure I’ll use will be the share of metro area residents who use transit to commute. If it’s at least 5%, the metro area gets 1 point. Second, if average commute time for drivers driving alone is less than average for large metros, the metro area gets 1 point. Cities that manage to score in both categories will get 0.5 points extra, so there are 2.5 points possible. The only city to rack up 2.5 points is Portland. Finally, if a city scored both on local transportation and on their airport, I award another 0.5 points.

Then we have some even more idiosyncratic points. First of all, I dock 0.5 points for cities on the west coast, on the assumption that Amazon wants some kind of geographic diversity. Second of all, I give Boston and Washington, DC one point each because each has been publicly identified in the press as somewhere Amazon execs would like to live. Third, I award half a point to any metro area mostly located in a state won by Clinton in 2016, taking this as a proxy for the “cultural fit” Amazon’s RFP desires.

To make a final tally, I divide each metro area’s points by the total number of points that were mathematically possible.

Here’s a map of each metro area’s score after we take these factors into account:

Metro areas that manage to rack up scores over 50% are highlighted in red. There are just 3 such metro areas: Chicago, Philadelphia, and Washington. Chicago earns points for its strong labor pool for Amazon, its transportation and logistical infrastructure, and its presence in a blue state. Washington earns points for its educated workforce, transportation infrastructure, being in blue states, and the special “Jeff Bezos has a house here” point. Philadelphia earns points for a combination of good housing supply and a good talent pipeline, along with decent scores for infrastructure.

But here’s the thing. These best-scoring cities barely crack 50% of possible points. Philadelphia tops out at 56%.

No city in America has a shoe-in bid for Amazon incorporating everything they could want right from the start.

This Time Is (Maybe) Different

Most site selection processes are fairly predictable, and decisions are often made well before bids are even solicited. Amazon may be different.

The scale of Amazon’s investment is so large that no city can absorb it without growing pains. Its RFP is sufficiently specific that no city can match every criteria. Its national presence is sufficiently diversified that no city has a home-court advantage. These are all the sorts of things that suggest that this site selection process may be different.

We know Amazon has been considering this for a long time. We know Boston and Toronto have been in the running in the past. We know they were not selected at those times. This sounds like a failed site selection process!

My guess is that Amazon is throwing the floodgates of submissions open because they’ve already looked at top-line data for most cities and came home unsatisfied. This doesn’t mean there are “no front runners;” doubtless certain cities did make very strong showings. But the point is, my guess is that folks at Amazon have realized that, on some level, they’re a “White Whale” in every literary sense of that phrase. They can inspire obsessive pursuit. If things go poorly, they can be the death of their pursuers; i.e. if they select a site but it ends up a bad match.

That means that what Amazon is really looking for is something else: that special something. They’re looking, or if they aren’t they should be looking, not for a city with X number of college graduates, but idiosyncratically optimal bids. A particularly good site on the ground. A unique university partnership. A local government that puts particular effort into supporting a bid.

In this sense, incentives may actually make-or-break the Amazon deal, despite the fact that incentives usually don’t matter much in site selection. The reason they could matter for Amazon is not that Amazon is short of cash, but that the fundamentals are bad everywhere for Amazon. There are a solid list of cities that can get to 30% of what Amazon wants. There are several that can make it to 40%. A few can crack 50%. Nobody breaks 60%. Nobody can supply Amazon’s fundamentals. No matter where Amazon goes, they will have to build their own fundamentals.

So what I suspect they’re actually looking for is a city willing to partner with them in that task: a city looking to grow with Amazon, and have its growth path decisively altered by Amazon. Incentive offers are a way to signal such a willingness.

I won’t predict where Amazon lands. I’d be willing to bet money that Amazon lands somewhere in one of the cities I’ve highlighted… but that’s not saying that much. If somebody wants to get a betting pool where we select, say, 10 or 15 cities, I think that’s reasonably predictively interesting. But the reality is that there really are no far-and-away leaders in such a difficult set of asks from Amazon.

Check out my Podcast about the history of American migration.

If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!

Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!

I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m married to a kickass Kentucky woman named Ruth.

My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.