Moving Toward Stagnation

I wanted to gather three recent posts from Free Exchange in one place. They amount to a sort of wonky restatement of some of the key arguments in The Gated City. First:

ED LUCE had in piece in yesterday’s Financial Times on America’s labour market, which attracted quite a lot of attention. Here’s one interesting snippet:

Finally, a growing share of whatever jobs the economy is still managing to create is in the least productive areas. Of the five occupations forecast by the Bureau of Labor Statistics to be the fastest growing between now and 2018, none requires a degree. These are registered nurses, “home health aides”, customer service representatives, food preparation workers and “personal home care aides”.

Manufacturing is nowhere in the top 20, and such jobs cannot replace the pay and conditions once typical of that sector. “The food preparation industry cannot sustain a middle class,” says Dan DiMicco, chief executive of Nucor, one of America’s two remaining big steel companies, whose company motto is “a nation that builds and makes things”.

Matt Yglesias notes that the emphasis on the importance of “manufacturing” is a bit foolish:

To understand this problem, you need to start with the fact that if I build a factory where people take fresh peas and put them in cans that’s a “manufacturing” facility full of manufacturing jobs and people who “make things.” But if I build a facility where people take fresh peas, mix them with some basil and a touch of mint, plus olive oil, parmigiano reggiano, and pine nuts then purée them to serve you a delicious pea pesto that’s a lowly service sector employment cite that couldn’t possibly generate good jobs. Similarly if I make pasta then dry it and stick it in boxes, I’m manufacturing. If I make fresh pasta and serve it to you on a plate with my pea pesto that’s services. The difference between manufacturing and services is not an ontological void between making things and not making things. It’s really a gap between putting things in boxes and not putting them in boxes. Like if you build a bookshelf and ship to a store and I buy it, that’s manufacturing. If I hire you to come to my house and install custom built-in shelves, that’s services.

I’m happy to see both Mr Yglesias and Kevin Drum note today that while the distinction between manufacturing and services is often meaningless, the distinction between tradability and non-tradability of products is most certainly not.

Tradable goods and services can, by definition, be consumed well away from the point of production. The international price of tradable products is therefore constrained within a fairly small range; you can try to sell a product for much more than its foreign equivalent, but don’t expect anyone to buy it. What this suggests is that real income differences across countries are largely attributable to differences in productivities within the tradable sector. This finding is associated with what economists call the Balassa-Samuelson effect, after economists Béla Balassa and Paul Samuelson.

In order to earn a higher wage than a worker in another country producing goods that trade at a more or less equal price, an employee must be more productive. The higher wage in the tradable sector will lead to a rising wage for workers in non-tradable sectors—that is, those producing non-transportable products like haircuts for local economies—as local firms must pay a competitive wage to attract employees. An overall higher level of income in an economy, in other words, is only possible thanks to higher productivity in the production of tradables.

The trouble, as Mr Luce rightly points out, is not necessarily that America is losing jobs in manufacturing. It’s that it is failing to create jobs in the tradable sector. Almost all net new job creation in America over the past 20 years has occurred in non-tradables: things like health care, for instance, or education. This is potentially a very serious issue. If America isn’t creating new jobs in the tradable sector, it is presumably because creation of such jobs is economically problematic: expected returns from worker outputs are less than the expected cost of hiring the worker. Put differently, it would seem that American productivity growth has not kept up with American labour costs across the economy as a whole.

Now, this isn’t necessarily the explanation for sustained high unemployment in America. Normally, we’d expect American wages to fall, either through nominal wage declines or a weakening of the exchange rate, until American labour costs are back in line with productivity and the market for labour clears. To explain unemployment, we probably need to look at a breakdown in that adjustment process. If the problem one is aiming to diagnose is one of prolonged stagnation in earnings, however, then this is an important dynamic to examine.


LET’S talk a little more about production of tradables and American stagnation (how’s that for an attention-grabbing lede?). Recently, I mentioned that productivity in the tradable sector is important, because it essentially governs the real incomes an economy can pay. And I noted that over the past two decades, virtually all of America’s net job creation has occurred in the non-tradable sector, which seems problematic. It’s probably useful to dig into this a bit more.

Mobility within the American economy is quite high. The rate of migration has declined in recent decades, but it is still quite common for households to live in multiple cities around the country over the course of a career. Because there are high levels of mobility within America, economists assume that real wages adjust across the economy so that the marginal resident of a city is indifferent between staying or moving out. Imagine a marginal resident of New York, for instance, who earns a high wage but also pays a lot in rent. If his wage drops or his rent rises (or if rents or wages change elsewhere) so that his purchasing power is reduced relative to what he might earn in, say, Dallas, then we assume he’ll probably move to Dallas. And if there’s a big gap, then we’d assume that a migratory flow between the two cities would occur until the marginal resident is once again indifferent. This isn’t a smooth, frictionless process in the real world, but it’s probably not a bad approximation for how things work.

Now, in order for the marginal resident to be indifferent between high-wage San Jose and low-wage Phoenix, the cost of living must be very different in the two places. As it happens, lots of things on the coasts are more expensive than they are across the Sunbelt, but the most striking and important gap is in housing costs. The median value of an owner-occupied home in San Jose is about 3 times that in Phoenix and almost 5 times that in Houston. The reason for this is fairly straightforward. Wages are much higher in the former metro than in the latter two. Population in San Jose should therefore grow until costs there rise to eliminate the gap in real living standards. This could occur through a rise in congestion costs or through population growth substantial enough to bid down nominal wages. As it happens, rich coastal cities tend to tightly restrict growth in housing supply, which quickly translates high demand into high home prices. Housing costs act as a lid of sorts, adjusting so that existing home supply is occupied, with the marginal resident indifferent between staying and going. Again, in practice, this isn’t a clean process. Last decade, home prices rocketed up along the coasts, and a stream of households flowed from the coasts to cheaper Sunbelt cities, all part of the mechanics of leveling out big real wage gaps.

While the marginal resident of Phoenix and San Jose is assumed to be indifferent between the two cities, however, there is still a real productivity gap. Housing costs across the Sunbelt have to be low to attract workers because wages are low, and wages are low because the productivity of the tradable sector in these cities is relatively low. That alone, however, shouldn’t impact the country’s ability to create jobs in the tradable sector. Productivity is lower in many Sunbelt cities, but so are wages.

Why, then, do we see very little net job creation in tradables, and lots in non-tradables? One possibility is that there are transfers across the economy which bid up wages in the non-tradable sector of growing cities. Consider this map.

(You can see the results of a similarly motivated analysis here.) What we see is that federal government spending results in large and persistent net transfers from some states to others. Moreover, very productive states like Massachusetts, New York, Washington, and California subsidise low productivity Sunbelt locales like Arizona and Florida. It’s not too hard to imagine how this might work. Sunbelt states are attractive to lots of different people, but retirees are well represented among migrants to the south. Retirees receive a lot of federal money through Social Security, Medicare, and other programmes. This produces net transfers from productive states which help bid up wages in non-tradable sectors—like health services, which is among the nation’s fastest growing employment categories—above the level that productivity in the tradable sector would normally permit. At that wage level, it’s difficult for firms in the tradable sector of these fast growing cities to profitably employ people; wage rates are above that justified by productivity.

That dynamic alone may go a long way toward explaining America’s labour market difficulties, but we can take the analysis a few steps further. There are a number of factors that make productive metropolitan areas an attractive location for firms, but economists increasingly emphasise the role of knowledge spillovers. A number of pieces of data point toward the growing importance of these spillovers. The relationship between large, skilled cities and high levels of productivity appears to be tightening. Research consistently finds an important spatial dimension to measures of idea dispersion. Wage figures also make the point; talented workers are enjoying bigger wage increases within large, skilled cities.

Changing technology seems to be driving these changes. New information and communication technologies are increasing the returns to ideas by expanding the markets over which they can be applied. Skilled cities, which help develop and disperse ideas, are therefore becoming more important. It is these ideas that are responsible for much of the value creation in the economy. And if we go back and look at the paper by Michael Spence and Sandile Hlatshwayo on net job creation, we see that while non-tradable sectors have been responsible for nearly all of the economy’s employment growth since 1990, the tradable sector has generated the bulk of the economy’s increase in value added. We can draw a line between idea creation, value added, and productive metropolitan areas, but this line does not continue on to employment growth.

Now if we focus in on the labour force in a place like Silicon Valley, we see that worker productivity is not uniformly dependent on these spillovers. Some individuals enjoy a slight bump in productivity when they locate to Silicon Valley. Others, however, will find that their productivity level—and their compensation—is hugely dependent on the existence of these spillovers. It seems probable that these relative dependencies on spillovers are related to skill levels and job types. An accountant with a specialisation in the tax difficulties confronted by online businesses is likely to benefit from locating in Silicon Valley, but might be nearly as productive in Phoenix. A computer systems engineer developing new business models based on the cloud may find his productivity and earning power significantly reduced if he is relocated to a metropolitan area with far fewer firms and workers operating at the technological frontier.

As costs rise in Silicon Valley, which type will move away first? It seems clear that the workers most involved in—and most dependent on—these productivity enhancing spillovers will be those for whom the real wage trade-off with cheaper cities is least attractive. Both kinds of workers face a similar drop in costs when they move, but the spillover-dependent worker faces a disproportionate decline in his nominal income.

What this means is that while population is flowing from high productivity to low productivity cities, this is not generating a proportional transfer in productivity. Migrating workers—even those who continue to work in a tradable sector—are those that were least involved in the process of idea creation in their old city, and they therefore contribute little to the development of a new spillover cluster in their new city. There is a relationship between population growth and productivity growth, but it is skill dependent, and the skills upon which it depends are very underrepresented in the migratory flows from the coasts to the Sunbelt. Migration to the Sunbelt is therefore failing to raise productivity in tradable sectors to a level sufficient to justify new hiring at prevailing wage rates.

The picture that emerges is one in which employment growth in high productivity, tradable industries is constrained at the rate of housing supply growth in skilled cities. And that rate is slow; for much of the past decade, Houston approved about ten times more new housing each year than San Jose. Value creation in high productivity cities continues, but a lot of that value is siphoned off through taxes and transfered to residents of low productivity cities, who use it to buy non-tradable services. That dynamic would seem to be the main mechanism through which America has been generating net job growth over the past two decades.

At least in this story, which might well be mistaken. Hopefully other analyses will shed more light on the picture.

And third:

I JUST wanted to add one additional, brief thought on yesterday’s post on migration, productivity, and job growth. It was pointed out to me on Twitter that federal government transfers are not the only mechanism through which value generated in productive, tradable-oriented cities is redistributed to less productive, non-tradable-oriented cities. Profits that accrue in one location but which are spent elsewhere might have the same effect. Or consider this example:

As technological progress raises the productivity of skilled, coastal cities it raises the demand to live in those cities and, because housing supply is limited, the price of housing in those cities. Much of the economic value of working in those places is capitalised into local home prices. With increasing frequency, older workers are leaving the workforce, cashing out of their expensive homes in productive places, and moving to the Sunbelt. Having relocated from, say, San Jose to Phoenix, the retiree can afford a grand home at a fraction of the price, and the rest of the gain from relocating becomes a stream of income used to fund consumption, including an expanding array of health services. That, in turn, bids up the wages within the non-tradable sector, making it difficult to produce tradable goods in a cost-effective way. And once again, national employment in tradables expands no faster than the pace of housing supply growth in highly productive cities. Value added can continue to rise in those cities, however, but much of it is redirected to employment growth in lower productivity non-tradable sectors. And so we get a couple of decades of stagnation in tradable employment and median wages.


  1. Luke Lea says:

    “the fastest growing between now and 2018, none requires a degree. These are registered nurses, “home health aides”, customer service representatives, food preparation workers and “personal home care aides”

    In other words, servants.

  2. Luke Lea:
    Pretty darn close!!! I don’t know how it fits into it, but where does the explosion in executive pay fit into this? It has to, somewhere.

  3. Lori says:

    Yes – the explosion in pay is so that they can easily afford all these servants!

  4. Matt says:

    Um . . . registered nurses do need degrees.

    Also, according to The New York Times, productivity has risen steadily while wages have stagnated, so I don’t see how you can assert that “In order to earn a higher wage than a worker in another country producing goods that trade at a more or less equal price, an employee must be more productive.”

  5. Derek says:

    The author should be more specific about the term “Sun Belt”. The analysis here applies well to Florida; it does not apply well to Texas, where the big cities (DFW, Houston, Austin, San Antonio) have big, successful and productive tradeable sectors (think companies like TI, Dell and the vast energy sector in Houston).

    I would be interested to read an analysis that includes cases like TX: low cost cities with relatively high productivity tradeable sectors. Of course, that might be behind the rapid population growth in TX over the past few decades (sometimes obscured by oil price swings).