Wages, Standards, and Robust Labor Markets

This week’s EconTalk was a very interesting episode with James Besson discussing his new book, Learning By Doing: The Real Connection Between Innovation, Wages, and Wealth.

Let’s start with a common view of how the Industrial Revolution proceeded. The conventional view, or at least the mental model I have, is that throughout the period wages steadily rose commensurate with technological advances. But Besson argues that when you take a detailed look at the data from the US and UK, you don’t see this pattern. For instance, if you look at US textiles, one of the posterboard examples of early industrialization, what you see are several decades of stagnant wages despite increasing size and output of the market. It’s not until after the Civil War that you start to see increases in wages. What can account for this?

Besson claims that it wasn’t until this time the textile market became standardized enough to create a robust labor market in textile workers. How can this be, if the sector had been growing for decades before?

The key lies in the idea of human capital, the idea that through training and education workers build up internal means of production. Often the term is used rather loosely, which can mislead. Think of the new textile worker, coming in from the rural farm. She doesn’t have any training or experience, and so can’t command a high wage. She is hired and trains at the closest textile mill, learning the complicated movements of the loom. After a few years she’s mastered the equipment and become the most productive worker in the mill. We can say she’s build up substantial human capital, far more than her coworkers. But still, she hasn’t gotten a raise, and if she asks for one, she is routinely denied.

Why? Well, her human capital is firm specific. There were no standards for looms or other textile technology at this time, so her knowledge ties her to the mill she works at. The mill across the street uses totally different machinery, so she’s no better than a fresh hire off the street.

It wasn’t until historical forces led to standards in the textile that a robust labor market developed, because now workers’ human capital was transferable across firms. The threat of leaving for competitors now had bite, and wages rose to match what the workers were actually worth.

This idea is of course very relevant today, as we think about decades of stagnant wage growth for many workers in the economy. If the labor market is too fragmented, if workers’ human capital is too tied up in firm specific technologies, then firms have the upper hand in setting wages.

A modern day example of barriers to workers in the labor market are non-compete agreements. These are agreements workers have to sign when joining a firm that relinquish the workers’ ability to work at a competitor firm for some period of time after leaving their company. The idea is to protect trade secrets, so you see them often in the technology sector. However, there is an obvious side effect in that they tend to keep workers settled where they are, since leaving will keep you shut out of your industry for some years.

Differential enforcement of these non-compete agreements across states has led to researchers estimating the effects they have on innovation. While today Silicon Valley is obviously the king of the technology world, this wasn’t pre-ordained. Boston’s Route 128 was a high tech sector rival to Silicon Valley in the 80s, but Massachusetts enforces non-compete agreements whereas California does not. Without the cross pollination of ideas and incentives for workers to innovate, Route 128 stagnated while Silicon Valley raced ahead.

Unfortunately, these agreements have spread beyond the world of technology; famously, the sandwich chain Jimmy John’s forces its workers to sign a non-compete agreement. That’s right, if you work a minimum wage sandwich job you may be shut out of the sandwich labor market for several years if you leave, which is of course patently ridiculous.

What about occupational licensing? On the one hand, if having clear standards leads to robust labor markets, then having a licensing process should be a good thing right? Well, there a big difference between a certification from a standards body that a worker meets certain specifications versus a legal requirement that a worker must meet certain specifications (which almost always require some years of schooling from particular institutions). Furthermore, since these licenses are different from state to state, they inhibit labor mobility since the human capital is tied to a particular state.

All in all this was a very good episode with a lot of food for thought, and I look forward to reading the book.


Why Is Infant Mortality Higher in the United States than in Europe?

It’s common knowledge among the Aaron Sorkin segment of the population that the United States has a high infant mortality rate compared to other developed nations. But if you were to ask why, you probably wouldn’t get a good answer; even for researchers the reason has been a bit mysterious. A new paper in the American Economic Journal of Economic Policy provides some answers, but like any good research leads to more questions as well.

Depending on who you ask, as of 2015 the US infant mortality rate (IMR, defined as the number of deaths of infants under one year old per 1000 live births) is around 5.6~5.9. In the EU, the rate is closer to 3.7. In the grand scheme of things we’re actually doing pretty decent; below is a plot of  IMR vs GDP per capita (PPP) for most countries in 2014.



We can see that there is a pretty clear correlation (on the log scale) between the two. Above the trend line is higher IMR than expected for a given per capita GDP, and below is lower. Most rich OECD countries are well below the line; we are slightly above. However, our neighborhood north of the line is mostly rich oil countries like United Arab Emirates, Saudi Arabia, Kuwait, Qatar, etc.; countries that were it nor for oil wealth would be far poorer and shifted to the left. In other words, for an advanced economy we seem to be pretty lonely above the trend line. So what’s going on?

There are three channels that can explain this difference. As is the case with crimes, countries have differing reporting standards for infant mortality, making cross country comparison difficult. For example, an extremely premature birth that doesn’t survive may be counted as a live birth in some countries or as a miscarriage in others, which alters the deaths/live births ratio. A second confounding problem is the incidence of low birth weight (LBW), which is associated with lower survival rates. The US has higher rates of low birth weight than other developed countries, so it could be possible that a higher IMR is a compositional effect, which has its own policy implications. Finally, the timing of deaths is often not available in data. A concentration of higher IMR early on or later in the first year again has very different policy implications. Early on could suggest substandard medical infrastructure, while a random or later distribution indicates some deeper issues.

The authors of ‘Why Is Infant Mortality Higher in the United States Than Europe?’, Alice Chen, Emily Oster, and Heidi Williams, were able to obtain microdata on the US, Austria, and Finland that allowed them to investigate all three of these problems. Particularly, the novel thing was getting access to detailed timing information, which allow distinction between neonatal (within the first month) and post neonatal (months 2-12) deaths.

With regards to reporting differences, some of the higher US IMR is indeed due to differences in reporting. After controlling for this, the average excess US IMR compared with these countries is reduced by about 40%. They also found that higher rates of low birth weight could account for another significant fraction of the US disadvantage, but not all.

The rest of the excess deaths were located in the post neonatal time period, and the graphs in the paper are really quite striking. In LBW children, the three countries perform similarly. In the first month of life with normal birth weight children the US performs comparably to Austria and Finland. But thereafter the probability of death in the US pulls away significantly (all subsequent images are from the paper).




This suggests the high US IMR is not due to any deficiencies in US medical care or facilities; if that were the case, we should expect elevated rates throughout the first year as well as with LBW births. Instead, these deaths are occurring at home.

A  natural follow up is to ask whether this phenomenon is randomly distributed throughout the population or is concentrated in space and/or socioeconomic status. Using the US Census Divisions, they find that the Northeast region performs best, whereas the East South Central (Alabama, Kentucky, Mississippi and Tennessee) performs significantly worse than the rest of the country.




As you may suspect, these areas have very different levels of income, and indeed it seems the main source of the effect is from socioeconomic status.



This research could find no ‘smoking gun’ as to the cause of increased mortality in the first year of life. Again, issues with reporting and coding prevent any clear conclusions. SIDS is hard to correctly diagnose, and the other major causes are the nebulous ‘Accidents’ and ‘Other’.

In the United States the focus has historically been on reducing instances of LBW,  but in light of this paper we can conclude that policies aimed at reducing premature births and/or low birth weight will only go so far. The authors suggest that supporting home nurse visits for new mothers, as is policy in several European countries (including the two analyzed here), could be effective in reducing the infant mortality rate through the mechanism of increased parental knowledge. Clearly this is an understudied topic and more research is needed, particularly into the cause of these deaths.


What’s the Big Idea, Again? (tl;dr)

The United States has a higher infant mortality rate than many other comparably rich countries. This is partly due to differences in reporting and higher incidence of low birth weight, but also due to higher rates of infant mortality in the 2-12th months of life. This phenomenon is concentrated both geographically and economically, particularly among the disadvantaged in areas like the Deep South. Areas like New England, with their higher share of the economically advantaged, see essentially no difference in IMR compared with other rich European countries.


What Do We Do?

I know many people are uncomfortable putting human life in monetary terms, and understandably so. But in certain situations it can be illuminating. The authors use the value of a statistical life, most commonly quoted at around $7 million, to put this in context. Given approximately 4 million births a year, our IMR is ‘costing’ us about $84 billion each year. Which on the flip side, suggests it would be worth spending at least $7,000 per child to reach Austrian levels of infant mortality, and even more if we just focused on disadvantaged births. If that’s the calculus, supporting home nurse visits is a no brainer. And if you think a human life is worth far more than $7 million, then home nurse visits are even more of a no brainer.

It has been argued in the past that in the United States we care deeply about the life of a child until the moment they are born. After that, they’re on their own. And I think this research bears that out to some degree. It’s also another symptom of the increasing levels of inequality here. The advantaged among us, with jobs with generous parental leave, with leisure time to research the dos and don’ts of pregnancy and early life, with the means to afford outside caregivers, are doing fine. Yes, tragedies still happen, but no more than anywhere else in the world. The disadvantaged, the ones who work until they can’t stand anymore, who can’t even take full use of their 12 weeks of unpaid leave because they have to get back to work to pay the bills, are struggling. And in too many instances, failing.

As a society we should be doing more to support disadvantaged mothers and their children. We can argue over what the best course is, whether it is in the form of easy access to family planning services, professional health care in the home, or through more fundamental issues like education or family stability. But I don’t think we can argue that we do enough already. If life begins at conception, then we must also remember it continues after birth.

Labor Markets, Trade, and the China Shock


During this election season, trade protectionism has reared its head across the political spectrum; no candidate seems willing to believe (or admit to believing) that free trade is good for the country. The closest thing to a defense anyone can seem to muster is to claim that our current trade deals are in fact bad deals, which at least leaves the implication that good trade deals exist. While public polling suggests free trade is more popular than you might think given the political rhetoric, a substantial fraction of Americans are suspicious of trade, despite overwhelming support of free trade among economists.  A National Bureau of Economic Research paper from earlier this year sheds light on why this difference of opinion exists.


The China Shock: Learning from Labor Market Adjustment to Large Changes in Trade’, by David Autor, David Gorn, and Gordon Hanson, looks at the economic impact of trade with China on US manufacturing. To simplify immensely, they find that the costs of trade are higher than previously thought (but importantly, still a net benefit to both nations). To fully appreciate their findings we will need to set the stage with a simple model of trade impacts, but first I’d like to define what is meant by ‘the China shock’.


The example of China is basically without historical precedent in having a very large, cheap pool of labor enter the manufacturing sector in a very short period of time. After the death of Mao in 1976, China embarked on a series of reforms throughout the 80s and 90s that liberalized portions of the economy and which culminated in the country joining the World Trade Organization in 2001. This ensured the Most Favored Nation status China had enjoyed from the US and Europe throughout the 90s was made permanent, as well as opening other global markets to Chinese exports.  From about 1990 to 2000, China’s share of world manufacturing exports more than doubled from 1.9% to 5%, and in the decade and a half since joining the WTO has more than tripled again to almost 20%.  This veritable explosion of manufacturing capacity in just over two decades, coupled with low tariff access to most developed countries’ markets, is the China shock.


According to the authors, on the eve of China’s accession to the WTO, there was a rough consensus among economists on the impacts of trade: (1) Trade had not been a major contributor to the decline in manufacturing employment or rise in wage inequality in developed countries; these trends predated the era of trade liberalization and major deals like NAFTA seemed to have had little additional impact. (2) Workers in regions that specialized in hard hit areas of manufacturing (Pennsylvania steel, Detroit auto, etc.) could move to other regions without much trouble if trade eliminated jobs in their area. (3) Labor market impacts of trade would not be confined solely to trade exposed industries; due to the ‘law of one price’ for skill, these impacts would be spread across all similarly skilled workers.  


Under this set of assumptions (more or less the neoclassical model of trade), the concentrated impact of a particular trade shock is absorbed by the national labor market. Think of heating a pot of water with a flame; while in the instant the water closest to the flame is hotter than the rest, soon the heat is spread throughout the entire vessel evenly. In this simple model, labor of a particular skill level is like the water molecules: highly mobile and totally identical.


With this in mind, the impact of a trade shock at the regional level should be as follows (after allowing for some time for adjustment):





Manufacturing employment should decrease because the local industry is no longer competitive, leading to a pool of labor seeking work. They have two options: they can leave the area to find similar work elsewhere in the country and/or or reallocate to the local non-manufacturing sector in jobs with a similar skill level. As such, we would not expect any differences in the regional unemployment rate or labor force participation. Finally, the change in local wages (for all workers, not only manufacturing) relative to any changes in national wages should be zero (this follows from the assumption of one price for skill).


Now, as you might imagine, actually estimating the real world impacts of a trade shock on a region is very tricky. I’ll leave the complicated bits to the actual paper, but in a nutshell what the authors did was exploit the variation in regional manufacturing and the variation in exposure to Chinese manufacturing to make these estimates. For example, Tennessee has a large concentration of furniture producers, which is a sector where the Chinese had substantial exports. In contrast, much of Alabama has lots of heavy industry which did not experience much competition from China. Even within industry sectors, there a substantial amount of variation in exposure to the China shock, which gives the authors an additional level of control to make more precise estimates. (In econometrics, having such variation is crucial. Imagine trying to estimate the impact of an additional year of education on income with data that only specifies that an individual has completed high school or college). To compare with the table from earlier, qualitatively what the authors found was this (all non-zero results were highly statistically significant):





It should not be surprising that the employment in trade exposed manufacturing decreased; this jibes with the earlier prediction.  What is surprising, though, is every other factor did the opposite of what was expected. Workers did not leave the region or shift to the non-manufacturing sector. They remained in the area and joined the ranks of the unemployed or left the labor force altogether. And the average change in wages (again, for all workers) in trade exposed regions was negative, not zero. To return to the analogy of heating a pot of liquid, it’s as if instead of water the pot were filled with peanut butter.


What can account for this? One explanation could be that the because size of the China shock was orders of magnitude greater than any other previous trade shocks, we are still in an adjustment period. Under this view, the neoclassical model of trade is essentially correct and the findings here are due to an exceptionally long adjustment time required for the labor market to return to equilibrium.


Another possibility for long adjustment time is that the rise in the US trade deficit has reduced the demand for sectoral reallocation among US tradables. Implicit in the neoclassical trade model is the assumption that countries are engaging in balanced trade. Simply put, countries trade goods and services for currency, which is then used to buy back other goods and services. Money is neutral, desirable only as a medium of exchange. So if manufacturing sectors A, B, and C in the US shrink because China has a competitive advantage, we should expect sectors X,Y, and Z in which the US has an advantage to expand. But what has actually happened is that the US has consistently run a very large trade deficit that ballooned during the 90s.




On a yearly basis we import billions of dollars more than we export. And China is particularly responsible for this growth, as the figure below demonstrates.


We can think of the trade deficit as reallocating demand across time. For a variety of reasons, China is willing to give us lots of stuff without us giving them anything but dollars in return. Future US consumption has been pulled into the present, and eventually we will have to run a trade surplus with China to make up for it. In the meantime, the depressed demand for US goods has slowed the reallocation within the tradables sector that neoclassical theory would predict.

While this is an interesting idea, and probably plays some part in what’s been happening, I consider the most likely explanation (and the larger story here) to be that US labor markets are not as flexible as we once thought. The neoclassical assumption of ease of labor mobility seems to be particularly under fire in recent research. As with the China shock,, research regarding the aftermath of the Great Recession also suggests that people find it hard to leave economically depressed areas.  Furthermore, income convergence among US states has declined due to rising housing costs in the highest productivity areas. For the hundred years from 1880 to 1980, people in the US migrated from poor areas to rich areas. Nowadays, if people move at all, they tend to go to areas where housing costs are the lowest (which tend to be poorer). In this vein, I highly recommend Ryan Avent’s The Gated City and/or Matt Yglesias’ The Rent is Too Damn High. Other research regarding the aftermath of the Great Recession also suggests that people find it hard to leave economically depressed areas.


The transferability of human capital across industries is probably lower than assumed in the neoclassical model as well. Frictions like occupational licensing, in addition to reducing interstate mobility, make it harder for people to switch to other kinds of relatively low skilled labor without spending significant time and money on certification (jobs requiring a license in some states include florist, barber, and scrap metal recycler).  


To sum up, while the costs of trade are more concentrated and more intense than was previously thought, it is still unambiguously the case that trade with China has been a net good for humanity. Hundreds of millions of Chinese have been lifted out of poverty, and Americans have been able to enjoy substantially cheaper goods. The real issue here is how to deal with the distributional impact of trade to ensure everyone can benefit, which I hope to discuss in a later post.