Below you’ll find summaries of all six new papers published in the Spring 2018 edition of the Brookings Papers on Economic Activity, along with—where applicable—links to related information and data available for download and use.

If you are a member of the media with a question about the research, would like to request additional data not made available here, or would like to interview an author(s), please contact Economic Studies Media Relations Manager Shannon Meraw at [email protected] or 202-797-6414.

This page was last updated: 3/8/2018

Automation has not reduced aggregate employment—rather, it has created more jobs than it has displaced—but the fruits of these innovations have not raised labor earnings by nearly as rapidly as they’ve raised productivity, according to new research by MIT’s David Autor and Utrecht University’s Anna Salomons.

“Is automation labor-displacing? Productivity growth, employment, and the labor share”
David Autor and Anna Salomons
Read the full paper | Read the blog | Download the data

Using nearly five decades of data from 28 industries in 19 countries, Autor and Salomons examine two questions. The first is whether rising productivity through technological progress ultimately diminishes employment. The authors find the opposite: productivity growth has been employment-augmenting in net throughout these decades. While the data make clear that industries experiencing large productivity gains pare their headcounts, the productivity spillovers accruing to customer industries (what the authors call “Costco effects”) and to overall consumer spending (what they call “Walmart effects”) more than offset these direct industry-level losses—so employment rises in net. While this conclusion is at odds with popular accounts, it is consistent with centuries of historical evidence as well as canonical economic wisdom.


Less expected is Autor and Salomons’ second finding: even as automation has enhanced the size of the economic pie, it has increasingly shrunk the size of labor’s slice. To be clear, automation does raise total worker earnings. But in recent decades, it has not raised earnings as rapidly as it has raised productivity—meaning that labor’s slice of the pie shrinks. And this finding holds whether automation is measured by productivity gains, by industry-level patenting flows, or by adoption of industrial robotics. Distinct from the case for employment, industry-level losses in labor’s slice of income are not fully offset by either “Costco” or “Walmart” effects, meaning that labor’s share falls in aggregate. These labor share-displacing effects of productivity growth were essentially absent in the 1970s, but became more pronounced in the 1980s and 1990s, and increased again in the 2000s.

Why has the relationship between automation, productivity growth, and labor’s share of national income shifted across decades? One hypothesis is that the rise of “winner-take-most” competition is reallocating production from small, labor-intensive firms to larger, more efficient, but more capital- and machinery-intensive “superstar” firms (e.g., from mom and pop retail outlets to big box stores, and then from big box stores to online megastores). A complementary hypothesis is that new waves of automation technologies are increasingly ‘task replacing’—meaning that they directly reallocate job tasks from workers to machines.

Whatever the underlying cause, this labor-displacing phenomenon is most pronounced in manufacturing, where labor’s slice of the pie has been falling for decades. But it is now present in every broad sector except, notably, in education and in healthcare, which remain stubbornly—or perhaps reassuringly—labor-intensive.

The slow-down in economic convergence across the U.S. and the concentration of long-term male joblessness in pockets of despair requires place-based policies that target jobless hotspots, according to new research published in the Brookings Papers on Economic Activity (BPEA) by Harvard’s Benjamin Austin, Edward Glaeser, and Larry Summers.

“Saving the heartland: Place-based policies in 21st Century America”
Benjamin Austin, Edward Glaeser, and Lawrence Summers
Read the full paper | See the maps on regional divides | Download the data

The rates of non-employment for working aged men 25-54 years old have almost tripled over the past 50 years, and today over ten percent of men have not been employed in the past 12 months. Many measures of well-being - such as mental health and opioid abuse - suggest that joblessness is a far worse outcome than low-income employment, which motivates a focus on employment rather than incomes.

America’s regions have long displayed enormous economic disparities, but for most of the 20th century, poorer states were catching up rapidly.  However, in recent decades the rate of convergence has slowed and pockets of long-term joblessness have emerged. Today, regional disparities in joblessness are large and enduring.  Only half of the men aged 25-54 in Flint, Michigan were employed in 2016 compared to  95 percent of men in Alexandria, Virginia.

The authors re-examine the traditional skepticism of American economists towards place-based policies. They argue that it is time to reconsider place-based policies because (i) the historic convergence of regions has slowed or reversed in recent decades (ii) increasing social problems are linked to a lack of jobs rather than a lack of income and place-based policies may do more than individually targeted policies to create jobs for the non-employed, and (iii) a modest body of evidence suggests that increasing the demand for labor has a materially greater impact in depressed areas.

The authors document evidence that increases in labor demand appear to have greater impacts on employment in areas where non-employment has been historically high, suggesting that subsidizing employment in such places could particularly reduce joblessness.  Pro-employment policies, such as a ramped up Earned Income Tax Credit, that are targeted towards regions with higher joblessness and more responsive labor markets, however financed, could plausibly reduce suffering and materially improve economic performance.

Other key findings:

  • America’s social problems, including joblessness, disability, opioid-related deaths and rising mortality, are concentrated in America’s eastern heartland, states from Mississippi to Michigan, generally east of the Mississippi and not on the Atlantic coast.
  • Over the past 40 years, migration has stopped flowing to high-income regions and declined more generally. Since 2007, the share of American residents who moved across counties has never exceeded 3.9 percent. The slowdown in migration may be contributing to the hardening of America’s geographic divisions.
  • Place-based tailoring of policies to particular locales may be more effective than simple large-scale transfers to distressed areas that do not take into account local circumstances.

Virtually all gains in spending on the social safety net for children since 1990 have gone to families with earnings. The poorest children, whose parents have no earned income, have seen a decrease in benefits from $48 billion in 1990 to $32 billion in 2015, according to new research from Hilary Hoynes of the University of California, Berkeley and Diane Schanzenbach of Northwestern University.

“Safety net investments in children”
Hilary W. Hoynes and Diane Whitmore Schanzenbach
Read the full paper | Read the blog | Download the data

Previous research shows having access to social safety net programs in childhood has positive long-term benefits in adulthood and society as a whole. These studies have shown early childhood shocks—access to nutrition, maternal stress, exposure to alcohol and tobacco, and environmental toxins and public health interventions—can have impacts on later-life health and labor market outcomes. However, spending on children has remained relatively flat at 2 percent of GDP, while spending on adults and the elderly has grown substantially, from 5 percent of GDP in 1975 to about 8 percent of GDP in 2016. In addition, within the funds allocated to children, an increasing share is going to children near and above the poverty threshold, with a decreasing share going to children living below the poverty threshold.

This paper looks at what groups of children are served by core childhood social-safety net programs, including Medicaid, EITC, CTC, SNAP, and AFDC/TANF, and how that has changed over time. It builds upon previous research by being the first to rely almost exclusively on administrative data to analyze data by poverty status and work status.

The authors find that in 1990, the vast majority of safety net spending on children was received by families receiving cash welfare. Following the reform of the welfare system under President Clinton in 1996, there has been a shift in spending toward work-contingent programs and away from traditional out-of-work assistance. Today there is minimal cash welfare spending and the vast majority of safety net expenditures are distributed via work tax credits and health insurance (as well as food stamps).  As a result, the share of safety net spending on families with incomes below the poverty level has fallen from 86 percent in 1990 to 53 percent in 2015. This has been replaced by an increase in the share going to families with income in 100 to 150 percent of the poverty line and to a lesser extent 150 to 200 percent of the poverty line. Looking at all social spending going to families without earnings, the picture is even bleaker: spending has fallen from almost 70 percent of spending in 1990 to 20 percent in 2015.

The research shows that welfare reform and the decline in cash assistance has been fully felt by those with the lowest incomes. More than half of the increased spending for the EITC and more than three-quarters of the increased spending for the CTC goes to those with income between 100 to 200 percent of the poverty line. Most of the gains in Medicaid spending are also going to those above 100 percent of the poverty line. However, SNAP spending has remained relatively consistent and important for low-income families.

The authors conclude these shifts in childhood social safety net spending result in less protection from negative social and economic shocks to disadvantaged families. Building the safety net around work is problematic when jobs are lacking, and spending on tax credits provides little protection in the case of an economic downturn. The research showing the long-term positive effects of access to social safety net programs in childhood suggests this vulnerability will result in worse adulthood health and economic outcomes.

A boom in nonbank mortgage lending has exposed the mortgage market to unacknowledged liquidity risk, according to new research by You Suk Kim, Steven M. Laufer, and Karen M. Pence of the Federal Reserve Board and Richard Stanton and Nancy E. Wallace of the University of California, Berkeley.

“Liquidity crises in the mortgage market”
You Suk Kim, Steven M. Laufer, Karen Pence, Richard Stanton, and Nancy Wallace
Read the full paper | Read the blog | Download the data

The financial crisis of 2008 resulted in increased regulation of the credit risk associated with mortgage lending, such as stronger underwriting standards. But according to the authors, liquidity risk associated with the nonbank mortgage sector was also a key driver of the crisis—and those same vulnerabilities are not only still present, but pose an even greater risk to the system today because the nonbank sector is an even larger part of the market. As of 2016, non-bank financial institutions originated close to 50 percent of all mortgages and 75 percent of mortgages with explicit government backing. The authors caution that, considering the substantial risks at-hand, “researchers—as well as many mortgage-market monitors and regulators—do not have the information needed to assess the risks of this sector.”

Nonbank mortgage originators and servicers—i.e., independent mortgage companies that are not subsidiaries of a bank or a bank holding company—are subject to far greater liquidity risks but are less regulated than bank-lenders and servicers. On the origination side, nonbanks are subject to liquidity risk because they are funded by warehouse lines of credit that can be quickly pulled in time of financial stress. On the servicing side, nonbanks are subject to liquidity risk because they have to advance funds to mortgage investors even when mortgage borrowers skip their payments.  Nonbanks cannot respond to liquidity strains by tapping the liquidity facilities available to banks, such as from the Federal Reserve System or the Federal Home Loan Bank System. These liquidity strains contributed to about half of nonbank mortgage originators—around 1,000 companies—going out of business during and after the financial crisis.

The risks are particularly acute for mortgages in securitizations guaranteed by Ginnie Mae. The servicing contract for these loans places substantial financial responsibility on the servicers, yet the limited data available suggest that the nonbanks that concentrate on servicing these loans have less ability to withstand liquidity strains.

The authors find that many nonbanks place a larger focus on refinance and lower-credit-quality mortgages than banks, putting them at a higher risk from macro-economic shocks. The focus on refinance mortgages makes these nonbanks particularly vulnerable to increases in interest rates. For each of the three largest nonbank mortgage lenders, refinances accounted for more than 70 percent of their 2016 originations. Another four of the 25 largest nonbank mortgage lenders relied on refinances for more than 90 percent of their total originations in 2016. Particularly striking is the concentration of refinance loans in the Ginnie Mae market, where 41 percent of all nonbank originations in 2016 were refinances, compared with 30 percent for banks.

The research suggests that mortgages originated by nonbanks are of lower credit quality than those originated by banks, making them more vulnerable to delinquencies triggered by a fall in house prices, through the higher costs of servicing delinquent loans. A larger fraction of nonbank originations are FHA or VA mortgages, which tend to be riskier than other types. Among mortgages in Ginnie Mae pools, the data indicate that as of the fourth quarter of 2017, 3.6 percent of mortgages originated by nonbanks were two or more months delinquent, compared with just 1.8 percent for bank-originated mortgages. The authors also find that borrowers with mortgages from nonbanks are more likely to have characteristics associated with a higher probability of default than borrowers with mortgages from banks. Nonbank borrowers have higher loan-to-value ratios on their mortgages, are more likely to be delinquent on their bills, are more likely to have lower incomes and wealth, are less likely to hold Bachelor’s degrees, and are more likely to be non-white. Due to the lower credit quality of loans being originated by nonbanks, a rise in defaults would probably hit nonbank lenders and servicers particularly hard, as happened in the years leading up to the financial crisis.

In the event of an adverse economic event, nonbanks have limited resources to draw upon and the government would probably bear the majority of increased credit and operational losses. Further, failure of these nonbanks could result in a considerable contraction in mortgage credit availability, which would have a disproportionate effect on lower-income and minority borrowers, who are more likely to receive mortgages from nonbank lenders.

The authors conclude: “Although various regulators are engaged in micro-prudential supervision of individual nonbanks, less thought is being given, in the housing-finance reform discussions and elsewhere, to the question of whether it is wise to concentrate so much risk in a sector with such little capacity to bear it, and a history, at least during the financial crisis, of going out of business. We write this paper with the hope of elevating this question in the national mortgage debate.”

In a new paper that considers several proposals for redesigning the current mortgage market to provide more efficient debt relief to homeowners in times of crisis, Columbia University’s Tomasz Piskorski and Stanford University’s Amit Seru stress the challenges posed by the varying nature of economic conditions across the country over time and how those conditions interact with housing markets.

“Mortgage market design: Lessons from the Great Recession”
Tomasz Piskorski and Amit Seru
Read the full paper | Read the blog

Seeking to aid homeowners during the financial crisis, the government intervened by lowering interest rates to historic levels and creating several debt relief programs like the Home Affordable Refinancing Program (HARP) and the Home Affordable Modification Program (HAMP). However, prior research suggests these efforts had mixed success. In particular, the rigidity of mortgage contract terms, as well as a variety of frictions in the financial intermediation sector, hindered pass-through of lower interest rates to struggling borrowers and efforts to restructure or refinance household debt.

The authors discuss a number of proposals for more efficient risk sharing between borrowers and lenders to lower the incidence of costly foreclosures and the severity of future housing downturns. These proposals include relying on the automatic indexation of mortgage payments to economic conditions, which has the potential to circumvent financial intermediary frictions and provide quick debt relief to struggling homeowners during economic downturns, as well as “post-crisis” debt relief policies.

The authors document that large variations in economic conditions across the country over time pose significant challenges to the effective design of new solutions. These differences not only include significant variation in obvious factors such as unemployment and house prices, but also significant variation across regions in other factors related to the details of mortgage contracts and the ability of financial intermediaries to provide debt relief. Such factors also turned out to play a very important role in the extent of the crisis and pass-through of debt relief across regions during the Great Recession. For example,  an area with many loans eligible for federal debt relief programs may not have gotten much benefit from the programs because local banks did not have capacity to implement such relief. Moreover, the pass through of low interest rates to households crucially depended on the local share of adjustable rate mortgages (ARMs), with areas with many ARMs benefiting more once national interest rate indices reached a historic low.

While the authors do not propose a single “bullet-proof” solution they discuss a number of lessons that can be drawn from this evidence. First, national level “one-size fits all” polices (e.g., monetary policy) are likely to be a quite blunt tool, as not all areas need the same amount of stimulus at the same time. Second, the authors stress the importance of using local rather than national economic conditions as a benchmark for mortgage contracts or debt relief polices; because housing and labor market downturns affect different parts of the country to such varying degree. For example, it makes sense to reduce mortgage payments more in the areas where local labor markets and housing markets experience a downturn, i.e. where unemployment rises and house price growth is negative.

Third, the authors stress that while such an indexing solution is quick, it may not be perfectly accurate. It requires knowledge of the signs and risk of a downturn before it happens. But the strength of the relationships—between defaults and house price growth and defaults and unemployment—varies over time, which would impact the benefits of such indexation schemes at any given point in time.

Their analysis highlights an important trade-off. While indexed mortgage contracts allow for quick debt relief, their implementation requires policymakers and academics have good “ex-ante” knowledge (i.e. knowledge before a crisis occurs) of the underlying risk structure and its relation to indices being used when designing such contracts. Furthermore, the authors present a simple framework, which illustrates that errors in beliefs about the structure of risk can significantly reduce the benefits of such solutions. Given the significant differences across regions, such errors are likely, especially as a major change in the nature of mortgage contracts or housing policy can on its own significantly alter relationships between market equilibrium outcomes in a potentially hard to quantify way. As a result, policymakers may have to also rely on “post-crisis” debt relief solutions that do not rely as much on a good ex-ante or pre-crisis understanding of the distribution of risk. However, post-crisis post solutions come with the possible cost of delaying debt relief and reducing its effectiveness due to the various frictions including political and financial intermediary constraints.

The authors conclude: “Our evidence suggests that effective mortgage design approach to debt relief requires more in depth analysis of the nature of relevant income and housing risk and its evolution across regions and borrowers. This could include the development of new and sufficiently granular indices on which such contracts could be based. The recent ‘big-data’ revolution is promising in this regard. Such analysis could also identify mortgage designs and polices that effectively implement debt relief across a range of possible environments.”

The 2017 Tax Cuts and Jobs Act is estimated to raise the GDP growth rate by 0.9 percentage points a year over the next two years, finds Harvard University’s Robert Barro and Jason Furman as part of a broader analysis of the long-term economic impact of the tax changes.

“Macroeconomic effects of the 2017 tax reform”
Robert J. Barro and Jason Furman
Read the full paper

The paper, which analyzes the macroeconomic impact of the tax changes and estimates the short-term impact on short-run economic growth, represents a reconciling of ideology between its authors.

The bulk of the author’s paper examines the long-run impact of the 2017 tax law. The authors find that the impact is very different for the law as written, as compared to what would happen if all of the provisions in effect in 2019 were made permanent. For the law as written, the long-run increase in productivity in the corporate sector would be 2.5 percent, which translates into an additional 0.4 percent increase in GDP after ten years—or an increase in the growth rate of 0.04 percentage point per year. If the 2019 provisions of the law—including expensing of equipment investment—are made permanent, long-run corporate productivity would increase by 4.7 percent. This translates into an additional 1.2 increase in GDP after ten years—or an increase in the growth rate of 0.13 percentage point per year. The paper also examines the sensitivity of these estimates to alternative parameters, including estimating that if interest rates rose as a result of fiscal crowd out they would be somewhat smaller—with the tenth year GDP increases being 0.2 percent and 1.0 percent for the two scenarios respectively.

The authors also use a regression-based methodology to assess the short-run impact of the tax bill, finding that it will reduce marginal tax rates on labor income by 2.3 percentage points, which historically would be associated with a 0.9 percentage point increase in the growth rate per year for 2018-19, largely reflecting labor supply—although the magnitude of the impact could vary from the historical pattern based on the details of the tax cut and the current economic situation.

The authors do not address welfare implications or issues of fairness with the bill, but they do address several qualitative effects that the neoclassical model of economics does not capture: quality of capital, spillovers associated with investment, international considerations, and the impact on government spending.

The authors conclude future changes to the tax law are inevitable due to the number of expirations in the law and the likelihood of closing the gap between projected revenue and spending with additional revenue. Additionally, the authors agree on the ideology that the only way to cut taxes is to cut spending, but disagree on the proper level of spending. The authors agree, however, that a tax system that provides expensing for all investment—potentially with other changes—could be very positive for economic growth going forward, while also raising more revenue over the longer run.