Erik Hurst
Erik Hurst
Studying cross-region variation in eco- Reserve policy may respond to aggregate nomic outcomes within a country is an shocks but not to local shocks. Ignoring increasingly common empirical research such general equilibrium factors can yield strategy in labor, public, and urban eco- estimates of local employment elasticities
nomics. 1 For example, local exposure to a given shock that are two to three times
Erik Hurst is a to trade with China has been shown to larger than the aggregate employment elas-
research associate in reduce local employment rates and wages. 2 ticity to the same shock. 7 Second, regional NBER’s Economic Fluctu-
comparisons ations and Growth, Public
Cross-state
cannot shed Economics, and Aging been used to
variation has
any light on Programs. He is the V.
shocks that Duane Rath Professor incidence of
estimate the
affect the of Economics at the local taxes. 3 entire econ-
University of Chicago’s Large local omy in the Booth School of Business
same way. and is a co-editor of programs
public work
Such shocks the Journal of Political
are “differ- Economy.
have been
enced away” His research spans ument the
used to doc-
when the dif- many fields, including existence of
ferent expe- empirical macroeconom-
riences of ics, labor markets, hous-
agglomera-
different ing markets, and consumer
tion econo-
mies in man-
regions are
compared. his Ph.D. in economics
behavior. He received ufacturing. 4 Figure 1
Much from the University of
Cross-
Note: Estimated wages are for white men, aged 40-44 who attended
of my current Michigan in 1999 and a
region variation
some college, working 40 hours per week
provides macro-
research uses a
B.A. in economics from economic researchers with richer informa- combination of local and aggregate data to Clarkson University in tion on economic fluctuations than aggre- learn about the drivers of aggregate busi- 1993. In 2007, he received
gate time series data at the national level. ness cycles and to explore the regional con- the TIAA-CREF Paul For example, during the Great Recession sequences of aggregate government poli- Samuelson Award for in the United States, some metropolitan cies. A combination of local and aggregate the best paper on life-
areas — Las Vegas, for example — experi- data, along with a structural economic long financial security enced larger declines in employment than model, often is needed to use local varia- for “Consumption vs. many others. Explaining this variation can tion to address macro questions. Expenditure” (with Mark
shed light on potential causes of the aggre-
Aguiar). In 2012, he was
gate recession. 5 Understanding the Causes
awarded the Ewing Marion
Drawing inferences about the aggre- of the Great Recession
Kauffman Prize Medal for
gate economy from data on regional vari-
Distinguished Research in
Despite aggregate employment rates Entrepreneurship.
ation is complicated by two issues, how-
ever. First, the way a regional economy falling substantially during the Great Erik currently lives in
responds to a given economic shock could Recession, aggregate real wage growth dur- Chicago with his wife and
be substantively different from the way ing the 2008-2010 period remained on two children.
a national economy responds because of its pre-recession trend. If employment fell both factor mobility and general equi- because of a labor demand decline, the librium forces. 6 Factors such as Federal employment decline during the recession
20 NBER Reporter • 2016 Number 3
NBER Reporter • 2016 Number 3 21
should have been accompanied by a decline in real wages. Many people who believe that the Great Recession was primarily caused by a lack of demand appeal to wages being “sticky” as the reason a decline in real wages during the Great Recession did not accompany
the sharp decline in employment. 8 [See
Figure 1.] In new work with Martin Beraja and Juan Ospina, I estimate the amount of wage stickiness using cross-
state variation. 9 Using a variety of data sources, we show that states with the largest rela- tive employment declines had the smallest relative wage increases. We con- struct state-level measures of real wages by combining state-level nominal wage data from the American Community Survey with state-level price indices constructed from scanner data. From this analysis, we estimate that wages are fairly flexible. While there is some stickiness to wages at the local level, real wages do respond to contempora- neous local labor demand shocks. [See Figure 2.]
To understand the broad causes of the Great Recession, we construct a model of local econo- mies that can aggregate to the national economy. The model allows for the trade of goods across local econo- mies, and for a national monetary authority that sets a common inter- est rate across local economies. We embed within the model four shocks:
a shock to households’ intertempo- ral consumption decisions, a shock to firms’ marginal products of labor, a shock to households’ current choices between market consumption and lei- sure, and a monetary policy shock. All but the monetary policy shock have both aggregate and local components.
Using data on both aggregate and local employment rates, prices, and nominal wages, we estimate
the shocks that drove both aggre- gate and local business cycles dur- ing the Great Recession period. We estimate the amount of wage sticki- ness using local data throughout the analysis. Our main finding is that the shock to household intertempo- ral consumption decisions — some- thing akin to a traditional Keynesian demand shock — explains at best only
40 percent of the employment decline during 2008–10 and essentially none of the persistently low employment
rate between 2010 and 2012. The degree of wage stickiness necessary for demand shocks to be the pri- mary cause of aggregate employment decline during the Great Recession is inconsistent with the flexibility of wages that we estimate from cross- state variation. Explaining aggregate wage patterns requires an aggregate labor supply shock. In our analysis, regional data are essential for identi- fying a key parameter that helps dis- tinguish between various shocks driv- ing aggregate business cycles.
Understanding the Decline in Employment Rates, 2000–15
What factors could have contrib- uted to an aggregate labor supply
shock during the Great Recession? A rising degree of skill mismatch could generate empirical patterns like those from a labor supply shock. For exam- ple, manufacturing workers may not
be able to fill jobs in the computer sector regardless of the wage being offered for the computer sector jobs.
In my work with Kerwin Charles and Matt Notowidigdo, we document the extent to which the secular decline in manufacturing employment during the 2000s contributed to the sharp
decline in employment rates experienced in the U.S. economy between 2000
and 2015. 10 Employment rates for men and women between the ages of 21 and
54 with less than a four- year college degree fell by roughly 2 percentage points between 2000 and 2007 and then fell by an additional 7 percentage points between 2007 and 2010. In 2015, employment rates for this group were still roughly 7 percentage points below the 2000 level.
The U.S. economy lost roughly 3.5 million man- ufacturing jobs during the 2000–07 period and
another 2 million manufacturing jobs during the 2007–10 period. In 2015, U.S. manufacturing employment was still roughly 5.5 million jobs below the 2000 level. The decline in manu- facturing employment in the U.S. dur- ing the 2000s was almost three times as large as the decline during the 1980s and 1990s.
A common question for struc- tural explanations of the employment declines since early 2000 is why so much of the decline was concentrated during the 2007–10 period. It is often argued that structural forces result in more gradual changes while cycli- cal factors can result in more abrupt changes. If manufacturing employ- ment declines are contributing to low employment rates in the U.S. econ-
Figure 2 Figure 2
Recession. Additionally, we show that more after QE in Dallas than in Las Charles, Notowidigdo, and I docu- employment rates were well above long- Vegas. Local refinancing booms were ment that the housing boom masked run trends in housing boom areas during correlated with local spending booms. the labor market effects of the declin- the early 2000s. At the national level, Collectively, the results show that effects ing manufacturing sector. As manufac- these two effects roughly offset. Part of monetary policy most helped regions turing declined, lower skilled men and of the masking we document occurred that needed help least. women were displaced, but the early because local areas were exposed differ-
While the Federal Reserve may not 2000s housing boom actually increased entially to manufacturing declines and
be independently interested in the dis- demand for construction workers, housing booms. However, we also docu- tributional aspects of monetary pol- mortgage brokers, real estate agents, ment that masking occurred at the indi- icy decisions, our paper highlights that and local services. This latter effect vidual level. If an individual who was the effectiveness of monetary policy in was due to a wealth or liquidity effect displaced from manufacturing in the stimulating consumption through home from rising house prices. As housing early 2000s lived in an area experiencing equity borrowing can be time-varying. prices increased, households increased
a housing boom, that person was more Monetary policy has a lower stimulus their spending on local goods. 11 Male likely to be re-employed than a displaced effect through this channel when the workers with less than a four-year col- manufacturing worker in an area with- economy is experiencing a housing bust lege degree moved sharply into con- out a housing boom.
and many homeowners are underwater. struction during the early 2000s.
By exploiting cross-region varia-
Simultaneously, female workers with tion, we find that the labor market was U.S. Mortgage Markets and less than a four-year college degree structurally weaker prior to the Great Cross-Region Redistribution moved sharply into real estate sales, Recession than had previously been rec- mortgage brokerage, and local services. ognized, and that the housing boom
The extent to which households can Using aggregate data, we docu- temporarily propped up labor market borrow to self-insure against regional ment that the decline in manufactur- statistics in aggregate data.
shocks depends crucially on how the ing employment for both men and
interest rate varies with regional eco-
women with less than a four-year col- The Regional Effects of
nomic conditions. Most economic
lege degree was almost completely U.S. Monetary Policy
models assume that regions within a offset by an increase in employment
monetary union share a common risk- in sectors spurred on by the housing
In addition to my work using adjusted interest rate. For most house- boom. In 2007, the housing market regional variation to learn about the holds, mortgages are the primary instru- collapsed and employment spurred by drivers of the aggregate macroecon- ment of borrowing. In a recent paper the housing boom collapsed with it. omy, I also have studied how aggre- with Ben Keys, Amit Seru, and Vavra, Manufacturing employment continued gate U.S. macro policies differentially
I show that the assumption of constant to decline throughout the recession. affect various regions of the country. risk-adjusted mortgage rates is well sup- Thus, a sharp downward trend in man- In recent work with Beraja, Andreas ported by data for mortgages securitized ufacturing employment coupled with Fuster, and Joe Vavra, we show that the by Government Sponsored Enterprises
a boom and bust in housing-related Federal Reserve’s policy of quantitative (GSEs). 13 Using loan-level data, we doc- employment caused aggregate employ- easing (QE) beginning in November ument that there is no spatial variation ment to remain relatively flat during 2008 disproportionately helped regions at all in mortgage rates and fees across the early 2000s and then to fall sharply of the country that were doing relatively U.S. metropolitan areas for loans secu- at the start of the recession.
well. 12 We document that right after ritized by the GSEs despite there being While the patterns are present in QE began, mortgage rates fell sharply. large ex ante predictable differences in aggregate data, by exploiting regional As rates fell, a mini refinancing boom default risk across those regions. All of variation the “masking hypothesis” can occurred. The increase in refinancing our results control for loan-level observ-
be illustrated clearly. Certain metro- was associated with individuals tapping ables like FICO score and loan-to-value politan areas — like Detroit — expe- into their home equity. However, the ratios. For example, GSE-secured loans rienced large manufacturing declines increase in refinancing was concentrated originated in 2007 in Las Vegas, con- and no housing boom. Other metro- in locations that had lower unemploy- ditional on borrower and loan observ- politan areas — like Las Vegas — experi- ment rates. The boom was much larger ables, faced the same mortgage rate as enced little manufacturing decline and in Dallas than it was in Las Vegas. The GSE-secured loans originated in 2007
a large housing boom. Exploiting cross- reason is that most homeowners in Las in Dallas, despite the Las Vegas loans,
22 NBER Reporter • 2016 Number 3
No. 21956 , February 2016. much higher predicted default probabil- “The China Syndrome: Local Labor
conditional on observables, having a 2 D. Autor, D. Dorn, and G. Hanson,
Return to text
ity, given that housing prices were falling Market Effects of Import Competition
8 See, for example, P. Krugman, to a greater extent there.
“Yellen, Wages, and Intellectual The patterns found within a sample Paper No. 18054 , May 2012, and The
in the United States,” NBER Working
Honesty,” blog post from August 25, of GSE loans differ markedly from the American Economic Review, 103(6),
patterns in a sample of otherwise similar 2013, pp. 2121–68.
Return to text
mortgages not securitized by the GSEs. Return to text
9 M. Beraja, E. Hurst, and J. Ospina,
“The Aggregate Implications of Regional sions aside from size. Loans above a cer- Lovenheim, “The Heterogeneous
These loans were similar in all dimen- 3 M. Harding, E. Leibtag, and M.
Business Cycles,” NBER Working Paper tain threshold are ineligible for securi- Geographic and Socioeconomic Incidence 21956 , February 2016. tization by the GSEs. Mortgage rates of Cigarette Taxes: Evidence from
Return to text
on these jumbo loans were higher in Nielsen Homescan Data,” American
10 K. Charles, E. Hurst, and M. areas where predicted mortgage default Economic Journal: Economic Policy,
Notowidigdo, “The Masking of the rates were higher. We show evidence 2012, 4(4), 2012, pp. 169–98.
Decline in Manufacturing Employment that the GSEs are bound by political Return to text
by the Housing Bubble,” Journal of
Economic Perspectives, 30(2), 2016, charging spatially different mortgage Economic Development, Agglomeration
constraints that prevent them from 4 P. Kline and E. Moretti, “Local
pp. 179–200; K. Charles, E. Hurst, rates conditional on borrower and loan Economies and the Big Push: 100
and M. Notowidigdo, “Housing Booms, characteristics.
Manufacturing Decline, and Labor If mortgage rates do not respond Valley Authority,” NBER Working
Years of Evidence from the Tennessee
Market Outcomes,” University of to local economic shocks that increase Paper No. 19293 , August 2013, and
Chicago Working Paper, 2016, http:// ex ante default risk, then households in The Quarterly Journal of Economics,
faculty.wcas.northwestern.edu/noto/ these regions face lower borrowing costs 129(1), 2014, pp. 275–331.
research/CHN_manuf_decline_hous- than they would if default risk were Return to text
ing_booms_mar2016.pdf. The prior
two papers are based on K. Charles, tion in borrowing costs may in turn off- Explains the 2007–09 Drop in
priced into interest rates. This reduc- 5 A. Mian and A. Sufi, “What
E. Hurst, and M. Notowidigdo, set some of the negative local economic Employment?” Econometrica, 82(6),
“Manufacturing Decline, Housing shocks that increased default risk in the 2014, pp. 2197–223. An earlier ver-
Booms and Non-Employment,” NBER first place. Conversely, those in regions sion of this paper circulated as “What
Working Paper No. 18949 , April 2013. with low default risk will face higher Explains High Unemployment? The
Return to text
borrowing costs than they would if this Aggregate Demand Channel,” NBER
11 A. Mian and A. Sufi, “House Price low default risk was priced into inter- Working Paper No. 17830 , February
Gains and U.S. Household Spending est rates. Thus, the constant interest rate 2012.
from 2002 to 2006,” NBER Working “policy” followed by the GSEs results in Return to text
Paper No. 20152 , May 2014.
state-contingent regional transfers. We 6 E. Nakamura and J. Steinsson,
Return to text
estimate that during the Great Recession “Fiscal Stimulus in a Monetary Union:
12 M. Beraja, A. Fuster, E. Hurst, and about $47 billion was transferred via Evidence from U.S. Regions,” NBER
J. Vavra, “Regional Heterogeneity and the mortgage market from regions with Working Paper No. 17391 , September
Monetary Policy,” Federal Reserve Bank smaller employment declines (above the 2011, and American Economic
of New York Staff Reports No. 731, June median) to regions with larger employ- Review, 104(3), 2014, pp. 753–92;
ment declines (below the median).
M. Beraja, E. Hurst, and J. Ospina,
Return to text
“The Aggregate Implications of Regional 13 E. Hurst, B. Keys, A. Seru, and
Business Cycles,” NBER Working Paper
J. Vavra, “Regional Redistribution
Through the U.S. Mortgage Market,” O. Blachard and L. Katz, “Regional
1 An influential example of this includes
No. 21956 , February 2016.
Return to text
NBER Working Paper No. 21007 ,
March 2015, and forthcoming in Activity, 23(1), 1992, pp. 1–76. “The Aggregate Implications of Regional American Economic Review. Return to text
Evolutions,” Brookings Papers on Economic 7 M. Beraja, E. Hurst, and J. Ospina,
Business Cycles,” NBER Working Paper
Return to text
NBER Reporter • 2016 Number 3 23
NBER News
Lisa Jordan Elected to NBER Board of Directors
At its September 2016 meeting, the NBER nomics, industrial relations, and race and gender eco- Board of Directors elected Dr. Lisa Jordan to a five- nomics. She currently serves as an adviser to the AFL- year membership term as an at-large director. Jordan CIO’s Commission on Race and is a past board member is director of education and membership develop- of the Labor and Employment Relations Association. ment for the United Steelworkers of America. She
Jordan graduated from the University of Notre and her staff create and coordinate the union’s educa- Dame with a Ph.D. in labor economics. She served tion program throughout North America and in the as a labor educator and assistant professor at the
Jordan United Kingdom. University of Illinois and at the University of Jordan is engaged in a range of applied economic Minnesota before becoming the director of the research. Her primary areas of interest are labor eco- School of Business at Brevard College.