The China Shock: Learning from Labor Market Adjustment to Large Changes in Trade David Autor David Dorn Gordon Hanson MIT and NBER Univ. of Zurich and CEPR UCSD and NBER May 6, 2016
The China Shock: Learning fromLabor Market Adjustment to Large Changes in Trade
David Autor David Dorn Gordon Hanson
MIT and NBER Univ. of Zurich and CEPR UCSD and NBER
May 6, 2016
Political Opposition to International Trade Is on the Rise
How Did We Get Here?
In the ’90s, freer trade—embodied in NAFTA & the WTO—werepitched as guarantors of future American prosperity
• Bill Clinton, NAFTA signing, Dec. 8, 1993
“I believe we have made a decision now that will permit us to createan economic order in the world that will promote more growth,more equality, better preservation of the environment, and a greaterpossibility of world peace.”
• The reality of how globalization would affect the US economyturned out to be a bit more complicated...
...And the Primary Complicating Factor Was China
02
46
810
Chi
na im
port
pene
tratio
n in
US
man
uf.
0
5
10
15
perc
ent
1991 1996 2001 2006 2011Year
China share of world manufacturing exportsChina import penetration in US manufacturing
The Economic Consequences of Trade
Economists have long known that international trade, whilegenerating aggregate welfare gains, also creates losers
• Paul Krugman and Maurice Obstfeld (1988)
“Owners of a country’s abundant factors gain from trade, butowners of a country’s scarce factors lose... international trade tendsto make low-skilled workers in the United States worse off—not justtemporarily, but on a sustained basis.”
• But until recently the consensus was that, in practice, trade justhadn’t mattered much for US labor-market outcomes
The Economic Consequences of Trade
The conventional wisdom among economists circa 2005
1 Trade had not been a major contributor to declining manufacturingemployment or rising wage inequality in the US
2 Workers employed in regions specializing in import-competingsectors could readily reallocate to other regions if displaced by trade
3 Any labor market impacts of trade would be felt by low-skill workersgenerally, not by trade-exposed workers specifically
The momentous impact of China’s recent export growth has helpedtopple this conventional wisdom
Agenda
1 What have we learned about the labor-market impacts of trade?
2 Should we be surprised by large earnings losses?
3 What’s the problem: Trade adjustment or trade itself?
4 Has the US policy response been effective?
5 Conclusions
What Are Impacts on Import-Competing Industries?
• It is perhaps unsurprising that jobs have been lost in industriessubject to more intense product-market competition from China
• Greater exit by manufacturing plants (Bernard Redding & Schott)
• Contractions in employment by plants that stay in operation(Acemoglu Autor Dorn Hanson & Price, Pierce & Schott)
• What has come as a surprise are the magnitudes of the impacts
• Every 1% increase in import penetration reduces industryemployment by 1.3%, with total manufacturing job losses fromtrade with China at 1 million (17% of the total for ’91-’11)
What Have Been Local Labor Market Consequences?
More surprising still is how US local labor markets have adjustedto the China trade shock (Autor Dorn & Hanson)
• Because regions vary greatly in what they produce...• Huntsville, AL, is (was) a textile manufacturer• Las Vegas, NV, is a mecca for tourism
• ... they were highly differentially exposed to greater importcompetition from China
• Among trade-impacted local labor markets in the ’90s and ’00s,Huntsville was in top quarter and Las Vegas in bottom quarter
Local Labor Market Exposure to Rising Import CompetitionCommuting zones by quartile of trade exposure
Adjustment to Trade in Local Labor Markets
Active margins of adjustment to greater import competition
• Reductions in manufacturing employment, increases innon-participation in the labor force and in long-run unemployment
Inactive margins of adjustment to greater import competition
• Little response of interregional migration to trade shocks
• Movement of labor across areas due to changes in labor demand isslow and incomplete (Blanchard & Katz, Glaeser & Gyourko, Yagan)
Local Labor Market Impacts of Import CompetitionEmployment impacts overall and by education
-‐0.6 % -‐0.6 % -‐0.6 %
-‐0.2 %
0.2 %
-‐0.5 %
0.2 % 0.1 %
0.3 %
0.6 %
0.3 %
0.8 %
-‐0.8 %
-‐0.6 %
-‐0.4 %
-‐0.2 %
0.0
0.2
0.4
0.6
0.8
All Educa4on Levels College Educa4on No College Educa4on
Imports from China and Employment Status of Working Age Popula4on within Commu4ng Zones (1990-‐2007)
Effect of an $1000 Per Worker Increase in Imports from China during 1990-‐2007 on Share of
PopulaBon in Employment Categories
Manufacturing Non-‐Manufacturing Unemployment Not in Labor Force
What Are Impacts on Specifically Trade-Exposed Workers?
Go back to ’91 and consider two observably similar workers(Autor Dorn Hanson & Song)
• Same age, gender, annual earnings, recent earnings growth, tenureat their employer, size of and wages paid by employer, etc.
• But one works in an industry that will, over the next two decades,be more exposed to the China trade shock
• Over the next 16 years, the initially more-exposed worker will have• Lower cumulative earnings• More churning between jobs• And higher uptake of disability insurance• With larger effects for low-wage workers
Impacts of Trade Exposure on Long-Run EarningsBottom versus top tercile workers
!35.0&
!30.0&
!25.0&
!20.0&
!15.0&
!10.0&
!5.0&
0.0&
5.0&
10.0&
15.0&
20.0&
Total&Effect& Ini5al&Firm& Different&Firm,&Same&Sector&
Other&Sector&
Cumula5ve&Earnings&Rela5ve&to&Baseline&
Impact&of&Trade&Exposure&on&Cumula5ve&Earnings&by&Sector:&&Workers&Ini5ally&in&Lowest&and&Highest&Earnings&Tercile&&
BoQom&Tercile& Top&Tercile&
Agenda
1 What have we learned about the labor-market impacts of trade?
2 Should we be surprised by large earnings losses?
3 What’s the problem: Trade adjustment or trade itself?
4 Has the US policy response been effective?
5 Conclusions
Durable Scars from Job DisplacementMuch Worse in Recessions
Average Earnings Losses of Displaced Workers as a Percent ofPre-Displacement Earnings
18 Brookings Papers on Economic Activity, Fall 2011
0
–5
–10
–15
Average earnings loss relative to control group earningsc
Thousands of 2000 dollars
Thousands of 2000 dollars
45
40
35
30
–2 –1–4 0 1 2 4 6 8 10 12 14 16 18
–2 –1–4 0 1 2 4 6 8 10 12 14 16 18
Average annual earningsb
In expansions
In recessions
In expansions
In recessions
Years
Years
Displacement year
0
–30
–40
–10
Average earnings loss as a percent of predisplacement earningsd
–2 –1–4 0 1 2 4 6 8 10 12 14 16 18
In expansions
In recessions
Percent
–20
Years
Figure 4. Earnings of Displaced Male Workers before and after Displacementa
(continued)Davis and von Wachter ’2011
Average 3-Year Earnings Losses of Displaced Workers:Much Larger in a Slack Labor Market
STEVEN J. DAVIS and TILL VON WACHTER 19
Source: Social Security Administration data, Bureau of Labor Statistics data, and authors’ calculations. a. Year labels indicate year of displacement; unemployment rate is that of the same year. b. Average earnings loss (including observations with zero earnings) in the third year of displacement
(year 3) for men 50 or younger with 3 or more years of prior job tenure, expressed as a fraction of average annual earnings in the years –4 to –1 before displacement in year 1. Losses are calculated from the administrative earnings data (W-2 earnings records) used in von Wachter and others (2011) and described in the text.
Earnings lossb (fraction of predisplacement earnings)
3 4 5 6 7 98
–0.101998
19991996
19951997
19942000
2001
2002
20032004
2005
19861987 1985
1988
1989
1984
1980
19811982
1983
1991
19921993
1990
–0.15
–0.20
–0.25
–0.30
Unemployment rate, all workers (percent)
Figure 5. Earnings Losses of Men in the Third Year of Displacement versus Unemployment Rate in the Displacement Year, 1980–2005a
Notes to figure 4:
Source: Authors’ calculations. a. In each panel the curve labeled “In recessions” shows average outcomes for workers displaced in
recession years from 1980 to 2005, and the curve labeled “In expansions” shows average outcomes for those displaced in expansion years in that period. When a given displacement year straddles recession and expansion periods, that year’s values are apportioned according to the number of months in each period (see the text for further details). Displaced workers are men 50 or younger who separate from their main job in a mass-layoff event and who have at least 3 years of prior job tenure. All averages are estimated using administrative data on W-2 earnings (following von Wachter and others 2011) and include observations with zero earnings.
b. Mean annual raw earnings before and after displacement of workers displaced in recessions and of those displaced in expansions.
c. Average earnings losses of displaced workers, as estimated from displacement-year regression models of annual earnings for displaced workers and control group workers. The regression models include controls for worker effects, a quartic polynomial in age, calendar-year effects, and an interaction of the latter with individual average earnings in the 5 years preceding displacement. See equation 1 and the accompanying discussion for further details.
d. Earnings losses in the middle panel expressed as a percent of displaced workers’ average annual earnings in the predisplacement baseline period.
Davis and von Wachter ’2011
When Workers Exit One Trade-Exposed Industry...They Often Enter Another
disproportionate share of manufacturing employment (Boundand Holzer 2000; Notowidigdo 2013). Consistent with limited mo-bility responses, Autor, Dorn, and Hanson (2013a) find littleimpact of regional trade exposure on changes in regional popula-tion. We revisit the issue of geographic relocation by extendingour analysis to consider whether workers initially employed inmore trade-exposed industries are more likely to change theirplace of residence during the sample period.
The SSA data provide a county code for the residence addressassociated with each individual record in a given year. Becauseindividuals may reside in one location and work in another, res-idence may be a noisy indicator of the employment location.However, because we designate geographic units to be CZs—which are defined precisely to be the regions within which
0.2
.4.6
.81
1.2
19911992
19931994
19951996
19971998
19992000
20012002
20032004
20052006
2007
Year
Conditional correlation in trade exposure between 1991 firm and firm in indicated yearCounterfactual with zero trade exposure in all firms except 1991 firm
Coeff
icien
t
FIGURE IV
Persistence of Trade Exposure since 1991
The graph plots regression coefficients and 90% confidence intervals ob-tained from 2!16 regressions that relate the 1991–2007 trade exposure of aworker’s industry in the year indicated on the x-axis to the 1991–2007 tradeexposure of the worker’s intial 1991 industry. The counterfactual data seriessets trade exposure to 0 for all firms except the worker’s initial employer. Itrefers to a hypothetical scenario in which no worker joins a trade-exposed firmafter separating from their initial firm, and so all persistence in trade exposureis due to workers who have not separated from their initial firm.
QUARTERLY JOURNAL OF ECONOMICS1828
at MIT Libraries on D
ecember 18, 2014
http://qje.oxfordjournals.org/D
ownloaded from
Autor, Dorn, Hanson and Song ’2014
Agenda
1 What have we learned about the labor-market impacts of trade?
2 Should we be surprised by large earnings losses?
3 What’s the problem: Trade adjustment or trade itself?
4 Has the US policy response been effective?
5 Conclusions
Is the Focus on Trade Adjustment a Misdirection?
Trade theory tells us that globalization’s impact is much greater on the wages of all non-college grads… not just a few dislocated manufacturing workers. The damage is widespread, not concentrated among a few... Trade theory says the result is a permanent, not temporary, lowering of wages of all “unskilled” workers.
Adjustment Versus Steady State
The Short and Medium Run: Trade Adjustment• Costs of job loss are sharp, steep, and scarring to displaced workers
• Concentration matters: A $10, 000 loss suffered by one worker is amuch greater social cost than a $1 loss suffered by 10, 000 workers
The Long Run: Prices and Wages Change• Lower prices of goods and services
• But also lower wages for workers made less scarce byglobalization—typically non-college workers
• Wage effects not limited to manufacturing workers—potentially allnon-college workers
Which is the Bigger Cost?
Adjustment Versus Steady State
Best existing estimates of impact of trade on wages 1980 – 1995
• Rising trade with Less Developed Countries (LDC) 1980 – 1995:
1 Less Developed Country imports to U.S. rose from 2.3 to 3.9percent of GDP
2 Reduced HS Dropout vs. HS Graduate wages by 0.6 – 1.2 pctpoints (∼= 10 percent of actual fall)
3 Raised College Grad by HS Grad wages by 0.7 – 1.4 pct points(< 10 percent of actual fall)
Borjas, Freeman, Katz ’1997
U.S. China Imports Rose Dramatically After 2001
0.3%
0.6%
1.0%
2.1%
2.6%2.7%
0.00%
0.25%
0.50%
0.75%
1.00%
1.25%
1.50%
1.75%
2.00%
2.25%
2.50%
2.75%
3.00%
1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
U.S.GoodsImportsfromChinaasaShareofU.S.GDP,1991- 2014
Adjustment Versus Steady StateExtending Borjas, Freeman, Katz ’97 to China Shock
Magnitude of China Import Expansion
• 1991 – 2001: 0.7% of GDP• HS Dropout vs. HS Grad wages: −0.26% to −0.53%• College Grad vs. HS Grad wages: +0.31% to +0.61%
• 2001 – 2014: 1.7% of GDP• HS Dropout vs. HS Grad wages: −0.64% to −1.28%• College Grad vs. HS Grad wages: +0.74% to +1.49%
• 1991 – 2014: 2.4% of GDP• HS Dropout vs. HS Grad wages: −0.90% to −1.80%• College Grad vs. HS Grad wages: +1.05% to +2.10%
Agenda
1 What have we learned about the labor-market impacts of trade?
2 Should we be surprised by large earnings losses?
3 What’s the problem: Trade adjustment or trade itself?
4 Has the US policy response been effective?
5 Conclusions
Policy Responses to Import Competition
On paper, Trade Adjustment Assistance is a main policy tool
• In principle, TAA helps workers hurt by trade retrain (throughextended UI) and resettle (through relocation allowances)
But the actual policy response has had little to do with TAA
• The increase in benefits uptake due to trade shocks is large...• Benefits rise in commuting zones by $58 per capita for every
increase in import exposure of $1000 per worker
• ...where uptake of Social Security Disability Insurance dwarfs TAA
• For every $1.00 increase in uptake of TAA due to greater tradeexposure, there is a $2.80 increase in uptake of SSDI
Benefits Uptake in Response to Import CompetitionIncrease in benefits varies sharply by program
$3.65&$8.40& $10.00&
$15.04&$18.27&
$57.73&
$0&
$10&
$20&
$30&
$40&
$50&
$60&
$70&
Unemployment&and&TAA&Benefits&
SSA&Disability&Benefits&
SSA&ReRrement&Benefits&
Other&Government&Income&Assistance&
Govt&Medical&Benefits&
Total&Benefits&
Dollar&C
hange&&
Imports&from&China&and&Change&of&Government&Transfer&Receipts&in&CommuRng&Zones&(1990L2007)&&
&Effect&of&an&$1000&Per&Worker&Increase&in&Imports&from&China&during&1990=2007&on&Dollar&Change&of&Annual&Transfer&Receipts&per&Capita&
Policy Options
Is the time right for higher trade barriers?
• Nothing in our results contradicts the conclusion that trade on netgenerates welfare gains for the United States
• General-equilibrium analyses show that US gains from trade withChina are positive (with impacts greater in long run than short run)
• Hsieh & Ossa, Caliendo Dvorkin & Parro, Galle Rodriguez-Clare Yi
• Recent evidence highlights the failure of social-insurance programsto help workers adjust to adverse labor-market shocks
• TAA and SSDI appear to weaken incentives to find newemployment, either in the short run or in the long run
Policy Options
What might effective alternative policies look like?
• Wage insurance
• May help workers leave declining regions and strengthen theirincentive to stay in the labor-force (certiainly relative to SSDI)
• Expanded Earned-Income Tax Credit
• May help insulate workers from reductions in labor demand, endurewage cuts at firms at risk of closure
• Gradual adjustment versus shock therapy
• Attrition versus layoffs. Sunset versus shutdown.
Agenda
1 What have we learned about the labor-market impacts of trade?
2 Should we be surprised by large earnings losses?
3 What’s the problem: Trade adjustment or trade itself?
4 Has the US policy response been effective?
5 Conclusions
Discussion
China trade impacts of future will look different from past
• The worst of the China shock is over
• China’s “transitional” growth is coming to an end, as it begins toconfront the challenges of being a middle-income nation
Donald Trump, Bernie Sanders not as surprising in retrospect...
• Inattention to negative labor-market consequences of trade havehelped catalyze opposition to globalization
• Trade boosterism has arguably been trade’s worst enemy in thepublic policy sphere
• Surprisingly, the political consequences of trade include movementto the extremes, which may worsen political disfunction