Importing Political Polarization? The Electoral Consequences of - PowerPoint PPT Presentation

Importing Political Polarization? The Electoral Consequences of Rising Trade Exposure David Autor David Dorn Gordon Hanson Kaveh Majlesi May 2016

Trade and Politics

Trade and Politics The impact of trade on US workers has become a touchstone issue in the 2016 presidential campaign • Both among Republicans “I would tax China on products coming in. I would do a tax, and the tax, let me tell you what the tax should be... the tax should be 45 percent.” Donald Trump • and Democrats “I voted against NAFTA, CAFTA, PNTR with China. I think they have been a disaster for the American worker.” Bernie Sanders

Widely Debated Hypothesis: Do the Economic Impacts of Trade Favor Ideologically Far-Left and Far-Right Politicians? Trump and Sanders Have a Point about Trade with China

Background: Rapid Growth of China’s Manufacturing Exports since 1990... 10 15 China import penetration in US manuf. 8 10 6 percent 4 5 2 0 0 1991 1996 2001 2006 2011 Year China share of world manufacturing exports China import penetration in US manufacturing

...Contributed to Decline in U.S. Manufacturing Economic Impacts of Import Competition from China • Closure of manufacturing plants (Bernard Jensen Schott ’06), declines in employment (Acemoglu Autor Dorn Hanson Price ’16; Pierce Schott ’16) in more trade-exposed industries • Lower lifetime incomes, greater job churning for workers in more trade-exposed industries (Autor Dorn Hanson Song ’14) • Lower employment, higher labor-force exit, higher long-run unemployment, greater benefits uptake in more trade-exposed local labor markets (Autor Dorn Hanson ’13)

Impact on Trade Legislation and US Politics Anti-trade views precede Trump and Sanders • Congressional representatives from trade-exposed regions are more likely to support protectionist trade bills (Feigenbaum Hall ’15; Che, Lu, Pierce, Schott, Tao ’15) and anti-China legislation (Kleinberg Fordham ’13; Kuk, Seligsohn, Zhang ’15) • Our work studies whether the impacts of trade exposure extend beyond voting on trade policy , and affect the ideological composition of Congress itself

Major Trend in U.S. Politics Increasing partisanship in the US Congress • Not due to a shift in vote shares going to the two major parties • GOP has bicameral majorities, but nat’l vote shares are close to even • Voter identification with parties has become weaker, not stronger, though persistence in county voting is greater • Rather, the change is more polarized behavior among legislators • Poole-Rosenthal DW-Nominate scores of roll-call votes • The ideological divide between the parties has been rising since the mid-1970s and is now at an all-time high • Although voters haven’t become more extreme, legislators have • Also visible in polarized speech patterns in Congress (Gentzkow Shapiro Taddy ’15)

Polarization in Congress: DW-Nominate Scores Mean Voting Behavior by Party in the House Mean Voting Behavior by Party in the Senate .8 .8 .6 .6 Mean DW-Nominate Score Mean DW-Nominate Score .4 .4 .2 .2 0 0 -.2 -.2 -.4 -.4 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 2012 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 2012 Year Year Democrats Republicans Democrats Republicans

Distribution of Democrats and Republicans on a 10-Item Scale of Political Values (Among Politically Engaged) Among the politically engaged Among the less engaged Source: Pew Research Center (2014).

Many Subtleties in Public Opinion (Gentzkow ’16) 1 No growth of extremism on average • Distribution of views on issues mostly single-peaked, relatively stable 2 Rising correlations: Views across issues; between issues and party • Less likely for people to hold liberal views on some issues, conservative views on others • Presidential votes increasingly predict citizens political views on taxes, redistribution, social policy, gun control, the environment, etc. 3 Politics has become more personal and hostile • More likely to see other party’s supporters as selfish and stupid • 27% of Dems, 36% of Repubs agree: opposite party’s policies “ are so misguided that they threaten the nation’s well-being ” (Pew ’14) • Less tolerant of cross-party marriage!

Explaining Polarization Literature is large but little consensus on causal mechanisms • Explanations shown to lack empirical support • Immigration, manipulation of blue-collar voters (Gelman et al. ’08) • Greater voter segregation, heterogeneity in voter attitudes (Glaeser Ward ’06, Ansolabehere Rodden Snyder ’08, Abrams Fiorina ’12) • Gerrymandering, changes in election structure or congressional rules (McCarty Poole Rosenthal ’09, Barber McCarty ’15) • Explanations supported by circumstantial evidence • Tax/regulatory reform (Bartels ’10, Hacker Pierson ’10) • Stronger ideological sorting of voters by party (Levendusky ’09) • Overall distribution of voter attitudes hasn’t changed but difference in distributions between Dem and GOP party members has • Media partisanship (DellaVigna Kaplan ’07, Gentzkow Shapiro ’11)

Subject of this Paper: Trade and Political Outcomes Has rising trade exposure in local labor markets contributed to greater political divisions in Congress? • Anti-incumbency effect • Incumbents punished for bad outcomes • Fair (’78), Margalit (’11), Jensen Quinn Weymouth (’16) • Party-realignment effect • Economic shocks change voter prefs — Leftward (Bruner Ross Washington ’11; Che Lu Pierce Schott Tao ’16) or rightward (Malgouyres ’14, Dippel Gold Heblich ’15) • Polarization effect • Economic shocks shift support from center to extremes • Failure of monotone likelihood ratio property: Dixit Weibull (’07), Baliga Hanany Klibanoff (’13), Acemoglu Chernozhukov Yildiz (’15)

Agenda 1 Measuring Electoral Outcomes 2 Exposure to Import Competition from China 3 Empirical Specification 4 Anti-Incumbent, Party Realignment Effects 5 Polarization Effects 6 Heterogeneity in Polarization Effects 7 1990s versus 2000s 8 Conclusions

Challenge: Mapping Political to Economic Geography Congressional districts can have extreme shapes that do not correspond to any definition of local labor market geography

An Extreme Example: District NC-12 NC-12 stretches over 100 miles and comprises parts of the Charlotte, Greensboro and Winston-Salem cities and Commuting Zones

An Extreme Example: District NC-12 The district closely follows Interstate 85, and at some points is barely wider than a highway lane

An Extreme Example: District NC-12 How should one deal with Davidson and Rowan counties, which both partly overlap with districts NC-5, NC-12, and NC-8?

Analysis at the County Level would be Problematic The Davidson and Rowan Cty residents cast votes in three different races; the sum of votes across these races is hard to interpret Data Structure: County-Level Analysis Geographic Source of Variables Local Labor Market Demographic Observations Shock Composition Election Outcomes Observation Weights? 1 Davidson Cty CZ Greensboro Davidson Cty NC-5?/NC-8?/NC-12? Total Votes? Population? 2 Rowan Cty CZ Charlotte Rowan Cty NC-5?/NC-8?/NC-12? Total Votes? Population?

Our Analysis is at the County-District Cell Level Incorporates the overlapping structure of economic geography (CZ/county) and political geography (district) Data Structure: County-District Cell Analysis Geographic Source of Variables Local Labor Demographic Election Observations Market Shock Composition Outcomes Observation Weight 1 Davidson Cty x NC-5 CZ Greensboro Davidson Cty NC-5 Cell Votes/District Votes 2 Davidson Cty x NC-8 CZ Greensboro Davidson Cty NC-8 Cell Votes/District Votes 3 Davidson Cty x NC-12 CZ Greensboro Davidson Cty NC-12 Cell Votes/District Votes 4 Rowan Cty x NC-5 CZ Charlotte Rowan Cty NC-5 Cell Votes/District Votes 5 Rowan Cty x NC-8 CZ Charlotte Rowan Cty NC-8 Cell Votes/District Votes 6 Rowan Cty x NC-12 CZ Charlotte Rowan Cty NC-12 Cell Votes/District Votes

Empirical Strategy We match local labor markets to congressional districts • Divide US into county-by-congressional-district cells • Attach each county to its corresponding commuting zone (CZ) • Weight each cell by its share of congressional-district votes • Result is a mapping of CZ shocks to district political outcomes • Use CZ trade shocks from Acemoglu Autor Dorn Hanson Price (’16) • Examine electoral outcomes over 2002 to 2010 • Because of redistricting, we can only examine intercensal periods • Helpfully, these are non -presidential election years • Our time period spans the rise of the Tea Party

Data Sources 1 Voting behavior of congressional representatives • DW-Nominate scores (Poole & Rosenthal ’85, ’91, ’97, ’01) • Estimated for each legislator in each Congress • Tag 2003-2005 score to winning legislator in 2002 election, 2011-2013 score to winning legislator in 2010 election 2 Vote shares by party in House elections • Dave Leip’s Atlas of US Presidential Elections • Vote counts for each party by county-district cell