Highway to Hitler Nico Voigtländer Joachim Voth - PowerPoint Presentation

1. Highway to Hitler Nico Voigtländer Joachim Voth ( (UCLA, NBER and CEPR) (UZH and CEPR)

2. MotivationImportant economic effects of major infrastructure projectsRailways in India, US, and Prussia; Chinese highwaysEffects on political outcomes less clear:Mixed evidence at the local/regional levelRole of large-scale, nationwide infrastructure projects unexploredYet, large infrastructure projects are a prominent policy tool in rising dictatorshipsStalin’s White-Sea-Baltic Canal Ferdinand Marcos’ large highway building program

3. This PaperCan large-scale infrastructure projects “buy hearts and minds” and contribute to institutional change? Consolidate a rising dictatorship?Under which circumstances does it ‘work’?Exploit Highway construction in Nazi Germany

4. BackgroundCritical juncture – Nazi regime not yet fully entrenched, looming crisis in the summer of 1934Increasing opposition, talk of “second revolution”Adolf Hitler, June 1934, Time Magazine: “Don't forget how people laughed at me 15 years ago when I declared that one day I would govern Germany. They laugh now, just as foolishly, when I declare that I shall remain in power!”Infrastructure spending and the entrenchment of dictatorship in Nazi GermanyAugust 1934 referendum: made Hitler dictatorStudy effect of highway construction on pro-Hitler votes

5. Highway Building and Elections Compare Nazi support in November 1933 election (before highway construction) and August 1934 referendum (after construction began)Highway building heavily exploited for propaganda

6. This Paper – ResultsStrong local effect of highway building: The closer communities were to the highway, the greater the increase in support for the Nazi regime IV strategy using least-cost paths confirms resultsMechanisms:Direct economic benefits unlikelyHighways particularly effective when combined with radio propagandaSuggests novel interaction: propaganda and visibility of projectsSignaling regime competence

7. Related LiteratureConsolidation of autocraciesElections increasingly common, including in autocracies (Simpser 2013; Egorov and Sonin 2014)Need for “enemies” (Schmitt 1926)Totalitarianism and the Rise of the Nazi PartyMass-society (Arendt, Ortega y Gasset)Can government spending “buy” political support? Empirical evidence mixed:Deficit spending before elections not reliably associated with electoral success (Brender and Drazen, 2008)Pork barrel effects are tiny (Stein and Bickers, 1994)Some positive results (Levitt and Snyder, 1997; Manacorda et al., 2011)

8. Outline Historical Background DataMain Empirical ResultsPossible Mechanisms

9. TimelineJan33Hitler comes to powerLast “free” elections Decision to build motorwaysMar33Construction beginsSep33Plebiscite Aug34Opening of first motorway May35ReichstagElectionsNov33Construction starts in earnest

10. Early Nazi electionsNovember 1933NSDAP only party on the ballot papers (+guests)Wins about 92% of the popular voteMassive intimidation (both elections)Large local variationSome places ~100%Others (Lübeck, Hamburg) ~ 64% , 73%August 1934President Hindenburg dies on August 2Plebiscite on August 17 to combine the position of president + chancellor=“Führer”About 90% in favor

11. A critical moment for the Nazi regimeSA (brownshirts) increasingly unruly, talk of 2nd revolution by their left wingEvans 2006: The moment was … critical for the regime. … enthusiasm of the ‘national revolution’ in 1933 had discernibly fallen off .... Social Democratic agents reported to the exiled party leadership in Prague that people were apathetic, constantly complaining, and telling endless political jokes about the Nazi leaders. Nazi meetings were poorly attended … The educated classes feared that the disorder caused by the stormtroopers might spill over into chaos or, worse, Bolshevism. Hitler also worried that conservatives around vice chancellor von Papen could join forces with the armyCrisis resolved by a) murder (Röhm-Putsch) and b) plebiscite

12. Highway construction fully rolled out in 1934

13. Propaganda Use

14. Outline Historical Background DataMain Empirical ResultsPossible Mechanisms

15. Autobahn Network in 1934Approved for constructionUnder constructionPlanned, not approved

16. Highway Construction and Planning - Overview

17.

18. Outline Historical Background DataMain Empirical ResultsPossible Mechanisms

19.

20. Main Empirical Results: OverviewElections before and after HW buildingPanelOLS results for main period (Nov’33-Aug’34)IV: Least Cost Paths

21. Change in Nazi support – before and after highway construction beganBinscatter – 25 equal-sized bins.

22. Nazi support before and after highway construction beganMagnitude: move from 1km to 100km distance  Nazi support decreases by 0.27 std, or 2pp.  opposition declines by ~20%

23. Panel results with city FE

24. Change in Nazi support, Nov’33-Aug’34Persuasion rates

25. IV AnalysisCalculate least-cost path (LCP) between terminal city pairs (38, from official list)Instrument actual road building with LCPsExclude terminal cities from analysisHigh degree of overlap3,276 towns in our sample1,465 within 20km of LCP1,276 (87.1%) also within 20km of planned HW819 (55.9%) with actual construction1,811 with >20km distance to LCP278 (15.4%) with actual construction

26.

27. IV ResultsTop-20 LCPsCompliers

28. Outline Historical Background DataMain Empirical ResultsPossible Mechanisms

29. Possible Mechanisms: OverviewIntimidation / fraud ×Direct benefits of highway construction ×Propaganda and signaling regime competence 

30. Intimidation/FraudPossible problem if correlated with highway constructionExamine empiricallyFilter out effect of voter turnoutTests for electoral fraud

31. Intimidation?Use narrow definition of Nazi support: pro-Nazi votes relative to actual voters (instead of eligible voters)

32. Electoral Fraud?Series of “election forensics” test (Hicken and Mebane 2015):2BL: Benford’s Law: empirical regularity that lower digits occur more often than higher digits in most sets of numerical dataExamples: city population sizes, length of rivers, physical constantsLastC: values of the final digit of the vote count should be distributed uniformly. C05s: Share of vote count ending in either 0 or 5 should be 0.2. P05s: Final digit of the rounded percentage of votes for the winning party. Is there an overabundance of zeros and fives?

33. Election Forensics Tests

34. Direct economic benefits?Relatively low car ownershipHighway network far from complete by 1934Effects not stronger where unemployment was high

35. Propaganda and signaling regime competence?Radio: most important propaganda tool (Adena et al. 2015)Signal strength

36. RobustnessSpatial correlationBalancing treatment and control groups Planning vs buildingHW under construction and those approved for constructionSample splitsPlacebos

37. Allowing for Spatial Correlation

38. Entropy Balancing

39. Matching

40. Planning vs. building

41. Under construction vs.approved for construction

42. Sample Splits

43. Placebos: distance to other transportation infrastructure

44. ConclusionsMain result: Robust association between Autobahn construction and decline in opposition to the nascent Nazi regimeProbably lower bound for aggregate effect on regime supportHelped to entrench the dictatorshipResults confirmed with least cost paths – probably causal

45. Why it workedPossible mechanism: Highway construction as a signal of government “competence” in the spirit of Rogoff (1990)Infrastructure spending can win local “hearts” when “minds” are led to associate it with visible economic progress in the aggregate

46.

47. Main result in one chartChange in Nazi support Nov’33-Aug’34Magnitude: Going from <10km to >60km: 0.47 std vote change of 2.4%, relative to 10% opposition

48. Manpower used in highway construction

49. Acemoglu, Egorov and Sonin (2013) predict populist bias of policy is greater when the value of remaining in office is higher for the politicianwhen there is greater polarization between the policy preferences of the median voter and right-wing special interestswhen politicians are perceived as more likely to be corruptwhen there is an intermediate amount of noise in the information that voters receivewhen politicians are more forward-lookingand when there is greater uncertainty about the type of the incumbent.

50. Autobahn building and increase in Nazi support

51. Use only LCPs between top-20 cities in terms of population

52. Persuasion Rates 

53. Persuasion Rates – Comparison

54. Compliers in the IV Analysis“Compliers”: within 20 km of LCPs and HW construction“non-compliers”: within 20 km of LCPs but no HW construction

55. Radio Signal Strength and Radio SubscribersBelow threshold of 20: reception possible, but only with high-quality AM receivers

56. Radio Signal Strength: Interaction over full rangeInsignificant effects for predicted radio coverage <20%Negative and significant effects for predicted radio coverage >20%