- Email: [email protected]
Bombs, homes, and jobs: Revisiting the Oswald hypothesis for Germany
Economics Letters 135 (2015) 65–68
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Bombs, homes, and jobs: Revisiting the Oswald hypothesis for Germany Nikolaus Wolf a,b,∗ , Paul Caruana-Galizia a a
Humboldt Universität zu Berlin, Germany
b
CEPR, United Kingdom
highlights • • • •
We re-examine Oswald’s hypothesis that homeownership increases unemployment. We exploit WWII bombing of Germany as a source of identification. Regional homeownership rates are significantly positively correlated with regional unemployment rates. Result holds with year and region fixed effects, instrumentation, and a set of controls.
article
info
Article history: Received 21 November 2014 Received in revised form 9 July 2015 Accepted 10 July 2015 Available online 23 July 2015
abstract Andrew Oswald (1996) hypothesized that homeownership restricts commercial development and labour mobility, increasing unemployment. Instrumenting homeownership with WWII Allied bombing for a German regional panel, we find homeownership has a large positive effect on unemployment, and homeownership decreases labour mobility. © 2015 Elsevier B.V. All rights reserved.
JEL classification: J01 J6 O18 R10 Keywords: Home ownership Unemployment Bombing Germany
1. Introduction Andrew Oswald hypothesized that homeownership restricts commercial land development and labour mobility, increasing unemployment (Oswald, 1996; Blanchflower and Oswald, 2013; Lerbs, 2011; Laamanen, 2013). Research on the topic has produced mixed results. We instrument homeownership with WWII Allied strategic bombing, which destroyed housing stocks, leading to state provision of rental housing. We then estimate unemployment equations on 87 German regions over 1998, 2002, and 2006. Reinforcing Lerbs’ (2011) OLS estimates, we find that when cor-
∗ Correspondence to: Spandauer strasse 1, Raum 411/2, Berlin 10178, Germany. Tel.: +49 30 2093 5715. E-mail address: [email protected] (N. Wolf). http://dx.doi.org/10.1016/j.econlet.2015.07.009 0165-1765/© 2015 Elsevier B.V. All rights reserved.
rectly identified higher homeownership increases unemployment, and that homeownership decreases labour mobility. 2. Data and results Researchers estimate an equation like: uit = α + β ownit + γ X it + ϑt + πi + µit
(1)
where u is unemployment rate (%) of region i in year t, own is homeownership rate (%), X controls for u’s other determinants, ϑ and π are region and year fixed effects terms, and α and µ are a constant and error term. The problem with (1) is that homeownership and labour market conditions are endogenous. We exploit WWII Allied ‘area bombing’ as an instrument. The United States Army Air Forces (USAAF) and
66
N. Wolf, P. Caruana-Galizia / Economics Letters 135 (2015) 65–68
Table 1 OLS unemployment rate (%).
Table 3 IV unemployment rate (%). All
All
All
−0.104**
0.074** [0.036] −0.015 [0.011] 0.035 [0.028] 31.764** [13.904] 0.246** [0.087] 0.039* [0.141]
Constant
12.198** [5.602]
−1.603 [8.605]
0.105** [0.040] −0.041 [0.017] 0.027 [0.062] 32.667 [21.587] 0.849*** [0.142] 0.125 [0.143] 0.042 [0.049] −13.617* [9.899]
Year FE Region FE N Within-R2
Yes No 247 0.486
Yes Yes 247 0.595
Yes Yes 164 0.743
Own
[0.036] −0.002 [0.002] 0.101** [0.039] 6.933 [11.02] 0.046 [0.079] 0.047 [0.133]
Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Owner t −1
Notes: Clustered robust standard errors in brackets. * p < 0.1. ** p < 0.05. *** p < 0.01.
Population density High school drops. Herfindahl % Age 17–24 % Age 45–64
Constant Year FE Region FE N R2
High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N F -stat. Under-ID Within-R2
Own
1 2002
2 2006
3 2006
4 All
5 All
0.009 [0.45] −0.014
0.091** [0.039] −0.038**
0.105** [0.042] −0.041**
−0.181*** −0.027
0.071** [0.036] −0.018
[0.021] −0.004
[0.017] 0.018
[0.017] 0.027
−0.131**
[0.016] 0.047
[0.060] −8.639 [18.462] −0.615** [0.194] −0.821** [0.404]
[0.061] 35.225 [21.16] 0.841*** [0.146] 0.092 [0.140]
[0.049] 10.159 [22.759] −0.305** [0.157] −0.525** [0.234]
[0.035] 50.49** [16.69] 0.310** [0.120] 0.147 [0.149]
−0.821** [0.345]
2.574*** [0.233]
[0.064] 32.667 [22.041] 0.849*** [0.145] 0.125 [0.146] 0.042 [0.049] 2.555*** [0.251]
No No 81 0.299
No No 83 0.458
No No 84 0.447
Owner t −1
Population density
bomb
Table 2 OLS first-differences unemployment rate (%).
Own
Own
[0.035]
[0.027]
−0.242
−1.608***
[0.216]
[0.172]
No No 164 0.291
Yes No 164 0.679
Notes: Clustered robust standard errors in brackets. * p < 0.1. ** p < 0.05. *** p < 0.01.
the Royal Air Force (RAF) strategically targeted civilian areas to demoralize Germans (‘dehousing’). Churchill protested that this would create ‘‘. . . a great shortage of accommodation for ourselves and our Allies. . . [and] some temporary provision would have to be made for the Germans themselves’’ (TNA: CAB 120/303). Between October and December 1944, 53% of air attacks targeted cities (Terraine, 1985, 675). All told, 20% of West Germany’s housing was destroyed; another 20% damaged (Voigtländer, 2009, 357). Only 43% of Bomber Command’s ordnance was expended on industrial cities (Bashow, 2014, 32). Vonyo (2012, 107) shows that distance from British airfields determined the extent of bombing. Along with refugee inflows, this created a 4.5 million home shortage by 1950 (Voigtländer, 2009, 358). Capital scarcity made government intervention necessary. By 1959, 50% of all new houses
Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N Within-R2
OLS Unemp
Second Stage Unemp
0.074** [0.036] −0.015 [0.011] 0.035 [0.028] 31.764** [13.904] 0.246** [0.087] 0.039* [0.141] −1.603 [8.605]
0.536* [0.301] −0.013 [0.013] 0.049 [0.056] 12.302 [23.91] 0.360 [0.103] 0.404 [0.272]
Yes Yes 247
0.595
Yes Yes 250 4.46** 2.96* 0.256
Reduced Form Unemp
First Stage Owners
−1.08e−07** [4.07e−08] 0.068* [0.035] −0.0131 [0.022] 0.029 [0.029] 28.919** [13.596] 0.225** [0.082] 0.084* [0.141] −2.567 [8.483]
−1.86e−07** [8.80e−08]
Yes Yes 247 0.603
Yes Yes 250 0.340
−0.001 [0.027] −0.061 [0.084] 30.689 [30.405] −0.149 [0.183] −0.571** [0.222]
Notes: Clustered robust standard errors in brackets. F -stat. in first stage is Kleibergen–Paap Wald rk F statistic. Under-ID is the Angrist–Pischke χ 2 test. * p < 0.1. ** p < 0.05. *** p < 0.01.
were built with public funds (Voigtländer, 2009, 358). Low homeownership persists because the dwellings are high quality, income limits are generous, and rental market regulation (Voigtländer, 2009, 360). We aggregated bomb tonnage by region, from the start of area bombing raids in March 1942 to the end of WWII, using Hastings (2013, 328–333) and USSBS (1947, 35, 46). Tonnage data are available for 97 areas, averaging 1.57 units per region. Mean tonnage is 9368 with a standard deviation of 19,901. We interacted these data with the inverse number of months from the last raid to our panel’s benchmark years, using Hastings (2013, 328–333) and the EAFHS (2014). This IV (bomb) captures bombing damage and bombed areas’ recovery time. The mean value of months since last raid is 637, with a standard deviation of 7. The IV’s coefficient of variation is 2.06. We use ‘Planning Regions’ (Raumordnungsregionen), which are functional labour markets. Lerbs (2011) provides unemployment and homeownership rate data. The Federal Statistics Office (Destatis, 2014) provides population density (persons per km2 ) to control for employment agglomeration; high-school dropouts (% of high-school graduates) to proxy low-skilled worker shares;
N. Wolf, P. Caruana-Galizia / Economics Letters 135 (2015) 65–68 Table 4 IV first-differences unemployment rate (%).
Own Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N F -stat. Under-ID Within-R2
bomb Own Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N Within-R2
67
Table 5 OLS net migration results.
OLS Unemp
Second Stage Unemp
0.071** [0.036] −0.018 [0.016] 0.047 [0.035] 50.49** [16.69] 0.310** [0.120] 0.147 [0.149] −1.608*** [0.172]
0.624* [0.353] −0.010 [0.016] 0.100 [0.069] 18.290 [30.649] 0.291** [0.126] 0.390 [0.272] −2.651 [0.790]
Yes No 164
0.595
Yes No 164 4.98* 3.25* 0.138
Reduced Form Unemp
First Stage Owners
−8.36e−09** [3.24e−09] 0.066* [0.040] −0.015 [0.013] 0.042 [0.035] 46.479** [17.853] 0.283** [0.095] 0.184 [0.161] −1.705 [0.184]
−1.50e−08*** [6.72e−09]
Yes No 164 0.749
Yes No 164 0.410
−0.010 [0.029] −0.104 [0.087] 50.534 [30.905] −0.014 [0.188] −0.368** [0.231] 1.693 [0.384]
Notes: Clustered robust standard errors in brackets. F -stat. in first stage is Kleibergen–Paap Wald rk F statistic. Under-ID is the Angrist–Pischke χ 2 test. * p < 0.1. ** p < 0.05. *** p < 0.01.
sectoral employment data, with which we construct a Herfindahl Index, to control for sector-specific shocks1 ; and regional workingage structures (17%–24% and 45%–64%), following Lerbs and Oberst (2014). The OLS estimation in Table 1’s first column pools the data with year fixed effects. own is negative, contrary to the hypothesis. Adding region fixed effects, own changes direction and retains significance, indicating time-invariant omitted variables. A lag of own in column 3 does not change the results, implying a contemporaneous effect. Table 2 shows first-differences of the variables to control for further time-invariance. For 2002, own is insignificant. For 2006, it is significant, and the same magnitude as that in columns 2 and
1 In German, Land- und Forstwirtschaft, Fischerei; Produzierendes Gewerbe ohne Baugewerbe; Baugewerbe; Handel, Gastgewerbe und Verkehr; Finanzierung, Vermietung, und Unternehmensdienstl; and Öffentliche und private Dienstleister. The index sums the squares of sectoral employment shares, ranging from 0 to 1, where higher values indicate greater specialization.
All
All
−188.451**
Constant
359.668 [951.889]
−266.560* [145.2] 1259.46** [402.61] −101.134 [126.85] −123 042.6 [78 646.6] −81.732 [1306.04] −127.14 [908.23] −176.716 [238.13] 8810.2 [11 269.3]
Year FE Region FE N Within-R2
Yes Yes 164 0.692
Yes Yes 164 0.693
Own Population density High school drops. Herfindahl % Age 17–24 % Age 45–64
[84.304] 1270.59** [403.915] −95.901 [125.188] −135 000.8* [77 760.62] −168.588 [1256.98] −50.358 [866.604]
Owner t −1
Notes: Clustered robust standard errors in brackets. All variables are firstdifferenced, except for Owner t −1 which is initial ownership level. * p < 0.1. ** p < 0.05. *** p < 0.01.
3 of Table 1. A one period lag of own in column 3 does not change this. Column 4 pools the data without year fixed effects, showing a negative and significant coefficient, as we have gone beyond a two period sample. With year fixed effects, own is again positive and significant. Still, a Durbin–Wu–Hausman test of homeownership’s endogeneity rejects the null (F (1, 83) = 4.71, p > F = 0.03), indicating that OLS is inconsistent. Table 3’s ‘Reduced Form’ column shows a negative effect of bomb on unemployment. The ‘First Stage’ column shows bomb has a larger direct effect on homeownership rates (R2 = 34%). Adding to the IV’s validity, the F -statistic on the excluded IV and under-ID statistics in the ‘Second Stage’ column are significant. The IV point estimate (0.54) is much larger than the OLS. The IV is imprecise because its first-stage correlation is weaker than the biased unmediated OLS correlation (R2 of 34% vs. 60%). The standard error grows from 0.036 to 0.301, but the t-ratio remains close: 2.10 and 1.80. Table 4 replicates the IV estimation with first-differences variables. The results are the same, apart from the IV coefficient growing to 0.62, yet retaining a t-ratio of 1.80, as in Table 2. 3. Mechanisms We aggregated county-level net migration data (inflows less outflows) from Destatis (2014) to proxy labour mobility. Table 5 shows that, after differencing, higher homeownership is associated with lower outflows relative to inflows. Table 5 does not rule out the second mechanism highlighted in Blanchflower and Oswald (2013) that homeownership restricts commercial land development, either through zoning regulations or NIMBYism. 4. Conclusion We support the hypothesis that homeownership impairs the labour market. This links two facts for Germany – unusually low homeownership and low unemployment – and has implications for Europe’s high homeownership, high unemployment countries (Spain, Italy). We leave this for future work.
68
N. Wolf, P. Caruana-Galizia / Economics Letters 135 (2015) 65–68
Acknowledgements We thank Alexandra Spitz-Oener, Michael Burda, and Andrew Oswald for helpful comments, and Felix Kersting for research assistance. Paul Caruana-Galizia was funded by a Macrohist fellowship from a European Commission Marie Curie programme. The usual disclaimer applies. References Bashow, D.L., 2014. In praise of bomber harris and area bombing. R. Can. Air Force J. 3, 28–56. Blanchflower, D.G., Oswald, A., 2013. Does high home-ownership impair the labor market? National Bureau of Economic Research No. 19079. Destatis (German Federal Statistical Office). 2014. Regionaldatenbank Deutschland. Online: https://www.regionalstatistik.de/genesis/online/logon (Accessed: 31.10.14). Eighth Air Force Historical Society (EAFHS). 2014. WWII 8th AAF COMBAT CHRONOLOGY. Online: http://www.8thafhs.org/combat1942.htm Accessed: 3.11.14. Hastings, M., 2013. Bomber Command. Zenith Military Classics, Minneapolis, MN.
Laamanen, J.-P., 2013. Home-ownership and the labour market: evidence from rental housing market deregulation. Tampere Economic Working Papers Net Series, No. 89. Lerbs, O., 2011. Is there a link between homeownership and unemployment? Evidence from German regional data. Int. Econ. Policy 8, 407–426. Lerbs, O., Oberst, C., 2014. Explaining the spatial variation in homeownership rates: Results for German regions. Reg. Stud. 48 (5), 844–865. Oswald, A., 1996. A conjecture on the explanation for high unemployment in the industrialized nations: part 1. University of Warwick Economic Research Papers. Terraine, J., 1985. The Right of the Line: The Role of the RAF in World War Two. Hodder and Stoughton, London. TNA: CAB 120/303, 1945. United States of America Air Force (USAAF) 1945a. The United States Strategic Bombing Survey: over-all report (European war). Washington, DC. United States of America Air Force (USAAF) 1945b. The United States Strategic Bombing Survey: Summary Report. Washington, DC. USSBS (United States Strategic Bombing Survey) 1947. The effects of strategic bombing on German morale [by] the United States Strategic Bombing Survey, Morale Division. Dates of survey: March–July, 1945. Washington. Voigtländer, M., 2009. Why is the German homeownership rate so low? Hous. Stud. 24 (3), 355–372. Vonyo, T., 2012. The bombing of Germany: The economic geography of war-induced dislocation in west German industry. Eur. Rev. Econ. Hist. 16 (1), 97–118.
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Bombs, homes, and jobs: Revisiting the Oswald hypothesis for Germany Nikolaus Wolf a,b,∗ , Paul Caruana-Galizia a a
Humboldt Universität zu Berlin, Germany
b
CEPR, United Kingdom
highlights • • • •
We re-examine Oswald’s hypothesis that homeownership increases unemployment. We exploit WWII bombing of Germany as a source of identification. Regional homeownership rates are significantly positively correlated with regional unemployment rates. Result holds with year and region fixed effects, instrumentation, and a set of controls.
article
info
Article history: Received 21 November 2014 Received in revised form 9 July 2015 Accepted 10 July 2015 Available online 23 July 2015
abstract Andrew Oswald (1996) hypothesized that homeownership restricts commercial development and labour mobility, increasing unemployment. Instrumenting homeownership with WWII Allied bombing for a German regional panel, we find homeownership has a large positive effect on unemployment, and homeownership decreases labour mobility. © 2015 Elsevier B.V. All rights reserved.
JEL classification: J01 J6 O18 R10 Keywords: Home ownership Unemployment Bombing Germany
1. Introduction Andrew Oswald hypothesized that homeownership restricts commercial land development and labour mobility, increasing unemployment (Oswald, 1996; Blanchflower and Oswald, 2013; Lerbs, 2011; Laamanen, 2013). Research on the topic has produced mixed results. We instrument homeownership with WWII Allied strategic bombing, which destroyed housing stocks, leading to state provision of rental housing. We then estimate unemployment equations on 87 German regions over 1998, 2002, and 2006. Reinforcing Lerbs’ (2011) OLS estimates, we find that when cor-
∗ Correspondence to: Spandauer strasse 1, Raum 411/2, Berlin 10178, Germany. Tel.: +49 30 2093 5715. E-mail address: [email protected] (N. Wolf). http://dx.doi.org/10.1016/j.econlet.2015.07.009 0165-1765/© 2015 Elsevier B.V. All rights reserved.
rectly identified higher homeownership increases unemployment, and that homeownership decreases labour mobility. 2. Data and results Researchers estimate an equation like: uit = α + β ownit + γ X it + ϑt + πi + µit
(1)
where u is unemployment rate (%) of region i in year t, own is homeownership rate (%), X controls for u’s other determinants, ϑ and π are region and year fixed effects terms, and α and µ are a constant and error term. The problem with (1) is that homeownership and labour market conditions are endogenous. We exploit WWII Allied ‘area bombing’ as an instrument. The United States Army Air Forces (USAAF) and
66
N. Wolf, P. Caruana-Galizia / Economics Letters 135 (2015) 65–68
Table 1 OLS unemployment rate (%).
Table 3 IV unemployment rate (%). All
All
All
−0.104**
0.074** [0.036] −0.015 [0.011] 0.035 [0.028] 31.764** [13.904] 0.246** [0.087] 0.039* [0.141]
Constant
12.198** [5.602]
−1.603 [8.605]
0.105** [0.040] −0.041 [0.017] 0.027 [0.062] 32.667 [21.587] 0.849*** [0.142] 0.125 [0.143] 0.042 [0.049] −13.617* [9.899]
Year FE Region FE N Within-R2
Yes No 247 0.486
Yes Yes 247 0.595
Yes Yes 164 0.743
Own
[0.036] −0.002 [0.002] 0.101** [0.039] 6.933 [11.02] 0.046 [0.079] 0.047 [0.133]
Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Owner t −1
Notes: Clustered robust standard errors in brackets. * p < 0.1. ** p < 0.05. *** p < 0.01.
Population density High school drops. Herfindahl % Age 17–24 % Age 45–64
Constant Year FE Region FE N R2
High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N F -stat. Under-ID Within-R2
Own
1 2002
2 2006
3 2006
4 All
5 All
0.009 [0.45] −0.014
0.091** [0.039] −0.038**
0.105** [0.042] −0.041**
−0.181*** −0.027
0.071** [0.036] −0.018
[0.021] −0.004
[0.017] 0.018
[0.017] 0.027
−0.131**
[0.016] 0.047
[0.060] −8.639 [18.462] −0.615** [0.194] −0.821** [0.404]
[0.061] 35.225 [21.16] 0.841*** [0.146] 0.092 [0.140]
[0.049] 10.159 [22.759] −0.305** [0.157] −0.525** [0.234]
[0.035] 50.49** [16.69] 0.310** [0.120] 0.147 [0.149]
−0.821** [0.345]
2.574*** [0.233]
[0.064] 32.667 [22.041] 0.849*** [0.145] 0.125 [0.146] 0.042 [0.049] 2.555*** [0.251]
No No 81 0.299
No No 83 0.458
No No 84 0.447
Owner t −1
Population density
bomb
Table 2 OLS first-differences unemployment rate (%).
Own
Own
[0.035]
[0.027]
−0.242
−1.608***
[0.216]
[0.172]
No No 164 0.291
Yes No 164 0.679
Notes: Clustered robust standard errors in brackets. * p < 0.1. ** p < 0.05. *** p < 0.01.
the Royal Air Force (RAF) strategically targeted civilian areas to demoralize Germans (‘dehousing’). Churchill protested that this would create ‘‘. . . a great shortage of accommodation for ourselves and our Allies. . . [and] some temporary provision would have to be made for the Germans themselves’’ (TNA: CAB 120/303). Between October and December 1944, 53% of air attacks targeted cities (Terraine, 1985, 675). All told, 20% of West Germany’s housing was destroyed; another 20% damaged (Voigtländer, 2009, 357). Only 43% of Bomber Command’s ordnance was expended on industrial cities (Bashow, 2014, 32). Vonyo (2012, 107) shows that distance from British airfields determined the extent of bombing. Along with refugee inflows, this created a 4.5 million home shortage by 1950 (Voigtländer, 2009, 358). Capital scarcity made government intervention necessary. By 1959, 50% of all new houses
Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N Within-R2
OLS Unemp
Second Stage Unemp
0.074** [0.036] −0.015 [0.011] 0.035 [0.028] 31.764** [13.904] 0.246** [0.087] 0.039* [0.141] −1.603 [8.605]
0.536* [0.301] −0.013 [0.013] 0.049 [0.056] 12.302 [23.91] 0.360 [0.103] 0.404 [0.272]
Yes Yes 247
0.595
Yes Yes 250 4.46** 2.96* 0.256
Reduced Form Unemp
First Stage Owners
−1.08e−07** [4.07e−08] 0.068* [0.035] −0.0131 [0.022] 0.029 [0.029] 28.919** [13.596] 0.225** [0.082] 0.084* [0.141] −2.567 [8.483]
−1.86e−07** [8.80e−08]
Yes Yes 247 0.603
Yes Yes 250 0.340
−0.001 [0.027] −0.061 [0.084] 30.689 [30.405] −0.149 [0.183] −0.571** [0.222]
Notes: Clustered robust standard errors in brackets. F -stat. in first stage is Kleibergen–Paap Wald rk F statistic. Under-ID is the Angrist–Pischke χ 2 test. * p < 0.1. ** p < 0.05. *** p < 0.01.
were built with public funds (Voigtländer, 2009, 358). Low homeownership persists because the dwellings are high quality, income limits are generous, and rental market regulation (Voigtländer, 2009, 360). We aggregated bomb tonnage by region, from the start of area bombing raids in March 1942 to the end of WWII, using Hastings (2013, 328–333) and USSBS (1947, 35, 46). Tonnage data are available for 97 areas, averaging 1.57 units per region. Mean tonnage is 9368 with a standard deviation of 19,901. We interacted these data with the inverse number of months from the last raid to our panel’s benchmark years, using Hastings (2013, 328–333) and the EAFHS (2014). This IV (bomb) captures bombing damage and bombed areas’ recovery time. The mean value of months since last raid is 637, with a standard deviation of 7. The IV’s coefficient of variation is 2.06. We use ‘Planning Regions’ (Raumordnungsregionen), which are functional labour markets. Lerbs (2011) provides unemployment and homeownership rate data. The Federal Statistics Office (Destatis, 2014) provides population density (persons per km2 ) to control for employment agglomeration; high-school dropouts (% of high-school graduates) to proxy low-skilled worker shares;
N. Wolf, P. Caruana-Galizia / Economics Letters 135 (2015) 65–68 Table 4 IV first-differences unemployment rate (%).
Own Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N F -stat. Under-ID Within-R2
bomb Own Population density High school drops. Herfindahl % Age 17–24 % Age 45–64 Constant Year FE Region FE N Within-R2
67
Table 5 OLS net migration results.
OLS Unemp
Second Stage Unemp
0.071** [0.036] −0.018 [0.016] 0.047 [0.035] 50.49** [16.69] 0.310** [0.120] 0.147 [0.149] −1.608*** [0.172]
0.624* [0.353] −0.010 [0.016] 0.100 [0.069] 18.290 [30.649] 0.291** [0.126] 0.390 [0.272] −2.651 [0.790]
Yes No 164
0.595
Yes No 164 4.98* 3.25* 0.138
Reduced Form Unemp
First Stage Owners
−8.36e−09** [3.24e−09] 0.066* [0.040] −0.015 [0.013] 0.042 [0.035] 46.479** [17.853] 0.283** [0.095] 0.184 [0.161] −1.705 [0.184]
−1.50e−08*** [6.72e−09]
Yes No 164 0.749
Yes No 164 0.410
−0.010 [0.029] −0.104 [0.087] 50.534 [30.905] −0.014 [0.188] −0.368** [0.231] 1.693 [0.384]
Notes: Clustered robust standard errors in brackets. F -stat. in first stage is Kleibergen–Paap Wald rk F statistic. Under-ID is the Angrist–Pischke χ 2 test. * p < 0.1. ** p < 0.05. *** p < 0.01.
sectoral employment data, with which we construct a Herfindahl Index, to control for sector-specific shocks1 ; and regional workingage structures (17%–24% and 45%–64%), following Lerbs and Oberst (2014). The OLS estimation in Table 1’s first column pools the data with year fixed effects. own is negative, contrary to the hypothesis. Adding region fixed effects, own changes direction and retains significance, indicating time-invariant omitted variables. A lag of own in column 3 does not change the results, implying a contemporaneous effect. Table 2 shows first-differences of the variables to control for further time-invariance. For 2002, own is insignificant. For 2006, it is significant, and the same magnitude as that in columns 2 and
1 In German, Land- und Forstwirtschaft, Fischerei; Produzierendes Gewerbe ohne Baugewerbe; Baugewerbe; Handel, Gastgewerbe und Verkehr; Finanzierung, Vermietung, und Unternehmensdienstl; and Öffentliche und private Dienstleister. The index sums the squares of sectoral employment shares, ranging from 0 to 1, where higher values indicate greater specialization.
All
All
−188.451**
Constant
359.668 [951.889]
−266.560* [145.2] 1259.46** [402.61] −101.134 [126.85] −123 042.6 [78 646.6] −81.732 [1306.04] −127.14 [908.23] −176.716 [238.13] 8810.2 [11 269.3]
Year FE Region FE N Within-R2
Yes Yes 164 0.692
Yes Yes 164 0.693
Own Population density High school drops. Herfindahl % Age 17–24 % Age 45–64
[84.304] 1270.59** [403.915] −95.901 [125.188] −135 000.8* [77 760.62] −168.588 [1256.98] −50.358 [866.604]
Owner t −1
Notes: Clustered robust standard errors in brackets. All variables are firstdifferenced, except for Owner t −1 which is initial ownership level. * p < 0.1. ** p < 0.05. *** p < 0.01.
3 of Table 1. A one period lag of own in column 3 does not change this. Column 4 pools the data without year fixed effects, showing a negative and significant coefficient, as we have gone beyond a two period sample. With year fixed effects, own is again positive and significant. Still, a Durbin–Wu–Hausman test of homeownership’s endogeneity rejects the null (F (1, 83) = 4.71, p > F = 0.03), indicating that OLS is inconsistent. Table 3’s ‘Reduced Form’ column shows a negative effect of bomb on unemployment. The ‘First Stage’ column shows bomb has a larger direct effect on homeownership rates (R2 = 34%). Adding to the IV’s validity, the F -statistic on the excluded IV and under-ID statistics in the ‘Second Stage’ column are significant. The IV point estimate (0.54) is much larger than the OLS. The IV is imprecise because its first-stage correlation is weaker than the biased unmediated OLS correlation (R2 of 34% vs. 60%). The standard error grows from 0.036 to 0.301, but the t-ratio remains close: 2.10 and 1.80. Table 4 replicates the IV estimation with first-differences variables. The results are the same, apart from the IV coefficient growing to 0.62, yet retaining a t-ratio of 1.80, as in Table 2. 3. Mechanisms We aggregated county-level net migration data (inflows less outflows) from Destatis (2014) to proxy labour mobility. Table 5 shows that, after differencing, higher homeownership is associated with lower outflows relative to inflows. Table 5 does not rule out the second mechanism highlighted in Blanchflower and Oswald (2013) that homeownership restricts commercial land development, either through zoning regulations or NIMBYism. 4. Conclusion We support the hypothesis that homeownership impairs the labour market. This links two facts for Germany – unusually low homeownership and low unemployment – and has implications for Europe’s high homeownership, high unemployment countries (Spain, Italy). We leave this for future work.
68
N. Wolf, P. Caruana-Galizia / Economics Letters 135 (2015) 65–68
Acknowledgements We thank Alexandra Spitz-Oener, Michael Burda, and Andrew Oswald for helpful comments, and Felix Kersting for research assistance. Paul Caruana-Galizia was funded by a Macrohist fellowship from a European Commission Marie Curie programme. The usual disclaimer applies. References Bashow, D.L., 2014. In praise of bomber harris and area bombing. R. Can. Air Force J. 3, 28–56. Blanchflower, D.G., Oswald, A., 2013. Does high home-ownership impair the labor market? National Bureau of Economic Research No. 19079. Destatis (German Federal Statistical Office). 2014. Regionaldatenbank Deutschland. Online: https://www.regionalstatistik.de/genesis/online/logon (Accessed: 31.10.14). Eighth Air Force Historical Society (EAFHS). 2014. WWII 8th AAF COMBAT CHRONOLOGY. Online: http://www.8thafhs.org/combat1942.htm Accessed: 3.11.14. Hastings, M., 2013. Bomber Command. Zenith Military Classics, Minneapolis, MN.
Laamanen, J.-P., 2013. Home-ownership and the labour market: evidence from rental housing market deregulation. Tampere Economic Working Papers Net Series, No. 89. Lerbs, O., 2011. Is there a link between homeownership and unemployment? Evidence from German regional data. Int. Econ. Policy 8, 407–426. Lerbs, O., Oberst, C., 2014. Explaining the spatial variation in homeownership rates: Results for German regions. Reg. Stud. 48 (5), 844–865. Oswald, A., 1996. A conjecture on the explanation for high unemployment in the industrialized nations: part 1. University of Warwick Economic Research Papers. Terraine, J., 1985. The Right of the Line: The Role of the RAF in World War Two. Hodder and Stoughton, London. TNA: CAB 120/303, 1945. United States of America Air Force (USAAF) 1945a. The United States Strategic Bombing Survey: over-all report (European war). Washington, DC. United States of America Air Force (USAAF) 1945b. The United States Strategic Bombing Survey: Summary Report. Washington, DC. USSBS (United States Strategic Bombing Survey) 1947. The effects of strategic bombing on German morale [by] the United States Strategic Bombing Survey, Morale Division. Dates of survey: March–July, 1945. Washington. Voigtländer, M., 2009. Why is the German homeownership rate so low? Hous. Stud. 24 (3), 355–372. Vonyo, T., 2012. The bombing of Germany: The economic geography of war-induced dislocation in west German industry. Eur. Rev. Econ. Hist. 16 (1), 97–118.