East Side Story: Historical Pollution and PersistentNeighborhood Sorting

Heblich (Bristol), Trew (St. Andrews) and Zylberberg (Bristol)

NIESR, 31.05

1/ 36

During the 19th century, pollution was extremely high in industrial cities, withwide within-city variations.

“In Manchester ... prevailing and strongest winds [blow] from thesouth west. This meant that when the dense sulphurous smokeleft Manchester’s tall chimneys it usually moved north east, andthis was to have a marked effect on the shaping of the city.”

Stephen Mosley (2008), The Chimney of the World.

“But it appears to me very remarkable that while so many of thetradesmen live out of town (for hardly live in Manchester) in otherdirections than those in which the vast mass of smoke iscarried, few reside in the outskirts on the eastern part of thetown.”

Reverend John Molesworth (1843).

2/ 36

I We locate all industrial chimneys on historical maps (1880–1900)of the 70 largest industrial areas in England.

I We use a pollution model to predict air pollution levels with 50mprecision.

I We geolocate historical Censuses records and construct indices ofneighborhood composition in 1817, 1881 and 1971–2011.

3/ 36

Factories that emerged during the industrial revolution generated a veryunequal distribution of pollution exposure, and:

1. [Sorting] The polluted neighborhoods were markedly poorer than othersin 1881, while they were similar in 1817.

2. [Persistence] These neighborhoods are still poorer in 2011.

4/ 36

Literature

Neighborhood sorting: We document sorting of a large magnitude along atemporary environmental disamenity.

I Banzhaf and Walsh [2008]; Chay and Greenstone [2005]; Depro et al. [2015]; Kuminoff et al.[2013]; Lee and Lin [2017]

I Hanlon [2016]; Chen et al. [2017]; Freeman et al. [2017]

Pollution exposure: Long run effect of environmental disamenities.I Performance: e.g., Graff Zivin and Neidell [2012]; Lavy et al. [2014];I Violent crime: e.g., Herrnstadt and Muehlegger [2015]; Heyes and Saberian [2015]I Health: e.g., Barreca et al. [2014]; Clay et al. [2016]; Beach and Hanlon [2016]

Dynamics of segregation and tipping points: Exogenous and temporarypollution shock has persistent but non-linear effect on income segregation.

I Card et al. [2008]; Durlauf [2004]; Guerrieri et al. [2013]; Rosenthal and Ross [2015]

Methodology: Exploiting historical maps and individual census data.I Hornbeck [2012]; Redding and Sturm [2016]; Siodla [2015]

5/ 36

Data

Summary

1. We locate chimneys using maps around 1880–1900.

2. We construct the associated atmospheric pollution.

3. We develop an algorithm to geo-locate individual observations fromhistorical census data in small spatial units (LSOA).

6/ 36

1. Chimney locations and geographic organisation of cities

Ordnance Survey Maps—25 inch to the mile, 1880–1900, the mostdetailed topographic mapping that covers all of England and Wales:

I road, railway, rivers and canals,I public amenities, i.e., schools, parks etc.,I industrial sites and associated chimneys.

We select:I 70 cities in England and Wales,I a quasi-exhaustive snapshot of the industry and associated pollution,I 60% of the total population in 1817 and 1881, and 66% of the total

population in 2011.

7/ 36

1. Town maps and chimney locations

Town maps—chimney symbols (about 5,000 industrial chimneys).

(a) Example 1. (b) Example 2.

8/ 36

1. Marking chimneysTown maps—marking and identifying chimneys.

Sources: Ordnance Survey Maps—25 inch to the mile, 1880–1900.

I Symbol: the red symbol X can easily be identified by any recognitionalgorithm.

I Id: the chimney 00007 belongs to Eastbrook Dye Works while 00006belongs to Britannia Saw Mills.

Output: about 5,000 chimneys across 5,000 map tiles and 70 cities (metal:1,613; textile production: 944, bricks: 262).

9/ 36

2. Modelling pollution: emissionWe construct pollution emission from a chimney in industry i:

Ei =Ci × Li

Chi

whereI Ci is an industry-specific measure of coal per worker [Hanlon, 2016],I Li is employment in industry i,I and Ci is the number of chimneys of type i.

Industry Weight EiBrewery 0.363Bricks 1.051Chemical 0.848Food processing 0.777Metal 1.000Mining 3.559Paper 2.478Shipbuilding 1.027Tannery 0.171Textile production 0.475Wood 0.106Works 0.100

10/ 36

2. Modelling pollution: dispersion

For each chimney, we predict pollution dispersion using ADMS 5:I wind directions,I and complex terrain.

(a) North England. (b) South England.

11/ 36

2. Modelling pollution: dispersion

For each chimney, we predict pollution dispersion using ADMS 5:I wind directions,I and complex terrain.

(a) Oldham (industrial). (b) Oldham (residential).

12/ 36

2. Modelling pollution: illustration

12/ 36

2. Modelling pollution: illustration

12/ 36

2. Modelling pollution: over-identification

Report of the Air Pollution Advisory Board, 1915,Manchester:

Deposits (1915) Our estimates

Station m. tons/ sq. km. µg/m3

Ancoats hospital 30.59 119.95Philips Park 22.59 74.49Whitworth Street 22.51 102.47Queen’s Park 20.18 70.00Moss Side 18.69 29.11Whitefield 15.53 11.92Fallowfield 13.24 17.69Davyhulme 12.68 6.93Cheadle 10.63 9.40Bowdon 6.25 0.02

m. tons/ sq. km.: monthly deposits of impurities in megatons per km2 .µg/m3 : micrograms of sulphur dioxide per cubic meter.

13/ 36

3. Census Data (geo-location)

1817: Baptism records over 1813–20 from Shaw-Taylor and Wrigley [2014] toreconstruct a quasi-census (834 parishes).

1881: Micro-census with coded occupations (parish code and address). Wedevelop a method to assign individual records to 2001 LSOAs:

I We match addresses with a registry of addresses: 25% perfect matches.I We infer the geo-references of unmatched entries given (i) their location

in the census books and (ii) their well-matched neighbors.geocode

1971–2011: Micro-censuses with coded occupations (2001 LSOAs).

14/ 36

Empirical strategy

We estimate the following equation (i: LSOA, p parish, c: city, t: time):

Yit = α+ βPi + γXi + δc + εict

with:I Yit: our measures of occupational structure in 1881 or 2011,I Pi is the treatment: non-linear function of sources, wind, topography.I Xi are controls:

I populatin and pre-treatment: employment in 1881, measures of occupationalstructure in 1817, (log) property tax in 1815,

I topography: elevation (min/max/mean), distance to canals,I pollution sources and amenities: distance to the closest chimney

(heavy/light), distance to town hall, distance to parks, LSOA area, LSOAarea within the city,

I coordinates: longitude, latitude.

I δc: city FE.

15/ 36

Two concerns:

1. Omitted variables:I counterfactual pollution imprints,I balance tests.

2. Reverse causality and strategic industry location upwind of poorneighborhoods,

I instrument with pollution sources as predicted by waterways,I instrument with the location of historical industrial districts, proxied by

1700–1800 steam engines [Kanefsky and Robey, 1980; Nuvolari et al.,2011].

With the IVs, we look at the effect net of endogenous industry locationdecisions (e.g., political decisions or market responses).

16/ 36

Descriptive statistics

Pollution Standard deviationVARIABLES Mean high low total between within

Air pollutionNormalized pollution -.034 .233 -.298 .928 .596 .560

Population measures1817∗Low-skilled workers .785 .788 .780 .110 .090 .072High-skilled workers .093 .083 .102 .086 .085 .068Farmers .122 .128 .117 .091 .065 .047Property tax (log) 9.93 10.04 9.82 1.25 1.04 .076

1881Low-skilled workers .600 .627 .574 .255 .149 .235High-skilled workers .278 .287 .268 .241 .125 .216Farmers .121 .085 .157 .199 .175 .180

2011Low-skilled workers .587 .604 .569 .173 .117 .123High-skilled workers .413 .395 .430 .173 .117 .123

Topography controlsMaximum elevation (m) 72.8 66.2 79.3 66.0 63.2 31.8Minimum elevation (m) 52.5 49.9 55.2 48.9 44.1 19.6Mean elevation (m) 62.3 57.8 66.7 55.8 51.6 23.4Distance canals (km) 6.12 5.47 6.76 14.9 19.1 1.30

Amenities controlsDistance town hall (km) 4.64 4.10 5.18 5.35 4.72 1.27Distance parks (km) 9.57 9.37 9.77 23.9 28.8 1.15Share LSOA within city .296 .404 .190 .417 .245 .296Area (square km) .939 .599 1.27 3.99 7.99 3.49Distance heavy (km) 2.48 1.92 3.03 6.36 8.81 .972Distance light (km) 5.31 4.93 5.69 13.5 17.6 1.15

DirectionEastings (km) 453 454 452 74.9 76.9 1.90Northings (km) 291 291 292 123 126 1.92

17/ 36

18/ 36

(a) Distance and Composition. (b) Direction and Composition | Amenities/Dis-tance.

There is an excess of 1.5 p.p. low-skilled workers North-East of the averagechimney.

19/ 36

Neighborhood sorting (19th century)

1. Benchmark results in 1881

Share of low-skilled (1) (2) (3) (4) (5) (6)Pollution .0440 .0428 .0409 .0377 .0354 .0337

(.0071) (.0072) (.0068) (.0066) (.0067) (.0069)[.1737] [.1689] [.1614] [.1491] [.1397] [.1332]

Observations 5,538 5,538 5,538 5,538 5,538 5,538Fixed effects (city) No Yes Yes Yes Yes YesControls (population) No No Yes Yes Yes YesControls (topography) No No No Yes Yes YesControls (amenities) No No No No Yes YesControls (lat./lon.) No No No No No Yes

Air pollution explains about 15% of the variation in neighborhood composition.

20/ 36

2. Validity

(a) 1817. (b) 1881.

(c) 1881 (parish FEs). (d) 1881 (parish FEs and residentialpollution).

21/ 36

2. Validity

We run a series of robustness checks to ensure that our estimates are notdriven by a non-random location of industries within cities:

1. balance test: Table .

2. counterfactual pollution imprints: Table .

3. IV: Table .

4. fixed effects, clusters, and sample selection: Table .

5. sensitivity to the exclusion of suburbs: Table .

6. other outcomes: Table .

22/ 36

Dynamics of Persistence (end of 20th century)

Benchmark results in 1971,..., 2011

Share of low-skilled workers 1971 1981 1991 2001 2011Pollution .0243 .0309 .0388 .0374 .0354

(.0046) (.0050) (.0063) (.0063) (.0057)[.1914] [.2203] [.2071] [.2298] [.2028]

Observations 5,535 5,538 5,538 5,538 5,538Fixed effects (city) Yes Yes Yes Yes YesExtended controls Yes Yes Yes Yes Yes

Past pollution explains up to 20 percent of neighborhood segregation today.

Robustness checks show...I similar patterns in regressions on deprivation indices (income, health,

education, crime, etc.)I a drop in house prices by 8% ( Prices ).

23/ 36

Non-linear dynamics of persistence

24/ 36

Non-linear dynamics of persistence

24/ 36

Non-linear dynamics of persistence

24/ 36

Non-linear dynamics of persistence

24/ 36

Non-linear dynamics of persistence

24/ 36

Non-linear dynamics of persistence: discussion

1. The graphs for 1971–2011 show two patterns:

I Persistence for extreme values of pollution.

I Reversion to the mean for intermediate values of pollution.

2. Results indicate that there is a tail effect generating tipping dynamics: theimpact of the initial distribution of pollution can be very large.

Between cities , Social housing

25/ 36

Local factors of persistence: bombs and homophily

Yit =2011∑τ=1971

βτPi × 1τ=t +2011∑τ=1971

βbτPi × 1τ=t × Di +

2011∑τ=1971

γτXi × 1τ=t + νi + δt + εit

(a) Bombings. (b) Local homophily [Schelling, 1971].

26/ 36

Local factors of persistence: bombs and homophily

Yit =2011∑τ=1971

βτPi × 1τ=t +2011∑τ=1971

βbτPi × 1τ=t × Di +

2011∑τ=1971

γτXi × 1τ=t + νi + δt + εit

(a) Bombings. (b) Local homophily.

27/ 36

Discussion

1. Potential Mechanisms

Consumption amenitiesI Public infrastructure (e.g., parks, hospitals, transports),I Private infrastructure (e.g., pubs),I Housing stock (Victorian or social housing, street width etc.).

Production amenitiesI Location of production,I Knowledge spillovers.

Other peer effectsI Schooling,I Crime,I Preferences (e.g., Whites/Blacks in the US).

Amenities , Composition , Social Housing

28/ 36

2. Policy implications

1. Economies in the process of structural transformation where pollutionpresents a major challenge should consider:

I the effects of pollution on neighborhood sorting,

I the long-run costs of the resulting spatial distribution of residents andactivities.

2. The success of urban redevelopment policies depends on tippingdynamics & concentration.

29/ 36

3. Next steps

30/ 36

An extension on urbanization

Mechanism:

I Pollution affects the relative return to the traditional sector rt/rm,

I which should foster the reallocation of land (Lt,Lm) and other factorstoward the modern sector.

31/ 36

Agricultural returns in 1877

Data: agricultural census (parish-level, one digitized wave).

Fragile crops Resistant crops LivestockPollution -.0518 -.0744 .1203 .1767 -.0437

(.0269) (.0339) (.1054) (.1047) (.1152)

Pollution × Limestone .1366 -.3411(.0642) (.4060)

Observations 1,462 1,462 1,462 1,462 1,252Fixed effects (city) Yes Yes Yes Yes YesControls Yes Yes Yes Yes Yes

Pollution affects the relative return to the traditional sector rt/rm.

32/ 36

Land use in 1930

Data: Dudley Stamp Land Utilisation map.

Panel AShare of urban Share of farming Unproductive

Pollution .0792 -.0750 -.0041(.0120) (.0108) (.0026)

Observations 5,764 5,764 5,764Fixed effects (city) Yes Yes YesControls Yes Yes Yes

Pollution fosters the reallocation of land (Lt,Lm) toward the modern sector.

33/ 36

Land use in 1930

Data: Dudley Stamp Land Utilisation map.

(a) Urban. (b) Farming.

34/ 36

Thanks!

yanos.zylberberg@bristol.ac.uk

35/ 36

Appendix IBanzhaf, H. S. and Walsh, R. P.: 2008, Do people vote with their feet? an empirical test of tiebout, American Economic Review

98(3), 843–63.

Barreca, A., Clay, K. and Tarr, J.: 2014, Coal, smoke, and death: Bituminous coal and american home heating, Technical Report 19881.

Beach, B. and Hanlon, W. W.: 2016, Coal smoke and mortality in an early industrial economy, Working paper.

Card, D., Mas, A. and Rothstein, J.: 2008, Tipping and the dynamics of segregation, The Quarterly Journal of Economics 123(1), 177–218.

Chay, K. Y. and Greenstone, M.: 2005, Does air quality matter? evidence from the housing market, Journal of Political Economy113(2), 376–424.

Chen, S., Oliva, P. and Zhang, P.: 2017, The effect of air pollution on migration: Evidence from china, Working Paper 24036, NationalBureau of Economic Research.

Clay, K., Lewis, J. and Severnini, E.: 2016, Canary in a coal mine: Infant mortality, property values, and tradeoffs associated with mid-20thcentury air pollution, Technical Report 22155.

Depro, B., Timmins, C. and O’Neil, M.: 2015, White flight and coming to the nuisance: Can residential mobility explain environmentalinjustice?, Journal of the Association of Environmental and Resource Economists 2(3), 439–468.

Durlauf, S. N.: 2004, Neighborhood effects, in J. V. Henderson and J. Thisse (eds), Handbook of Regional and Urban Economics, Vol. 4,Elsevier, pp. 2173–2242.

Freeman, R., Liang, W., Song, R. and Timmins, C.: 2017, Willingness to pay for clean air in china, Working Paper 24157, National Bureauof Economic Research.

Graff Zivin, J. and Neidell, M.: 2012, The impact of pollution on worker productivity, American Economic Review 102(7), 3652–73.URL: http://www.aeaweb.org/articles?id=10.1257/aer.102.7.3652

Guerrieri, V., Hartley, D. and Hurst, E.: 2013, Endogenous gentrification and housing price dynamics, Technical report.

Hanlon, W. W.: 2016, Coal smoke and the costs of the industrial revolution, Working paper.

Herrnstadt, E. and Muehlegger, E.: 2015, Air pollution and criminal activity: Evidence from chicago microdata, Working paper.

Heyes, A. and Saberian, S.: 2015, Air pollution causes violent crime, Working paper.

Hornbeck, R.: 2012, The enduring impact of the american dust bowl: Short- and long-run adjustments to environmental catastrophe,American Economic Review 102(4), 1477–1507.

Kanefsky, J. and Robey, J.: 1980, Steam engines in 18th-century britain: A quantitative assessment, Technology and Culture21(2), 161–186.URL: http://www.jstor.org/stable/3103337

Kuminoff, N. V., Smith, V. K. and Timmins, C.: 2013, The new economics of equilibrium sorting and policy evaluation using housingmarkets, Journal of Economic Literature 51(4), 1007–1062.

35/ 36

Appendix II

Lavy, V., Ebenstein, A. and Roth, S.: 2014, The impact of short term exposure to ambient air pollution on cognitive performance and humancapital formation, Working Paper 20648, National Bureau of Economic Research.URL: http://www.nber.org/papers/w20648

Lee, S. and Lin, J.: 2017, Natural amenities, neighborhood dynamics, and persistence in the spatial distribution of income, The Review ofEconomic Studies forthcoming.

Nuvolari, A., Verspagen, B. and von Tunzelmann, N.: 2011, The early diffusion of the steam engine in britain, 1700–1800: a reappraisal,Cliometrica 5(3), 291–321.

Redding, S. J. and Sturm, D. M.: 2016, Estimating neighborhood effects: Evidence from war-time destruction in london, Technical report.

Rosenthal, S. S. and Ross, S.: 2015, Change and persistence in the economic status of neighborhoods and cities, in J. V. H.Gilles Duranton and W. C. Strange (eds), Handbook of Regional and Urban Economics, Vol. 5 of Handbook of Regional and UrbanEconomics, Elsevier, pp. 1048–1120.

Schelling, T. C.: 1971, Dynamic models of segregation, Journal of mathematical sociology 1(2), 143–186.

Shaw-Taylor, L. and Wrigley, E. A.: 2014, Occupational structural and population change. Chapter 2 in Floud, R., Humphries, J. andJohnson P., The Cambridge Economic History of Modern Britain: Volume 1, Industrialisation 1700-1860 . Cambridge University Press.

Siodla, J.: 2015, Razing san francisco: The 1906 disaster as a natural experiment in urban redevelopment, Journal of Urban Economics89, 48–61.

Williamson, J. G.: 1980, Earnings inequality in nineteenth-century britain, The Journal of Economic History 40(3), 457–475.

36/ 36

A static model of neighborhood sorting (1/2)

Two neighborhoods in a city j ∈ {W,E}, no moving costs.

A portion γ of workers earn θl; 1− γ earn θh > θl.

Workers maximize,

V(j, `) = A(j, `)c(j, `) subject to c(j, `) + R(j, `) = θ, (1)

where A(j, `) is the amenity level in location ` of neighborhood j, c(j, `) isconsumption and R(j, `) is rent.

A(j, `) = a(j) + x(`, j). (2)

I a(j) is air quality in neighborhood j;I x(`, j) is a location amenity at location ` within neighborhood j

(distributed uniformly over [0, 1]).

36/ 36

A static model of neighborhood sorting (2/2)Amenities and Neighborhood Sorting: a(E) ≤ a(W), γ = 1

2 .

R

A

Rl = θl

θh

Rh

A∗

A(W)

A(E)low-skilled high-skilled

36/ 36

A dynamic and quantitative model (1/2)

Workers are infinitely-lived and choose their location (no moving costs) ineach period to maximize,

V(j, `, t) = A(j, `, t)c(j, `, t) subject to c(j, `, t) + R(j, `, t) = θ,

where A(j, `, t) is the amenity level in location ` of neighborhood j incalendar year t.

A(j, `, t) = a(j, t) + x(`, j) + d(j, t). (3)

I a(j, t) is air quality in neighborhood j (exogenous!);I x(`, j) is a location amenity at location `,I d(j, t) is the endogenous amenity at the neighborhood level.

36/ 36

A dynamic and quantitative model (2/2)

We assume:

d(j, t) = (1− δ)d(j, t − 1) + e(j, t) + b(j, t) (E)

where:

1. 1 − δ captures the inertia of the AR(1) process,

2. there are two endogenous perturbations:I e is continuous in average income,I b is a tail effect.

36/ 36

Calibration and estimation (1/2)

1. We first calibrate observable parameters:

Parameter Value Rationaleθh High income 2 Williamson [1980], highest quartile to the lowestθl Low income 1 Williamson [1980]γ̃ Low-skill share 0.50 Normalizationα Pollution sensitivity 0.102 Correlation pollution/occupation in 1881d Initial amenity 1 Normalization

2. We estimate the equation (E) in the data using:I the within-city residuals of low-skill share between 1881–1971,I the within-city residuals of atmospheric pollution for the 4,519

neighborhoods.

We treat neighborhoods as independent observations (the residuals arecleaned from city-FEs).

36/ 36

Calibration and estimation (2/2)

Parameter Description Estimate Standard errorφe

1 Coefficient for the continuous effect 0.11 0.04φe

2 Curvature for the continuous effect 0.89 0.08φb

1 Coefficient for the tail effect 0.10 0.06φb

2 Curvature for the tail effect 1.45 0.30S̄ Tail point 0.76 0.08δ Depreciation factor 0.08 0.03

These estimates are such that:I δ = 0.08 implies that half of the gap between neighborhoods would be

bridged after only 9 years,I the continuous effect is positive but too small to generate persistent

sorting,I the tail effect is large and convex. The tipping point is 0.76.

36/ 36

Model fit

We simulate the model over the period 1971–2011.

Data Baseline SH-LSpread in 1971 .0550 .0555 .0555Spread in 2011 .0278 .0235 .0281Correlation ρ2011,1971 .4337 .4010 .4331

To improve the model fit, we re-estimate with social housing accounting forthe 1979 liberalization (“SH”).

36/ 36

The 1979 liberalization Back

36/ 36

Pollution Standard deviationVARIABLES Mean high low total between within

Air pollutionNormalized pollution -.034 .233 -.298 .928 .596 .560

Population measures1817∗Low-skilled workers .785 .788 .780 .110 .090 .072High-skilled workers .093 .083 .102 .086 .085 .068Farmers .122 .128 .117 .091 .065 .047Property tax (log) 9.93 10.04 9.82 1.25 1.04 .076

1881Low-skilled workers .600 .627 .574 .255 .149 .235High-skilled workers .278 .287 .268 .241 .125 .216Farmers .121 .085 .157 .199 .175 .180

2011Low-skilled workers .587 .604 .569 .173 .117 .123High-skilled workers .413 .395 .430 .173 .117 .123

Topography controlsMaximum elevation (m) 72.8 66.2 79.3 66.0 63.2 31.8Minimum elevation (m) 52.5 49.9 55.2 48.9 44.1 19.6Mean elevation (m) 62.3 57.8 66.7 55.8 51.6 23.4Distance canals (km) 6.12 5.47 6.76 14.9 19.1 1.30

Amenities controlsDistance town hall (km) 4.64 4.10 5.18 5.35 4.72 1.27Distance parks (km) 9.57 9.37 9.77 23.9 28.8 1.15Share LSOA within city .296 .404 .190 .417 .245 .296Area (square km) .939 .599 1.27 3.99 7.99 3.49Distance heavy (km) 2.48 1.92 3.03 6.36 8.81 .972Distance light (km) 5.31 4.93 5.69 13.5 17.6 1.15

DirectionEastings (km) 453 454 452 74.9 76.9 1.90Northings (km) 291 291 292 123 126 1.92

36/ 36

Robustness checks: balance test Back

Panel A: Share of low-skilled (1817)(1) (2) (3) (4) (5) (6)

Pollution .0077 .0115 .0070 .0021 .0048 .0061(.0100) (.0142) (.0146) (.0145) (.0187) (.0182)[.0715] [.1063] [.0654] [.0201] [.0451] [.0567]

Observations 559 559 559 559 559 559Fixed effects (city) No Yes Yes Yes Yes YesControls (population) No No Yes Yes Yes YesControls (topography) No No No Yes Yes YesControls (amenities) No No No No Yes YesControls (lat./lon.) No No No No No Yes

Panel B: Property tax (log, 1815)(1) (2) (3) (4) (5) (6)

Pollution .3254 .4828 .2205 .1668 -.0283 -.0319(.0843) (.1277) (.0926) (.0952) (.1079) (.1107)[.2623] [.3892] [.1777] [.1344] [-.0228] [-.0257]

Observations 532 532 532 532 532 532Fixed effects (city) No Yes Yes Yes Yes YesControls (population) No No Yes Yes Yes YesControls (topography) No No No Yes Yes YesControls (amenities) No No No No Yes YesControls (lat./lon.) No No No No No Yes

36/ 36

Robustness checks: counterfactual pollution Back

Share of low-skilled workers (1) (2) (3) (4)Pollution .0442 .0355 .0427 .0377

(.0082) (.0113) (.0139) (.0137)[.1746] [.1403] [.1684] [.1487]

Mirror Pollution -.0120 -.0260 -.0273 -.0262(.0068) (.0145) (.0146) (.0145)[-.0474] [-.1027] [-.1080] [-.1035]

Static Pollution .0209 .0253 .0228(.0197) (.0210) (.0209)[.0828] [.0831] [.0900]

Placebo Industry -.0136 -.0114(.0153) (.0152)[-.0536] [-.0452]

Residential Pollution .0138(.0046)[.0546]

Observations 5,538 5,538 5,538 5,538Fixed effects (city) Yes Yes Yes YesExtended controls Yes Yes Yes Yes

Note: there is a correlation between the mere proximity to industries andneighbourhood composition, but pollution explains most of this covariance.

36/ 36

Robustness checks: IV Back

First stage Pollution(1) (2) (3) (4)

Pollution (waterways) .2020 .1834(.0346) (.0384)

Pollution (steam engines) .1700 .1937(.0325) (.0399)

Second stage Share of low-skilled workers (1881)(1) (2) (3) (4)

Pollution .0936 .0695 .1011 .0801(.0296) (.0358) (.0379) (.0402)[.3695] [.2743] [.3991] [.3160]

Observations 4,830 4,557 2,519 1,860F-statistic 33.91 22.81 28.00 23.49OLS coefficient .0216 .0208 .0221 .0193Sample Canal>250m Canal>500m Mill<10km Mill<5kmFixed effects (city) Yes Yes Yes YesExtended controls Yes Yes Yes Yes

36/ 36

Robustness checks: suburbs Back

Share of low-skilled (1) (2) (3) (4)Pollution .0339 .0350 .0361 .0425

(.0068) (.0075) (.0085) (.0093)[.1340] [.1383] [.1425] [.1677]

Observations 5,122 3,521 2,467 1,583Sample Hall<10km Hall<5km LSOA ∩ city LSOA ∩ city>0.5Fixed effects (city) Yes Yes Yes YesExtended controls Yes Yes Yes Yes

36/ 36

Robustness checks: other outcomes Back

MigrantsVARIABLES Low-skilled, all All England/Wales CommonwealthPollution .0242 .0314 .0282 .0033

(.0051) (.0072) (.0071) (.0019)[.1192] [.1103] [.1101] [.0433]

Observations 5,538 4,312 4,312 4,312Fixed effects (city) Yes Yes Yes YesExtended controls Yes Yes Yes Yes

36/ 36

Robustness checks: house prices and transactions (2001–2011)Back

Nationwide Land registryHouse prices House prices Transactions

VARIABLES (1) (2) (3) (4) (5) (6)Pollution -.1035 -.0852 -.1116 -.0642 -.0473 -.0878

(.0168) (.0121) (.0161) (.0120) (.0233) (.0230)[-.1684] [-.1385] [-.2030] [-.1168] [-.0892] [-.1656]

Observations 5,226 5,226 5,538 5,538 5,538 5,538Fixed effects (city) Yes Yes Yes Yes Yes YesExtended controls Yes Yes Yes Yes Yes YesControls (house ch.) No Yes No Yes No Yes

36/ 36

Robustness checks: social housing/migrant shares (1971–2011)Back

Effect of pollution on ... 1971 1981 1991 2001 2011Social housing .0035 .0191 .0316 .0260 .0283

(.0138) (.0123) (.0097) (.0078) (.0073)[.0136] [.0653] [.1278] [.1190] [.1404].287 .358 .297 .260 .232

Owners -.0021 -.0086 -.0251 -.0247 -.0312(.0077) (.0089) (.0074) (.0065) (.0065)[-.0082] [-.0318] [-.1035] [-.1073] [-.1399]

.429 .494 .580 .583 .535

Migrants (New Commonwealth) .0067 .0147 .0143 .0172 .0253(.0034) (.0046) (.0046) (.0056) (.0075)[.1030] [.1730] [.1718] [.1812] [.2101].041 .060 .064 .085 .128

Migrants (Other) .0004 -.0003 -.0008 -.0006 .0033(.0014) (.0012) (.0008) (.0011) (.0014)[.0081] [-.0054] [-.0116] [-.0098] [.0439].034 .035 .043 .053 .075

Observations 5,534 5,538 5,538 5,538 5,538Fixed effects (city) Yes Yes Yes Yes YesExtended controls Yes Yes Yes Yes Yes

36/ 36

Robustness checks: fixed effects, clusters and sample selectionBack

Panel A: Fixed effects Share of low-skilled workers (1881)(1) (2) (3)

Pollution .0387 .0377 .0393(.0091) (.0115) (.0118)[.1527] [.1490] [.1551]

Observations 5,538 5,538 5,538Fixed effects Parish Ward MSOAPanel B: Clusters Share of low-skilled workers (1881)

(1) (2) (3)Pollution .0337 .0337 .0337

(.0058) (.0066) (.0086)[.1332] [.1332] [.1332]

Observations 5,538 5,538 5,538Clusters MSOA Ward CityPanel C: Sample Share of low-skilled workers (1881)

(1) (2) (3)Pollution .0297 .0642 .0343

(.0070) (.0126) (.0069)[.1172] [.2536] [.1353]

Observations 3,661 4,395 5,247Excluding... London NW NE

36/ 36

Figure: House transaction prices (y-axis) and pollution (x-axis) across neighborhoods– average and evolution between 1995 and 2011.

(a) Average (b) Evolution

Notes: The left (resp. right) panel represents the relationship between the (logarithm of the) average transactionprices between 2000 and 2011 (resp. in 1995, 2000, and 2011) and our (standardized) measure of past pollution.

We consider the residuals of all measures once cleaned by city Fixed-Effects, geographic and topographic controls.For the sake of exposure, we group neighborhoods, create 100 bins of neighborhoods with similar past pollution and

represent the average house prices within a pollution-bin. The lines are locally weighted regressions on allobservations.

36/ 36

Back

(a) Distribution (high pollution). (b) Distribution (low pollution).

(c) Persistence (high pollution). (d) Persistence (low pollution). 36/ 36

Non-linear dynamics of persistence: role of social housing

Back

36/ 36

The Mill of the World

Between 1700-1870, the UK went through a period of economic growth andstructural transformation:

I with a large take-off around 1830: Description

I essentially fostered by the textile industry.

This transformation had been accompanied by a huge rural-urbanmigration which was strongest between 1780 and 1850.

Growth rates in cities

36/ 36

The Chimney of the World

In parallel, coal became the main energy source with a large accelerationaround 1850 due to the adoption of large boilers.

I Coal use: Figure .

I Industry share: Figure .

It leads to high levels of pollution: Figure .

36/ 36

Timeline

1760–1800: start of the industrial revolution.

1800–1840: small acceleration and large migration to the cities.

1840–1900: rapid economic growth, stable urban population and take-off forpollution.

1900–1915: stabilisation.

1930–1968: de-industrialisation and first Clean Air Act of 1956 followingLondon fog.

1968–today: end of coal pollution after more restrictive Clean Air Acts of1968.

36/ 36

Population growth (1/2)

Descriptive Statistics

Population Growth Rates in cities, 1801-1891.

Notes: The figure plots the average decadal population growth rate for the period 1801-1891 in our sample cities.

36/ 36

Population growth (2/2)

Descriptive Statistics

Huge rural-urban migration flows between 1800 and 1850:

Population Manchester Salford Liverpool Leeds

Period1801 75,281 18,179 82,000 53,000

1851 303,382 85,108 376,000 172,000

1861 338,722 102,449 443,938 207,165

1871 351,189 124,801 493,346 259,201

Source: Historical censuses (1801-1871).

36/ 36

Economic growth

Descriptive Statistics

Output growth, 1700-1870:

Period Industry Services Total Population

1700-1760 0.49 0.71 0.63 0.32

1760-1780 1.00 0.66 0.81 0.62

1780-1801 2.18 1.40 1.54 0.97

1801-1830 2.59 1.79 1.69 1.43

1830-1870 3.01 2.58 2.40 1.18

Source: British Economic Growth, 1270-1870 (Broadberry et al, 2010).

36/ 36

Pollution intensity

National Air Quality Standards:I Primary standards: 12 µg/m3.I Secondary standards: 15 µg/m3.

Coal Consumption , Coal Use

36/ 36

Energy Back

Coal consumption (shares), domestic, industrial and mining, 1560–2001.

Notes: The figure illustrates the coal consumption share for domestic use (solid), industrial use in manufacturing andiron and steel production (dash), mining (dash-dot) and other sectors as composite of railways building, ship buildin,

gas and electrics. Sources: Mitchell, 1988.

36/ 36

Coal output and consumption Back

Coal consumption in million tons, 1560–2001.

Notes: The figure illustrates the increase in coal consumption over the period 1560–2001. The figure is based onWarde (2007) who reports coal consumption in petajoule. To convert numbers from petajoule to tons, we use aconversion factor of 1:34,140. The two solid grey lines indicate the years 1817 and 1881 for which we have detailedoccupational information within cities. The dashed grey lines mark the introduction of the 1956 Clean Air Act and thestricter 1968 Clean Air Act. Sources: Warde, 2007.

36/ 36

Geo-locating individuals in census data (1881) Back

A list of entries in a parish is:

id i folio f address

block n break

1. f1 5, Chruch Street, Clifton, Bristol

n1

......

...

45. f1 23, Church Street, Clifton, Bristol

n1

46. f1 18 Ambrose Vale Cliftonwoods

n2 B1

......

...

78. f1 Ambrose villas, Clifton, Bristol

n2

79. f2 Ambrose villas, Clifton, Bristol

n2 B2

36/ 36

Geo-locating individuals in census data (1881) Back

A list of entries in a parish is:

id i folio f address block n

break

1. f1 5, Chruch Street, Clifton, Bristol n1

......

...

45. f1 23, Church Street, Clifton, Bristol n146. f1 18 Ambrose Vale Cliftonwoods n2

B1

......

...

78. f1 Ambrose villas, Clifton, Bristol n279. f2 Ambrose villas, Clifton, Bristol n2

B2

36/ 36

Geo-locating individuals in census data (1881) Back

A list of entries in a parish is:

id i folio f address block n break1. f1 5, Chruch Street, Clifton, Bristol n1

......

...

45. f1 23, Church Street, Clifton, Bristol n146. f1 18 Ambrose Vale Cliftonwoods n2 B1

......

...

78. f1 Ambrose villas, Clifton, Bristol n279. f2 Ambrose villas, Clifton, Bristol n2 B2

36/ 36

Geo-locating individuals in census data (1881) Back

1. We run a fuzzy-matching process (20% perfect match, 35% goodenough).

2. We locate all matched households within their parish.3. We then define a new id, e.g., a folio number, and run:

4. We associate all households with the same id to the matched LSOA.Notes: sensitivity tests to id definition, fuzzy threshold, iteration in the algorithm.

36/ 36

What are the potential channels? Back

Panel A: Deprivation indicesIndex Income Empl. Educ. Health Housing Crime Environ.

Pollution .0627 .0681 .0491 .0781 .0341 .0003 .0273 .0559(.0119) (.0135) (.0116) (.0120) (.0083) (.0050) (.0076) (.0131)[.2435] [.2407] [.1783] [.2704] [.1346] [.0012] [.1133] [.2271]

Observations 5,538 5,538 5,538 5,538 5,538 5,538 5,538 5,538Panel B: Education and crime

Private Student Disadvantaged School Anti-social Drug-rel. ViolentSchool (KS2) Scores (KS2) Students (KS2) VA (KS2) Behaviors Burglary Crimes Crimes

Pollution -.0014 -.0049 .0073 -.0000 .0042 .0314 .0092 .0654(.0027) (.0012) (.0013) (.0001) (.0044) (.0084) (.0021) (.0120)[-.0110] [-.1040] [.0887] [-.0007] [.0327] [.0925] [.1186] [.1854]

Observations 5,538 5,538 5,538 5,538 5,538 5,538 5,538 5,538Panel C: Housing quality

Building Building Building Year of Square1900 1970 2000 construction meters Bedrooms Flats Detached

Pollution .0011 -.0227 .0008 1.293 -2.171 -.0337 .0714 -.0253(.0109) (.0117) (.0053) (1.537) (.7340) (.0165) (.0117) (.0054)[.0043] [-.1027] [.0047] [.0396] [-.0866] [-.0542] [.2649] [-.1761]

Observations 5,538 5,538 5,538 5,226 5,226 5,226 5,538 5,538Panel D: Amenities

Parks Entert. Church Hospital Public Justice Transport BotanicalPollution .0139 -.0144 .0155 -.0010 .0153 .0028 -.0001 -.0064

(.0117) (.0103) (.0080) (.0018) (.0108) (.0035) (.0077) (.0033)[.0317] [-.0357] [.0549] [-.0125] [.0389] [.0242] [-.0003] [-.0392]

Observations 5,538 5,538 5,538 5,538 5,538 5,538 5,538 5,538

36/ 36