The carbon footprint of suburbanization: Evidence from French - - PowerPoint PPT Presentation

French data Empirical strategy Preliminary results Main results CO2 predictions

The carbon footprint of suburbanization: Evidence from French household data

Camille Blaudin de Th´ e1 and Miren Lafourcade2

1Systra 2University Paris-Sud (RITM) / University Paris-Saclay & Paris School of Economics (PSE)

November 30, 2016 PSE-TSE-MEEM Seminar

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Motivation

Suburbanization is a long-standing trend (1950-...) ⇒ Accelerated growth in real per capita income since WWII ⇒ Decades of low energy prices, steep decline in transport costs ⇒ Preference for living in detached single-family homes... But urban sprawl has come at many costs (market failures) : ⇒ Consumption of undeveloped land & reduction of bio-diversity ⇒ Growth of GHG emissions due to commuting patterns... → Road transport : ∼ = 35% of anthropogenic CO2 emissions → Personal driving : ∼ = 60% of road emissions in France → Rising share (= other sources), despite improved vehicle fleet ⇒ But strong resistance to carbon taxes related to the use of fuel... ⇒ Spatial policies as complements (substitutes ?) to other CO2 mitigating macro-policies such as carbon taxes ?

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Motivation (cont’d)

Vast body of literature since Newman & Kenworthy (1989) ⇒ Negative relation between density and transport-related energy ⇒ Key contributions include : Bento et al. (2005), Browstone & Golob (2009), Glaeser & Kahn (2010), Zheng et al. (2011)... ⇒ But causal impact of urban form or sorting / endogeneity issues ? → People who like driving sort into areas of particular density → Density correlated with unobserved variables affecting driving What this paper adds to the literature ⇒ First attempt to correct for both sorting and endogeneity issues → Not only density : job-housing distance, transport access... ⇒ Europe is seldom investigated, whereas sprawl increases rapidly → French MAs have expanded by 20% in the last decade... → SEEID studies (Lemaˆ ıtre and Kleinpeter or Perrissin, 2009...) ⇒ Policy prospects : is there an optimal or a sub-optimal city ? → Ranking of cities wrt CO2 emissions of a “marginal” household → Relationship between MA-size and driving emissions : ∩-curve

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Household data

The fuel consumption associated with personal driving ⇒ French survey on “Family expenditures” (INSEE, 2006) : → 10,215 households (25,364 individuals), 7,800 urbanites ⇒ Exhaustive coverage of consumption expenditures : → 2 weeks purchases, 900 budgetary items → Gasoline, diesel and LPG expenditure... → Converted into volumes with related fuel prices in 2006 ⇒ Regular or extraordinary resources (gifts, lottery, inheritance,...) ⇒ Socioeconomic characteristics of all household members : → Nb of children, workers, job seekers, retired... → Age, gender, diploma, occupation (head of the household) ⇒ Dwelling characteristics : → Municipality code (statistical disclosure)

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Descriptive statistics

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Core-base geographical unit of observation

⇒ 352 French Metropolitan Areas (MAs)

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Urban form

The metrics of urban sprawl : the three D’s ⇒ Density : Population increases faster at the urban fringe → Growth of city centers : +0,3%/year (1999-2006) → Growth of inner (+0,6%) & outer (+1,3%) suburbs → Density gradients fall with distance to city-centers... → Census data : number of inhabitants/jobs per km2 ⇒ Design : The centrality of the urban space decreases as... → Workplaces and residence places develop further away → Commuting distances between residence and CBD increase... → “Great-circle” distance between the residence and the CBD ⇒ Diversity : Transport networks shape cities → Tentacular cities and “leap frog” sprawl → Possibly countered by transport networks → Public transit substitutes to personal driving → Public transport increases both local/global connectedness → GIS BD-TOPO (NGI) : all transports but buses

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Econometric specifications

Fueli(k) = α0 + α1Parisk + α2ISk + α3OSk + α4Mk + α5Rk + Xiθ + ǫi(k) Fueli(k) = α+β ln densP

k +δ ln distCBD k

+γ ln densT

k + x λx ln TPk(x)+Xiθ+εi(k)

⇒ Fueli(k) : (Log)Fuel volume of household i living in municipality k ⇒ densP

k : population density in the municipality of residence k

⇒ distCBD

k

: distance between k and the CBD of the MA ⇒ densT

k : density of train/metro stations in municipality k

⇒ TPk(x) =

• k′=k,∈MA

densk′(x) distkk′

: connection of k within the MA → Number of train/metro stations (as many as transit lines) → Length of the road network (weighted by traffic levels) ⇒ Xi : Socioeconomic characteristics of household i : → Family composition : nb of (non-)working adults and children → Age, gender, diploma, occupation of the household-head ⇒ If distance to CBD doubles, fuel cons. varies by ln 2 × δ = 0, 7 × δ

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions in log Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Preliminary results : to sum up

Fuel consumption and individual characteristics ⇒ Affluent households and (working) families consume more fuel... ⇒ The impact of a working-adult is 1.5-fold that of a non-working adult, and 15-fold that of an underage child ⇒ Households headed by elderly and women consume less fuel The impact of suburbanization on fuel consumption ⇒ Moving a household living in a non-Parisian city-center to : → The Parisian pole would save 60 gallons/year → Paris “intra-muros” would save 135 gallons/year → The inner suburb of Paris would save 40 gallons/year → The inner suburb of another city would add 40 gallons/year → An outer / multipolar suburb would add 90 gallons/year → A rural area would add 80 gallons/year ⇒ A household living at the fringe of a MA instead of a city-center would bear an extra-consumption of about 6 fuel tanks per year

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Sorting and endogeneity issues

Econometric issues ⇒ Omitted variables (bus...) / Reverse causality (density...) ⇒ Instrumental Variables econometric techniques : → Length of the royal road departmental network in 1837 pro-rated on the basis of the surface area of each municipality (Source : Statistique g´ en´ erale de la France, Tome 2, 1837) → Lagged market potentials (pop. dens.) (Sources : INED, 1936) → Lagged density of deaths by municip. (Source : Census, 1962) ⇒ Sorting of (non-)motorized households ⇒ OLS / IV restricted to the sample of motorized households ⇒ Heckman two-step estimation : → 15% of households do not have a car... → Selection equation : probability to own (at least) a car

P

• Cari(k) ≥ 1
• = α+β log densk+δ log distCBD

k

+γ log densT

k + x λx log TPk(x)+Yiθ+µi(k)

→ Outcome equation : Same as before - underage children + 1/Mill’s ratio

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Main results

The impact of urban form on fuel consumption ⇒ Washes out the effect of all residence-types... ⇒ Doubling distance to CBD increases the household fuel consumption by up to 20 gallons/year, against 45 gallons in the US ⇒ Doubling pop. density in the residence reduces the household fuel consumption by up to 18 gallons/year, against 82 gallons in the US ⇒ Doubling dens. of train stations in the residence (MA) reduces the household fuel consumption by up to 30 gallons/year ⇒ Doubling dens. of road length in the MA increases the household fuel consumption by up to 10 gallons/year ⇒ Impact of Density overestimated (15%) due to endogeneity ⇒ Impact of Design underestimated (90%) due to endogeneity ⇒ Even larger endogeneity bias for Diversity ⇒ Smaller sorting bias : ranges from 10% to 30%

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

The ranking of French MAs

Driving emissions of the sample mean-household ⇒ Representative household (mean household of the BdF survey) → Motorized, 1,03 work. ad., 0,83 non-work. ad., 0,62 children → Head : Man, 49 years-old, intermediate profession, CAP/BEP ⇒ Fuel consumption of this household for each municipality :

• Fuelm(z) =

α+ β ln densP

z +

δ ln distCBD

z

+ γ ln densT

z + x

λx ln TPz(x)+X θ+ εm(z) ⇒ Fuel consumption of this household for each MA : →

z∈MA(

Fuelm(z) × Number of motorized householdsz/PopMA) ⇒ CO2 driving emissions of this household for each MA : → 1 gallon US of gasoline ≃ 19.2 lbs of CO2 (2.3kg/l) → 1 gallon US of diesel ≃ 21.7 lbs of CO2 (2.6kg/l) → 1 gallon US of LPG ≃ 14.2 lbs of CO2 (1.7kg/l) → 1 gallon US of fuel ≃ 20.6 lbs of CO2 (average weighted share of each type of fuel in the total fuel consumption of the BdF survey)

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

CO2 “carprint” of the sample mean-household (in lbs per year)

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

The greenest MAs (CO2 “carprint” of the sample mean-household in lbs / year)

Name of the MA OLS R IV R OLS&IC R IV&CI R Pop 2006 Paris 3510 1 3524 1 3492 1 3507 1 11,769,424 Fourmies 3907 2 3810 2 3706 2 3608 3 16,324 Caudry 3965 3 3826 3 3745 3 3605 2 14,322 Montereau-Fault-Yonne 4175 6 4055 4 3949 5 3828 5 26,109 Bolbec 4262 7 4128 5 4066 9 3932 8 15,750 Lille 4146 4 4155 6 3990 6 3998 9 1,164,717 Menton 4155 5 4157 7 3820 4 3820 4 68,826 Hendaye 4319 9 4196 8 4018 7 3894 7 14,993 Yvetot 4400 11 4212 9 4237 14 4048 10 15,432 Villerupt 4423 13 4231 10 4062 8 3869 6 19,019 Tergnier 4446 15 4333 11 4237 13 4123 14 23,383 Grenoble 4328 10 4334 12 4203 12 4208 21 531,439 Boulogne-sur-Mer 4416 12 4341 13 4174 11 4098 11 133,195 Aulnoye-Aymeries 4503 19 4341 14 4284 17 4121 13 19,649 Saint-´ Etienne 4299 8 4351 15 4103 10 4155 16 318,993 Nemours 4513 20 4372 16 4311 20 4168 18 18,429 Sedan 4535 23 4372 17 4337 25 4173 19 31,395 Landerneau 4525 22 4382 18 4272 16 4129 15 14,927 Lun´ eville 4564 27 4396 19 4326 23 4158 17 27,549 F´ ecamp 4555 26 4419 20 4326 24 4189 20 30,233 Le Havre 4430 14 4429 21 4303 19 4303 26 290,826 Berck 4553 25 4435 22 4347 27 4229 22 24,648 Creil 4486 16 4441 23 4481 38 4436 40 101,982 Longwy 4604 31 4444 24 4267 15 4107 12 40,650

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

The dirtiest MAs (CO2 “carprint” of the sample mean-household in lbs / year)

Name of the MA OLS R IV R OLS&IC R IV&CI R Pop 2006 Chamonix-Mt-Blanc 6274 352 6257 352 5899 352 5881 352 13,127 Bourg-St-Maurice 6234 351 6154 351 5897 351 5817 351 10,357 La Bresse 6009 350 5989 350 5739 348 5719 349 12,851 Bressuire 6007 349 5965 349 5752 349 5710 347 18,225 Sallanches 5935 346 5954 348 5692 346 5711 348 43,413 Les Herbiers 6001 348 5910 347 5722 347 5630 344 14,833 Annemasse 5901 343 5910 346 5591 333 5599 343 244,178 La Roche-sur-Yon 5865 335 5878 345 5626 338 5638 345 107,584 Niort 5894 342 5873 344 5783 350 5762 350 134,927 Arles 5782 327 5864 343 5577 330 5658 346 54,712 Sarlat-la-Can´ eda 5947 347 5845 342 5677 344 5575 342 18,022 Saint-Louis 5883 337 5841 341 5614 335 5571 340 89,549 Auch 5892 339 5819 340 5632 339 5559 338 36,934 Cahors 5918 344 5816 339 5655 340 5553 337 40,175 Sabl´ e-sur-Sarthe 5875 336 5794 338 5614 336 5533 334 30,193 Lannion 5827 332 5793 337 5594 334 5560 339 63,425 Loud´ eac 5892 340 5792 336 5571 329 5471 327 14,217 Bergerac 5791 328 5779 335 5489 317 5475 328 76,179 Saint-Gaudens 5924 345 5773 334 5669 343 5517 332 27,175 Ancenis 5885 338 5767 333 5658 342 5540 335 19,308 Ussel 5892 341 5743 332 5690 345 5540 336 14,074 Sainte-Maxime 5779 326 5739 331 5457 311 5416 321 17,920 Fos-sur-Mer 5815 331 5734 330 5655 341 5574 341 15,734 Oloron-Ste-Marie 5854 334 5710 329 5620 337 5476 329 22,382

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

MA-size and the CO2 “carprint” of the sample mean-household : a bell-shaped curve ?

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

MA-size and the CO2 “carprint” of the sample mean-household : without Paris

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

Conclusion

⇒ A household living in a MA with the density of Lyon instead of that of Toulouse would save about 2 tanks per year, which is small... ⇒ But, a household living in a MA with the characteristics of Paris instead of those of Chamonix would halve its annual fuel consumption... ⇒ Larger savings if compaction coupled with job-housing centralization, improved rail routes, low pressure for road construction ⇒ Bell-shaped relation between driving emissions and pop. size : → Small MAs : people do not drive far to ge to work → Big MAs : density/public transit strong enough to reduce mileage → Medium MAs : worst of the world ? ⇒ Specific anti-sprawl policies ? → Public transportation networks (metro/train stations) → Collective denser (and energy-efficient ?) housing → Price signal : urban taxes (which base/amount ?) ? ⇒ Still, CO2 driving emissions are 5-fold lower in France than in the US...

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions

“On devrait construire les villes ` a la campagne car l’air y est plus pur” “We should build cities in the countryside, because the air is cleaner there” Quotation credited to Alphonse Allais

Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions back Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar

French data Empirical strategy Preliminary results Main results CO2 predictions Miren Lafourcade (University Paris-Sud / Saclay & PSE) PSE-TSE-MEEM Seminar