Energy and Emissions from US Population Shifts ... - ACS Publications

Subscriber access provided by University of South Dakota

Policy Analysis

Energy and Emissions from US Population Shifts and Implications for Regional GHG Mitigation Planning Rachel Marie Hoesly, H. Scott Matthews, and Chris Hendrickson Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.5b02820 • Publication Date (Web): 09 Sep 2015 Downloaded from http://pubs.acs.org on September 10, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 24

Environmental Science & Technology

1

Energy and Emissions from US Population Shifts and Implications for Regional

2

GHG Mitigation Planning

3

Rachel Hoesly*1, H. Scott Matthews1,2, Chris Hendrickson1,2

4 5

1Carnegie

6

TOC Art

2Carnegie

Mellon University, Department of Civil and Environmental Engineering Mellon University, Department of Engineering and Public Policy

7 8

Abstract

9

Living in different areas is associated with different impacts; the movement of people to and from

10

those areas will affect energy use and emissions over the US. The emissions implications of state-to-

11

state migration on household energy and GHG emissions are explored. 3 million households move

12

across state lines annually, and generally move from the North East to the South and West.

13

Migrating households often move to states with different climates (thus different heating and

14

cooling and needs), different fuel mixes, and different regional electricity grids, which leads them to

15

experience changes in household emissions as a result of their move. Under current migration

16

trends, the emissions increases of households moving from the Northeast to the South and

17

Southwest are balanced by the emissions decreases of households moving to California and the

18

Pacific Northwest. The net sum of emissions changes for migrating households is slightly positive

19

but near zero; however, that net zero sum represents the balance of many emission changes.

20

Planning for continued low carbon growth in low carbon regions or cities experiencing high growth

21

rates driven by migration is essential in order to offset the moderate emissions increases

22

experienced by households moving to high carbon regions.

23

Background

ACS Paragon Plus Environment


24

In recent decades, many cities and metro areas have compensated for the absence of

25

comprehensive national climate planning by creating Climate Action Plans (CAPs), which include

26

conducting GHG emissions inventories, establishing reduction targets, developing and

27

implementing reduction strategies, and monitoring results. To date, 32 states have completed

28

climate action plans and more than 1,000 cities have committed emission reduction initiatives by

29

becoming members of ICLEI Local Governments for Sustainability 1,2. These cities face many

30

challenges 3,4, but it seems especially difficult for cities to achieve overall emission reduction goals

31

alongside increasing populations and energy demand 1,2,5. The redistribution of population

32

throughout the US contributes to changes in energy and emissions within local boundaries, which

33

are partially documented in local GHG inventories as some portion of net emission increases or

34

decreases due to population gains and losses. Moreover, since regions differ by climate, fuel mix,

35

housing stock, and other characteristics, the household energy use and emissions of people that

36

relocate within the US may change and contribute to net changes in energy use and emissions at the

37

national level.

38

Studies have shown that household energy use and GHG emissions vary widely across the US for

39

many reasons 3,4,6-8. Residential energy use and household transportation both vary differently over

40

states. Average annual vehicle miles traveled (VMT) per capita varies by state from less than 9,000

41

VMT/person in Pacific Northwest states, to over 11,000 VMT/person in states characterized by

42

open spaces like Colorado, Mississippi, and Wyoming 9. Residential energy, which accounts for 21%

43

of total primary energy consumption and 17% of total GHG emissions in the US 10,11, is influenced by

44

many factors including lifestyle choices, household income, and the use of energy efficient

45

appliances and lighting 12. Residential energy use is dominated by heating and cooling, which

46

accounts for 48% of energy use in homes 13, and is highly correlated with heating degree days

47

(HDD) and cooling degree days(CDD), a measurement of the difference between mean outdoor

48

temperature and room temperature14. Average household energy use is highest in the Midwest and

49

Northeast (as defined by the US Census), 112 and 108 MBTU/year respectively, followed by the

50

South and West (76 and 73 MBTU/year respectively)15.

51

Variations in residential GHG emissions arise from differences in total energy demand combined

52

with regional variation in fuel mix and carbon intensity of the regional electricity grid. In Florida,

53

almost 30% of household energy is used for space cooling, resulting in large electricity

54

consumption (90% of all energy consumed by Florida homes)16. While in Colorado, electricity use is


Page 2 of 24

Page 3 of 24


55

low because it is not commonly used for space heating, cooling, or water heating, even though total

56

household energy use is higher than the US average 17. The electricity grid mix varies across the US

57

by state and region, and despite the difficulty associated with estimating local grid factors, state

58

average electricity emission factors in different regions vary from less than 0.3 kg CO2/kWh the

59

Pacific Northwest, where states employ large amounts of hydro power, to 0.8 kg CO2/kWh in states

60

that rely heavily on coal 18.

61

In 1960, just over 30% of the country lived in the warmest states, but by 2010 that share had grown

62

to 43%; while the percentage of population living in coldest states decreased from 60% to 48% 19.

63

This has led to a decrease in population weighted average HDD and an increase in CDD, which,

64

while driving the growth of air conditioning and electricity use in the South, has contributed to the

65

flattening of per capita energy use in the US overtime 13,19-21.

66

The American population continues to be mobile, and while the numbers of migrating Americans

67

have decreased since 2000, 35 – 40 million people change address in the US every year 22. In 2005,

68

of these moves, 57% of households stayed within their original county of residence, 20% moved to

69

a different county in the same state, 19% moved across state lines, and 4% moved to or from

70

outside the US 23. Figure 1a, using data from the International Revenue Service 24, shows the annual

71

average net migration from 2005 – 2010 into states in number of households. Annual state-to-state

72

migration patterns have been relatively consistent since 2004 (with the exception of the Gulf Coast

73

area). Households continue, on average, to move from the Northeast to the South and West.

74

While previous studies have shown that the long-term population shift has had effects on US

75

residential energy use19,20, migration should also influence emissions. This analysis explores how

76

inter-state migration contributes to net changes in GHG emissions within the US by estimating GHG

77

emissions of migrating households in their origin and destination states and calculating the annual

78

change in household emissions as a result of moving to a different state.

79

Data

80

The major data required to conduct this analysis were residential energy data, migration and

81

demographic data, electricity grid emission data, and household transportation data.



82

6 years of mobility data, over the period 2005 – 2010, from the US Internal Revenue Service (IRS)

83

area-to-area migration data sets are used 24. The IRS Statistics of Income (SOI) Division reports

84

migration data based on the change of address fields in annual tax returns. It provides yearly

85

estimates of the number of households and persons moving, origin and destination locations, and

86

aggregate and median income for each migratory flow. IRS migration data excludes those who do

87

not file taxes and may underrepresent the poor and the elderly as well as those granted extensions

88

to file by the IRS, like those with very high income 25. Although a comparison of IRS data with other

89

migration data sets from the American Community Survey (ACS) and the Current Population Survey

90

(CPS) shows that the population of tax filers may migrate more frequently than non filers, this does

91

not affect reporting of migration trends and IRS data remains the most comprehensive data set for

92

the US, reporting paired flows for both state and county 26. In order to protect the privacy of

93

taxpayers, the IRS suppresses data on flows with small numbers of households. While origin and

94

destination state are provided for all households migrating within the US, the origin and destination

95

county is only disclosed for roughly 60% of migrating households.

96

Additional household information is drawn from the 2009 1-year sample ACS 27. The ACS,

97

conducted by the US Census Bureau provides annual household (and individuals within those

98

households) demographic data, including recent migratory status, renter or owner status, as well as

99

limited self-reported utility bill data.

100

Residential energy use data is available from the US Energy Information Administration’s (EIA’s)

101

2009 Residential Energy Consumption Survey (RECS) 28. RECS contains data on household energy

102

use sampled from over 12,000 households in 16 states, representing 27 domains (groups of states)

103

and is considered the most comprehensive household energy survey for US homes. While RECS

104

summary tables include summary statistics of household energy use and expenditures of US

105

households by domains and characteristics, RECS microdata contains survey data from all

106

individual households sampled along with representative weights used to estimate representative

107

population distributions. RECS data used in this analysis includes annual residential energy use by

108

fuel type including electricity, natural gas, liquefied petroleum, kerosene and fuel oil.

109

The changes in household transportation emissions for migrating households are calculated using

110

state average estimates for VMT per household. State estimates were derived from total annual

111

VMT estimates per state, available from the Federal Highway Administration’s (FHWA) Highway


Page 4 of 24

Page 5 of 24


112

Statistics Series 29, and census estimates of the total number of households per state. VMT estimates

113

were calculated for 2005 and 2010, shown in Equation 1.

= ÷

114

Equation 1

115

Where:

116

!" #$%&'$() *

is the average annual VMT/household in state i;

117

+,,-./ 012*

is the total VMT in state i for all vehicles;

118

3-4567 89 :8-;6ℎ8/=;* is the total number of households in state i.

119

Total annual vehicle-miles, reported by the FHWA, is the compilation of data that are both collected

120

from individual states and calculated by the FHWA using state-provided data in the Highway

121

Performance Monitoring System (HPMS). Rather than actual odometer readings from vehicles,

122

these values are usually based on annual average daily traffic (AADT) and centerline length for

123

reported AADT sections. In some cases, vehicle miles are based on fuel use or supplemental traffic

124

counts 30.

125

Similar to household energy use data, microdata exists describing household transportation

126

behaviors over different states from the DOT’s National Household Transportation Survey 31. A

127

discussion of why such microdata is not used in this analysis is in the Methods section.

128

Methods

129

Historical migration data and energy use survey data were used to individually estimate the

130

residential energy use and GHG emissions of all migrating households over a given year in their

131

origin and destination states. In this analysis, the expected change in annual emissions for a

132

migrating household is the difference between annual emissions from residential energy use from

133

living in the previous (origin) and new (destination) state.



134

From the IRS migration data, 6 51x51 origin-destination (OD) matrices were created, containing the

135

number of households that moved from each state (and the District of Columbia) to every other

136

state over each year from 2005 - 2010. This analysis uses the average annual flow, or the average of

137

each cell over the 6 OD matrices. The 2,550 non-zero flows range from less than 20 households for

138

small flows (Idaho to Delaware), to over 30,000 households (New York to Florida).

139

The energy use of migrating households in their origin and destination state was estimated using a

140

process of resampling energy use data for single households from the 2009 RECS microdata. The

141

RECS microdata was split into 27 groups, by RECS domain, as defined by RECS where some

142

domains represent multiple states. Using representative household weights provided by RECS, 27

143

empirical distributions were established from which to sample household energy use data. Only 1

144

year of energy data, the 2009 RECS microdata, is used for all 6 years of mobility data. RECS samples

145

are also available for 2005; however, this survey samples less than half the number of households

146

from fewer states than the 2009 sample.

147

Two methods were considered for estimating the relationship between residential energy use of a

148

single household in its origin and destination states: first by percentile of residential energy use,

149

then by household income. Both methods yielded similar results, shown in the Supplementary

150

Information (SI). Results shown in the body of this report reflect estimation by percentile of

151

residential energy use.

152

Changes in household energy use were estimated using Monte Carlo analysis to draw pairs of

153

correlated random uniform numbers, which were used to match household energy use in origin

154

and destination states. For a single household, a pair of correlated random uniform numbers

155

between 0 and 1 were created to represent a household’s percentile of residential energy use in

156

both origin and destination state. Residential energy use percentile was measured by estimating

157

energy use as total annual residential fuel use (all residential fuels). We assumed these values are

158

highly correlated because in most cases, it is likely that a migrating household’s energy habits, such

159

as willingness to use energy to provide comfort (like space heating or air conditioning) relatively

160

compared to other households in the same state would be fixed. In other words, it is unlikely a

161

household would jump from 95th percentile energy use in its origin state to 5th percentile in its

162

destination state. Percentile pairs were generated so that they would have a correlation of about

163

0.8; sensitivity of this assumption is discussed further in the limitations section and SI. Correlated


Page 6 of 24

Page 7 of 24


164

percentiles were created by randomly generating correlated random normal variables then

165

transforming them to uniform variables using cumulative Gaussian distribution functions.

166

Using the generated correlated percentile pairs, 2 observations were drawn from the RECS

167

microdata to approximate energy use behavior of the household in both the origin and destination

168

state. Each household sampling draws complete data for a household entry in RECS including

169

annual consumption of electricity, natural gas, kerosene, fuel oil, and liquefied petroleum gases.

170

When using income to estimate the energy use, a household was randomly sampled from origin

171

state distribution then matched with a household in the destination state data with similar income.

172

Complete data for each household was then sampled and retained for analysis.

173

Next, GHG emissions, measured in CO2eq, for the household in the origin and destination state were

174

calculated using life cycle GHG emission factors for natural gas 32,33, fuel oil, liquefied petroleum gas

175

(propane) and kerosene, detailed in the SI. Emissions from electricity were calculated using the

176

state emission factors and regional emission factors reported by EIA 34, which are comparable to

177

eGRID emissions factors reported by the US Environmental Protection Agency (EPA)35. Emission

178

factors are measured in CO2eq and detailed in the SI. A discussion of sensitivity from electricity

179

emission factors is discussed in the Limitations section.

180

The estimated emissions, >4*? , of migrating household, n, in origin state, i, is ? ? >4*? = >*, × >B* + D >*,E × BE E

181

Equation 2

182

Where:

183

? >*, is household n’s annual electricity use in origin state, i, measured kWh;

184 185 186 187

>B* is the electricity grid emission factor in origin state, i, in mt/kwh;

? >*,E is household n’s annual energy use in origin state, i, of fuel k, which includes natural gas, fuel oil,

liquid petroleum, and kerosene; and BE is the carbon emission factor for fuel, k.

188



189

Page 8 of 24

Similarly, the estimated emissions of migration household, n, in destination state, j, is ? ? >4F? = >F, × >BF + D >*,E × BE E

190 191

Equation 3 ? The estimated change in emissions, for household n, >4G(H ? >4G(H = >4F? − >4*?

192

Equation 4

193

This process is repeated roughly 3 million times for each of the households that move between

194

states in a given year using origin and destination states according to the Origin-Destination matrix

195

mentioned above. The energy use and emissions for each household in its origin and destination

196

states as well as the expected emissions change the household experiences from the move is

197

retained for later analysis.

198

Household transportation emissions for migrating households were estimated using average point

199

estimates to approximate distributions of actual household transportation emissions changes

200

experienced by households. Many households will experience emissions changes different than the

201

average estimates; aside from regular variation in samples, these differences are likely driven by

202

factors that vary within states, such as urbanity. We believe that, unlike residential energy use,

203

there is no clear correlation between the driving habits of a household in its origin and destination

204

state. Annual VMT is dominated by trips to work and how work trips change from origin to

205

destination state will vary widely over different types of moves; estimating a migrating household’s

206

proximity to work place, as well as attempting to estimate a relationship between household

207

driving behavior in origin and destination states beyond state averages, are outside the scope of

208

this analysis. Therefore, the use of microdata distributions is unnecessary and point estimates will

209

sufficiently describe the difference in driving behavior and resulting emissions changes between

210

states in the US. Weinberger and Goetzke showed that recent migrators retain some transportation

211

behaviors from their origin location36. This analysis may overestimate changes in transportation

212

emissions and is discussed further in the limitations section and SI.


Page 9 of 24


213

Average fuel use per household for each state was calculated using total Annual VMT estimates per

214

state, which are available for the years 2005 and 2010 from the US DOT 29,37, census estimates for

215

the number of households per state, and the average fuel economy of the US passenger vehicle fleet

216

measured in mpg (22.1 and 23.3 mpg for 2005 and 2010 respectively) 38 according to Equation 5.

217

As fuel economy varies with levels of urbanity 39, the average fuel economy of vehicle fleets also

218

likely vary by state; this data however, is not readily available. Average vehicle fuel economy can be

219

derived from NHTS microdata; however, the sample size for many states is quite small, less than

220

300 vehicles, compared to a select few states which have more than 15,000 samples like California,

221

Florida, and New York 31. The statistical difference in fuel economy between many states will likely

222

be indistinguishable from zero.

J* =

28K./ +,,-./ 012* ÷ LM 9-6/ 6N8,84O 28K./ :8-;6ℎ8/=;*

223

Equation 5

224

Where:

225

J*

226

A 51x51 fuel-change matrix was created which describes the change in annual fuel use of an

227

average household moving from one state another described in Equation 6:

is the average fuel use per household in state i.

∆J*F = JF − J* 228

Equation 6

229

Where:

230

∆J*F is the average change in fuel use for a household moving from state i to state j. ∆J*F is located in

231

row i and column j of the fuel change matrix.

232

The fuel-change matrix was multiplied by the OD migration matrix in order to obtain estimates of

233

total change in fuel use for all households in each migration flow. GHG emissions are estimated



234

from fuel estimates using a gasoline emission factors, 2.3 kg CO2/L 40 plus 20% to account for up

235

stream lifecycle emissions 6.

236

Results

237

Residential Energy and Emissions

238

Results shown in this section reflect emissions changes from migrating households for the average

239

annual migratory flow from 2005-2010. Annual total emissions changes over this period do not

240

vary widely and there is no increasing or decreasing trend. The variability from using different

241

years of migration flow data is smaller than uncertainty from other modeling choices, discussed

242

further in the limitations section.

243

The aggregate estimated residential emissions changes from households migrating state to state

244

was very small but slightly positive, ranging from 30,000 – 200,000 mtCO2eq. Compared to annual

245

energy related CO2 emissions estimates from the residential sector, on the order of 1 billion

246

mtCO241,42, this represents less than 0.1%, and an even lower percentage on the basis of total US

247

GHG emissions. However, this near-zero value is the summation of many households experiencing

248

both emissions increases and decreases. 1.54 million of migrating households, just over half,

249

experience emissions increases equal to 6 million mtCO2eq. The remaining 1.51 million migrating

250

households experience decreases equal to -5.8 million mtCO2eq.

251

115 million households in the US 41 emit 1.1 billion mtCO2 emissions annually from residential

252

energy use, and the average household in the US emits approximately 9.5 mtCO2 per year.

253

Households that experienced emissions increases, on average, increased their emissions by +3.9

254

mtCO2eq per year, or 40% of the average US household. While many households experience an

255

emissions change close to zero, the range of emissions changes expected is wide. The 5th and 95th

256

percentile of all household emission changes were –8.3 mt and +8.3 mtCO2eq/household/year

257

respectively. In extreme cases, a migrating household may experience an emissions change that is

258

almost equivalent to the emissions change of adding or taking away an additional average US

259

household.

260

Aggregate US emissions changes from households moving in the US annually net close to zero;

261

however, aggregate emission changes over regions and states are not trivial. Figure 1b shows the

262

sum of expected residential emission changes from all households moving to each state. While the


Page 10 of 24

Page 11 of 24


263

figure reflects emission changes of large migratory flows into each state, it does not reflect

264

migration out of states nor the changes in emissions within a state over a given year. Negative sums,

265

shown in blue, indicate that on average households move to that state from areas where they were

266

likely to emit more GHGs. Some households moving to these blue colored states, will likely

267

experience emissions increases, but the sum of emissions changes from all household moving to

268

those states are negative, and represents a net emissions decrease. Positive state emissions sums,

269

shown in red, indicate that on average households move to that state from areas where they are

270

likely to emit less GHGs. Migratory flows to California, the Pacific Northwest, and the North East are

271

responsible for net emissions decreases while flows to Texas, Georgia, and most of the Mid West

272

and Mountain regions are responsible for net emissions increases.

273



274 275 276 277 278

Figure 1a) Net migration into states. Average of annual migration flows from 2004 – 2010, estimated by the IRS Area to Area Migration Data b) Sum of expected residential emission changes for all households migrating to each state from residential fuels (electricity, natural gas, fuel oil, kerosene, and propane) and c) Sum of expected transportation emission changes for all households migrating to each state d) Sum of total household emissions 43 (residential and transportation) for households migrating to each state.

279

State emission sums for incoming households vary from -1.3M mtCO2e to 217,000 mtCO2e. This

280

illustrates that the net zero emission sum for the US is the addition of many household emission

281

increases and many household emissions decreases. California is almost solely responsible for net

282

zero emissions balance. The 1.3M mt net emission decreases for the almost 80,000 households

283

moving to California every year is larger than the sum of emissions changes of all other states with

284

net negative emission sums. California’s emission sum is almost 4 times larger than the absolute

285

value of the next largest sum, Texas, approximately 217,000 mt. The moderate emissions sums of

286

households moving to red colored states are balanced by many households moving to California

287

and Washington; thereby decreasing their emissions significantly. A table showing the emissions

288

change sums for all states is included in the SI. Some migration flows have both large size and

289

emission sums (e.g. California to Texas), other large flows do not have large emissions

290

consequences (e.g. New York to Florida), and other small flows have larger consequences (e.g.

291

California to Colorado). The size and emissions sums for select flows are shown in the SI.


Page 12 of 24

Page 13 of 24


292



293 294 295 296

Figure 2 The distributions of expected change in residential emissions for all households moving to selected states. Each curve is the distribution for households moving from 49 other states and DC (origin state) to a) Arkansas b) New Mexico and c) California (destination state), colored by origin state Census region. Additional states shown in SI.

297

Regional energy stories are further illustrated by examining the distributions of expected changes

298

in household emissions for each migratory flow. Figure 2a shows the distributions of the changes in

299

household emissions for all households moving to Arkansas by origin state. The 50 curves in the

300

figure represent the 31,000 households moving to Arkansas from the other 49 states and DC,

301

colored by US Census Region. It shows the regional nature of migration flows and the similarity of

302

state household emission profiles for geographically close states. Many curves peak at or close to

303

zero. However, curves representing origin states in the Pacific region, shown in orange, are shifted

304

right, indicating that households emit more emissions when living in Arkansas than in states in the

305

Pacific region. The 5th and 95th percentile of emission changes for all households moving to

306

Arkansas is -8 and 9.5 tCO2eq/household/year respectively, but the overall range is -20 to +20

307

mtCO2eq/household/year.

308

When examining similar graphs for other destination states, three general stories emerge, as shown

309

in Figure 2. Most states, like Arkansas (2a) are symmetrical and centered at zero. Most curves peak

310

at and are centered around zero. In these states, an equal number of households experience

311

emissions increases as emissions decreases, which result in net emission sums close to zero. States

312

with carbon intensive electricity grids like New Mexico (2b), have distributions that are shifted

313

right of zero, indicating that most household moves result in an emissions increase. Some states are

314

shifted left like California (2c), where most households emit fewer emissions in their destination

315

state than their origin state. Similar figures for other states are shown in the SI. Also included in the

316

SI are regional radial flow plot, such as that shown in the TOC art, which further illustrate the

317

emission flows between US census regions.

318

Differences in state residential emissions profiles are dominated by differences in state electricity

319

grids but dampened (equalized) by the addition of natural gas and other fuels. Shown in the SI,

320

state emission sum maps that only reflect electricity emissions are very similar to those reflecting

321

the carbon intensity of electricity grids. States with the largest magnitude emission sums tend to be

322

states with either extremely low or high intensity carbon grids, like the Washington, California and

323

Kentucky, or states with the highest population and largest migration flows like California and


Page 14 of 24

Page 15 of 24


324

Texas. This highlights the importance of managing the carbon intensive electricity grids in areas

325

that continue to experience high net migration, like Texas and Florida.

326

Household Transportation Energy and Emissions

327

VMT per household in 2005 and 2010 varies widely over states, shown in the SI. Annual average

328

transportation emissions per household vary from 8.4 mtCO2eq to in New York to 21 mtCO2eq in

329

Wyoming, compared to the US average household of 12.7 mtCO2eq/year. VMT per household

330

decreases in almost all states from 2005 to 2010, which is consistent with decreasing VMT in the US

331

over the past 10 years9,18. Figure 1c shows the sum of household transportation emissions for

332

households moving to each state. Emission sums for household moving to individual states range

333

from -454,000 mtCO2eq (New York) to +345,000 mt (Georgia) and sum to +453,000 mt over the US,

334

which is similar to the total sum of residential emission changes. Like residential energy,

335

households experience a range of emissions changes. Average households moving from

336

Massachusetts to Wyoming experience an emissions increase of 11 mtCO2eq/year/household,

337

while average households moving from New York to California experience an emissions increase of

338

2.7 mtCO2eq/year. The absolute value of all emissions changes experienced by households

339

migrating in a given year is almost 6.3 million mtCO2eq; however, emission increases and decreases

340

balance to a relatively small value.

341

Total Household Energy and Emissions

342

Figure 1d shows total emissions changes for migrating households by destination state. Emission

343

sums for some states are dominated by either residential energy or transportation emissions, while

344

sums of other states are balanced by both. A complete table of emissions sums for all states is

345

shown in the SI. Households moving to California experience modest transportation emission

346

increases; however, they are far outweighed by the emissions savings from mild climates and a low

347

carbon electricity grid. The absolute value of California’s sum remains almost 3 times larger than

348

the next largest emissions sum, Georgia, who’s migrating households experience a sum of +499,000

349

mtCO2eq. Residential emissions decreases experienced by all households moving to Florida are

350

almost completely canceled by increases in transportation emissions. In Northeastern and Pacific

351

Northwestern states, households experience decreases in both household residential and

352

transportation emissions, exacerbating the emissions savings they experience as a result of a move.

353

Households moving to Texas, New Mexico, Wyoming, and most southern states, experience

354

increases in both residential energy and transportation emissions.



355

Limitations and Uncertainty

356

Sensitivity and uncertainty in this analysis also comes from electricity grid emission factors. Studies

357

have show that significant uncertainty exists within different methods of estimating grid emission

358

factors at regional and sub-regional levels 18,44. Sub-region electricity trading occurs between states,

359

which adds the uncertainty in the differences between state level electric grid estimates within the

360

same region44,45. Using state grid estimates may inflate the importance of emission implications of

361

cross state migration flows between states that often trade electricity, or that are member of the

362

same interconnect. The SI shows summary statistics and figures for household emissions changes

363

by destination state when using state, regional, and the average of state and regional emission

364

factors. Most states show similar results; however, a few states that have both large regional

365

migration flows and large differences between state emission factor estimates and regional

366

emission factor estimates show different results. Distributions of household emission changes for

367

these destination states, including Arizona and Idaho, tend to be shifted right but remain similar in

368

shape. While using state emission factors does exacerbate the magnitude of emissions changes for a

369

few states, it does not contribute to large uncertainty in aggregate US emission totals. The range of

370

expected aggregate US emissions changes increases from 36,000 mt +/- 13,600 mtCO2eq to

371

136,000 tons +/- 13,600 mtCO2eq when switching from state emission factors to regional emission

372

factors. Both numbers remain less than 1% of total annual US residential emissions. NERC regions

373

were not used in this analysis. The use of NERC regions would likely dampen the importance of

374

migration between western states, as most of the Pacific and mountain regions are part of the

375

Western Interconnection.

376

Additionally, this analysis used average emission factors (AEFs) rather than marginal emission

377

factors (MEFs), which are more appropriate for short term (on the scale of hours) analysis; further

378

discussion of this choice is shown in the SI. AEFs that follow attributional LCA (ALCA) paradigms,

379

such as those used in this analysis can be misleading within policy contexts, and researchers

380

suggest the use of consequential LCA (CLCA) paradigms (Plevin et al). While differences in statically

381

calculated AEF’s can reliably measure household emissions differences for small numbers of

382

marginal households moving between states, the marginal annual emissions emitted by the first

383

additional households moving to a new state is likely not equal to the marginal annual emissions

384

emitted by the 50,000th household. Marginal emissions of that household would reflect additional

385

energy technologies deployed to satisfy their marginal energy demand. CLCA would account for

386

this, but it use is outside the scope of this analysis. As marginal emissions tend to increase with


Page 16 of 24

Page 17 of 24


387

marginal energy use, net emissions savings of high migration flows from high to low carbon states

388

may be smaller and net emissions increases from low to high carbon states may be larger if CLCA

389

were employed. While, the specific quantitative results of this study remain uncertain in light of this

390

limitation, the qualitative conclusions of this study remain and can highlight the importance of

391

certain policies. Varying emissions increases and decreases for migrating households will likely

392

balance over the entire US, and planning for low carbon, sustainable growth remains especially

393

important in low carbon areas with high migration rates from other states.

394

Energy data for this analysis is drawn from RECS data, which is representative of the US population;

395

however, analysis of ACS data shows characteristic differences between populations of recently

396

migrated households versus populations of longer-term resident households in their destination

397

states. On average, households that identify in the survey as having moved to that state from

398

another state in the past 3 years have fewer people, smaller family incomes, live in homes with

399

fewer rooms, and have smaller self reported utility bills than longer-term resident households. The

400

magnitudes of these trends vary widely by state. There is no data to either track changing

401

demographic characteristics over the course of a move, or verify if households likely to move in the

402

future are also smaller, younger, etc. compared to households in their origin state.

403

Three factors, unaccounted for in this model, effect why average household characteristic

404

differences in destination states exist: the general characteristics of migrating households (average

405

US moving rate by age increases through early twenties, then declines)46; types of moves (for

406

example retirement moves which involve downsizing); and the energy use behavior of migrating

407

households (households may retain behavior from origin states)36. These categories affect both

408

how this model should identify migrating households from US representative samples and how to

409

relate origin and destination state energy use. They also bias our model in inconsistent and

410

competing ways. While general trends can be supported by studies or data, adjusting the model for

411

these issues at the state level would be largely unsupported by state specific data. This reasoning is

412

explained further in the SI, but we believe that any accuracy gained from adding average or

413

uninformed modeling complexities required to account for these issues would likely be clouded by

414

the additional bias/uncertainty. These combined effects may dampen the changes in household

415

energy use changes from migration, but the main conclusions from this paper will remain the same.

416

In this analysis, we assume that a household’s energy use in its origin state relative to other

417

households in its origin state will be highly correlated with its energy use in its destination state



418

relative to other households in its destination state. As correlation increases, distribution of

419

household emission changes by migration flow become taller and narrower (shown in the SI)

420

because more households are expected to experience emissions changes closer to zero; however,

421

both state and US aggregate emission sums remain similar regardless of correlation values between

422

0 and 1. Uncertainty is also associated with IRS migration data, however we believe that this is

423

much smaller than those driven by the energy use of migrating versus non-migrating households.

424

While estimates presented in this analysis are uncertain for many reasons, we are confident that

425

aggregate emission changes from migration in the US are close to zero but driven by many

426

household increases and decreases grouped regionally. Even with the uncertainty present in this

427

analysis it remains important to manage the electricity grid in areas that expect high immigration

428

rates.

429

Discussion

430

If current migration trends continue, we are unlikely to see significant annual emissions increases

431

from household mobility in the US alone. However, these results may shift as the carbon intensity of

432

electricity grids evolve in the future. Many state are planning the decarbonization of their electricity

433

production by creating Renewable Portfolio Standards (RPSs). These policy designs vary widely

434

across states, but represent some measure of how regional electricity grids may change in the

435

future. Texas, Georgia and Missouri contributed to the largest emissions increases from migration

436

from 2005 – 2010. While Georgia has no RPS, in 2008, Missouri set an RPS goal of 15% by 2021.

437

California, which contributed the most to emissions decreases from migration, has one of the most

438

aggressive goals, 33% by 2020. Washington and Oregon, which were also responsible for large

439

emissions decreases, have goals of 15% by 2020 (a moderately aggressive goal) and 25% by 2525

440

respectively47. Most states in the South Atlantic, however, do not have RPS goals and this region

441

may see the smallest changes in electricity grid emissions while other states work to reduce grid

442

carbon emissions. This may cause emission increases from households moving to this region to be

443

exacerbated in the future.

444

While state-to-state migration doesn’t significantly contribute to annual changes in US emissions,

445

they may become more important when evaluated over longer periods. In areas where migrating

446

households often experience emissions increases, regional and local policies to encourage

447

population growth without targeted residential GHG emission mitigation policies could encourage


Page 18 of 24

Page 19 of 24


448

locking in emission increases into infrastructure. Living in different regions has impacts that extend

449

beyond GHG emissions. Within the scope of this analysis and other GHG footprint analyses, emitting

450

1 mtCO2 in Florida has the same effect as emitting 1 mtCO2 in California. The same is not true for

451

other relevant metrics like water use or particulate matter (PM) emissions. This analysis showed

452

the large emissions benefit of households moving to California and experiencing emissions

453

decreases that balanced the moderate emissions increases of many other destination states.

454

However, an analysis including water use would likely produce very different results, reflecting the

455

population shift away from unstressed water regions like the North East to the regions with the

456

highest stress water regions like California and the South West.

457

An analogous study by the authors explored migration between counties within Pennsylvania

458

highlighting the shift between rural, urban, and suburban counties48. Conclusions on the emissions

459

effects of urban-rural migration are similar: many households experience emissions changes but

460

net emissions sum to close to zero, and the emissions increases of households moving to higher

461

emissions areas (rural and suburban) are balanced by households moving to urban centers. This

462

analysis considered annual migration within the US; however, both migration and natural growth

463

rates drive population growth and associated emissions. The total US population growth rate in

464

2013 slowed to .71%, or less than 2.3 million people, which is the slowest growth rate since the

465

1930’s49. From 2005 – 2012, the number legal immigrants to the US varied between 1 and 1.2

466

million people50, which represents half the annual growth of the US population. However, the

467

growths of individual states and its drivers vary widely. Utah’s growth rate is almost twice that of

468

the national rate, driven by a high birth rate, while North Dakota’s growth rate, driven by the

469

growing oil industry, is 4 times that of the national growth rate. States that historically show large

470

growth rates, like Florida, Arizona, and Nevada, have slowed to moderate growth rates about 1.2%.

471

However, Florida’s growth is driven almost solely by migration, as it’s births and deaths are almost

472

equal49. Overtime, US energy and emissions consequences of immigration and natural growth will

473

far outweigh that of internal migration, but migration measures change, rather than flat increases.

474

Similar analyses for other counties, or international migration as a whole may yield varying results.

475

From 2005 – 2010, Europe experienced that largest net migration, 8.21 M people, followed by

476

North America, 6.06M people, while South Asia was the largest sender, loosing 8.7 M people

477

dominated by migration out of India51. The US, Canada, and Russia, all large net receivers of

478

migrants, each have per capita emissions more than twice the global average, and the European



479

Union, also a large receiver, has per capita emissions just over the global average; India’s per capita

480

emissions, however, are less than half the global average52. High migration rates into

481

counties/regions with high per capita emission, and high migration rates out of low per capita

482

emitters suggest that international migration flows will contribute to net increases of emissions.

483

In a paradigm dominated by regional emissions planning, it is important to evaluate how

484

interactions of many regional emissions goals and plans fit into national emissions mitigation

485

progress. Some cities have reached reductions below business as usual projections or baseline

486

inventory estimates even with growing population. However, some areas receive a high volume of

487

migrants that produce more GHGs as a result of their move, so regional emissions reductions do not

488

necessarily translate to national emissions reductions. Conversely, emissions increases within some

489

of these communities may not necessarily be driving total emissions increases in the US. Growth is

490

usually associated with emissions increases, but when growth is driven by migration from higher

491

carbon regions, low carbon growth can result in net emission decreases, even with population

492

growth. While achieving regional reduction goals anywhere contributes to lower carbon future,

493

emphasizing regional mitigation efforts for higher carbon areas with quickly growing populations

494

driven by migrators from low carbon areas becomes more important for realizing national level

495

emission reductions. Maintaining net emissions form migration both between and within states is a

496

balancing act in the short term. Achieving ultimate emissions reductions in communities that

497

expect increasing populations will prove to be quite challenging. Planning for low carbon growth is

498

particularly important 2 types of regions or cities: areas that play an important role in balancing net

499

emissions such as California and the pacific north west, and higher carbon areas that are have or

500

will see high migration rates from low carbon places such as Arizona, Georgia and Texas. High

501

growth rates driven by migration, without low carbon planning for growth could leave to emissions

502

increases rather than the net balances seen in the recent past.

503 504 505 506 507 508 509 510 511 512 513

Acknowledgements The authors acknowledge the intellectual and financial support of a Steinbrenner Institute’s U.S. Environmental Sustainability PhD fellowship sponsored by the Colcom Foundation. Supporting Information Additional figures, tables, and discussion are available in the Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org. Corresponding Author Information Rachel Hoesly Address: 5825 University Research Court Suite 3500, College Park, MD 20704


Page 20 of 24

Page 21 of 24

514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563


Phone: 301-314-6721 Email: [email protected] References (1) (2)

(3)

(4) (5)

(6) (7) (8)

(9) (10) (11) (12)

(13)

(14) (15)

(16) (17) (18)

ICLEI USA. ICLEI USA Members http://www.icleiusa.org/about-iclei/members (accessed Dec 4, 2013). US Environmental Protection Agency. State and Local Climate and Energy Program http://www.epa.gov/statelocalclimate/state/state-examples/action-plans.html (accessed Dec 4, 2013). Blackhurst, M.; Scott Matthews, H.; Sharrard, A. L.; Hendrickson, C. T.; Azevedo, I. L. Preparing US community greenhouse gas inventories for climate action plans. Environ. Res. Lett. 2011, 6, 034003. Dhakal, S.; Shrestha, R. M. Bridging the research gaps for carbon emissions and their management in cities. Energy Policy 2010, 38, 4753–4755. Hoesly, R.; Blackhurst, M.; Matthews, H. S.; Miller, J. F.; Maples, A.; Pettit, M.; Izard, C.; Fischbeck, P. Historical Carbon Footprinting and Implications for Sustainability Planning: A Case Study of the Pittsburgh Region. Environ. Sci. Technol. 2012, 46, 4283– 4290. Glaeser, E. L.; Kahn, M. E. The greeness of cities: Carbon diaxide emissions and urban development. Journal of Urban Economics 2010, 67, 404–418. Hillman, T.; Ramaswami, A. Greenhouse Gas Emission Footprints and Energy Use Benchmarks for Eight U.S. Cities. Environ. Sci. Technol. 2010, 44, 1902–1910. Min, J.; Hausfather, Z.; Lin, Q. F. A High-Resolution Statistical Model of Residential Energy End Use Characteristics for the United States. Journal of Industrial Ecology 2010, 14, 791–807. Puentes, R.; Tomer, A. The Road… Less traveled: An analysis of vehicle miles traveled trends in the US. 2008. US Energy Information Administration. Annual Energy Outlook 2008; DOE/EIA0383(2008); US Department of Energy: Washington DC, 2008. US Energy Information Administration. Annual Energy Outlook 2011; DOE/EIA0383(2011); US Department of Energy: Washington DC, 2011. Lima Azevedo, I.; Morgan, M. G.; Palmer, K.; Lave, L. B. Reducing U.S. Residential Energy Use and CO2 Emissions: How Much, How Soon, and at What Cost? Environ. Sci. Technol. 2013, 47, 2502–2511. US Energy Information Administration. Heating and cooling no longer majority of U.S. home energy use http://www.eia.gov/todayinenergy/detail.cfm?id=10271&src=%E2%80%B9%20Con sumption%20%20%20%20%20%20Residential%20Energy%20Consumption%20Sur vey%20(RECS)-b1 (accessed Dec 4, 2013). Quayle, R. G.; Diaz, H. F. Heating degree day data applied to residential heating energy consumption. Journal of Applied Meteorology 1980, 19, 241–246. US Energy Information Administration. CE1.1 Summary totals and intensities, U.S. homes; 2009 Residential Energy Consumption Survey; US Department of Energy: Washington DC, 2012. US Energy Information Administration. Household Energy Use in Florida; US Department of Energy: Washington DC, 2013. US Energy Information Administration. Household Energy Use in Colorado; US Department of Energy: Washington DC, 2013. Weber, C. L.; Jaramillo, P.; Marriott, J.; Samaras, C. Life Cycle Assessment and Grid



564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613

Page 22 of 24

Electricity: What Do We Know and What Can We Know? Environ. Sci. Technol. 2010, 44, 1895–1901. (19) US Energy Information Administration. Population shifts across U.S. regions affect overall heating and cooling needs http://www.eia.gov/todayinenergy/detail.cfm?id=8810 (accessed Dec 5, 2013). (20) Michael Sivak. Energy-demand consequences of the recent geographical shift in the metropolitan population of the US. Cities 2009, 26, 359–362. (21) Samson, J.; Berteaux, D.; McGill, B. J.; Humphries, M. M. Demographic Amplification of Climate Change Experienced by the Contiguous United States Population during the 20th Century. PLoS ONE 2012, 7, e45683. (22) Ihrke, D. K.; Faber, C. S.; Koerber, W. K. Geographical Mobility: 2008 to 2009; 2011; pp. 1–20. (23) US Census Bureau. GEOGRAPHICAL MOBILITY BETWEEN 2004 AND 2005; US Census Bureau, 2005; pp. 1–3. (24) Internal Revenue Service Statistics of Income Division. Internal Revenue Service AreaTo-Area Migration Data. (25) Gross, E. US Population MIgration Data: Strengths and Limitations; 2012; pp. 1–4. (26) Molloy, R.; Smith, C. L.; Wozniak, A. Internal Migration in the United States. Journal of Economic Perspectives 2011, 25, 173–196. (27) US Census Bureau. 2009 American Community Survey1-year Public Use Microdata Sample. (28) US Energy Information Administration. 2009 RECS Public Use Microdata, 2013. (29) Federal Highway Administration. Table VM-2: Annual Vehicle-Miles by Functional System; US Department of Tranportation, 2011. (30) Federal Highway Administration. Highway Statistics Series User's Guide http://www.fhwa.dot.gov/policyinformation/statistics/2010/userguide.cfm (accessed 2014). (31) Federal Highway Administration. The National Household Transportation Survey; US Department of Tranportation, 2009. (32) Venkatesh, A.; Jaramillo, P.; Griffin, W. M.; Matthews, H. S. Uncertainty in Life Cycle Greenhouse Gas Emissions from United States Natural Gas End-Uses and its Effects on Policy. Environ. Sci. Technol. 2011, 45, 8182–8189. (33) Advanced Resource International Inc; IFC International. GREENHOUSE GASLIFECYCLE EMISSIONS STUDY: Fuel Life-Cycle of U.S. Natural Gas Supplies and International LNG; 2008; pp. 1–68. (34) US Energy Information Administration. Updated State-level Greenhouse GasEmission Coefficients for Electricity Generation 1998-2000; US Energy Information Administration, 2002; pp. 1–9. (35) US EPA. eGRID 9th edition Version 1.0 Year 2010 Summary Tables; 2014; pp. 1–13. (36) Weinberger, R.; Goetzke, F. Unpacking Preference: How Previous Experience Affects Auto Ownership in the United States. Urban Studies 2010, 47, 2111–2128. (37) Federal Highway Administration. Table VM-2: Annual Vehicle Miles by Functional System; US Department of Tranportation, 2005. (38) US DOT Bureau of Transportation Statistics. Table 4-23: Average Fuel Efficiency of U.S. Light Duty Vehicle; US Department of Tranportation. (39) Transportation Research Board. Driving and the Built Environment: The Effects of Compact Development on Motorized Travel, Energy Use, and CO2 Emissions ; 298; Transportation Research Board: Washington DC, 2009. (40) US Energy Information Administration. Voluntary Reporting of Greenhouse Gases Program Fuel Carbon Dioxide Emission Coefficients; US Department of Energy.


Page 23 of 24

614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639


(41) (42)

(43) (44) (45) (46) (47) (48) (49)

(50) (51) (52)

US Census Bureau. USA QuickFacts http://quickfacts.census.gov/qfd/states/00000.html (accessed Mar 26, 2014). US Energy Information Administration. Carbon Dioxide emissions grow in the residential sector http://www.eia.gov/todayinenergy/detail.cfm?id=1970 (accessed Mar 26, 2014). US Census Bureau. 2014 TIGER/Line Shapefiles (machine readable data file), 2014. Marriott, J.; Matthews, H. S. Environmental Effects of Interstate Power Trading on Electricity Consumption Mixes. Environ. Sci. Technol. 2005, 39, 8584–8590. Hawkes, A. D. Energy Policy. Energy Policy 2010, 38, 5977–5987. Ihrke, D. Reason for Moving: 2012 to 2013; P20-574; US Census Bureau, 2014; pp. 1– 15. Center for Climate and Energy Solutions,. Renewable and Alternative Energy Portfolio Standards http://www.c2es.org/node/9340 (accessed Sep 22, 2014). Hoesly, R. Implications of Mobility, Population Shifts, and Growth for Metropolitan Energy and Greenhouse Gas Emissions Planning, Carnegie Mellon University, 2014. Toppo, G.; Overberg, P. U.S. population growth slows to just 0.71% http://www.usatoday.com/story/news/nation/2013/12/30/census-statepopulation-estimates-growth/4248089/ (accessed Sep 21, 2014). Office of Immigration Statistics, D. O. H. S. Yearbook of Immigration Statistics: 2012; US Department of Homeland Security: Washington, DC, 2013; pp. 1–124. Abel, G. J.; Sander, N. Quantifying Global International Migration Flows. Science 2014, 343, 1520–1522. Ge, M.; Friedrich, J.; Damassa, T. 6 Graphs Explain the World's Top 10 Emitters http://www.wri.org/blog/2014/11/6-graphs-explain-world%E2%80%99s-top-10emitters (accessed August 10 2015).


Ce n

in unta

tra

l

West−

Mo

West North Central Ea s

t−

No

rth

East North 24 Central vironmental Science Page&24 Technolo of Mountain d

n gla

n wE New NeEngland

tic Mid Mid AAtlan tlan+c

Paciﬁc uous

Contig

ACS Paragon Plus Environmen So uth

Atl a

c South Antitlan+c

uth

So

us

tiguo

st−

Ea

East South Central West South Central West−

Paciﬁc Non Con+guous

Energy and Emissions from US Population Shifts ... - ACS Publications

Recommend Documents