[PPT] - Planning Research using InfoUSA/InfoGroup Data By Qisheng Pan, PowerPoint Presentation

SLIDE 1

Planning Research using InfoUSA/InfoGroup Data

By Qisheng Pan, Ph.D. Professor Department of Urban Planning and Environmental Policy Texas Southern University Houston, Texas February 7, 2018

SLIDE 2

Case 1: The Impacts of Light Rail on Residential Property Values: A Case Study of the Houston METRORail Transit Line

SLIDE 3

Empirical studies are not agreed on the impacts of rail transit

facilities on residential property values.

Houston light rail transit, METRORail, opened to the public in

January 2004. It is expected to provide commuters and transit dependent population better access to economic opportunities and social activities. However, it is not known if this transit line will have desirable impacts on various residential property values.

This study employs multi-level regression (MLR) to identify the

effects of Houston’s METRORail on the corridor, using hierarchical data at two levels, i.e., individual property level and traffic analysis zone (TAZ) level. We compare the results from MLR and Ordinary Least Square (OLS) in order to try to corroborate the findings.

Intro trodu ductio ction

SLIDE 4

Though transit rail systems, especially heavy or commuter rail,

have been found to have significant positive impacts on residential property values in some cities, the estimates of values and effects

f transit station proximity are widely different as reported in the

literature (Huang 1994).

Due to slower speed, smaller service area, and lower portion of

residential land nearby, properties near light rail stations may accrue lower premiums than those near heavy rail and commuter rail stations (Hess and Almeida 2007).

Studies of light rail report mixed results. Landis et al. (1994) found

that light rail stations have positive effects on home prices to San Diego Trolley (+$2.72/meter), negative effects in San Jose (- 1.97/meter), and indiscernible effects in Sacramento.

Pr Previo ious us St Studie ies s on the Impacts cts of Light ht Rail

SLIDE 5

The study area is defined as the traffic analysis zones

(TAZs) in Harris County.

It includes the 7.5-mile light rail line and 16 stations.
The light rail service areas are classified at distances of 1/4,

1/2, 1, 2, and 3 miles from the light rail stations.

Residential properties are grouped by their locations in the

various service areas.

St Study y Ar Area

SLIDE 6

The values and physical characteristics of properties are obtained from

property transaction data in InfoUSA’s 2007 residential database, which records sales price, home size, and home age for each of the properties in Harris County. Property sales price is converted to comparable values using Consumer Price Index (CPI) data obtained from the Bureau of Labor Statistics (BLS).

There are 2,028,880 property records in the InfoUSA database for Harris

County in 2007. After removing redundant records, limiting transactions after 1983 (the oldest year Houston CPI data are available from BLS); selecting records with sale price information, there are 529,734 properties utilized in this study, including 800 properties located in the ¼-mile radius, 1,890 within ¼-and ½-mile, 5,390 within 1/2-and 1-mile, 12,250 within 1 and 2 miles, and 16,292 within 2 and 3 miles from light rail stations (Figure 1).

Data

SLIDE 7

Figure 1. Properties in the study area

SLIDE 8

Distance to Light Rail Stations Sales Price ($/sqft) Home Year Less than1/4 Mile 55.24 31 1/4-1/2 mile 70.98 35 1/2 to 1 mile 68.66 40 1-2 mile 70.19 44 2-3 mile 69.91 45 Beyond 3 miles 39.23 27 Regional Average 41.31 28

Table 1. Sales price and age of properties at different distances to light rail stations

SLIDE 9

Most previous studies use traditional linear regression, e.g. OLS to

estimate the relationship between property values and impact factors (Bowes and Ihlanfeld 2001). They ignore the different spatial levels of some variables. However, some data like property values are available at the individual point level while others are available at a more aggregated level, such as census tract, TAZ, city and county, etc.

This study employs a two-level MLR model to separate the data at

individual level and group level (i.e. TAZ in this case). The results from both MLR and OLS models are compared.

To eliminate the effects of possible multicollinearity, this study tests

multiple models with different sets of variables and employs MLR and OLS to examine the effects of light rail on property values, step by step.

Model

SLIDE 10

Employment by industrial sector by traffic analysis zone

(TAZ) for the study area are available from the Census CTPP 2000 Part 2 data set.

Census CTPP 2000 also provides geographic maps of the

TAZs and average travel time for OD pairs.

GIS maps are utilized to identify employment centers,

calculate job accessibility, identify spatial relationships between variables, and assist computation.

SLIDE 11

The regression results from both OLS and MLR models show that

the physical characteristics of properties, including home size and home age, are the major contributors to the variance of home sales price (the R-square is about 0.32, and Pseudo R-square is 0.35, see Model 2 in Table 2).

For the light rail line, distances to rail stations and bus stops all

have significant effects on the residential property values. Based

n the coefficients estimated by MLR, light rail service has

positive effects while bus stop has negative impacts on property

values. OLS reports that both light rail and bus stop services have

significant positive effects on property values. However, their effects are modest in terms of R-square (+0.04) of OLS and Pseudo R-square values (+0.02)of MLR, see Model 3 in Table 2.

Results lts

SLIDE 12

Table 2. The impacts of light rail system on property values (Model 1-3)

Model 1 Model 2 Model 3 Model MLR OLS MLR OLS MLR OLS Fixed eff. Level 1 Intercept 4.0774 *** 4.21874 ***

0.2949

***

2.6274

***

0.3010

***

2.6826

*** lghomesize 0.6511 *** 0.9319 *** 0.6439 *** 0.9399 *** lgage

0.1586

***

0.0895

***

0.1567

***

0.1258

*** Raillineop 0.0773 *** 0.0556 *** RailQMI 0.1830 *** 0.6551 *** RailHMI 0.2416 *** 0.5608 *** Rail1MI 0.1909 *** 0.4925 *** Rail2MI 0.1303 *** 0.4933 *** Rail3Mi 0.0426 *** 0.3936 *** BusStQMI

0.0186

*** 0.1410 *** lgdistcbd lgdistmed Level 2 Popdens JobDens JobAccess Random eff. Level 1 Intercept 0.3794 *** 0.3183 *** 0.3168 *** Level 2 Intercept 0.3252 *** 0.1422 *** 0.1323 *** R Square 0.3188 0.3587 Pseudo R Square 0.3464 0.3626

Obs. of Level 1

529734 529734 529734

Obs. Of Level 2

1427 1427 1427

SLIDE 13

Table 3. The impacts of light rail on property values (Model 4-6)

Model 4 Model 5 Model 6 Model MLR OLS MLR OLS MLR OLS Fixed eff. Level 1 Intercept 0.2887 ***

1.9382

***

0.3481

***

2.6848

*** 1.1827 ***

1.4906

*** lghomesize 0.6448 *** 0.9522 *** 0.6443 *** 0.9398 *** 0.6449 *** 0.9403 *** lgage

0.1579

***

0.1545

***

0.1568

***

0.1221

***

0.1577

***

0.1415

*** Raillineop 0.0771 *** 0.0505 *** 0.0773 *** 0.0550 *** 0.0772 *** 0.0507 *** RailQMI

0.0490

0.2241 *** 0.0876 * 0.2787 ***

0.1184

**

0.0782

*** RailHMI 0.0373 0.1068 *** 0.1995 *** 0.5290 *** 0.0144 0.0573 *** Rail1MI 0.0090 0.0657 *** 0.1642 *** 0.4887 ***

0.0019

0.0587 *** Rail2MI 0.0097 0.1544 *** 0.1164 *** 0.4886 *** 0.0072 0.1441 *** Rail3Mi

0.0135

0.1551 *** 0.0368 *** 0.3801 ***

0.0141

0.1340 *** BusStQMI

0.0245

***

0.0154

***

0.0193

*** 0.1358 ***

0.0243

***

0.0070

*** lgdistcbd 0.0121

0.0637

*** 0.0227

0.0622

*** lgdistmed

0.2529

***

0.1989

***

0.2888

***

0.2296

*** Level 2 Popdens 0.0057 ***

0.0028

***

0.0121

***

0.0115

*** JobDens 0.0031 *** 0.0057 *** 0.0027 *** 0.0043 *** JobAccess

0.0004

***

0.0001

*** Random eff. Level 1 Intercept 0.3168 *** 0.3168 *** 0.3168 *** 0.0006 Level 2 Intercept 0.1092 *** 0.1271 *** 0.1023 *** 0.0041 R Square 0.3723 0.3619 0.3763 Pseudo R Square 0.3954 0.3700 0.4052

Obs. of Level 1

529734 529734 529734

Obs. of Level 2

1427 1427 1427

SLIDE 14

Case 2: Economic Losses from a Hypothetical Hurricane Event in the Greater Houston Area

SLIDE 15

1. Introduction
Hurricanes Katrina and Rita have caused tremendous socio-economic

losses on the Gulf Coast region as well as throughout areas of the United States.

Federal Emergency Management Agency (FEMA) has developed a

Hurricane Model in its HAZUS package to estimate direct losses of hurricanes, however, there is no module in the hurricane model to calculate indirect and induced impacts of hurricanes on industry purchases and household consumption.

This study proposes a systematic way to estimate direct, indirect, and

induced effects of hurricanes and also allocate them to small impact analysis zones. This approach is helpful for planners and decision makers to evaluate and forecast the economic losses of future hurricanes in their region so as to allocate manpower and resources for disaster planning more cautiously and efficiently.

SLIDE 16

2. Methodology
This study employs the damage functions in HAZUS for

estimating the property damage of a hypothetical hurricane on buildings, contents, and inventories by industrial sectors.

It transforms the property damage to the direct losses of

employment and productions, calculates total indirect and induced effects using a regional input-output model.

It allocates direct, indirect, and induced effects to impact

analysis zones in the region.

SLIDE 17

Building damage state probabilities by occupancy class by census tract Hurricane Model Storm track, wind speed, central pressure, radius to max winds, and other parameters for a hurricane Building recovery time for different damage states by

ccupancy class

Commercial building and service interruption time by occupancy class by census tract Small area employment data by industrial sector Job losses due to business interruption Dollar/job ratios Output losses due to business interruption

Figure 1. Procedures to Calculate Direct Impacts of a Hurricane Event

SLIDE 18

3. Empirical Study
This study creates a scenario for a hypothetical major

hurricane event in the Houston-Galveston area.

It estimates building damage for this designated hurricane

using the Hurricane Model in the FEMA’s HAZUS-MH package.

It calculates direct, indirect and induced effects by industrial

sector in small areas.

It highlights the most vulnerable geographic areas in the

region.

SLIDE 19

Figure 2. Hurricane Data Available in the HURREVAC 2000 Database

SLIDE 20

Figure 3. Storm Track and Advisory Points for the Hypothetical Hurricane Event

SLIDE 21

To estimate the impacts of a hurricane, this study utilizes the 2005

InfoUSA business databases for Houston-Galveston Area Council (H- GAC) region, which were obtained from the Houston Geographic Data Committee (GDC)

The data includes the number of jobs by Standard Industrial Classification

(SIC) code and North American Industry Classification System (NAICS) code for the 207,690 business establishments in the H-GAC region.

These businesses are represented as points in the data file compatible to

Geographic Information System (GIS) format, which allows point-level employment data to be easily superimposed on other geo-referenced data.

The employment by business location is aggregated to census tracts and

the SIC codes are aggregated to a small number of industrial sectors to facilitate later analysis. Employment Data from InfoUSA

SLIDE 22

Figure 4. Total Losses of Jobs from the Hypothetical Hurricane Event

4. Results

SLIDE 23

Table 1. Output and Job Losses from the Hypothetical Hurricane Event

Output ($1,000s) Jobs Direct Indirect Induced Total Direct Indirect Induced Total City of Houston 5,188,013 2,425,319 2,813,600 10,426,932 40,201 15,858 30,010 86,069 Brazoria County 790,277 147,893 1,231,385 2,169,555 5,577 920 13,134 19,631 Chambers County 7,827 25,637 952 34,416 56 158 10 225 Fort Bend County 238,171 293,828 191,906 723,905 1,334 1,550 2,047 4,931 Galveston County 962,184 131,934 2,420,944 3,515,061 8,218 920 25,822 34,960 Harris County 7,209,271 3,359,015 4,614,983 15,183,269 54,365 21,858 49,223 125,447 Liberty County 456 28,473 52 28,980 3 160 1 163 Montgomery County 17,225 252,563 4,992 274,780 117 1,369 53 1,540 Waller County 44 17,374 17,417 103 104 Sum of Eight Counties 9,225,454 4,256,717 8,465,213 21,947,383 69,671 27,039 90,290 187,000 Regional Leakages 3,724,930 1,529,657 3,104,553 8,359,140 25,302 10,236 32,981 68,519 REGIONAL TOTAL 12,950,384 5,786,373 11,569,766 30,306,523 94,973 37,275 123,271 255,518 Source: Author Calculations.

SLIDE 24

Table 2. The 20 Cities with the Highest Losses of Output from the Hypothetical Hurricane Event

Output ($1,000s) Jobs Index City Name Direct Indirect Induced Total Direct Indirect Induced Total 1 Houston 5,188,013 2,425,319 2,813,600 10,426,932 40,201 15,858 30,010 86,069 2 Pasadena 448,527 94,412 481,558 1,024,496 3,200 514 5,136 8,850 3 Galveston 260,244 44,374 534,269 838,886 2,669 326 5,699 8,694 4 Texas City 265,844 31,137 370,795 667,776 1,888 196 3,955 6,039 5 League City 123,143 16,008 498,867 638,017 1,515 153 5,321 6,989 6 Pearland 139,341 20,954 336,790 497,084 1,179 147 3,592 4,918 7 Friendswood 68,162 8,368 314,083 390,614 794 71 3,350 4,216 8 Alvin 130,112 16,142 135,064 281,318 1,012 98 1,441 2,551 9 Deer Park 105,656 19,103 152,131 276,890 651 128 1,623 2,402 10 La Marque 93,233 10,260 162,432 265,925 374 48 1,733 2,155 11 La Porte 95,386 20,500 146,844 262,730 559 134 1,566 2,259 12 Webster 175,369 15,536 58,020 248,925 1,994 128 619 2,741 13 Stafford 105,282 72,000 5,612 182,894 406 303 60 769 14 Sugar Land 50,268 106,632 17,381 174,281 301 449 185 936 15 Hitchcock 41,856 4,448 109,253 155,556 197 21 1,165 1,383 16 Missouri City 46,421 24,405 60,133 130,959 350 172 641 1,163 17 The Woodlands 9,493 110,853 2,333 122,679 54 435 25 514 18 Channelview 64,023 21,823 36,172 122,018 282 118 386 786 19 Manvel 17,234 2,114 96,745 116,093 201 18 1,032 1,250 20 Seabrook 40,568 5,592 59,073 105,233 226 31 630 887 TOTAL 7,468,172 3,069,980 6,391,157 16,929,309 58,054 19,348 68,168 145,570

SLIDE 25

Case 3: Economic Losses from Hurricane IKE in the Greater Houston Area

SLIDE 26

When Hurricane IKE made its US landfall in Galveston, Texas at

2:10am CDT on September 13, 2008, its wind speed was 110 mile per hour (MPH), the maximum of a Category 2 hurricane. It also created a maximum of 15-foot storm surge at the Texas coast.

Hurricane IKE caused a variety of physical damages on its way

right crossing the heart of the Houston region. High wind ripped away roofs, tumbled walls, broke windows, and tore off exterior materials of houses and buildings. Severe damages were found in many tall buildings in the center of the city.

Though the magnitude of the hurricane was less than the

expectation, it still caused significant property damage and economic losses to the Gulf Coast region and directly and indirectly throughout much of the U.S.

Intro trodu ductio ction

SLIDE 27

Economic losses of residents and businesses have extended for a

long period time after Hurricane IKE. The widespread damages

n residential houses, commercial buildings, transport facilities,

and utility service have taken the regional much longer than usual to recover.

The Houston Chronicle on October 21, 2008 reported that over

half of the 2,000 apartment complexes in Houston were hardly hit by Hurricane IKE and about 150 of these apartment complexes were severely damaged.

Overall three quarter of households in Houston had suffered from

power cuts. On September 25, 13 days after the strike of the hurricane, there was still about one quarter of the residents in the city having no electricity.

SLIDE 28

An important step of disaster relief is to have a quick estimate of

hurricane damage losses, which may help government agencies to identify most cost-effective options for disaster mitigation and allocate manpower and resources for disaster relief more efficiently.

The report for Hurricane IKE became available in the early of

2009, about four months after the Hurricane IKE made its US landfall.

National Hurricane Center (NHC) at Miami has released Tropical

Cyclone Report for major hurricanes striking the U.S. These estimates of uninsured losses is based on empirical studies and the method to calculate the total losses by doubling the insured damage costs seems too arbitrary. In addition, the NHC’s reports have no information about the spatial distribution of the damage losses in small areas.

SLIDE 29

Figure 2. Storm Track of Hurricane IKE and Hypothetical Hurricane “RITA”.

SLIDE 30

Figure 3. Estimated Wind Speeds of Hurricane IKE by HAZUS-MH.

SLIDE 31

Figure 4. Estimated Property Damage Losses of Hurricane IKE

SLIDE 32

As Pan et al. (2008) suggested, building damage states are applied to

estimate business interruption times and employment data is utilized to calculate the business interruption losses, which is denoted by Equation (1).

The business interruption times are calculated by the building

recovery time for the corresponding building damage states and the building interruption time multipliers, both available from HAZUS manual (FEMA 2006, 2006b). They are converted to an annual ratio to be consistent to employment data and the annual data used in the input-output model. Employment data come from the InfoUSA 2008 business database, which provides the number of jobs by standard industrial code (SIC) for Houston-Galveston Area Council (H-GAC) region.

Employment data enters Equation (1) with the building recovery

times to yield the economic losses of jobs by sector by census track, which are the direct impacts. Dollar per job ratios are utilized to covert economic job losses to losses in dollar value.

SLIDE 33

An economic input-output model, i.e. IMPLAN in this study, is

employed to calculate indirect and induced effects using the direct impacts in dollar or job. Multiple programs are developed to bridge the employment sectors and the sectors in IMPLAN 2000 sectors.

The indirect impacts of the IMPLAN input-output model are not

listed by impact analysis zone. Equation (2) with employment data by census tract is utilized to allocate total indirect impacts.

Similar to indirect impacts, the induced effects as the output of the

IMPLAN input-output model are the regional total without spatial

information. A journey home-to-work O-D matrix and a journey

home-to-work O-D matrix from the H-GAC 2000 base year model are used to allocate the induced effects to the 2954 H-GAC TAZs, which are further migrated to the 866 census tract.

SLIDE 34

The results of direct, indirect, and induced effects measured by both

dollar value and job for City of Houston and eight H-GAC Counties are listed in Table 2.

It reports $11.6 billion of output losses and 97,374 persona-years of

employment losses, which are combined with the $20.6 billion of property damage to make Hurricane IKE the third most destructive and costliest hurricane in the U.S. (lower than the losses in Hurricane Andrew of 1992 and Hurricane Katrina of 2005).

Table 2 also shows that regional leakages, i.e. spillovers in the

indirect and induced effects, are very large ($8.1 billion of output and 55,449 person-years of employment), which reflects the fact that the local component of the manufacturing and mining sectors is very high, with heavy reliance on trades with outside the region.

SLIDE 35

Table 2. Estimated Output and Job Losses from Hurricane IKE

Output ($1,000s) Jobs Direct Indirect Induced Total Direct Indirect Induced Total City of Houston 491,661 260,449 499,804 1,251,914 6,616 2,649 5,265 14,530 Brazoria County 48,813 14,217 90,361 153,391 762 152 994 1,908 Chambers County 38,379 4,883 21,610 64,871 629 46 234 909 Fort Bend County 1,038 33,870 157,695 192,602 9 375 1,659 2,044 Galveston County 351,852 24,824 106,740 483,416 5,595 251 1,186 7,033 Harris County 1,058,963 361,334 975,820 2,396,117 14,366 3,710 10,293 28,369 Liberty County 7,160 1,160 12,625 20,946 116 12 140 268 Montgomery County 19,727 19,346 75,906 114,979 338 205 809 1,351 Waller County 13 554 3,455 4,022 6 38 44 Sum of Eight Counties 1,525,945 460,187 1,444,211 3,430,343 21,816 4,757 15,353 41,926 Regional Leakages 3,422,245 1,785,626 2,926,258 8,134,128 14,515 9,731 31,201 55,449 TOTAL 4,948,190 2,245,813 4,370,469 11,564,472 36,331 14,488 46,554 97,374

Source: Author Calculations.

SLIDE 36

Among the eight counties in the H-GAC region, Harris County

accounts for over two third of the losses in output and employment. Galveston County, where Hurricane IKE made landfall, ranked second by losing $483 million of output and over 7,000 jobs. Brazoria County, located in the south of the region, ranked third in terms of direct output and job losses. The spatially allocated output losses are shown in Figure 5.

Table 3 shows the tops 20 cities in the region ranked by their losses
f output in Hurricane IKE. City of Houston, the biggest city in the

region, has the largest impacts on the list, losing about $1.3 billion

utput and 14,530 jobs, which are about seven times larger than the

second city on the list, City of Pasadena, and also one third of the regional losses excluding the leakage. City of Galveston, regardless

f its size, ranks the third on the list and losses $144.9 million of
utput and 1,856 jobs. All the rest cities among the top ten are located

in the south of the region along the storm tract.

SLIDE 37

Figure 5. Estimated total losses of output from Hurricane IKE Event

SLIDE 38

Table 3. The 20 Cities with the Highest Losses of Output from Hurricane IKE Event

Output ($1,000s) Jobs Index City Name Direct Indirect Induced Total Direct Indirect Induced Total 1 Houston 491,661 260,449 499,804 1,251,914 6,616 2,649 5,265 14,530 2 Pasadena 124,673 9,580 39,337 173,590 1,705 90 421 2,216 3 Galveston 125,630 5,441 13,838 144,909 1,642 57 158 1,856 4 Baytown 95,548 2,907 13,825 112,280 1,233 30 150 1,414 5 Webster 88,283 1,919 3,115 93,317 1,164 21 35 1,220 6 League City 59,064 2,446 16,874 78,385 1,287 32 189 1,509 7 Texas City 37,280 3,413 13,850 54,542 528 32 154 714 8 La Porte 31,690 2,406 10,403 44,499 496 24 111 631 9 Pearland 27,060 2,462 14,363 43,885 429 27 155 610 10 Deer Park 25,011 2,428 9,823 37,262 359 24 105 487 11 The Woodlands 5,091 7,720 21,781 34,592 68 76 224 369 12 Atascocita 21,382 1,360 9,794 32,536 376 17 105 498 13 Friendswood 22,402 1,103 8,357 31,862 395 13 92 500 14 Sugar Land 386 8,413 22,810 31,608 3 82 237 322 15 Channelview 19,993 2,228 7,791 30,012 294 19 82 395 16 Missouri City 201 3,195 20,194 23,590 2 41 211 255 17 Alvin 11,725 1,639 4,742 18,106 177 15 51 243 18 Conroe 3,937 3,850 8,292 16,079 67 43 90 200 19 Nassau Bay 12,771 487 701 13,959 155 5 8 168 20 Aldine 7,171 2,233 4,524 13,929 94 22 48 164 TOTAL 1,210,961 325,679 744,218 2,280,858 17,091 3,318 7,890 28,299

Source: Author Calculations.

SLIDE 39

CONCLUSIONS AND DISCUSSIONS

Disaster relief plans have to base on a careful study on damage states

and economic losses in each impact zone, which are not limited to the areas under direct attack.

To gain a better estimate of the economic losses from natural

disasters, this paper develops a systematic method with a general framework that is a combination of disaster model, economic input-

utput model, and spatial allocation model.
The method complements the HAZUS Hurricane Model with a more

reasonable estimation of direct output and job losses and fills the gaps by developing the functions to estimate indirect and induced losses. It also utilizes GIS spatial analysis functions to allocate the losses to multiple types of impact analysis zones or political jurisdictions.

SLIDE 40

This method is implemented to investigate the economic losses of

Hurricane IKE in the Houston-Galveston region. It reports that Hurricane IKE has caused significant economic losses to the Houston region. Its property damage and the business interruption losses have made Hurricane IKE the third costliest one in the U.S.

f all time.
The property damage and economic losses are not limited to the

corridor along the storm track. The losses have widely extended from the south coast in the Galveston Bay to the Montgomery County in the north of the region. Some areas such as the ones in Fort Bend County receive few direct impacts but a fair amount of induced impacts, which makes the total impacts large.

The results are in line with the observation from the real world.

The quick and proper estimate of hurricane damage losses will help policy maker and planners to identify most cost-effective

ptions for disaster and mitigation and allocate manpower and

resources for disaster relief more efficiently.

SLIDE 41

SLIDE 42

Pan, Q. (2015) "Estimating the Economic Losses of Hurricane IKE in the Greater

Houston Region," Nat. Hazards Rev., Vol.16, Issue 1, http://dx.doi.org/10.1061/(ASCE)NH.1527-6996.0000146

Pan, Q., H. Pan, M. Zhang, and B. Zhong (2014) “The Effects of Rail Transit on

Residential Property Values: A Comparison 1 Study on the Rail Transit Lines in Houston and Shanghai,” Transportation Research Record (TRR), Journal of the Transportation Research Board, No. 2453, pp. 118-127. http://trrjournalonline.trb.org/doi/abs/10.3141/2453-15?journalCode=trr