PTC’15 Networked Planet
DOES THE INTERNET MAKE PEOPLE STAY HOME? EVALUATING THE IMPACT OF THE INTERNET ON VEHICLE TRANSPORTATION AND MOTOR FUEL CONSUMPTION
Xue “Snow,” Dong PhD candidate, Pennsylvania State University, USA
ABSTRACT The striving of nations to build an Internet-based information society is based on the belief that positive economic impacts will be achieved via the Internet. What about the environmental impacts of the Internet? This paper investigates the determinants of vehicle transportation and motor fuel consumption, with the specific aim of quantifying the effects of the Internet on vehicle transportation and motor fuel consumption. A panel data from 50 U.S. states in the period of 2001-2012 (with gaps) was analyzed using fixed effects least square dummy variable (LSDV) model and random effects feasible generalized least squares (GLS) model. Results show that Internet usage is negatively related to vehicle transportation and motor fuel consumption, and therefore the development of the Internet indeed decreases fuel mileages.
KEYWORDS: the Internet, vehicle transportation, motor fuel consumption, environment
1. INTRODUCTION The burning of fuel results in carbon dioxide emissions, which contribute to greenhouse effect and global climate change. According to the US National Broadband Plan, the transportation industry is the second largest consumer of energy and the second highest emitter of greenhouse gases (Federal Communications Commission [FCC]). Internationally, transportation energy use accounts for approximately one fourth of the total energy consumption in OECD countries (Uusitalo & Djerf, 1983).
1
PTC’15 Networked Planet
The impact of the Internet on vehicle transportation, energy use, and the environment has been discussed for a while, although the number of academic research papers in this area is very small. Most of the existing studies suggested a positive impact of the Internet/broadband on less consumption of the planet’s resources and less spewing of polluting emissions thanks to e-commerce, telecommuting, teleconferencing, online shopping, and downloading files, all of which cut transportation and fuel mileage (Fuhr & Pociask, 2007; Fuhr & Pociask, 2011; Mcmahon, Thomas & Kaylor, 2012). For instance, Mcmahon, Thomas, and Kaylor (2012) found that broadband technologies could directly reduce motor fuel use in the transportation system in two ways. The first is to reduce the time people spend in cars burning fuel in congested traffic; the second is to decrease the number of vehicle trips. And according to Katsumoto (2002), ICT would reduce energy consumption by 3.6% in Japan in the year of 2010. Is that true? Does the Internet really reduce vehicle transportation through telecommuting, teleconferencing, and online shopping, and therefore cut down on motor fuel consumption? A few scholars have different opinions. Teppayayon, Hohlin, and Forge (2009) contended that there actually might be an increase in consumption of goods and services enabled by broadband, such as more transport due to online shopping and customized delivery. The picture is ambiguous and the two opposite standpoints are intriguing. However, none of the previous research on this topic is empirical; all of them are prospective simulation studies. Looking for the correct answer to this question not only improves our understanding of how telecommunication technologies change our life, but also is beneficial to the society and policy makers. To bridge the gap, the current study attempts to empirically examine whether the Internet decreases or increases vehicle transportation and motor fuel consumption. 2. LITERATURE REVIEW 2.1 ENERGY IMPACTS FROM INFORMATION TECHNOLOGIES Data for the environmental impacts of information technologies are scarce, partly because it is very complicated and next to impossible to accurately measure. Existing literatures provide a diverse picture of positive and negative environmental effects. Most scholars agree on the positive environmental impacts of information technologies. Fuhr and Pociask (2007; 2011) believed that a considerable amount of travel is displaced through telecommuting, teleconferencing, and downloading. Broadband services provide data, video, and voice communications, permitting employees to remotely work from home. By making it easier to work from home and use video-conferencing or email rather than travel to meetings, the amount of commuting and business travels is reduced. This not only decreases the number
2
PTC’15 Networked Planet
of local trips but also long distance plane and train travel. Downloadable content of books, music, videos, and software all reduce the transport costs between manufacturers, warehouses, and retail stores (Fuhr & Pociask, 2011). Fuhr and Pociask (2011) predicted that all of these phenomena would reduce travel congestion as well as motor fuel consumption. They predicted that in the next 10 years, telecommuting would cut 247.7 million tons of greenhouse gas emissions due to less driving (Fuhr & Pociask, 2011). On the other hand, some scholars hold different thoughts that telecommunication technologies might have negative effects on the environment and natural resources. Teppayayon, Hohlin, and Forge (2009) posited that there are both direct and indirect negative effects of broadband on environmental sustainability. Direct effects include increased electricity use from information and communications technology (ICT) equipment and waste. According to the statistics provided in the National Broadband Plan, ICT industries account for 120 billion kilowatt-hours (kWh) of electricity use annually—approximately 3% of all U.S. electricity. Indirect effects relate to increased consumption of goods and services enabled by broadband (Teppayayon, Hohlin, & Forge, 2009). The authors argued that the “rebound effects” could even cancel out the environmental benefits of broadband (Teppayayon, Hohlin, & Forge, 2009). Mokhtarian (2002) argued that the relationship between vehicle travel and telecommunications are complementary rather than substitutive, “although direct, short-term studies of that impact focusing on a single application (such as telecommuting) have often found substitution effects, such studies are incomplete and likely to miss the more subtle, indirect, and longer-term complementarity effects that are typically observed in more comprehensive studies.” (p. 53-54). In a later study conducted by him and Choo (2006), they supported Mokhtarian’s standpoint with empirical results. One thing worth noticing is the assumed environmental impacts of the Internet are not limited to vehicle transportation and motor fuel use. The Internet/broadband is believed to be an important part of the solution to solve energy and environmental challenges as mentioned in the US National Broadband Plan (FCC). Beyond transportation, it is widely believed that telecommuting can also lead to less office construction, less paper use, and less pollution. Fuhr and Pociask (2011) estimated that telecommuting can reduce greenhouse gas emissions over the next 10 years by approximately 588.2 tons, of which 247.7 million tons is due to less driving, 28.1 million tons is due to reduced office construction, and 312.4 million tons because of less energy usage by businesses (Fuhr & Pociask, 2011). Moreover, according to the National Broadband Plan (FCC), broadband investment will also improve energy innovation. For example, broadband will make energy data readily accessible to consumers so that consumers are enabled to save energy at home; digital innovations – like real-time traffic information systems and broadband-enabled navigation tools – can enable more efficient
3
PTC’15 Networked Planet
route-planning and driving for commuters. Due to the complex and multifaceted nature of the relationship between the Internet and the environment, the current study only addresses the impact of the Internet on vehicle transportation and motor fuel consumption. 2.2 FACTORS AFFECTING VEHICLE TRANSPORTATION AND MOTOR FUEL CONSUMPTION Concerns over motor vehicles’ environmental impact have been around since the invention of vehicles. In 2012, the transportation sector accounts for 26.71% of U.S. energy consumption and 28% of U.S. greenhouse gas emissions (USDOT; USEPA). Vehicle transportation also contributes to problems like oil dependence and congestion. Scholars have been attempting to examine the links between lifestyles, travel, vehicle use, and fuel use for decades. Numerous studies have scrutinized factors that may influence vehicle travel and motor fuel use. 2.2.1 TRAVEL COST It is commonly agreed that the level of fuel price is critical in determining fuel mileage. Pickrell and Schimek (1998) found in their regression model that a 10% increase in gasoline price reduces American households’ vehicle miles traveled (VMT) by 1.9% to 3.2%. The result echoed with Nordhaus’s calculation for 6 OECD countries in 1977, which estimated a price elasticity of -0.36 for transportation in those countries, which means a 10% of increase in gasoline price would lead to a decrease of VMT by 3.6%. Fuel efficiency, operationalized as MPG, is also related to travel and fuel consumption. Small and Dender (2007) identified the rebound effect for motor vehicles that improved fuel efficiency actually stimulates additional travel. Schipper, Steiner, Figueroa, and Dolan (1993) used panel data from the United States and OECD countries to conduct a cross-sectional comparison, in which they concluded that fuel prices have a significant impact on per capita automobile fuel use, fuel use per vehicle, and fuel use per kilometer. That means, with higher fuel prices, people would cut on fuel use by driving less and choosing more fuel-efficient cars. They also found that income and income-related factors influence car ownership and choice of car models (Schipper, Steiner, Figueroa, & Dolan, 1993). 2.2.2 SOCIOECONOMIC AND DEMOGRAPHIC FACTORS Previous studies have identified socioeconomic and demographic characteristics as key explanatory factors for VMT, such as income (Buehler, 2010; Simma & Axhausen, 2001; Pickrell & Schimek, 1998; Noland, 2001), gender (Frondel & Vance, 2009; Buehler, 2010), education (Frondel & Vance, 2009), employment/unemployment (Frondel & Vance, 2010; Buehler, 2010; Hymel, 2014; Washington State Department of Transportation, 2010), and car ownership (Simma & Axhausen, 2011; Washington State Department of Transportation, 2010).
4
PTC’15 Networked Planet
For instance, Pickrell and Schimek (1998) estimated the income elasticity of daily VMT ranges from 0.35 to 0.37, which means a 10% of increase in income would result in a 3.5%-3.7% rise in VMT. Hymel (2014) identified different factors that are correlated with VMT in the state of California, including fuel prices, economic factors (unemployment rate, per-capita income, and median household income), motor vehicle registrations and licensing. Using a more rigorous linear regression analysis, Hymel (2014) found fuel prices to be significant with a long-run (over one year) elasticity of -0.16, per capita income with a long-run elasticity of 0.3, and unemployment rate with an elasticity of 0.008. Washington State Department of Transportation (2010) took into consideration many potential factors, including fuel consumption, gas prices, Washington motor vehicle registrations, Washington unemployment rate, Washington personal income, population density, etc to identify an accurate VMT forecast model. The optimal econometric VMT forecast model includes three independent variables: Washington employment, Washington motor vehicle registrations, and Washington gas prices. 2.2.3 SPATIAL PATTERNS Population density is considered to be one of the major factors influencing VMT, and the factor has been examined in existing works (Fang, 2008; Su, 2011; Karathodorou, Graham, & Noland, 2010; Newman & Kenworthy, 1999; Buehler, 2010). Road density (lane miles divided by area) is another factor affecting transportation – increased road density leads to an increase in VMT (Noland, 2001; Su, 2010; Karathodorou, Graham, & Noland, 2010). 2.2.4 THE INTERNET While a number of studies have examined the effects of demographic and socioeconomic characteristics, population density, road density, and fuel price on vehicle transportation and motor fuel use, the effect of the Internet on vehicle transportation and motor fuel use has not been examined empirically. 3. DATA AND MEASURES This study used panel data for the 50 U.S. states in 7 years: 2001, 2003, 2007, 2009, 2010, 2011, and 2012. The years are not consecutive due to the availability of data. Internet use question was only asked in the US Current Population Survey in those 7 years between 2000 and 2013. District of Columbia was left out because of missing data. As a result, a total of 50×7 = 350 observations were pooled for statistical analyses. In this study, vehicle transportation is operationalized as vehicle-miles traveled (VMT). VMT data for the 50 states from 2001–2012 were obtained from the U.S. Department of Transportation Federal Highway Administration (FHWA). Note that VMT is a measure of the extent of travel on all public roads (including interstate, freeways, arterials, collectors, and local roads) in each state. The first dependent variable of the study VMT per capita (vmtpc) was
5
PTC’15 Networked Planet
calculated by dividing VMT by the state population. The unit of vmtpc used in the dataset is in thousand vehicle-miles. Motor-fuel use data for the 50 states from 2001-2012 were obtained from the U.S. Department of Transportation Federal Highway Administration (FHWA). Motor-fuel use includes private and commercial and public use on both highways and non-highways. Motor-fuel use per capita (mfuelusepc) was calculated by dividing total motor fuel use by state population and the unit is in gallons. Because the two dependent variables are both continuous, linear regression models were developed to determine the factors affecting them separately. Based on previous research that examined vehicle transportation and fuel consumption, ten independent variables were included in the model. They are: internet use (internet), gasoline price (gasprice), vehicle registration per capita (vehiclepc), road density (roaddens), gender (gender), education (edu), disposable income per capita (incomepc), unemployment rate (unemp), population density (popdens), and MPG (mpg). The specifications, sources, and unit of each dependent and independent variable are listed in Table 1. Table 1. Variables and Measurement Variable type
Variable
Label
Specification
Unit
Sources U.S. Department of Transportation,
Depende nt
VMT per capita
vmtpc
total VMT in each state divided
in thousand
Federal Highway
by state population
vehicle-miles
Administration U.S. Department of
variable Motor-fuel use per capita
mfuelusepc
total motor-fuel use in each
Transportation,
state divided by state
Federal Highway
population
in gallons
Administration
percentage of people with access to the Internet (both dial-up and broadband) at Internet use
internet
home
Current Population in percent
Survey
motor gasoline sales through Independ
retail outlets prices; There is
ent
no gas price data by state for
variable
Gasoline price
gasprice
Vehicle registration per
vehiclepc
the year 2011 and 2012;
U.S. Energy
national average is used as a
Information
proxy.
in U.S. dollars
Administration
vehicle registration in each
U.S. Department of
state divided by state
Transportation,
6
PTC’15 Networked Planet
capita
population
Federal Highway Administration U.S. Department of Transportation,
Road density
roaddens
road length in each state
in miles per
Federal Highway
divided by
sq. miles
Administration
state area
U.S. Census Bureau; Current Gender
gender
percentage of male
in percent
percentage of people with Education
edu
bachelor's degree and above
Population Survey Current Population
in percent
Survey U.S. Department of
Disposable
Commerce, Bureau
income per capita
incomepc
disposable income per capita
U.S. dollars in
of Economic
thousands
Analysis U.S. Department of Labor, Bureau of
Unemployment
unemp
Population density
popdens
unemployment rate
in percent
Labor Statistics
state population divided by
population per
Current Population
state area
sq. mile
Survey U.S. Department of Transportation, Bureau of
MPG
mpg
average U.S. light duty vehicle
miles per
Transportation
fuel efficiency
gallon
Statistics
Ten independent variables were employed in predicting the dependent variables VMT per capita (vmtpc) and motor-fuel use per capita (mfuelusepc). The specification of the models are as following: vmtpcit = β1 + β2internetit + β3gaspriceit + β4vehiclepcit + β5roaddensit + β6genderit + β7eduit + β8incomepcit + β9unempit + β10popdensit + β11mpgit + Wit.
(1)
Where Wit = ci + Uit. mfuelusepcit = β1 + β2internetit + β3gaspriceit + β4vehiclepcit + β5roaddensit + β6genderit +β7eduit + β8incomepcit + β9unempit + β10popdensit + β11mpgit + Wit. (2) Where Wit = ci + Uit. To detect whether there is multicollinearity problems among independent variables, the variance inflation factor (VIF) test was run using STATA. The result is listed in Table 2. Various rules of thumbs for acceptable VIF levels have been proposed. The most common
7
PTC’15 Networked Planet
standard is a value of 10 as the maximum level of individual VIF (Hair, Anderson, Tatham, & Black, 1995; Klein). Since there is no value larger than 10, there is no serious multicollinearity problems among the ten independent variables. Table 2 VIF Results Variable
VIF
1/VIF
internet
9.65
0.10
popdens
8.12
0.12
roaddens
7.82
0.13
incomepc
6.53
0.15
mpg
5.07
0.20
gasprice
3.78
0.26
edu
3.48
0.29
unemp
2.68
0.37
gender
2.06
0.49
vehiclepc
1.55
0.65
Mean VIF
5.07
A Wooldridge test for autocorrelation in panel data was conducted to determine whether there is a first-order autocorrelation for equation (1) and equation (2). The result for equation (1) F (1, 49) = 9.479, p = 0.034 indicated there is an autocorrelation problem; the result for equation (2) F (1, 49) = 3.852, p = 0.0554 indicated there is no autocorrelation problem. Next, a modified Wald test for groupwise heteroskedasticity in fixed effect regression model was run in STATA and got chi2 (50) = 2028.33, p = 0.000 and chi2 (50) = 9492.92, p = 0.000 for equation (1) and equation (2) respectively, which indicated there are heteroskedasticity problems and the variances of dependent variables vary across observations. With autocorrelation and heteroskedasticity problems, OLS estimators are unbiased but inefficient. Thus, OLS with cluster-robust standard errors was adopted to produce robust standard error estimates for linear panel models (Hoechle, 2007). 4. MODEL SELECTION AND RESULTS Using panel data, I first ran fixed effects least square dummy variable (LSDV) model, then random effects generalized least squares (GLS) model, and last the Hausman test for the exogeneity of the unobserved error component to decide which model is more appropriate for the equation (1) and equation (2) (McManus, 2011). The assumption for fixed effects least square dummy variable (LSDV) model is each subject has its own time-invariant characteristics (omitted variables) and they are correlated with the independent variables. Fixed effects least square dummy variable (LSDV) model removes the effect of time-invariant characteristics from the independent variables so I can assess the
8
PTC’15 Networked Planet
independent variables’ effects. By adding the dummy for each state (statedm) I am able to estimate the pure effect of the ten independent variables by controlling for the unobserved heterogeneity. The equations for fixed effects least square dummy (LSDM) variable model become: vmtpcit = β1 + β2internetit + β3gaspriceit + β4vehiclepcit + β5roaddensit + β6genderit + β7eduit + β8incomepcit + β9unempit + β10popdensit + β11mpgit + β12statedm1it + β13statedm2it + β14statedm3it+ … β60statedm49it + Uit. (3) mfuelusepcit =β1 + β2internetit + β3gaspriceit + β4vehiclepcit + β5roaddensit + β6genderit + β7eduit + β8incomepcit + β9unempit + β10popdensit + β11mpgit + β12statedm1it + β13statedm2it + β14statedm3it+ … β60statedm49it + Uit.
(4)
The results of the fixed effects least square dummy variable (LSDM) model are shown in Table 3 and Table 4. Seven independent variables are significant predictors of VMT per capita (vmtpc) at 95% confidence level. They are Internet use (internet) (p = 0.008), gasoline price (gasprice) (p = 0.038), road density (roaddens) (p = 0.011), gender (gender) (p = 0.000), unemployment rate (unemp) (p = 0.057), population density (popdens) (p = 0.005), and MPG (mpg) (p = 0.037). F (10, 49) = 11.13, p = 0.0000 indicates the model has a good fit. R-squared within for the ten independent variables is 0.2667; Adjusted R-squared for ten independent variables plus the state dummy variables is 0.9579. Table 3. Fixed Effects Least Square Dummy Variable Model for VMT Per Capita
Six independent variables are significant predictors of motor fuel use per capita (mfuelusepc) at 95% confidence level. They are Internet use (internet) (p = 0.003), gasoline price (gasprice) (p = 0.01), gender (gender) (p = 0.041), disposable income per capita (incomepc) (p = 0.017), unemployment rate (unemp) (p = 0.005), and population density (popdens) (p = 0.039). F (10, 49) = 31.25, p = 0.0000 indicates the model has a good fit. R-squared within for the ten
9
PTC’15 Networked Planet
independent variables is 0.3957; Adjusted R-squared for ten independent variables plus the state dummy variables is 0.9597. Table 4. Fixed Effects Least Square Dummy Variable Model for Motor Fuel Use Per Capita
The assumption for using random effects model is that the omitted variable is not correlated with the independent variables and the omitted variable is a random variable. Under this assumption, fixed effects methods are inefficient as they throw away information on between-individual variations (Greene, 2003). Random effects model use feasible generalized least squares (GLS) estimation to estimate parameters, which are a weighted average of between and within estimators. The results for equation (1) and equation (2) are shown in Table 5 and Table 6 below. Table 5. Random Effects Feasible GLS Model for VMT Per Capita
According to the random effects model, seven independent variables are significant predictors
10
PTC’15 Networked Planet
of VMT per capita (vmtpc) at 95% confidence level. They are Internet use (internet) (p = 0.003), gasoline price (gasprice) (p = 0.026), road density (roaddens) (p = 0.009), gender (gender) (p = 0.000), unemployment rate (unemp) (p = 0.032), population density (popdens) (p = 0.000), and MPG (mpg) (p = 0.011). Wald chi2 (10) = 143.09, p = 0.0000 indicates the model has a good fit. R-squared overall for the ten independent variables is 0.3127. Table 6. Random Effects Feasible GLS Model for Motor Fuel Use Per Capita
Six independent variables are significant predictors of motor fuel use per capita (mfuelusepc) at 95% confidence level. They are Internet use (internet) (p = 0.001), gasoline price (gasprice) (p = 0.006), vehicle registration per capita (vehiclepc) (p = 0.020), disposable income per capita (incomepc) (p = 0.007), unemployment rate (unemp) (p = 0.002), and population density (popdens) (p = 0.011). Wald chi2 (10) = 310.25, p = 0.0000 indicates the model has a good fit. R-squared overall for the ten independent variables is 0.3493. To choose from fixed effects and random effects models for the two equations, Hausman specification tests were run to see whether E(Xjit ci) = 0. The null hypothesis is that the unique errors (ci) are uncorrelated with the independent variables (Xjit) (Greene, 2003). The Hausman test result for equation (1) is chi2 (11) = 65.63, p = 0.0000 and for equation (2) is chi2 (11) = 44.14, p = 0.0000. Both null hypotheses are rejected. I conclude that random effects model is inconsistent and fixed effects model for both equations is preferred. 5. CONCLUSION The findings of the study are provocative in the sense that the Internet does have a significant impact on people’s travel behavior and motor fuel use. In accordance with Fuhr & Pociask’s (2007; 2011) point of view, usage of the Internet will cut on travel mileages and fuel consumptions. For VMT per capita, the coefficient of the Internet use is -0.023 (p = 0.008), meaning if the percentage of people who have the Internet access at home grows by ten
11
PTC’15 Networked Planet
percent and all other factors being fixed, an average person’s annual vehicle miles traveled will decrease by about 235 miles (note: unit of VMT per capita is in thousand vehicle-miles). For motor fuel use, the coefficient of the Internet use is -1.77, which tells us that ten percent of increase in the percentage of population who have the Internet access at home (with all other factors fixed) will lead to a decrease of around 17.7 gallons of motor fuel consumption per person. To put the number into a context, the national average gasoline consumption per person in the United States is 315 gallon in 2014 (The World Bank). If the Internet penetration rate grows by 10%, each person will cut on approximately 5.6% of his/her motor fuel use. This result can be explained by the functions of the Internet. With the Internet, some people do not need to commute to work everyday but work from home; teleconferencing and email communications save the trips to physical meetings; downloading music, software, and videos cut on the trips to fetch them; e-commerce minimizes times of visiting a “brick” store; and social media like Facebook and Twitter connect families and friends virtually. Moreover, millions of websites, apps, games, online activities, and all sorts of information can be attractive and time-consuming, preventing people from going out and travel. Other significant predictors found in this paper resonate with previous studies. Gas price is a significant predator of both VMT per capita and motor fuel use per capita. Gas price is negatively related to the two dependent variables, with the coefficients being -0.22 (p = 0.008) and -38.1 (p = 0.003) respectively. A one-dollar increase in gas price per gallon will lead to a decrease of 220 vehicle miles traveled and 38 gallons of fuel use for each person per year. Unemployment is negatively related to both VMT per capita and fuel use per capita with the coefficients being -0.048 and -4.5. One percent increase of unemployment rate will decrease VMT per capita by 48 miles and decrease fuel use per capita by 4.5 gallon, holding other factors unchanged. This is understandable since unemployment means people do not need to commute to work everyday. Population density is negatively associated with VMT per capita and fuel use per capita. Metropolitan areas with high population density in most cases have better public transit systems than remote areas so people do not need to drive. MPG is a significant and positive predictor of VMT per capita but not significant for fuel use per capita. The interesting positive relationship between MPG and VMT per capita can be explained by the rebound effect of higher vehicle efficiency (Small & Dender, 2007). Small and Dender (2007) called it “rebound effect” because they noticed that improved fuel efficiency (higher MPG) actually causes more travel. Disposable income is positively related to motor fuel use per capita but not significantly related to VMT per capita. It makes sense because rich people with higher disposable income tend to
12
PTC’15 Networked Planet
choose bigger size vehicles such as SUV. They are less concerned about vehicles’ fuel efficiency. Road density is positively related to VMT per capita. If a state builds more roads, people tend to drive more. And more roads means less congestion, the motor fuel wasted in traffic jam maybe lessened. Gender is negatively related to VMT per capita and motor fuel use per capita. A population with more males tends to have a lower VMT per capita and a lower fuel use per capita, which indicates on average females travel more miles than males. In the U.S., there has been a sharp decline in vehicle miles traveled for males and the decline for women is not as great (Lowy, 2012). And according to Federal Highway Administration (2011), women made significantly more trips than men in 2009. Due to data availability, the current study utilized data for 50 U.S. states in seven years between 2001 and 2012. If a longer period of data that include the inception of the Internet are available, the result would be more interesting in showing the impact of the Internet on vehicle transportation and motor fuel use and the validity of the study would be enhanced. Another limitation of the current study is that the fact that a growing number of vehicles are using alternative energy sources other than fuel, such as solar and electricity, was not considered. Last but not least, there are other motivational factors influencing vehicle transportation and motor fuel use, like the availability of public transit systems, culture, lifestyles, etc. However, these factors are difficult to operationalize and quantify thus they are left out from the regression equation. As is mentioned previously, the environmental impact of the Internet is complex and multifaceted. Energy consumption associated with vehicle travel is just one side of the story. I hope the findings of the paper will offer intriguing clues for future research. The bottom line is we are now aware that the Internet does make people stay home or at least stay off vehicles.
13
PTC’15 Networked Planet
REFERENCES Buehler, R. (2010). Transport policies, automobile use, and sustainable transport: A comparison of Germany and the United States. Journal of Planning Education and Research, 30(1), 76. Choo, S. & Mokhtarian, P. (2007). Telecommunications and travel demand and supply: Aggregate structural equation models for the US. Transportation Research 4(18). Fang, H., 2008. A discrete–continuous model of households’ vehicle choice and usage, with an application to the effects of residential density. Transportation Research Part B, 42, 736–758. Federal Communications Commission [FCC]..National Broadband Plan. United States. Retrieved from http://www.broadband.gov/plan/ Federal Highway Administration (2011). Summary of travel trends: 2009 national household travel survey. Retrived from http://nhts.ornl.gov/2009/pub/stt.pdf Frondel, M., & Vance, C. (2010). Driving for fun? Comparing the effect of fuel prices on weekday and weekend fuel consumption. Energy Economics, 32, 102-109. Fuhr, J. P., & Pociask, S. (2007). Broadband services: Economic and environmental benefits. Retrieved from http://www.theamericanconsumer.org/2007/10/31/broadband-services-economic-and-environmental-benefits/ Fuhr, J. P., & Pociask, S. (2011). Broadband and telecommuting: helping the U.S. environment and the economy. Low Carbon Economy, 2(1), 41. Hoechle, D. (2007). Robust standard errors for panel regressions with cross-sectional dependence. Retrieved from http://fmwww.bc.edu/repec/bocode/x/xtscc_paper.pdf Greene, W. H. (2003). Econometric analysis. Prentice Hall. Hair, J. F. Jr., Anderson, R. E., Tatham, R. L. & Black, W. C. (1995). Multivariate Data Analysis (3rd ed). New York: Macmillan. Hymel, K. (2014). Factors influencing vehicle miles traveled in California: Measurement and analysis. Retrieved from http://www.csus.edu/calst/FRFP/VMT%20Trends%20-%20Hymel%20-%20Final%20Report.pdf Karathodorou, N., Graham, D. J., & Noland, R. B. (2010). Estimating the effect of urban density on fuel demand. Energy Economics, 32(1), 86–92. doi:10.1016/j.eneco.2009.05.005 Katsumoto, S. (2002). Information and communications technology and the environment: An Asian perspective. Journal of Industrial Ecology, 6(2). Klein, T. Quantitative research methods session 5: Precision of OLS estimators, multiple regression models, multicollinearity, F-tests for goodness of fit [PDF document]. Retrieved from http://www.klein.co.uk/downloads/MPhil/Michaelmas/lecture5.pdf Lowy, J. (2012). More women drivers than men on U.S. roads now. Associated Press. Retrieved from http://www.usatoday.com/story/money/cars/2012/11/12/women-drivers-men-licenses-roads/1700185/ McMahon, K., Thomas, R. L., & Kaylor, C. (2012). Broadband, Sustainability, and the Environment. Planning Advisory Service Report, 41. McManus, P. A. (2011). Introduction to regression models for panel data analysis [PowerPoint slides]. Retrieved from http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CDgQFjAA&url=http%3A%2F%2Fw ww.indiana.edu%2F~wim%2Fdocs%2F10_7_2011_slides.pdf&ei=tv2yUOyCGeSzygHsooH4Bw&usg=AFQjCNGU LHQVllr-AGSAbZYx2fOrP6s2_A&sig2=Nv6lONqmab5fjYsfd10mhw Mokhtarian, P. (2002). Telecommunications and travel. Journal of Industrial Ecology, 6(2).
14
PTC’15 Networked Planet
Newman, P., & Kenworthy, J., (1999). Costs of automobile dependence: global survey of cities. Transportation Research Record 1670, 17–26. Noland, R., (2001). Relationships between highway capacity and induced vehicle travel. Transportation Research Part A, 35, 47–72. Nordhaus, W. D., (1977). The demand for energy: an international perspective. In: Nordhaus, W. D. (Ed.), International Studies of the Demand for Energy. North-Holland Publishing, Amsterdam. Pickrell, D., & Schimek, P. (1998). Trends in personal motor vehicle ownership and use: Evidence from the NPTS. 1995 Nationwide Personal Transportation Survey. Schipper, L., Steiner, R., Figueroa, M., & Dolan, K. (1993). Fuel prices and economy: Factors affecting land travel. Transport Policy, 1(1), 6-20. Simma, A., & Axhausen, K. (2001). Structures of commitment in mode use: A comparison of Switzerland, Germany and Great Britain. Transport Policy, 8. Small, K. & Dender, K. (2007). Fuel Efficiency and Motor Vehicle Travel: The Declining Rebound Effect. Energy Journal, 28(1), 25-51. Su, Q. (2011). The effect of population density, road network density, and congestion on household gasoline consumption in U.S. urban areas. Energy Economics, 33(3), 445–452. doi:10.1016/j.eneco.2010.11.005 Teppayayon, O., Bohlin, E., & Forge, S. (2009). Will broadband networks make the world greener? Evaluating pros and cons of broadband development. Communication & Strategies. The World Bank. Road sector gasoline fuel consumption per capita. Retrieved from http://data.worldbank.org/indicator/IS.ROD.SGAS.PC United States Environmental Protection Agency [USEPA]. Sources of greenhouse gas emissions. Retrieved from http://www.epa.gov/climatechange/ghgemissions/sources/transportation.html United States Department of Transportation [USDOT]. U.S. energy consumption by the transportation sector. Retrieved from http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/html/table_04_0 4.html_mfd Uusitalo, L. (1983). Determinants of gasoline consumption. Journal of Economic Psychology, 4(1), 149. Washington State Department of Transportation (2010). Vehicle Miles Traveled (VMT) statewide forecast model. Retrieved from http://www.wsdot.wa.gov/NR/rdonlyres/380A1F61-EC09-478D-990C-4AA9B9292AFE/0/VMTForecastWorkGroup SummaryMay2010final.pdf
15
