METHODOLOGY

• Home Footprint
• Estimating Energy Use
• Intelligent Bill Input
• Carbon Emissions
• Travel Footprint
• Personal Vehicles
• Flights
• Public Transports and Taxis
• Work Footprint
• "I work at home" or "I am unemployed"
• "I work in a building that doesn't manufacture stuff"
• "I work in a building that manufactures stuff"
• "I am a student or teacher"
• Shopping Footprint

Home Footprint

In cases where a user's energy bill history is not available, we adopt the Lawrence Berkeley National Labs Home Energy Saver weather tape model wherein the country is divided into 285 unique climatic regions. We assign users to a weather tape based on the home zip code they provide on signup. Each weather tape contains a default average energy use for home appliances, home lighting, home heating, home water heating, home cooling, and miscellaneous electricity use. The weather tape model effectively accounts for regional climatic variations as well as any regional variations in energy use habits that may be present. It alone is insufficient to produce a reasonable estimate of energy use and carbon emissions, however, since there are many other variables that affect energy use.

To produce a viable personalized energy use calculation in the absence of bills, we use the weather tape average as a starting point, and proceed to scale the default estimates based on a number of different factors. Perhaps most significant is housing type: we ask users whether they live in an apartment in a large building (more than 4 units), an apartment in a small building or townhouse, a freestanding house, or a mobile home, and scale electricity use and home fuel use accordingly (with slightly different scaling factors for each fuel based on Department of Energy Residential Energy Consumption Survey (RECS) data. We also ask what fuels are used for home heating, water heating, clothes drying (if applicable), and the user's oven. The estimated energy use for each category is scaled differently based on the fuel type to reflect the average efficiency of different heating and water heating equipment (e.g. natural gas furnaces tend to have a higher annual fuel utilization efficiency [AFUE] than oil furnaces). We further scale fuel and electricity use for heating, water heating, and cooling by asking users the age of their equipment and adjusting accordingly. Home heating and cooling requirements can be further refined by asking the age of the user's house, and using that information combined with the housing type and weather tape to create a proxy estimate for insulation values. Energy use for cooling is scaled by asking the cooling equipment type (central AC unit, window AC(s), electric heat pump, fan, or no cooling).

Home energy use can be further refined by asking if the user has a garage (heated or not heated), attic, or basement. The square footage of the household also provides an important scaling factor for home energy use, and the number of rooms can serve as a proxy for square footage in cases where the user may not know the former. In cases where users chose not to input room number or square footage, we create a default estimate based on housing type. The number of members of the household also scales energy use independently of the square footage based on a regression analysis of RECs microcensus data for each region of the country. Finally, home electricity and fuel use is scaled based on whether or not the user works at home, as people who work at home tend to use considerably more energy on weekdays than people who work away from home.

These questions should be structured in a format that minimizes the requirement that users must input all of these values. To that end, the only question that is essential for users to answer is the housing type. We fill in all other questions with the defaults most applicable for the user's weather tape and housing type which, in most cases, are natural gas for home heating and water heating, a central AC unit, electricity for clothes drying, a standard square footage for each housing type, etc. This approach allows users to get a ballpark estimate of home electricity and fuel use with minimal inputs, and refine these inputs by answering numerous additional questions. Ultimately, however, the best estimate of home energy use will have some remaining uncertainties, and is not a substitute for obtaining monthly electricity, natural gas, and fuel oil bill information for users.

Using energy bills is somewhat complicated due to potentially strong annual variation in home energy use. While this is not a serious issue when a full year of past energy bills are available, this may not always be the case, especially for users who have recently moved or when users are manually inputting bills instead of simply providing us with their utility account number. We have developed an smart bill calculator that requires only a single month's bill input (though it allows for multiple months) and, based on the user's state of residence and heating and cooling equipment and fuel types, estimates annual electricity, natural gas, and/or fuel oil use. For example, a user with a window AC unit that lives in Texas would likely have higher summer electricity use than winter electricity use, and our smart bill calculator takes this into account when estimating the annual bill if the user inputs a summer month. Likewise, a user with a natural gas furnace in, say, New York would have up to an order of magnitude larger natural gas use in the winter than in the summer, and a large natural gas bill during the winter would yield a reasonable annual use estimate based on our model.

The model also assigns the bill intelligently across the user's home energy use categories. For example, if a user has a natural gas furnace, water heater, and dryer, the natural gas bill is divided across the three based on an analysis of the average natural gas use division for the region of the country in which the user resides. In the Northeast, for example, a proportionally larger share of energy tends to be used for heating than water heating, while the opposite is true in the South. Electricity use is somewhat less precisely modelable, simply because there are so many distinct appliances that use electricity. For the time being, we simply scale our electricity user simulations for all the home categories (heating, water heating, cooling, appliances, lighting, and misc.) by the ratio of the observed bills to the simulated bills.

To effectively translate bills into kilowatt hours, therms of natural gas, and gallons of fuel oil used, we need up-to-date energy price data for each user. For electricity, since this differs on the utility level, we need a way to assign each user to a specific utility. We created a database of all of the utilities serving each zip code in the country, and populate a list of potential utilities for each user based on their home zip code. When they select a utility, we look up the latest monthly rate when it is available (as the Department of Energy's Energy Information Agency (EIA) only publishes monthly rates for about 500 of the 3500 utilities in the country, though they include most of the largest regulated ones). If a monthly rate is not available for the user's specific utility in the past three months, we use the latest monthly average rate for the user's state as a proxy. For natural gas and fuel oil, state-level price data is taken from the EIA for the lastest month.

Because local generation sources are connected to the larger grid, it is impractical to determine an individual's electricity fuel mix based on their proximity to specific generators. Rather, our footprint calculator uses NERC subregion level emission factors based on fuel mix and generation efficiency data from the EPA's eGRID. Emission factors also include transmission losses based on data from the EIA and indirect emissions associated with the fuel-cycle, plant construction, and plant decommissioning of natural gas, nuclear, oil, coal, solar, wind, biomass, geothermal, and hydro. Estimates of fuel cycle and plant construction and decommissioning emissions are based on P.J. Meier's "Life-Cycle Assessment of Electricity Generation Systems and Applications for Climate Change Policy Analysis" (2003). Direct emissions from home natural gas and fuel oil use are calculated based on emission factors from the EPA and estimated fuel-cycle emissions from Meier (2003).

Travel Footprint

To determine a user's travel footprint, we ask questions about the user's personal vehicles, flights, vehicle rentals, taxis, and public transportation. For personal vehicles, users input the year/make/model of the car into a database from the EPA's National Vehicle and Fuel Emissions Laboratory that provides the car's fuel efficiency in miles per gallon. Dividing the annual mileage of the car by the average fuel efficiency in miles per gallon yields gallons of gasoline consumed. The calculator then divides the gallons of gasoline by the average number of passengers in the car to yield per person gallons of gasoline. The number of gallons used per year is converted to pounds of carbon dioxide using conversion factors from the Technical Guidelines Voluntary Reporting of Greenhouse Gases (DOE, 2006). For users who know their own vehicles actual miles per gallon, they can choose to overwrite the default fuel economy of their vehicle with their actual fuel economy. This number (in miles per gallon) simply replaces the value assigned from the EPA year/make/model database.

Users can also input their annual number of short flights (0 to 300 miles), medium flights (301 to 1000 miles), long flights (1001 to 3000 miles), and flights outside the US (extended flights, over 3000 miles). To convert the number of flights into carbon dioxide emissions, we assign an average length in miles to each class of flights: short flights are 200 miles, medium flights 700 miles, long flights 2000 miles, and extended flights 5500 miles. In addition, for each flight class there is an emissions factor in pounds of carbon dioxide per flight mile derived from the World Resources Institute, GHG protocol initiative. By multiplying the average flight length by the emissions factor, and summing for all the flights, the calculator derives the "flying" component of the Travel footprint.

Users can further refine the "driving" component of the Travel footprint by describing the number of days she rents a car each year, and specifying what type of car is typically rented (choices are small car, midsize/sedan, minivan, SUV/pickup, hybrid SUV, and hybrid car). To calculate the associated consumption of gasoline, we take the number of rental car days and multiply by an average daily driving load of 50 miles (number based on rental packages from various rental car companies). This yields annual rental car miles. The calculator then divides by the average fuel efficiency for a car in the class (derived by observational studies of EPA mileages of various cars in the class) to yield annual gallons of gasoline consumed for rented cars.

The user can also refine the Travel footprint by answering questions to define the "other" component. Specifically, the user can input how much she spends on busses/taxis/commuter trains/subways, train travel between cities, and ferries/water taxis. For each of these three categories, there are corresponding multiplication factors that relate user-inputted dollars spent to both emissions of carbon dioxide based on data from Carnegie Mellon University Economic Input-Output Lifecycle Assessment (EIOLCA) program. By multiplying the dollars spent by the respective EIOLCA multiplication factor, and summing across the three spend categories, the calculator determines the "other" component of the Travel footprint.

Work Footprint

The Work footprint is calculated in a number of different ways based on the user's occupation. Users get to choose from the following:

• "I work at home."
• "I work in a building that manufactures stuff."
• "I work in a building that doesn't manufacture stuff."
• "I am a student or teacher."
• "I am unemployed."

Based on the user's response, she is directed down one of a number of paths, described below. The user is also asked to indicate the zip code in which she works, since some users may live in one zip code and commute to work in another.

For both of these responses, a user's work footprint is zero. An unemployed user does not work, so by definition must have a work footprint of zero. For a user that works at home, the fuel consumed in the course of this work will be included in the bills entered in the Home function questions, and will thus be part of the Home function. In cases where users do not enter bills, the default home energy use simulations are scaled to estimate extra energy use associated with working at home.

If a user indicates that she works in a non-manufacturing commercial field, she is prompted to describe the type of building she works in with the following choices: school, supermarket or grocery store, restaurant, hospital, doctor or dentist office, hotel or motel, retail store, professional or administrative office, social space, police or fire department, place of religious worship, post office or copy center, dry cleaners/laudromat/beauty parlor, auto service or gas station, warehouse or storage facility. Each of these responses corresponds to one of the building types described in the EIA's Commercial Building Energy Consumption Survey (CBECS, 2003). This survey provides per worker electricity and natural gas consumption for each of these building types.

CBECS also assigns average per worker consumption of electricity and natural gas based on the census of the commercial building. A census is a geographical division, with nine censuses in the nation, each consisting of a varied number of states with a similar geography. For each census, we derived a multiplication factor that relates average consumption of electricity and natural gas to average consumption for the entire nation. As such, when a user reports his state, we can assign him to a census and multiply the per worker consumption based on his building type by the census multiplication factor. This outputs a census- and building-modified per worker consumption of electricity and natural gas. Since these are the only required inputs, we can convert these physical units of fuel to emissions of carbon dioxide and energy consumption using the same NERC subregion-level multiplication factors described earlier in the Home function.

Although these questions are enough to output an estimated Work footprint, the user will be able to refine his Work footprint by answering any or all of the following six questions:

• The square footage of the building
• The age of the building
• The number of floors
• The number of people working in the building
• The hours of operation for the building
• The building's exterior material

The CBECS survey provides per worker consumption of electricity and natural gas for workers in the different building characteristics outlined in each of these. For each response we generate a multiplication factor that relates the building type with the overall average, and then multiply it by the census- and building-modified per worker average. Since these are independent multiplication factors, we can just sequentially multiply by them in any order. Moreover, if a user does not know the response to a question, or leaves it blank for any other reason, we simply do not multiply by any factor and the per worker consumption does not change.

If a user selects that they work in a building that manufactures things, she is then prompted to describe the manufacturing subsector of his facility. The choices for this question are: food, beverage and tobacco products, textile mills, textile product mills, apparel, leather products, wood products, paper, printing-related support, petroleum and coal products, chemicals, plastics and rubber products, nonmetallic mineral products, primary metals, fabricated metal products, machinery, computer and electronic products, electrical equipment, transportation equipment, furniture and related products, miscellaneous. Each of these categories corresponds to a subsector in the EIA's Manufacturing Energy Consumption Survey (MECS, 2002). MECS gives the total consumption, consumption per employee, electricity consumption, and natural gas consumption, broken down by region (there are four regions in the nation, and each comprises at least two censuses). From this data, we can derive per worker electricity and natural gas consumption for each region, and assign the user to one of the regions by knowing the user's work state. We then adjust the per worker numbers to account only for non-process consumption. In other words, we do not assign to the user the electricity and natural gas that is used in the manufacturing process, but only the electricity and natural gas that is used for the benefit of the facility's workers, such as for HVAC, lighting, on-sight transportation, etc. Thus, with only the worker's state and subsector, we can output per worker consumption of electricity and natural gas along with the overall Work footprint that is the sum of these two.

As with other footprint components, a user can return and refine her Work footprint by answering two more questions about her manufacturing job. Specifically, within certain subsectors, there are more specific industries. For instance, if a user selects the subsector "food", she may refine his industry to wet corn milling, sugar, fruits and vegetable canning, or I don't know/none of these. By selecting an industry, a user is assigned to a more specific category on the MECS survey, although the same data is available for the industry and it is manipulated in the same way. If a user selects "I don't know/none of these", we simply carry the calculation forward with the data from subsector rather than the more specific industry data Not all subsectors have industries within them, so for those subsectors there is not question asking for the specific industry.

Lastly, a user is also asked to describe the number of workers in her manufacturing facility. From the MECS survey, we can generate multiplication factors within each industry and subsector relating consumption for each facility size to the average consumption across all facilities. So, if a user is able to select the facility size, we can multiply the consumption of electricity and natural gas by this multiplication factor to further refine the Work footprint. Once again, we can convert to carbon dioxide emissions using the NERC subregion-level conversion factors used above.

If the user selects this statement, they are further asked to clarify whether they are a student or teacher, and in what level of schooling (kindergarten, elementary school, middle or high school, college or graduate school). Based on the response to this question, there a few pathways we can take.

If a user is a teacher in kindergarten through high school, they are actually treated in the same way as those users who "work in a building that doesn't manufacture stuff." In this pathway, outlined above, the user is normally prompted to describe his building. However, in the case of teachers, we can assign the building type to "school." Using this response, and the work state, we can utilize CBECS data to yield per worker consumption of electricity and natural gas.

In addition, as with the non-manufacturing questions outlined above, the user can refine his footprint by answering questions to describe the school's square footage, construction year, number of floors, number of employees, weekly operating hours, and exterior wall material. The resulting CBECS-derived multiplication factors can refine the user's Work footprint.

This pathway also utilizes the same CBECS pathway utilized above and in non-manufacturing buildings. However, that data outputs electricity and natural gas per employee, so our calculator adds another multiplication factor to the student pathway which accounts for the larger number of students as compared to just workers. This larger number will decrease the per student consumption of electricity and natural gas, as the total consumption is spread out over a wider range of students. Here a conscious decision is made to assign less consumption to students than teachers, as teachers are assigned per worker values, while students are assigned a value that is per (worker + student). This decision was made because students spend less time in the school than teachers do, and have a less direct financial stake and smaller choice to be in the school in the first place.

As above, we take the CBECS data for education buildings in the appropriate census division (based on user state). Now, the calculator multiplies by a factor relating number of worker to total number of workers and students. This factor is derived from the National Center for Education Statistics, which provides student to teacher ratios for kindergarten, elementary school and secondary school, as well as student to administrative staff ratio, all broken down by state. By combining these data, the calculator derives a ratio of workers to workers and students, which when multiplied by the per worker electricity and natural gas consumption, providing electricity and natural gas consumption per workers plus students. These outputs, electricity and natural gas, are the subcategories for a student's footprint, and when summed, provides the overall Work footprint.

As with the non-manufacturing questions outlined above, the user can refine his footprint by answering questions to describe the school's square footage, construction year, number of floors, number of employees, weekly operating hours, and exterior wall material. The resulting CBECS-derived multiplication factors can refine the user's Work footprint.

Students and teachers in college or grad school are treated as equals, in contrast to students and teachers at any other level of schooling. The reason there is no difference between students and teachers in college relates to the fact that both spend comparable amounts of time in the school buildings, and both choose to be in the buildings for either current employment or training for potential future employment. In this category, our team researched published emissions inventories from dozens of colleges in the United States, inventories that took into account all buildings on a university campus. These college reports were grouped into four regions, and the average carbon dioxide emissions per community member at the college was calculated. As such, a student or teacher in college or graduate school is assigned one of these average footprints, which are subsequently broken down into the subcategories of electricity, on-campus sources, and other.

Shopping Footprint

The Shopping footprint is meant to capture the indirect emissions associated with the manufacture and distribution of the products we purchase on a daily basis. To break down a typical user's spending into discrete categories, our calculator begins with 2005 consumer spend data from the U.S. Bureau of Labor Statistics (BLS), which details average spending by Americans in 13 broad categories:

• Food and alcohol, which includes food at home, food away from home, and alcoholic beverages
• Housing, owned dwellings, rented dwellings, other lodging, utilities fuels and publics services, household operations, household supplies, household furnishings and equipment
• Apparel and services
• Transportation, which includes vehicle purchases, gasoline and motor oil, other vehicles expenses, and public transportation
• Healthcare
• Entertainment
• Personal care products and services
• Reading
• Education
• Tobacco products and smoking supplies
• Miscellaneous (not included)
• Cash contribution (not included)
• Personal insurance and pensions, which includes life and other personal insurance and pensions and social security

For each of these categories, the description from the BLS survey was used to assign a reference category from Carnegie Mellon University's EIOLCA program. This process provides multiplication factors to convert the dollars spent in each of these categories to the corresponding emissions of carbon dioxide and energy consumption. Certain categories were omitted: utilities fuels and publics services were omitted because these are included in the Home footprint, education was omitted because it is included in the student's work footprint, gasoline/motor oil and public transportation were omitted because these are included in the Travel footprint, and miscellaneous and cash contributions were omitted because of difficulty in defining these for the user and in assigning an EIOLCA reference category.

In order to derive a Shopping footprint, the calculator multiplies the amount spent in each spend category by the corresponding EIOLCA multiplication factor and a value to adjust for inflation based on the BLS Consumer Price Index. To assign spending in each category without asking the user, the calculator utilizes data from the BLS survey, which provides average consumer spending for each of these categories, broken down by income range of the consumer. This is based on the user's household's combined annual income or, when not provided, the U.S. average household income for 2005. Based on the user's reported income, we can assign the average spending for her family in each of the spend categories.

The user is also asked to report whether she is a vegetarian, vegan, or omnivore. We use BLS survey data to estimate food expenditure in each major food category (cereals and breads, chicken and fish, red meat, dairy products, fruits and vegetables, and sugars and sweets) based on income level. We derive the estimated calories consumed for each food type based on the average calories per dollar for that food type based on data from Weber and Mathews (2008). For users who are vegan, we replace all red meat, chicken/fish, and dairy calories with an equal division of grains and breads and fruits and vegetable calories. For vegetarian users, we divide red meat and chicken/fish calories equally between fruits and vegetables, grains and breads, and dairy. The emission factors per calorie and per inflation-adjusted dollar are taken from Weber and Mathews (2008).

If a user chooses to refine her Shopping footprint, she can describe the specific amount of spending in each of the subcategories. There is also an additional subcategory, credit card spending, which is incorporated into the miscellaneous subcategory since purchasing any product with a credit card as opposed to cash leads to additional emissions of carbon dioxide and energy consumption. To allow maximal flexibility for users, they can enter weekly, monthly, or yearly spending for each of the spend categories, and the calculator can annualize these numbers.