Implementable Strategies for Shifting to Direct Usage-Based Charges for Transportation Funding (2009)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

9 2. FORECASTING VMT FEES If one potential benefit of making a transition to VMT fees is maintaining and increasing revenues to support transportation, an essential first step to identifying VMT-fee options is considering the levels of revenue to be raised. As will be shown in this chapter, the growth of national VMT levels has outpaced growth in fuel consumption over the past several decades. Generally speaking, the amount of fuel-tax revenue raised per mile of travel has declined over time. To forecast future trends, the research team considered factors that may influence future VMT, available VMT-forecast data sources, and methods that states might use to forecast VMT- fee revenue. A forecast of the future revenue stream resulting from an initially revenue-neutral replacement of federal fuel taxes with VMT fees beginning in 2015 illustrates such a method. 2.1. Past Trends in VMT and Fuel Consumption Over the past several decades, as population, incomes, and the number of cars in the United States have increased, so too has fuel consumption risen. VMT, however, has grown at a much faster rate. This is due, in large part, to federally mandated corporate average fuel economy (CAFE) standards, first enacted in 1975, which have required gradual increases in the average fuel economy of cars and trucks sold in the United States (BEES 2002). As the average fuel economy of vehicles on the road has increased, the fuel tax revenue per mile of travel has decreased correspondingly. This is not the only factor to undermine fuel taxes – inflation is the other main culprit – but its effects are significant. Figure 2.1 illustrates the percentage growth in VMT and fuel consumption from 1980 to the present. As the figure demonstrates, fuel consumption has increased by about 50 percent over this period, while VMT has almost doubled. Should this trend continue, as many expect that it will, the efficacy of fuel taxes for raising sufficient transportation revenue will continue to decline in the coming years.

10 Figure 2.1. Historical Growth in VMT and Fuel Consumption -20% 0% 20% 40% 60% 80% 100% 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 Pe rc en ta g e G ro w th s in ce 1 9 8 0 VMT Fuel Consumption Sources: FHWA (2007 Table 5.2.1), ORNL (2008 Table 2.7) 2.2. Factors Influencing Future VMT Forecasting future VMT, and in turn VMT fee revenue, is inherently difficult. Aggregate travel is affected by a broad range of factors, including the health of the economy, the size of the population, the price of fuel, the average fuel economy of the fleet, and the supply of roadways (which in turn relates to the level of congestion on the roads). Some of these factors can be predicted with relative confidence; for example, population trends are relatively stable and are not subject to major sudden swings. On the other hand, oil prices are notoriously difficult to predict, and have historically fluctuated greatly over short time periods. This makes it quite challenging to accurately forecast future VMT. Forecasting VMT is further complicated because the relative importance of these factors may vary depending on the type of travel. For example, passenger VMT may be more sensitive to changes in the price of fuel, while truck VMT may be more sensitive to changes in the economy. Commute trips may be less elastic than shopping trips, because people have less leeway to forego trips to work even if gas prices rise considerably. Since not all VMT reacts in the same way to changes in the underlying factors, it can be hard to predict the overall impact. Finally, there could be major paradigm shifts over the longer-term future that change these relationships in ways that would be difficult to predict. For years it was almost an article of faith

11 that VMT rose annually—until 2008, when oil prices spiked and VMT declined for the first time since records have been kept. This added new data to our understanding of the relationship between fuel prices and VMT, which has historically been relatively inelastic. Similarly, while most retired people drive considerably less than when they worked, the Baby Boom generation is the first to grow up with widespread dependence on automobiles, and they are living longer and healthier lives than their predecessors Predicting their behavior based on that of previous generations may thus prove misleading. The imposition of VMT fees—or, alternatively, more aggressive pricing of carbon emissions—would also change the cost of travel, in turn producing an entirely new variable that would affect VMT. The degree of the effect would depend on the specific rate structure—a policy decision rather than an independent factor. It is well beyond the scope of this research to assess such potential paradigm shifts, but these examples imply that forecasts based on previous experience are uncertain at best. The preceding points suggest three important considerations relevant to the forecasting of future VMT-fee revenue. • First, the base VMT estimates should account for, among other possible factors, changes in population, in the economy, in fuel efficiency, and in the price of fuel. • Second, given inherent uncertainty in the underlying factors that influence VMT, reliance on point estimates of future VMT is not advisable. Rather, it would be helpful to examine alternate future scenarios to understand the potential range of future VMT, and in turn the range of revenue. • Third, for any given forecast of future VMT, it is important to include a feedback mechanism to understand how the imposition of VMT fees may change the cost of travel, and in turn total VMT. This may be less crucial for a simple per-mile charge set to replace fuel taxes on a revenue-neutral basis, but it could have a considerable effect, for example, if mileage-based congestion tolls were widely applied (these could significantly alter the cost of travel, and in turn VMT and revenue). 2.3. VMT Forecast Data Sources This research does not develop VMT forecasts, but rather analyzes existing forecasts. Two federal agencies prepare national forecasts of VMT: the Federal Highway Administration (FHWA) and the Energy Information Administration (EIA). As discussed above, long-term forecasting is risky and there is no guarantee that these VMT forecasts, used in this research to develop estimates of future VMT-fee revenue, will prove to be accurate. But these represent the most authoritative VMT forecasts available, and the federal government currently relies on these forecasts to support policy decisions and revenue projections. FHWA VMT Forecasts. FHWA develops 20-year VMT forecasts based on data from the Highway Performance Monitoring System (HPMS), which samples traffic flow on about 116,000 road segments throughout the country. States are responsible for reporting sampled data on the segments within their jurisdiction and forecasting how the traffic flows on those segments will change in future years. FHWA then integrates current and future flow predictions on the segments and uses the information to forecast future VMT across the entire road network. The

12 most recent forecasts from FHWA, available in the 2006 Conditions and Performance Report (FHWA 2006), are several years out of date. While FHWA typically provides its forecasts aggregated for over all vehicles at the national scale, it has in the past provided estimates broken down by state and classified by urban and rural roads (GAO 2002). When considering the rigor of FHWA VMT forecasts, there are several important points to make. First, the road segments included in HPMS may not constitute a representative sampling of all roads. While the segments remain constant, sample bias can creep in over time if traffic volumes on some segments increase or decrease more rapidly than the average. States are supposed to update samples if they are not capturing a stratified random sample of functional classes and traffic volume. There is thus some difficulty in interpolating between travel on the sampled segments and travel across the entire road network. Second, FHWA does not provide strict guidance to the states in terms of the methodology for forecasting future travel on the road segments, except to state that they “should come from a technically supportable State procedure or data from MPOs or other local sources [and that] HPMS forecasts for urbanized areas should be consistent with those developed by the MPO at the functional system and urbanized area level” (FHWA 2005, p. IV-38). While some states may develop sophisticated models that incorporate predicted changes in population, the economy, fuel prices, and fuel economy, others may simply perform linear extrapolation based on recent trends. These issues make it difficult to suggest, as a general rule, that all states could rely on HPMS data to develop accurate statewide VMT forecasts. EIA VMT Forecasts. EIA also develops VMT forecasts, as part of an integrated suite of models (the National Energy Modeling System, or NEMS) aimed at predicting future energy consumption and prices. EIA’s forecasts are informed by FHWA numbers, but also take into consideration predicted trends in population, the economy, fuel prices, and fleet-wide fuel economy. Moreover, the influence of these factors within the models is allowed to vary for different types of vehicles (e.g., trucks vs. passenger cars), as theory would suggest. Despite this rigor, there are two challenges to the use of EIA’s VMT predictions within the context of forecasting VMT revenue at the state level. First, though EIA’s VMT forecasts are broken down by different vehicle classes (light-duty household vehicles, fleet vehicles and freight trucks), they are not disaggregated to the state level; rather, they are only available at the national level. Second, although EIA may evaluate different modeling scenarios during the analysis stage, they routinely publish point estimates for future VMT rather than ranges (EIA 2009). In addition to VMT forecasts, NEMS is also used to forecast fuel consumption. Figure 2.2 graphs EIA’s NEMS forecasted growth for both VMT and fuel consumption in percentage terms from a base year of 2008. Note that despite the distinctions in methodology, FHWA’s HPMS-based VMT forecasts are roughly similar to those of EIA. A key distinction, however, is that the most recently published FHWA data are several years out of date and thus do not account for the recent downturn in fuel consumption resulting from the spike in fuel prices followed by the severe economic recession. As a result, FHWA VMT forecasts are slightly higher than EIA VMT forecasts, though they follow a parallel trajectory. Figure 2.2 shows only the EIA forecasts, as they are more recent.

13 Figure 2.2. EIA Forecasted Growth in VMT and Fuel Consumption -20% 0% 20% 40% 60% 80% 100% 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 Pe rc en ta g e G ro w th F o re ca st f ro m 2 0 0 8 VMT Fuel Consumption Sources: EIA (2009 Tables 45, 60, 65, and 67) Comparing the two forecasts in Figure 2.2, one can see that growth in VMT is expected to continue to outpace growth in fuel consumption over the forecast horizon. This reflects the expectation that more fuel-efficient conventional vehicles, along with alternative fuel options, will achieve greater market penetration in the years to come. 2.4. Options for Forecasting State VMT-Fee Revenue The process of forecasting VMT-fee revenue can be broken down into two stages: forecasting VMT, then applying a VMT-fee rate (which may in turn influence VMT). The research team first considered several options for developing state VMT forecasts using the data sources discussed above, and then applied VMT fees to forecast revenue. Leveraging the above data sources, there are several methodologies that states could pursue to forecast VMT at the state level. The forecasting methodology should, if possible, account for the range of factors that may influence VMT, and also allow for the generation of multiple future scenarios to reflect uncertainty. • HPMS-based forecasts. States with sophisticated modeling capacity could utilize HPMS data as a basis for developing their own VMT forecasts. Currently, states are required to

14 forecast VMT on the specific sampling segments within the HPMS system, not for the road network as a whole, so this would constitute an extension of current practice. Though representing a significant amount of work, an advantage here is that states would be able to perform multiple modeling runs to examine whatever scenarios they would like to consider. • Extrapolation from national EIA forecasts. The idea here would be to begin with EIA national forecasts and then extrapolate state forecasts based on the current percent of national VMT that occurs within each state. The current breakdown of VMT by state can be found in Table VM-2 of the Highway Statistics Series (FHWA 2007). As an example, 2007 data indicate that of 3.049 trillion VMT nationwide, Alabama accounted for 61.41 billion VMT, or 2.03 percent of the national total. To derive future VMT forecasts in Alabama, then, the state would simply multiply EIA’s national forecasts by 2.03 percent for each year over the forecast horizon. To enable states to consider alternate future scenarios, EIA could publish multiple forecasts based on different assumptions about future changes in population, the economy, fuel prices, and fleet-wide fuel efficiency. If this is not possible, states could simply make upward or downward adjustments by, say, 10 percent to examine the revenue implications. Note that inherent in this approach is the assumption that the share of national VMT for a given state will remain roughly constant in the coming years. If states have reason to believe that their share of VMT will either increase or decline relative to the nation as a whole, they could choose to examine either more optimistic (from a revenue perspective) or pessimistic VMT forecast scenarios. • Forecasting state VMT through use of EIA models. A final option that might be considered would be for EIA to share its modeling structure with states so that they could perform their own analyses at the state level. (The research team did not explore whether this would be possible). The models are complex, so this would require considerable investment on the part of states; on the other hand, this would also provide states with significant modeling flexibility and rigor. Once VMT forecasts have been generated, the next step is to predict future VMT-fee revenue. This must account for two factors: (a) how the VMT fee will be structured, and in turn how that will influence the cost of driving, and (b) how VMT will be affected by changes in the cost of driving. The latter can be determined by examining the elasticity of travel demand with respect to changes in the cost of driving. Considerable research on this question has been conducted; see, for example, the reviews presented by Litman (2008) and Goodwin, Dargay, and Hanly (2004). One useful measure is the elasticity of VMT with respect to changes in the price of fuel, which may be used to evaluate, for instance, the effect on VMT that would result from increasing or decreasing fuel taxes (see, for example, Sorensen 2006). In similar fashion, the effect of replacing fuel taxes with VMT fees could be approximated as a change in the price of fuel. Based on our calculations (described in more detail at the end of this chapter), it appears that an initially revenue-neutral replacement of fuel taxes with flat VMT fees would have very little dampening effect on VMT over the forecast horizon, even though VMT is expected to grow faster than fuel consumption during this period. This is because changes in the price of fuel have only a modest effect on changes in VMT (our review of the research suggests that the expected elasticity is about -0.29 over the long term), and fuel taxes only account for a small share of the price of fuel. On the other hand, if the VMT fee structure included some form of congestion

15 pricing, this could increase the cost of travel significantly, and the effect on VMT would likely be much more dramatic. Calculating such impacts would have required much more detailed modeling, beyond the scope of the current research project. 2.5. A Simple Illustration of VMT Revenue Forecasting To illustrate the potential effects of replacing fuel taxes with VMT fees, the research team developed several forecasts of federal VMT-fee revenue and compared them to forecasted motor fuel tax revenue. The work began with setting the VMT fee for the first year, 2015 (the near-term transition year envisioned in this research) to a level that would generate revenues for that year equal to those forecast for motor-fuel tax revenues. Based on EIA fuel-consumption forecasts for 2015, federal revenue from motor fuel taxes of 18.4 cents per gallon for gasoline and 24.4 cents per gallon for diesel would total about $35.7 billion in 2015. (All dollar values in this section are in unadjusted 2009 dollars.) Assuming that all vehicles would pay the same per-mile rate under a system of VMT fees, the fee would need to be set at 1.1 cents per mile ($35.7 billion divided by 3.23 trillion VMT in 2015) to be initially revenue neutral. Forecasts of future VMT revenues were then made under four scenarios: • VMT fees remain at 1.1 cents per mile. • VMT fees remain at 1.1 cents per mile, but VMT grows at a rate 10 percent lower than projected by EIA. • VMT fees remain at 1.1 cents per mile, but VMT grows at a rate 10 percent higher than projected by EIA. • VMT fees are set at 0.8 cents for cars and 3.4 cents for trucks, reflecting the differences in current contributions to fuel tax revenues. Obviously, many more scenarios would be possible. For example, the base per-mile rate might be increased in future years, or additional forms of pricing – varying the per mile rate by time and location, or varying the rate for trucks based on vehicle weight – might be introduced. Such scenarios would, however, have required much more complex modeling beyond the scope of the study. The research team therefore focused on VMT-fee scenarios with a simple rate structure that remains constant over time. EIA fuel consumption forecasts were also used to project motor fuel tax revenues. Here two scenarios were created: • Fuel taxes remain at current levels over the entire forecast period • Fuel taxes are increased by five cents per gallon (for both gasoline and diesel) in 2015 and then remain constant over the forecast period (note that the value of five cents was selected, after some trial and error calculations, to achieve roughly the same revenue as a 1.1-cents- per-mile VMT fee by 2030). Comparisons of the four VMT fee and two fuel tax revenue forecasts are provided in Table 2.1 and Figure 2.3. The analysis shows that in the absence of a fuel tax increase, an initially revenue-neutral switch to VMT fees will likely produce approximately 20 percent more revenue by 2030, even if growth in VMT is 10 percent lower than the median projected. Increasing fuel

16 taxes by five cents per gallon (to 23.4 cents for gasoline and 29.4 cents for diesel) could produce revenue in 2030 similar to that forecast for VMT fees. Table 2.1. Illustrative National VMT-Fee Revenue Forecasts Scenario 2015 Revenue (2009 $B) 2030 Revenue (2009 $B) Growth 2015 - 2030 Motor fuel taxes at current rates $35.7 $39.2 10% Motor fuel taxes with five-cent increase $43.1 $47.2 10% VMT fees (flat fee, 1.1 cents) $35.7 $47.4 33% VMT fees (flat fee, 10% lower VMT) $35.7 $46.1 29% VMT fees (flat fee, 10% higher VMT) $35.7 $48.7 37% VMT Revenues (cars less than trucks) $35.7 $47.7 34% Sources: Computed by authors based on data from EIA (2009 Tables 45, 60, 65, and 67) Figure 2.3. National VMT Fee Revenue Forecasts Sources: Computed by authors based on data from EIA (2009 Tables 45, 60, 65, and 67)

17 As Figure 2.3 shows, a 1.1-cent VMT fee would results in approximately $7 to $9 billion in additional revenue in 2030, without adjusting for inflation. It is beyond the scope of this paper to project future expenditures, so it is difficult to say whether this would be sufficient to meet the country’s transportation needs. However, given that the needs may well be greater, and that VMT may decline due to any number of factors as discussed above, it is entirely possible that instituting a revenue-neutral shift from fuel taxes to VMT fees will not prove sufficient to meet future needs. Fees may therefore need to be raised in the future, presenting the same political challenges as raising fuel taxes. Alternatively, VMT fees could be varied – for instance, charging a higher rate for driving more heavily polluting vehicles or for driving in congested areas during peak periods – to raise additional revenue while simultaneously addressing other policy goals. The research team also calculated an additional adjustment to EIA’s forecasted VMT to account for the fact that over time, driving would become more expensive with a VMT fee than with a gas or diesel tax. VMT would most likely decrease slightly, all things being equal, as driving becomes more expensive. The calculation began with AAA’s estimate of the average fuel price in late 2008 of $2.30 per gallon. Of this, 18.4 cents, or roughly 8 percent, corresponds to the federal fuel tax (note that EIA’s fuel price forecasts suggest that this will increase only slightly – by a few cents per gallon – over the forecast horizon). To account for how the replacement of the federal fuel tax with VMT fees would affect that cost over time, 18.4 cents was subtracted out the of federal fuel taxes and added back the cost of the VMT fee, expressed in per-gallon terms given expected fleet fuel economy. This led to what might be viewed, in effect, as a slight increase in the cost of gas (more precisely, a change in the cost of gas minus federal fuel taxes plus federal VMT fees). Finally, a long run elasticity of -0.29 was applied, suggested by the studies examined by Litman (2008), to account for the degree to which this effective change in the cost of fuel would reduce VMT. In aggregate, the effect was negligible, dampening VMT and in turn VMT-fee revenue by only about 0.2 percent by 2030. Given that this change was so small, it is not incorporated into the VMT-fee revenue estimates in Table 4.1 and Figure 4.1.