Welfare and Equity Impacts of Gasoline Price Changes

The impacts on public transit ridership of changes in gasoline prices and service levels have been studied, while the combined effects of gasoline price changes under different levels of transit service have not. This paper discusses a consumer welfare calculation based on a binary mode choice model for commuters in idealized corridors with varying public transportation levels of service. Welfare losses are seen to be greatest for commuters in corridors with poor public transit options, and losses increase with rising gas prices. Low-income commuters are seen to suffer more welfare loss in corridors with low-performing transit options than in corridors with well-performing public transit systems. This simple model points to the need for more research regarding the impact of high gas prices on low-income households’ commute behavior and access to jobs. Introduction In the Phoenix metropolitan area, for most trips, door-to-door travel times by public transportation can be three to five times as long as by automobile. The transit mode share for work trips there is less than half of that of the United States Journal of Public Transportation, Vol. 13, No. 3, 2010 2 —2.3 versus 4.8 percent (U.S. Census Bureau 2005-2007). Even travelers from most of the poorest households are “captive” drivers, having few other reasonable travel options. A total of 6.2 percent of households have no vehicles, much lower than metropolitan areas of similar size, such as Baltimore (11.4%), Philadelphia (13.6%), and Boston (12.5%) (U.S. Census Bureau 2005-2007). Generally speaking, then, the increase in gas price likely represents a more significant welfare loss from the population in Phoenix than in the other cities, since the choice to avoid payment is not a reasonable option for most. An interesting set of question arises: How would losses in cities with poor public transit options compare to the losses for commuters in cities with better transit options? Would low-income and high-income commuters suffer similar losses? Would low-income households suffer less in a city with better transit options? This paper explores these questions using an existing binary mode choice model to analyze different commuting mode choice scenarios with changing gasoline prices. Before proceeding to our analysis, the context for this area of questioning is further discussed. Gas Prices and Low-Income Travel Concern for low-income workers’ access to jobs has been a central one in urban research over the past half-century, especially as it relates to metropolitan decentralization (the spatial-mismatch hypothesis), welfare reform, and access to transportation. Access to jobs is indeed found to be influenced by access to transportation. Some work emphasized the role public transit systems could have in providing needed access to overcome the spatial-mismatch problem (Sanchez 1999), while many studies questioned these conclusions. Taylor and Ong (1995) and Gurley and Bruce (2005) emphasized the importance of automobile access in explaining job accessibility, renaming the spatial mismatch as one of “automobile mismatch.” Others confirmed the “automobile mismatch” conclusion and questioned public transit’s effectiveness for job accessibility compared to the automobile (Ong and Blumenberg 1998; Wachs and Taylor 1998). Cervero et al. (2002) and Sanchez et al. (2004) found that public transit access was largely insignificant in affecting employment likelihoods for former welfare recipients. Welfare and Equity Impacts of Gasoline Price Changes 3 Key to studies: 1: Cervero et al. 2002; 2: Gurley and Bruce 2005; 3: Ong and Blumburg, 1998; 4: Ong, 2002; 5: Sanchez, 1999; 6: Sanchez et al., 2004; 7: Taylor and Ong, 1995 (Energy Information Agency, 2010) Figure 1. Dataset year for spatial mismatch studies focused on transportation mode superimposed on annual average unleaded gasoline prices, 1975 to 2009. The review of the Spatial Mismatch research by Ihlanfeldt and Sjonquist (1998) shows that none of the studies incorporated out-of-pocket costs as an element of transportation costs; costs were either spatial or temporal. Rogers (1997) found that the results for employment access predictions are sensitive to the specification of accessibility models, however. If this is the case, could a rise in gasoline prices add significantly to time costs that were thought to be the main component of travel costs? This issue may not have mattered during the periods of historicallylow gasoline prices, but prices are unlikely to remain as stable or as low in the future (EIA 2009). Figure 1 shows the dataset years for the job access research superimposed on annual average gasoline prices in the U.S. Note that gasoline prices were below $2 per gallon (2008 dollars) from 1985 to 2005, the period during which a bulk of the job access research was performed. Thus, the question should be asked: How might rising marginal costs of automobile operation affect job access? Clearly, ownership costs are significant barriers to Journal of Public Transportation, Vol. 13, No. 3, 2010 4 overcome for low-income households. Now, with rising or volatile gasoline prices, marginal costs may become more significant and affect the ability to use vehicles for commuting. This adds a new dimension to the mismatch problem of accessibility cost and may impact employment outcomes, resulting policy emphases, and the “automobile mismatch” conclusion. These issues are explored in this paper using several choice scenarios to model the impacts gasoline prices may have on commute mode choice and the welfare of low-income commuters. Study Approach Some hypotheses concerning the interaction between welfare, transit service levels, income, and fuel prices can be stated a-priori: under rising fuel prices, welfare losses in places with poor transit options will be greater than in places with good transit, low-income populations will suffer more as a share of income, and low-income households will suffer less in places with better transit options. While these conclusions may seem obvious, no studies have addressed these simple questions. In this paper, these interactions between transit service, income and fuel prices are explored by developing commute scenarios and comparing their modeled welfare changes. For an example choice utility function and representative 10-mile commute, the choice model calculates choice and welfare changes under changing gasoline prices. Three models are set up for three public transit “levels of service,” representing, loosely, a commute trip in a corridor with few reasonable transit options; a corridor with reasonable transit options compared to driving where access, travel times, and out-of-pocket costs are competitive; and a corridor where public transit access and travel times are significantly faster than driving options. First, the specific performance assumptions and choice model are presented. Next, the scenarios are evaluated for commuters of different income levels to compare how welfare impacts differ for them under the different level of transit service scenarios. Before proceeding to the scenarios, previous work concerning the interactions between welfare, fuel prices and mode choice is reviewed. Background The National Research Defense Council (David Gardiner & Associates 2007) alluded to the idea of connecting transit quality with the impacts of fuel price changes when they sought to identify which U.S. states’ drivers were most “vulnerable” to oil dependency, measured by the share of the residents’ incomes spent on Welfare and Equity Impacts of Gasoline Price Changes 5 gasoline. The most vulnerable states tended to be more rural, such as Mississippi, or had large urban areas with few public transit options, such as Georgia and Arizona. The least vulnerable states, such as New York and Massachusetts, have large cities with well-performing public transit systems. While such aggregate measures lose the detailed connection between mode choice and transportation characteristics, they point to a connection between urban form, transit quality, and gasoline price impacts on economic welfare. The two underlying issues of interest here are the interactions between gasoline price and mode choice, and the estimation of welfare changes resulting from these price and mode choice changes. The impacts of various cost factors such as parking, fuel, and transit fare prices on transit ridership are well studied (Bhat el al. 2009; Litman 2004; Mattson 2008; Taylor and Fink 2003; Wang and Skinner 1984). Cross-elasticity estimates for transit ridership due to gasoline price differ in the short and long timeframes and by type of transit technology (Mattson 2008). Estimates of short-run elasticities typically fall below 0.15, while longer-run estimates ranged from 0.12 to 0.4 (Mattson 2008). The issue of gas price effects on ridership within a context (though unspecified) of transit quality is brought up indirectly by two recent studies by Currie and Phung (2007) and Haire and Machemehl (2007). Currie and Phung (2007) estimated the ridership elasticity with respect to gas price based on national total ridership data while removing new system expansions from their dataset. They find ridership elasticities for bus, light rail and heavy rail to be 0.04, 0.27 and 0.17, respectively. Using a different approach, Haire and Machemehl (2007) estimate the same three elasticities (actually correlations) to be 0.24, 0.07, and 0.27. Instead of using national data, they focus on five large cities: Atlanta, Dallas, Los Angeles, San Francisco, and Washington, D.C. It may be difficult to determine exactly why such opposing results were found, but they do point to some interaction between transit quality and mode choice under changing fuel prices. Looking at the results for bus, the cities’ in the Haire and Machemehl (2007) study have substantial bus systems with service levels which may enable a realistic alternative for large segments of the population, resulting in a larger choice response to gas price changes. For the national data used in Currie and Phung (2007), it may be that bus systems do not, nationally, offer good choice options, and so elasticities were found to be especially low. Understanding the differences in the rail elasticities would take a more specific analysis of the systems studied. Journal of Public Transportation, Vol. 13, No. 3, 2010 6 Numerous studies have estimated welfare impacts from price and choice changes in transportation policy realms (Hau 1987; Mannering and Hamed 1990; Niskanen 1986; Small and Rosen 1981; Pines and Sadka 1984). Several studies use analytical welfare calculations to find that welfare in general falls as prices rise. Pines and Sadka (1984) developed a simple analytical urban commute model that combines gas prices and congestion tolls to show that increasing gas prices reduce welfare, and that congestion tolls should be reduced in order to remain optimal. Similar results are found for the case of Iranian domestic gasoline consumption under rising gasoline prices (Ahmadian et al. 2007). Hau (1987) looked at different transit levels of supply and their effects on consumers’ welfare, but did not test changing gas prices as an independent variable. Methodology The approach here is to model user economic welfare before and after gasoline price changes. The calculation is made for three corridor “scenarios” representing different levels of public transit service relative to automobile level of service. The welfare changes of commuters of different income levels are calculated and compared under the three scenarios. Note that this approach is not based on empirical or analytical work, but uses an existing choice model to analyze idealized choice scenarios and welfare impacts. These welfare calculations are described here, followed by the construction of the three corridor scenarios. Welfare Calculation In the microeconomic model of mode choice, consumers of transportation derive satisfaction, or “utility,” from each of the mode choices available to them. For consumer n, the utility derived from mode choice i, can be represented as Vin(Xin, Zn,), where V is called the indirect utility function, Xin are attributes of the mode and the particular trip (such as fare or travel time), and Zn are consumer’s socioeconomic characteristics (such as age or income) (Ben-Akiva and Lerman 1985). The “compensating variation” (CV) is a standard estimate of welfare change resulting from a policy change (Hanemann 1999). The logit discrete choice formulation conveniently contains the expected maximum utility derivable from a choice set through the “log-sum” (denominator) term, , for consumer n, Welfare and Equity Impacts of Gasoline Price Changes 7 where i is the index of choices in the choice set. The standard derivation of the CV within the logit discrete choice formulation effectively calculates the difference between the expected utilities with and without a policy intervention (Small and Rosen 1981). Here, the expected CV for consumer n is: