Would Congestion Pricing Harm the Poor? Do Free Roads Help the Poor?

Abstract

Congestion pricing could reduce urban congestion, but might disproportionately benefit the affluent and burden the poor. We show that this common concern also applies to free roads. Free urban highways primarily subsidize richer people, and the resulting congestion creates pollution that disproportionately burdens poorer people. Furthermore, the poor drivers burdened by peak-hour road pricing would be a small minority of total peak-hour drivers and a minority of the poor. These facts suggest that the revenue generated by pricing could compensate any poor drivers harmed. Free roads, in contrast, generate no revenue to compensate the people they harm.

Keywords

environment transportation sustainability ethics omission bias

Introduction

Congestion tolls seem to exemplify the tension between efficiency and equity. Traffic congestion, a perpetual problem of urban life, is an inefficiency. Congestion occurs because roads are underpriced—they are free to use even when demand is high. Underpriced goods often suffer shortages, and congestion is essentially a shortage of road: at peak times drivers want more road space than there is space available. As a result they must wait, and in waiting they lose time, cause pollution, and increase the risk of crashes (Downs 2004; Edlin and Karaca-Mandic 2006).

Charging market prices to access roads—for instance, charging more at 8 a.m. than 8 p.m., and more on Monday than Sunday—could correct this inefficiency. But because market prices ignore people’s ability to pay, the poor would pay the same toll as the rich, and solving the efficiency problem might create an equity problem. Thus while transportation scholars generally support pricing as a way to reduce congestion, elected officials and the public often worry that tolls would burden low-income drivers. On these grounds opponents have disparaged priced roads as “Lexus Lanes” for the rich and described pricing as a tax on the poor (e.g., Rutten 2009; Zint 2013; Tobar 2015). Anthony Downs, in his influential book on congestion, summarized this dilemma when he attributed the political infeasibility of tolls to “the egalitarian American desire not to provide any relative advantage to high-income travelers” (2004, 334–35).

Downs’s statement contains two implicit assertions: that Americans strongly value equity, and that priced roadways would conflict with that value. Both assertions can be questioned. Much evidence suggests that Americans, compared to citizens of other countries, are quite tolerant of inequities and intolerant of attempts to remedy them (e.g., Piketty and Saez 2003; Alesina and Glaeser 2006). This article, however, focuses on the second assertion: that congestion pricing would be an inherent affront to equity. We find this assertion misguided: neither logic nor existing evidence suggests that equity and pricing must be incompatible.

To begin, and perhaps to state the obvious, few equity agendas in other areas of social policy suggest that all goods be free. Almost no one, for example, suggests that all food be free because some people are poor. Society instead identifies poor people and helps them buy food. It is thus unclear why all roads should be free because some drivers are poor. This question is all the more relevant in the United States, where most drivers are not poor, many poor people do not drive, and most miles are driven by the affluent, because driving involves other substantial expenses (vehicles, gasoline, and insurance) that are easier for the affluent to afford. While half of nonpoor households drove more than ten thousand miles in 2009, only 30 percent of poor households did the same (NHTS 2009).

Put another way, equity concerns about congestion pricing often assume, implicitly or explicitly, that free roads are an important subsidy to the poor. But free roads might be more accurately described as a subsidy to the affluent that some poor people enjoy. Unpriced roads are a government benefit subject to a reverse means test: people with enough money to pay for vehicles, fuel, and insurance can use valuable urban land for free. Because both vehicle ownership and driving rise with income, the benefits of free roads might flow more up than down.

The idea that priced roads will harm the poor further rests on the assumption that most poor people are poor drivers, and on the assumption that free roads (and the congestion they cause) will not harm poor people. While the first assumption is plausible, the second is less so, because not all of congestion’s costs fall on people in vehicles. Certainly the largest, best-known, and most visible costs of congestion are costs drivers impose on each other: wasted time and fuel and, to a lesser extent, crashes (Schrank and Lomax 2014; Edlin and Karaca-Mandic 2006; Anas and Lindsey 2011).¹ But congestion also creates air pollution, and the burden of this pollution often lands on people who live and work near congested roads. If these near-road residents have low incomes and drive less than the average resident, then low-income people may bear more of unpriced driving’s burdens while garnering fewer of its benefits.

In this article we first demonstrate, using the ten most congested urban areas in the United States, that pricing roads, by reducing air pollution, might benefit low-income people who live near congested roads. We then show that unpriced roads are an inequitable and inefficient form of government redistribution, because most of their benefits accrue to the affluent. Our contributions are twofold. First, we add to the large research literature on the fairness of congestion pricing, and explicitly connect that literature to research about the health effects of living near congested roads. Many scholars have documented the relationship between congestion and pollution, and have shown that the costs of pollution fall disproportionately on the disadvantaged (Houston et al. 2004; Ponce et al. 2005). To our knowledge, however, none have tied this information directly to equity discussions about pricing. Similarly, while numerous studies have shown that congestion pricing can reduce air pollution (e.g., Banister 2008; Beevers and Carslaw 2005; Mitchell 2005; Mitchell, Namdeo, and Milne 2005), discussions of pricing’s fairness have largely neglected this finding and focused instead on tolls’ potential harm to low-income drivers (Arnott, de Palma, and Lindsey 1994; Cain and Jones 2008; Levinson 2010; Ecola and Light 2009; Schweitzer and Taylor 2008). In short, while research tying congestion to pollution often has an explicit equity focus, research about pricing’s equity rarely has a pollution focus. Bridging these literatures is our first contribution.

Second, and more broadly, in examining the distributional aspects of free roads we link congestion pricing to a common problem in judging the equity of policy proposals: a tendency to assume that the status quo is fair. This tendency, which psychologists sometimes call omission bias, manifests as an implicit belief that failures to act (omissions) carry less weight than actions themselves (commissions), even if those failures to act result in equal or greater harm (Baron and Ritov 1995; Ritov and Baron 1995, 2004). As a result, people strictly scrutinize harms that arise from changing the status quo, and downplay or overlook harms that arise from the status quo itself.

Omission bias may play a large role in responses to congestion pricing. Tolls are more visible, literally and figuratively, than air pollution. Congestion prices would be new and noticeable, while pollution is quietly normal. When a government tolls a road and burdens low-income drivers, the harm arises from something government does. When a government fails to price a road and people living near it become sick from pollution as a result, that is a harm the government allows. (Philosophers who examine omission bias, for this reason, often call it the problem of “doing vs. allowing” [Howard-Snyder 2011].) Psychologists have shown that people are more attuned to doing harm than allowing it, but it is not obvious that these instinctive reactions are appropriate roadmaps to policy. Harms are no less real for being less visible, and preventing a visible harm from occurring is not inherently more important than halting an invisible harm that is ongoing.

The remainder of the article first reviews existing debates about pricing and equity and then reviews the literature linking congestion to air pollution and its health consequences. From there we describe our methods and data, and present our results. The article’s final section offers a brief discussion of pricing’s other equity implications, and then concludes.

Congestion and Fairness

Tolling’s Burden on Poor Drivers

Congestion pricing creates equity concerns because tolls are regressive, meaning their burden rises as income falls. Congestion charges are designed to move traffic freely, and prices are thus set to maximize the road’s performance—for example, maintain speeds of 55 mph—without any consideration for driver income. London’s congestion charge, for instance, is about $15 per vehicle, regardless of who is in the vehicle, and thus accounts for a larger share of income for poor drivers than rich.

But regressive charges are not necessarily unfair, because fairness and equity (which we use here interchangeably) have no single definition (e.g., Brand 2015). Indeed, common conceptions of equity often conflict with each other, and regressive charges can satisfy one definition of fairness but violate another. This tension is evident in discussions about the fairness of congestion pricing. Pricing conforms to the “user pays” or “benefits received” principle of fairness, which states that people should pay for the resources they consume and not shunt the costs—in time, pollution, or crash risk—onto others. Under the user pays principle, congestion tolls are simply fees for a private good, albeit a publicly owned private good—the road. Drivers pay market prices to use roads and in exchange receive roads that do not suffer shortages. The prices are regressive, but from a fairness perspective this regressivity is fairly meaningless, since many other aspects of driving (such as buying gas and vehicles) also require the purchase of market-priced private goods. In this view, if the poor cannot afford these purchases, it is because they are burdened by insufficient income, not incorrect prices.

Yet if tolling conforms to the benefits-received principle of fairness, it conflicts with the “vertical equity” or “ability to pay” principle, which says those who have more should pay more. This is the same logic that underlies progressive redistribution, and levying a $15 toll on both the rich and the poor violates this principle. It is worth noting, however, that keeping roads free also violates this principle; rich drivers pay the same (nothing) as poor drivers. (Because no one pays anything, free roads do conform to a third notion of fairness—that of horizontal equity, which says those who have the same should pay the same.)²

In a sense, then, congestion pricing’s equity debate boils down to whether road space should be allocated according to the user-pays or the ability-to-pay principle. While there is no obviously correct answer to this question, it is worth noting that roads are public infrastructure, and most infrastructure is allocated via the user-pays principle. Governments regularly charge regressive user fees for water, electricity, and heating fuel (all at least as important as roads) but few people consider water, gas, or electric meters affronts to equity. For that matter, much of the transportation system is financed regressively: cities use property taxes, sales taxes, and gas taxes (all regressive) to pay for streets, roads, and transit. Vehicle registration fees are regressive, and so too are transit fares. Indeed, since transit riders, unlike drivers, are disproportionately low-income, planners wishing to help low-income travelers might be better off reducing transit fares instead of keeping roads or other vehicle infrastructure free (Manville 2014).

The income disparity between drivers and transit riders has led some scholars to argue that congestion pricing’s equity impacts could be blunted if cities spent the toll revenue on transit. Levinson (2010) reviews this idea in some depth, but the basic logic is that regressive charges can have a progressive net impact (or “balanced budget incidence”) if the revenue is collected from a group that is on average richer and spent on one that is on average poorer.³ Tolling could transfer income from richer motorists to poorer transit riders.

The trouble with this idea is twofold. First, while low-income people are more likely to ride transit, not all transit riders have low incomes, and not all transit spending helps low-income riders. Bus riders are poorer than the population overall, but rail riders are richer, and transit agencies spend disproportionately on rail (Taylor and Morris 2015). Thus simply spending revenue on “transit” may not be progressive. For that matter some observers contend that new transit spending only sometimes benefits riders, since transit workers capture some of the increased revenue in higher wages (Winston and Shirley 1998).

Even if one discounts this concern, the second and more fundamental issue is that spending toll revenue on transit—even buses—solves the wrong problem. Funneling revenue to transit does not address the fundamental concern that pricing harms poor drivers. Bus riders may be disproportionately low-income, but they suffer no direct harm from pricing and may even benefit from it, if less congested roads make bus travel faster (Small 2005). Poor drivers subject to high tolls are not helped when their toll payments are transferred to bus riders. Dedicating toll revenue to transit can make pricing progressive, but progressivity in and of itself is an unusual goal. Cities that simply want to maximize pricing’s progressivity could spend their revenue on the poorest people possible—on services for the chronically homeless, for example, or aid for impoverished people in very poor countries. Such spending would almost certainly have social benefits, but would do nothing to mitigate equity problems created specifically by tolls.

Addressing the equity burdens created by pricing probably demands not a transfer across modes (from drivers to transit) but within driving itself (from richer drivers to poorer). Here again public utilities may provide an example. Like roads, utilities are government-owned (or government-regulated) monopoly infrastructure, and they often charge regressive prices to recover costs and discourage overuse. These prices almost undoubtedly enhance efficiency: the nation’s heat infrastructure and electricity grid are far more reliable than its road system. Despite an aging power grid, blackouts—temporary shortages of electricity—remain quite rare. The states with the least reliable electric grids, such as New York, Pennsylvania, and New Jersey, average just over 200 minutes of shortage-induced power outages per year (Eto and LaCommare 2008). In contrast, a typical large urbanized area suffers hundreds of minutes of congestion—a temporary shortage of road—every day. This vast difference in reliability is at least in part a result of using prices to allocate electricity but not to allocate roads.

The potential equity costs of these efficiency-enhancing prices, however, are high. The prices charged by utilities can be substantial. They also vary widely across different types of utility. The US Energy Information Agency (US EIA 2016) estimates that between 2012 and 2015, average annual heating expenditures for households using oil were 3 to 3.5 times as high as they were for households using natural gas. But utilities that sell oil do not take some of their revenue and transfer it to users of natural gas (or vice versa). Rather each utility identifies the low-income users of its own product and uses some revenue to subsidize those users. In principle, similar programs could be enacted for road pricing. Governments, rather than transferring revenue across transportation modes, from drivers to transit riders, can transfer it within drivers, from nonpoor to poor.⁴

Of course, such an approach might face political obstacles. Americans tend to be suspicious of redistribution (Alesina and Glaeser 2006) and might resist proposals that place new and visible charges on drivers and devote most of the resulting revenue to the poor. If relatively few low-income people use congested roads, however, governments could compensate poor drivers with only a portion of the toll revenue, leaving most of it to be spent on projects that higher income drivers support (e.g., King, Miller, and Sanders 2007).

The existing literature on low-income travel provides only some guidance on how many poor people would be exposed to tolls. Poor Americans are less likely than others to drive, but driving is still their primary mode of transportation (Pucher and Renne 2003). Yet poor people are also less likely to work regularly, and those who work may be less likely to commute at peak hours and in peak directions. Low-wage workers may also be more likely to work nights, weekends, and swing shifts (Blumenberg and Manville 2005). Poor people, in short, may be less likely to travel overall, and less likely to travel at times and in directions when tolls are high.

Pollution and Equity

Although roadway proximity is valuable, living in a busy road’s shadow can be a disadvantage, because people near roads are exposed to traffic noise and air pollution. For this reason (and also because many freeways were originally built through poor areas that remained poor), low-income people may be more likely to live near freeways, while affluent people will live further away but nevertheless close enough to access them.

It is by now well established that people living close to freeways are exposed to more vehicular air pollution than people living further away (e.g., Houston et al. 2004; Rioux et al. 2010). While automobiles have become steadily cleaner in recent decades, they still emit substantial amounts of pollution, and even vehicles with clean engines emit more when caught in traffic than when flowing freely (between 45 and 65 mph), as a result of engine inefficiencies caused by frequent braking and acceleration. Cars and trucks emit over 75 percent of the carbon monoxide in the United States, over half of the nitrogen dioxide, and over half of air toxics such as benzene (EPA 2014). Automobiles also account for considerable shares of hydrocarbons and ground-level ozone. Motor vehicles are responsible for only about 10 percent of particulate matter, but vehicular particulates arguably cause disproportionate harm. Unlike larger particulate sources (such as rural power plants), vehicles emit particulates at ground level and near many people.

Vehicular pollutants are generally most toxic closest to the point of emission, and decline sharply in concentration about 200 meters from roads (Brugge, Durant, and Rioux 2007; Hu et al. 2009; Karner, Eisinger, and Niemeir 2010). Researchers have linked exposure to these pollutants, in both the short and long term and at relatively small doses, to a wide variety of health problems. Vehicular air toxics are the largest cause of air pollution–related cancers in the United States (Union of Concerned Scientists 2014), while other car-based pollutants can cause respiratory disease, cardiac disease, premature birth, and low birth weight (Table 1).

Table 1.

Motor Vehicle Pollutants and Their Known Health Impacts.

Pollutant Group	Sources	Scale	Known Health Effects
Carbon monoxide	Engines	0–1,000 meters	Headaches, dizziness, nausea; death at high levels; low birth weight and premature birth during maternal exposure
Air toxics (benzene, acetylene)	Engines	Very local (<100 meters)	Eye irritation, cancer, asthma
Particulate matter (fine, ultrafine, black carbon)	Diesel engines	0–300 meters	Respiratory diseases and infections, cardiovascular disease
Nitrogen dioxide	Engines	300–400 meters	Sudden infant death syndrome, eye irritation, upper respiratory tract infections, bronchial irritation, exacerbated asthma
Hydrocarbons (ground-level ozone)	Photochemical reactions	Regional	Eye and throat irritation, exacerbated respiratory disease

Source: Houston et al. (2004) and Brugge, Durant, and Rioux (2007).

Vehicular pollutants are more dangerous to people in older housing, which has a higher “indoor-outdoor ratio” and thus lets more outdoor contaminants inside. Pollutants are also more dangerous to the old, the young, and the unborn (Medina-Ramon and Schwartz 2008). Ground-level ozone poses a large threat to the elderly, while carbon monoxide can cause preterm birth. Preterm birth can be fatal—in 2010 it was the leading cause of infant mortality in the United States—and can cause lasting damage in those infants who survive it. Infants born preterm grow up to score lower on childhood tests of intellectual and social development and have worse educational and employment outcomes into their thirties and forties (Case, Fertig, and Paxson 2005; Currie and Hyson 1999). Exposure to pollution as a child (as opposed to in utero) has been similarly linked to long-lasting socioeconomic consequences (Currie 2009). In combination, the evidence suggests that early exposure to vehicular pollution does not just burden people in poverty but may also actually create and perpetuate poverty, by undermining the health, mental endowments, and life chances of the very young and unborn. Air pollution, to use Currie’s (2011) phrase, creates “inequality at birth.”

Vehicular air pollution is not, to be sure, the sole cause of poor health among low-income people, and congestion is not the sole cause of vehicular air pollution. Low-income people are more likely to smoke, more likely to lack health insurance, and more likely to suffer from poor nutrition. All these factors degrade personal health. And much vehicular air pollution arises not from congestion but from a small share of poorly maintained vehicles that emit large amounts of pollution regardless of traffic conditions (Wang et al. 2015).

Nevertheless, the link between congestion, pollution, and illness appears to be both causal and consequential (Lin and Yu 2008; Knittel, Miller, and Sanders 2011). Perhaps the most convincing evidence of congestion’s health consequences comes from Currie and Walker (2011) who used the introduction of electronic tolls on the New Jersey Turnpike to examine congestion’s impact on infant health. Electronic tolling eliminated bottlenecks at many toll plazas, reducing congestion by 85 percent and substantially reducing emissions as well. A regression discontinuity analysis showed that tolling also reduced premature births and low birth weights, by 11 and 12 percent, within a kilometer of the booths. The largest decreases occurred closest to the roadways.

Currie and Walker estimate that a nationwide effort to reduce congestion by the same amount could save up to $450 million in preterm births alone. Given that premature birth is only one of many health problems caused by vehicular pollution, the total health benefits of reducing congestion could be much larger. Levy, Buonocore, and von Stackelberg (2010), for example, estimate that peak-hour particulate matter was responsible for $31 billion of mortality-related damages in 2007. By way of comparison, the Texas Transportation Institute’s (TTI’s) estimate for congestion costs due to lost time and wasted fuel that year was $80 billion. Estimates of this sort always have limitations (e.g., Levy, Buonocore, and von Stackelberg are unable to separate overall peak-hour particulate damages from damages caused specifically by congestion) but even accounting for such imprecision the numbers are large enough to warrant attention. Congestion’s pollution damages appear significant, and should play a role in discussions about pricing’s equity impacts, particularly if vehicular pollution poses more danger to lower income people who are less likely to drive.

Data and Empirical Approach

We have two main empirical goals. First, we estimate the socioeconomic characteristics of people who live close enough to freeways to suffer from vehicular air pollution. Second, we examine the socioeconomics of peak-hour freeway drivers. In combination, these examinations show us which groups help create congestion-based air pollution, which groups suffer from it, and the extent to which low-income drivers would be harmed if freeways were tolled.

We describe our approach to these tasks in turn. For both tasks we assume that only freeways will be tolled, and that tolls will be the highest (and perhaps only implemented) during the morning (6 a.m. to 10 a.m.) and the evening (3 p.m. to 7 p.m.) peaks. In part, these assumptions let us work within our data limits and keep our scope manageable, but we also think these assumptions are reasonable (though we relax some of these assumptions later on). Of course freeways are not the only and sometimes not even the most congested roads, but by virtue of their limited access they may be the most likely candidates for tolling. And while tolls may not be zero in nonpeak hours, they will likely be much lower than at peak times. Traveling 10 miles westbound on the dynamically priced SR-91 toll lanes in Orange County, for example, costs over $4 between 6 a.m. and 9 a.m. but just over $2 for the rest of the workday and less than $1.50 in the evening and overnight. The lanes also cost much less in nonpeak directions: at 8 a.m., going east for 10 miles costs $2.25, as opposed to $4.55 going west.⁵

Our final assumption is that the pricing program will be dynamic and well designed. For pricing to reduce pollution over the long term, it must reduce congestion over the long term. If tolling programs do not adjust prices quickly enough to meet demand, or if they exempt high-emission vehicles, the benefits we discuss will either not materialize or not endure. Thus the programs we have in mind bear less similarity to flat-fee area charges like those in London or Stockholm, or to charges like the one in Milan, which exempted some high-polluting two-wheeled vehicles (Percoco 2013). Instead we envision a dynamic freeway charging system more akin to that in Singapore, or along the I-15 in San Diego County, where the price changes every 6 minutes to maintain traffic flow.

Estimating the Socioeconomics of Freeway-Adjacent People

We examine the ten most congested urbanized areas (UAs) in the United States, as measured by the TTI. We use UAs because congestion is an urban problem. Table 2 shows population, poverty, and freeway data for these UAs. These places are generally less poor than the average US urban area—congestion is after all associated with economic success—but because of their large populations they nevertheless have many poor people. Half of these areas have fewer lane-miles of freeway per person than the US average, and in all but one (Philadelphia) drivers use freeways more intensively than the US urbanized area average. The table also shows, for each UA, congestion’s estimated contribution to carbon dioxide emissions. We include this figure not because carbon dioxide has consequences for people who live near freeways (it does not) but because its emissions are correlated with emissions of many localized pollutants, and thus give a sense of peak-hour vehicular pollution. These areas emit, on average, more than 100 pounds more CO₂ per commuter than US UAs as a whole.

Table 2.

Ten Congested Urbanized Areas, 2010.

	Population (1,000s)	Share in Poverty	Freeway Lane-Miles	Freeway Lane-Miles per 1,000 Persons	Daily VMT per Freeway Lane-Mile (1,000s)	Carbon Emissions per Peak-Period Commuter (lb)
Atlanta	4,537	0.15	2,930	0.65	16.7	456
Boston	4,190	0.10	2,911	0.69	14.5	523
Chicago	8,612	0.14	3,400	0.39	17.9	433
Houston	4,962	0.16	3,270	0.66	16.6	455
Los Angeles	12,191	0.16	5,770	0.47	22.9	432
New York	18,376	0.14	7,600	0.41	15.5	557
Philadelphia	5,441	0.13	2,600	0.48	13.1	457
San Francisco	3,296	0.12	2,533	0.77	20.3	498
Seattle	3,075	0.11	2,220	0.72	14.5	440
Washington, DC	4,602	0.08	2,180	0.47	18.1	621
Ten Urbanized area average	6,928	0.13	3,541	0.57	17	487
All US Urbanized area average	448	0.16	264	0.59	13.9	376

Source: US Census Bureau (2014) and Schrank and Lomax (2014).

Note: VMT = vehicle miles traveled.

These ten UAs are composed of 47,451 populated Census block groups. Across these groups, the simple correlation between the share of people in poverty and the distance to the nearest freeway is −0.19, suggesting that block group poverty is in fact higher closer to freeways. The mean share in poverty for block groups within 2,000 feet of freeways is 17 % for block groups between 2,000 and 30,000 feet away, 12 % and for block groups more than 30,000 feet away, 6 percent.⁶

Our interest, however, is in the total population exposed to freeway pollution, not the average population attributes of block groups near freeways. Thus in the remainder of the article, we do not analyze block groups but instead use block group data to build, for each UA, a single population of people vulnerable to freeway pollution. We measure this vulnerability using distance from the freeway. In so doing we ignore some people who will be vulnerable to pollution as a result of proximity to congested arterials, and also ignore poor people who do not live near busy roads. Our goal, again, is not to identify all people who suffer from pollution, nor all poor people. Our goal is to estimate the characteristics of people who might benefit from reduced pollution if freeways were priced.

The distance at which pollutants are dangerous varies by pollutant, and by atmospheric conditions such as wind and air temperature (e.g., Zhou and Levy 2007; Zhu et al. 2002). Hydrocarbons affect entire urban regions, while air toxics and particulate matter are most harmful within 1,500 feet of freeways (Table 1). Damage from carbon monoxide is most severe within 200 meters, although Curry and Walker’s evidence suggests its effects can be felt much further away. We err on the conservative side and draw a 1,000-foot (roughly 300-meter) buffer around each freeway in the UA, recognizing we may overestimate exposure to some pollutants and underestimate exposure to others. We validate our results by repeating the analysis using 750- and 1,250-foot buffers; we discuss this check later on.

To determine the socioeconomic composition of the population within the buffer, we match the buffer areas to Census block groups. One problem with this approach is that block group size varies greatly, from less than one-thousandth of a square mile to over 4 square miles, meaning that in the largest block groups a substantial share of the land area, and perhaps the population, is not within 1,000 feet of the freeway.⁷

To address this problem we test three different definitions of freeway proximity. Our first definition included only those block groups where 75 percent or more of the land area was occupied by the freeway or its buffer. Our second, less-stringent definition included block groups where the freeway or its buffer passed through the block group’s centroid. Our last definition categorizes a block group as freeway-adjacent if a freeway or its buffer intersected the block group at any point. For each definition, we use the block groups to aggregate American Community Survey (ACS) data on various socioeconomic indicators, and create freeway-adjacent populations.

These different definitions of freeway proximity unsurprisingly yield different results, although substantial differences exist only between the third definition and the first two. All definitions suggest that only a small minority of the urbanized population lives close to freeways, and the first two suggest that a very small minority does. Our most restrictive definition yields 1,157 freeway-adjacent block groups, containing about 1.6 percent of the population in these ten UAs. The second yields 2,224 freeway-adjacent block groups (5 percent), and the third, 7,619 block groups (17 percent). Because the first and second definitions yield similar results, in what follows we present only the first and third, which we call, respectively, “freeway dominated” and “freeway intersected.”⁸

Estimating the Socioeconomic Status of Drivers Subject to Tolls

Exposure to tolls is a function of mode, time, and route. To pay high freeway congestion charges, people must drive on freeways at peak hours and in peak directions. People who do not drive at peak times will not pay high tolls, nor will people who drive at peak hours but not on freeways, nor will people who drive east when most traffic is heading west, and so on. Unfortunately, no single data source allows us to isolate these behaviors at the UA level. As such, we use data from both the ACS Integrated Public Use Microdata Sample (PUMS; Ruggles et al. 2011) and the 2009 National Household Travel Survey (NHTS 2009). PUMS data let us examine, for each area, the socioeconomics of people who leave for work during the morning peak. But the PUMS has three disadvantages. First, the data only permit examination of Metropolitan Statistical Areas (MSAs), not UAs. Our geographies are therefore not perfectly commensurate: while our estimates of pollution exposure are restricted to urban residents, our PUMS tabulations will include the rural residents of each MSA as well. Rural residents suffer little congestion, and rural households at all income levels are more likely to drive. In these ten MSAs, 17 percent of urban households are carless, compared to 4 percent of rural households. Similarly, only 59 percent of poor urban commuters drive to work, compared to 83 percent of poor rural commuters. As such, including rural areas will likely overestimate the share of low-income drivers traveling at the peak.

Second, PUMS data will also overestimate the share of poor commuters harmed by tolls, because the data include no information on drivers’ routes, directions, or attitudes. Not every peak-hour commuter will pay tolls: some commuters can change their schedules, some can change their routes, some may not use freeways to begin with, and some may not travel in peak directions. For that matter, some low-income drivers might value fast travel, and be willing, at least in some circumstances, to pay additional charges to avoid congestion. The PUMS offers no insight into these possibilities.

The PUMS’s final disadvantage is that it only tracks commutes, which are a minority of personal vehicle trips. Commutes are important to document, because they are often less flexible than other trips, but the absence of data on other trips means PUMS data will likely understate the share of low-income people traveling at peak periods. Some poor people will drive at peak times even if they are not commuting. Overall, then, PUMS data are likely to overestimate the share of poor commuters exposed to tolls, but understate the share of poor drivers exposed.

The detailed travel information in the NHTS helps overcome some of these limitations. The NHTS examines all travel, not just commutes, and identifies trips that occur on freeways. The survey also asks workers if they can change their schedules, and about attitudes toward congestion and the price of travel. The disadvantages of the NHTS are also threefold. First, the NHTS only surveyed households with landline telephones. As a result, some low-income households are omitted, because lower-income people are more likely to lack landlines (Blumberg and Luke 2014). The direction of any bias this omission implies is unclear: some households without landlines may be so poor that they also lack vehicles, or drive only rarely. Others, however, may rely solely on mobile phones for convenience and still drive at peak hours. It is thus possible, but not certain, that the NHTS underestimates low-income peak-hour driving.

Second, the NHTS does not permit valid inferences at any geographic scale below the Census region (US Department of Transportation 2011). As a result, our NHTS analysis examines all US urban residents, not just residents of the ten most congested UAs. As a result, we likely overestimate the low-income driving share, since low-income residents of the ten most congested UAs are less likely to drive than low-income urban residents overall. Across all UAs, the average share of low-income people driving to work is 81 percent, compared to 61 percent in the ten UAs we examine. Similarly, 9 percent of all urban households are carless, compared to 13 percent across the ten most congested UAs.

A final weakness of the NHTS is that, like the PUMS, it includes no information on the direction of travel, which might also overstate the share of low-income people exposed to tolls.

Results

Table 3 shows that freeway-dominated places are substantially poorer than their UAs at large, and much poorer than places without freeways. Across the freeway-dominated areas, poverty averages 20 percent, compared to 13 percent in areas without freeways. In New York, the poverty rate in freeway-dominated places is almost double that in places without freeways, and in Atlanta, Boston, and Seattle it is at least twice as large. In total, freeway-dominated areas appear to be locations of concentrated disadvantage. They are 0.4 percent of the land area in these UAs, and hold 1.6 percent of the total population, but almost 4 percent of the poor.

Table 3.

Share of People in Poverty by Freeway Proximity.

	Urbanized Area	Freeway Dominated	Freeway Intersected	No Freeway	Difference (%)
Atlanta	0.15	0.32	0.19	0.13	146
Boston	0.10	0.22	0.08	0.11	107
Chicago	0.14	0.21	0.14	0.13	55
Houston	0.16	0.17	0.20	0.16	12
Los Angeles	0.16	0.23	0.18	0.16	43
New York	0.14	0.25	0.16	0.13	83
Philadelphia	0.13	0.18	0.12	0.13	41
San Francisco	0.12	0.14	0.12	0.11	23
Seattle	0.11	0.22	0.14	0.11	107
Washington, DC	0.08	0.11	0.08	0.08	34
Average	0.13	0.20	0.14	0.13	65

Source: American Community Survey 2008–2012.

Note: “Freeway dominated” denotes population in Census block groups whose land area is at least 75% occupied by a freeway or by a 1,000-foot buffer on either side. “Freeway intersected” denotes population in block groups that touch a freeway or its buffer in any way. “Difference” is percentage difference between the freeway-dominated group and the no freeway group. Some percentage differences appear incorrect because of rounding.

With the exceptions of Philadelphia and Washington, DC, freeway-dominated areas are also more likely to be nonwhite and more likely to be black (Table 4). On average, areas dominated by freeways are 62 percent nonwhite and 20 percent black, compared to areas without freeways, which are 49 percent nonwhite and 16 percent black. These differences are much bigger in San Francisco and Seattle, where the share black in freeway-dominated areas is almost 80 percent larger than in areas without freeways.

Table 4.

Race and Ethnicity by Freeway Proximity.

	Percent Nonwhite					Percent Black
	Urbanized Area	Freeway Dominated	Freeway Intersected	No Freeway	Difference (%)	Urbanized Area	Freeway Dominated	Freeway Intersected	No Freeway	Difference (%)
Atlanta	0.54	0.79	0.63	0.49	62	0.35	0.47	0.43	0.31	51
Boston	0.27	0.36	0.20	0.27	32	0.08	0.07	0.03	0.08	−16
Chicago	0.48	0.67	0.50	0.47	44	0.18	0.33	0.24	0.17	92
Houston	0.65	0.76	0.69	0.64	19	0.18	0.27	0.18	0.18	49
Los Angeles	0.71	0.86	0.78	0.69	24	0.07	0.09	0.08	0.07	32
New York	0.52	0.74	0.57	0.51	44	0.17	0.17	0.15	0.17	−1
Philadelphia	0.37	0.36	0.31	0.37	−1	0.21	0.17	0.16	0.22	−21
San Francisco	0.62	0.74	0.71	0.59	27	0.09	0.14	0.12	0.08	79
Seattle	0.34	0.41	0.41	0.32	29	0.06	0.10	0.08	0.05	87
Washington, DC	0.56	0.49	0.55	0.56	−12	0.27	0.21	0.25	0.27	−21
Average	0.51	0.62	0.53	0.49	27	0.17	0.20	0.17	0.16	33

Source: American Community Survey 2008–2012.

Note: “Freeway dominated” denotes population in Census block groups whose land area is at least 75% occupied by a freeway or a by 1,000-foot buffer on either side. “Freeway intersected” denotes population in block groups that touch a freeway or its buffer in any way. “Difference” is percentage difference between the freeway-dominated group and the no freeway group. Some percentage differences appear incorrect because of rounding.

Children and the elderly are more vulnerable to airborne pollutants than others. Table 5 shows that young children are actually less likely to live in freeway-dominated places than in places without freeways (though Los Angeles is a notable exception), and the share elderly, while larger, is not substantially so. Across the ten UAs, 10 percent of the population in freeway-dominated areas is elderly, compared to 7 percent in areas without freeways.

Table 5.

Age of Population and Housing, by Freeway Proximity.

	Atlanta	Boston	Chicago	Houston	Los Angeles	New York	Philadelphia	San Francisco	Seattle	Washington, DC	Average
Children 5 Years and Younger
Urbanized area average	0.07	0.06	0.07	0.08	0.07	0.06	0.06	0.06	0.07	0.07	0.07
Freeway dominated	0.06	0.04	0.07	0.04	0.08	0.07	0.06	0.05	0.05	0.05	0.06
Freeway intersected	0.08	0.05	0.06	0.09	0.07	0.06	0.06	0.06	0.07	0.07	0.07
No freeway	0.07	0.06	0.07	0.08	0.07	0.06	0.06	0.06	0.06	0.07	0.07
Difference (%)	−13	−23	3	−51	24	13	4	−7	−26	−24	−10
People 65 years and older
Urbanized area average	0.09	0.13	0.12	0.08	0.07	0.06	0.06	0.06	0.07	0.07	0.08
Freeway dominated	0.08	0.10	0.12	0.08	0.08	0.11	0.11	0.12	0.12	0.09	0.10
Freeway intersected	0.08	0.05	0.06	0.09	0.07	0.06	0.06	0.06	0.07	0.07	0.07
No freeway	0.07	0.06	0.07	0.08	0.07	0.06	0.06	0.06	0.06	0.07	0.07
Difference (%)	13	80	78	−5	30	69	84	111	93	28	58
Housing built before 1940
Urbanized area average	0.03	0.38	0.24	0.03	0.13	0.3	0.23	0.26	0.11	0.09	0.18
Freeway dominated	0.04	0.55	0.38	0.03	0.14	0.38	0.44	0.24	0.23	0.05	0.25
Freeway intersected	0.03	0.29	0.20	0.05	0.12	0.27	0.20	0.17	0.13	0.05	0.15
No freeway	0.03	0.39	0.24	0.02	0.13	0.30	0.23	0.28	0.11	0.10	0.18
Difference (%)	31	42	58	38	4	28	93	−16	110	−44	35

Source: American Community Survey 2008–2012.

Notes: “Freeway dominated” denotes population in Census block groups whose land area is at least 75% occupied by a freeway or a 1,000-foot buffer on either side. “Freeway intersected” denotes population in block groups that touch a freeway or its buffer in any way. “Difference” is percentage difference between the freeway-dominated group and the no freeway group. Some percentage differences appear incorrect because of rounding.

People in older housing units are also more vulnerable to air pollution. Table 5 also shows that with the exceptions of San Francisco and Washington, DC, freeway-dominated areas have more housing stock built before 1940 (25 and 18 percent, respectively). This simple observation might actually understate the vulnerability of near-road housing stock, because some evidence suggests that near-freeway homes are not just older but also in worse physical condition. The 2009 American Housing Survey shows that 5 percent of all housing in urban areas is degraded enough to let in substantial amounts of outside air (US Census Bureau 2014; e.g., has holes in the roof or foundation, or holes in walls that expose the living areas to the outside). For housing built before 1940, this figure is 11 percent. For pre-1940 housing near major transport infrastructure, it is 13 percent.⁹

Table 6 shows that people living near freeways drive less than other residents. Across all ten UAs, per capita vehicle ownership is 21 percent lower in freeway-dominated places, and households in these places are almost twice as likely to be vehicle-free as households in places without freeways. In many areas, the differences are much larger. Households in the freeway-dominated area of Boston are almost three times as likely to be carless, and in Atlanta and Seattle almost four times as likely. Every freeway-dominated place also has lower rates of automobile commuting than places without freeways. The evidence thus suggests that the people likely to suffer the pollution costs of unpriced freeways are not just disproportionately poor and minority, but also disproportionately unlikely to benefit from unpriced roads.

Table 6.

Vehicle Ownership and Commute Behavior by Freeway Proximity.

	Atlanta	Boston	Chicago	Houston	Los Angeles	New York	Philadelphia	San Francisco	Seattle	Washington, DC	Average
Share of households without vehicles
Urbanized area average	0.07	0.14	0.13	0.06	0.09	0.32	0.15	0.15	0.08	0.11	0.13
Freeway dominated	0.20	0.40	0.21	0.10	0.11	0.54	0.27	0.15	0.26	0.21	0.25
Freeway intersected	0.09	0.14	0.12	0.09	0.08	0.33	0.13	0.12	0.12	0.10	0.13
No freeway	0.05	0.14	0.13	0.06	0.09	0.31	0.15	0.16	0.07	0.12	0.13
Difference (%)	264	191	62	68	25	73	86	−3	266	84	112
Vehicles per capita
Urbanized area average	0.63	0.59	0.57	0.60	0.58	0.43	0.58	0.59	0.71	0.62	0.59
Freeway dominated	0.35	0.33	0.46	0.28	0.44	0.21	0.40	0.52	0.50	0.50	0.40
Freeway intersected	0.55	0.53	0.48	0.52	0.51	0.34	0.52	0.56	0.61	0.54	0.52
No freeway	0.55	0.50	0.47	0.52	0.53	0.34	0.48	0.54	0.62	0.50	0.51
Difference (%)	−36	−33	−1	−47	−17	−39	−17	−5	−19	0	−21
Share commuters driving alone
Urbanized area average	0.82	0.70	0.73	0.82	0.77	0.51	0.75	0.62	0.73	0.67	0.71
Freeway dominated	0.66	0.38	0.60	0.80	0.74	0.26	0.51	0.60	0.49	0.51	0.55
Freeway intersected	0.78	0.74	0.75	0.80	0.77	0.49	0.76	0.66	0.68	0.67	0.71
No freeway	0.83	0.71	0.73	0.82	0.77	0.52	0.76	0.61	0.74	0.67	0.71
Difference (%)	−21	−46	−17	−3	−4	−49	−32	−1	−34	−25	−23

Source: American Community Survey 2008–2012.

Finally, Table 7 shows that our findings are not sensitive to drawing the freeway buffer at 750 feet or 1,250 feet. Populations closer to freeways remain poorer, more nonwhite, and less likely to own vehicles than areas not close to freeways. (The table only compares freeway-dominated to no-freeway areas.) Poverty in most cases is highest close to freeways, and vehicle ownership lowest.

Table 7.

Selected Characteristics of Populations within 750 and 1,250 Feet of Freeways.

	Poverty			Share Nonwhite			Share of Households without Vehicles
	No Freeway	Within 1,250 Feet	Within 750 Feet	No Freeway	Within 1,250 Feet	Within 750 Feet	No Freeway	Within 1,250 Feet	Within 750 Feet
Atlanta	0.13	0.19	0.36	0.49	0.64	0.80	0.05	0.09	0.19
Boston	0.11	0.21	0.25	0.27	0.34	0.36	0.14	0.36	0.50
Chicago	0.13	0.21	0.23	0.47	0.67	0.73	0.13	0.21	0.25
Houston	0.16	0.29	0.30	0.64	0.86	0.89	0.06	0.16	0.13
Los Angeles	0.16	0.22	0.24	0.69	0.84	0.86	0.09	0.11	0.12
New York	0.13	0.22	0.26	0.51	0.72	0.76	0.31	0.51	0.58
Philadelphia	0.13	0.21	0.15	0.37	0.40	0.31	0.15	0.28	0.23
San Francisco	0.11	0.14	0.15	0.59	0.72	0.76	0.16	0.15	0.20
Seattle	0.11	0.21	0.24	0.32	0.41	0.42	0.07	0.22	0.33
Washington, DC	0.08	0.12	0.12	0.56	0.52	0.42	0.12	0.19	0.29
Average	0.13	0.20	0.23	0.49	0.61	0.63	0.13	0.23	0.28

Source: American Community Survey 2008–2012.

Socioeconomics of Peak-Hour Drivers

Having established that freeway pollution could disproportionately burden poor people who are less likely to drive, we turn now to the question of who drives at peak hours—who emits peak-hour pollution and who would pay peak-hour tolls. Table 8 presents the PUMS commute data. The table first shows that poor commuters are less likely to drive than nonpoor commuters (61 percent compared to 78 percent), and that poor commuters who drive are less likely than nonpoor drivers to leave during the morning peak (54 percent compared to 65 percent). But the table also shows that the majority of poor workers do drive, and that the majority of those driving do leave during the morning rush. In total, across the ten MSAs, about one-third of poor commuters (54 percent of 61 percent) drive at peak hours. This figure is lower than the comparable figure for nonpoor commuters (about 50 percent), but nevertheless nontrivial.

Table 8.

Commute Behavior of Poor and Nonpoor Workers, Ten Congested Metropolitan Areas, 2011.

	Atlanta	Boston	Chicago	Houston	Los Angeles	New York	Philadelphia	San Francisco	Seattle	Washington, DC	Average
Driving commuter share
Poor (%)	78	47	63	86	68	30	56	60	65	59	61
Nonpoor (%)	81	77	80	92	86	56	82	74	80	76	78
Share of drivers leaving during morning rush
Poor (%)	52	55	47	55	55	59	55	57	58	48	54
Nonpoor (%)	66	70	59	62	64	70	68	67	63	66	65
Share of total commuters driving during morning rush
Poor (%)	41	26	30	47	37	18	31	34	38	28	33
Nonpoor (%)	53	54	47	57	55	39	56	50	50	50	51
People in poverty
Share without vehicles (%)	15	27	24	13	17	54	36	16	16	23	24
Average household income ($): no vehicles	13,530	13,290	11,050	11,130	14,270	15,680	11,290	14,320	10,130	13,800	12,849
Average household income ($): peak drivers	20,350	30,780	22,720	20,320	28,100	25,930	22,920	28,340	32,740	30,920	26,312
Share of total population (%)	17	13	15	17	17	14	14	12	12	8	14
Share of peak drivers (%)	5	3	4	6	6	3	3	3	4	2	4
Household income >$150,000
Share of total population (%)	10	17	12	12	13	16	12	20	14	24	15
Share of peak drivers (%)	23	32	21	21	22	31	31	33	25	39	28

Source: US Census Microdata 2011, Integrated PUMS (Ruggles et al. 2011).

At the same time, these figures also suggest that more than 65 percent of poor commuters would not be burdened by morning peak tolls. Moreover, poor people are less likely to be commuters in the first place. Many low-income people do not work, and these nonworkers include some of the most vulnerable poor. Only 51 percent of poor adults aged 20 to 64 years in these UAs are in the labor force, and only 36 percent are employed. Overall labor force participation in the United States over this same time, in contrast, was 64 percent, and more than 90 percent of labor force participants were employed.¹⁰

Furthermore, many poor households do not own automobiles, an attribute that strongly suggests both low income and little driving. The table shows that on average 24 percent of the poor in these areas live in households without vehicles. In every MSA, at least 10 percent of the poor live in carless households. These vehicle-free households are substantially poorer than poor households with peak-hour drivers. Average income in carless households is just under $13,000, while average income in poor households with peak drivers is more than twice that, at $26,000.

The table’s bottom rows contrast commuters from households in poverty with those from households making more than $150,000 per year. In the ten MSAs we examine, the poor on average account for 14 percent of the population, but only 4 percent of peak-hour drive commutes. Households earning more than $150,000, in contrast, are 15 percent of the population but 28 percent of peak-hour drive commutes. And again, because these data include rural residents, and make no distinction about travel routes or directions, they may overstate the share of poor drivers who would pay tolls.

For commuters, then, the benefits of free peak-hour roads appear to flow overwhelmingly to the affluent. Even in estimates that bias the share of poor commuters upward, the poor are disproportionately underrepresented on roads at peak times, while the rich are overrepresented. Furthermore, poor drivers who commute at peak hours are among the best-off of the poor. This fact does not make them less deserving of assistance, but reinforces the idea that free roads are a weak form of rich-to-poor redistribution—they do least for those who have least.

Table 9, drawn from the NHTS, shows that the same pattern holds if we examine all peak-hour freeway trips, not just commutes. Across all UAs, the poor are less likely than the rich to drive, although a majority of the poor’s peak-hour trips (68 percent) are still by car. For all income groups, however, only a minority of morning peak trips occur on freeways, and drivers from households earning more than $150,000 are almost twice as likely as poor drivers to use freeways during the morning peak (38 percent and 22 percent, respectively). Furthermore, the table reinforces the idea that the poor are underrepresented in peak travel: the poor are 19 percent of the population but only 11 percent of peak freeway driving trips, while households earning more than $150,000 a year are 21 percent of the population but account for 31 percent of peak freeway drives. Although not shown in the table, the mean trip distance for poor drivers using interstates at peak hours is 16 miles, for nonpoor drivers 17 miles, and for households earning more than $150,000 per year 19 miles. The rich therefore use freeways at peak hours not just more often but for longer distances. These facts imply, again, that the subsidy in the unpriced road flows more up than down, and that the rich contribute more than the poor to congestion and congestion-based pollution.

Table 9.

Characteristics of Trips and Travelers, All US Urbanized Areas, 2009.

	Poor (%)	Nonpoor (%)	Household Income >$150,000 (%)
Share of trips taken by automobile	73	84	84
Share of morning peak trips taken by automobile	68	83	84
Share of morning peak trips that use interstates	22	33	38
Share of total population	19	81	21
Share of total morning interstate peak drive trips	11	89	31
Primary purpose of interstate peak-hour trip
Work	34	39	38
School	10	9	9
Medical/dental	5	2	2
Shopping	14	13	11
Characteristics of travelers
Flexible work start time	43	51	63
Can work from home	10	18	29
Congestion biggest concern	23	24	30
Price biggest concern	32	33	25
Currently pay toll	6	9	12

Source: NHTS 2009, Person and day-trip files. Weekday urbanized area trips of 75 miles or less.

Tolling the road, of course, would change the composition of its users. If wealthier people with higher average values of time are more likely to value the fast travel that tolling offers, then some low-income people could be replaced on the freeway by higher income people. The consequences of such a displacement are difficult to predict and depend on the extent to which it occurs, the second-best options available to displaced drivers, and the relative emissions of displaced vehicles compared to those that replace them. (For example, higher-income people might be more likely to drive higher-emission vehicles such as light trucks, but low-income people may be more likely to drive older, gross-polluter vehicles.)

The NHTS offers only limited evidence on these questions, but its data suggest that poor drivers are less likely to be concerned about congestion and more likely to be concerned about price. (The poor are also less likely to be currently traveling on a tolled route.) In this way, low-income people might have more reason to avoid tolled roads. But they may also be less able to avoid tolls. Only 40 percent of poor commuters have flexible work start times, compared to 63 percent of nonpoor commuters (and a flexible start time does not guarantee avoiding tolls—starting at 9:30 a.m. rather than 9:00 a.m. still requires rush-hour travel). And only 10 percent of poor commuters have the option of working from home, compared to 29 percent of nonpoor workers. On the other hand, compared to the nonpoor, poor peak-hour freeway drivers are slightly less likely than the nonpoor to be en route to work, school, or medical appointments, suggesting some of their trips may be easier to reschedule. In sum, while the evidence suggests that tolled roads would probably displace some low-income drivers, the data are limited, so the extent and full impact of this displacement are ambiguous.

As a final step, we relax the assumption that freeway tolls would only be levied at peak hours. In many highly congested areas, tolls could be high throughout the day. But examining all daily freeway trips by different income groups does not substantially change the story. Roughly half (49 percent) of all drive trips by poor households in UAs use a freeway. But these trips account for only 12 percent of all daily freeway trips. In contrast, 48 percent of all trips taken by households earning over $150,000 per year in UAs use a freeway. But these trips account for 29 percent of all freeway trips.

Extensions and Implications

While our data are imperfect, overall they suggest that poor drivers constitute a small share of overall drivers and would pay a small share of overall tolls. If this conclusion is valid, toll revenue could be used to offset any burden imposed on poor drivers. It is not our purpose here to detail such a redistribution program—many others have discussed ways to redistribute toll revenue (e.g., King et al. 2007)—but for illustration we can sketch one, using Los Angeles as an example. Los Angeles is useful because multiple scholars have estimated the revenue potential of its freeways, were they tolled.¹¹ In 1996, for example, Deakin and Harvey estimated that such tolling would net $3 billion, rising to $7 billion by 2010. Small (1992) also estimated that tolling Los Angeles’s freeways would net $3 billion in 1991. Adjusted for inflation (not increased congestion), in 2015 terms this is about $5.3 billion, while the Deakin and Harvey estimate is about $7.6 billion.

Suppose this revenue was used to give a $150 per month transportation allowance to each of the Los Angeles area’s 1.3 million poor adults. The allowance could be neutral with respect to mode: recipients could use it to pay tolls, transit fares, or—if they walk everywhere—as cash. Such a program would help low-income people who need to drive, but simultaneously provide an incentive to drive less for anyone who valued money more than mobility. A program of this sort would cost about $2.4 billion—less than half the net revenue of the low-end estimate of tolling.

Moreover, this program would be broad, targeting all poor adults, not just poor drivers. One could imagine a narrower, less expensive, and more generous program specifically targeting poor drivers who paid tolls. Qualifying drivers could receive special toll transponders that gave them discounted toll rates (much as low-income households get discounts on utilities through lifeline pricing), or they could receive rebates through the tax code on their toll payments, in the same way low-income workers receive rebates through the Earned Income Tax Credit. The details could vary in a number of ways: our point, again, is only that such redistribution would be neither impossible nor unprecedented.

One might object that our analysis only considers the short-term consequences of congestion tolls—who would pay and who might benefit from cleaner air. We have indeed ignored the longer-range equity impacts of pricing, but while these are inherently less certain we expect they are also on balance progressive. For example, regardless of how the toll revenue is spent, pricing that reduces congestion could help public transportation. Freed from some traffic, buses can run faster. This improved service (along with the increased price of driving) can increase ridership and revenue and let fares fall (Small 2005).¹² Because bus transit is used disproportionately by low-income people, these benefits would likely be progressive. Reducing congestion should also reduce carbon emissions (Barth and Boriboonsomsin 2008). In the short to medium term, carbon reduction is likely progressive, as it will mainly benefit low-income people in low-income countries decades from now, who even in discounted terms would be poorer than drivers in affluent countries today (Schelling 1995; IPCC 2014).

Some urban economists believe pricing freeways would channel population and investment back into cities (e.g., Brueckner and Helsey 2011). The equity implications of such reinvestment are ambiguous. Reinvestment would primarily benefit landowners, who are richer than the average resident. But rising property values could also help city governments finance services for their on-average poorer populations (e.g., Gyourko 1998). Moreover, if local governments currently fight congestion indirectly and inefficiently through policies that restrict density (Manville, Beata, and Shoup 2013), priced roads could help liberalize land use regulation and make density easier to provide. More density, in turn, could increase the supply and variety of housing, which could lower housing prices and disproportionately benefit low-income renters (Rosenthal 2014).

A last concern is that freeways would simply push congestion and pollution onto adjacent roads. If congestion falls on a freeway but reappears on the streets nearby, then pricing may not make freeway-adjacent low-income neighborhoods better off, and might even make them worse. While this concern is reasonable, the extent to which freeway pricing would exacerbate traffic and pollution on nearby roads is not clear-cut. The ambiguity stems, first, from the fact that pricing will not only push vehicles off freeways but also pull them on. This “pull” factor occurs because congestion is its own deterrent to road use. Some drivers avoid congested facilities precisely because they are congested, and congested roads, because their average speeds are low, carry relatively few vehicles. The paradox of congestion, in other words, is that when too many vehicles enter a road, fewer vehicles can actually use it (Parry 2008; Varaiya 2005; Chen and Varaiya 2002). This paradox implies that pricing a relatively small number of vehicles off a road can move many more vehicles over the road—that not just speed but also volume will rise. The net effect of pricing, therefore, will be more rather than fewer vehicles on freeways, and more vehicles tolled on than priced off.

Nevertheless, even if their numbers are relatively small, some vehicles will be priced off the road. These displaced vehicles could exacerbate congestion and pollution near the freeway. But this outcome is not automatic. For displaced drivers to congest roads near freeways, they have to choose to drive on those roads, and the roads must be sufficiently busy that additional vehicles will cause or exacerbate congestion. Neither condition will always hold. If nearby streets are largely empty, they can absorb new vehicles without suffering delay. Furthermore, displaced drivers may not use local streets. They might instead choose to travel on the freeway at a different time when the price is lower, or switch to transit, or simply forego their trips. Even if displaced drivers do replace freeway trips with trips on local streets, they may not use local streets near freeways. Because freeways are routes, not destinations, no driver must use a freeway.¹³ Many drivers choose to, and many will travel some distance to do so, because the freeway offers the best path to their destination. Indeed, one reason freeways get so congested is that drivers from many different origin points converge on them to reach many different destinations.

This logic suggests that drivers displaced from the freeway may not go near the freeway at all, because the second-best route to their destination might look very different from a route involving the freeway. If this is the case, then displaced drivers could be dispersed widely across the region, not concentrated in the low-income corridors near the freeway itself (Downs 2004).

Our point is not that freeway traffic would never spill onto nearby streets, only that it is not obvious that such spillover would dramatically worsen congestion. As a final insurance policy, cities could dedicate some toll revenue to traffic calming in freeway-adjacent neighborhoods. Traffic calming is a relatively low-cost way to discourage vehicles tolled off the freeway from driving down nearby streets instead.

Conclusions

Our goal in this article was to estimate the equity impacts of freeway congestion pricing in a way that accounted both for pollution and for the distribution of road space under the status quo. Before concluding, we re-emphasize that our data set has limits, and these limits subject our estimates to substantial uncertainty. No single data set permits a direct answer to our research questions, so we use multiple sources of information, all of which have shortcomings. The PUMS tells us only about commutes, the NHTS might undersample the poor, and no survey tells us about travel direction. Nevertheless, the balance of evidence suggests that free roads primarily benefit affluent drivers, and that the congestion these disproportionately affluent drivers create imposes costs on low-income people living nearby. At the same time, there is no denying that introducing tolls would burden some poor drivers.

The equity question in pricing would be easier to answer if it involved choosing between one alternative that was plainly fair and another that was plainly not. Unfortunately, this is not the case. We are instead confronted with a decision about what type of unfairness we are willing to tolerate, which makes for a more difficult ethical puzzle.

One way to approach this puzzle is to imagine a counterfactual world where all freeways are priced, and where some revenue is used to mitigate burdens on poor drivers. Then imagine an attempt to depart from this status quo and make all priced roads free. In this scenario, supporters of free roads would need to introduce and defend a program that would primarily benefit the affluent. The program would eliminate one potential harm to the poor (high peak-hour tolls), but increase another (air pollution near roads). And where the poor harmed by congestion pricing would receive not only a benefit (driving on an uncongested road) but also compensation (from the toll revenue), the poor harmed by free roads would get no benefit (they suffer for living near the road, not for using it) and no compensation (without tolls, there is no revenue to redistribute). Finally, the low-income people who would benefit from free roads would be, on average, not just financially better off than those who would be harmed but also less vulnerable. Free roads benefit low-income adults with enough resources to drive, while pollution’s largest and longest-lasting harms fall on children and the unborn.

On equity grounds, such a proposal would warrant skepticism and scrutiny; it would harm the disadvantaged and reward the more advantaged. Today, however, this situation is not a proposal but the status quo, so it is a departure from this scenario, not its introduction, that arouses suspicion. People worry that prices, rather than their absence, will create equity burdens. We have shown that this concern has merit. Congestion prices could harm some low-income drivers, and dedicating toll revenue to offset that burden would be essential. But it is not only priced roads that could harm the poor while helping the rich. Free roads do the same. Consistency would seem to require equal concern in both scenarios. People who worry about harms to the poor when roads are priced, and not when roads are free, may be worried more about the prices than the poor.

Footnotes

Acknowledgements

The authors thank the Cornell Institute for Social Sciences for funding, Patrick Braga for research assistance, and Mike Smart and Dan Kuhlmann for valuable comments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We were funded by the Cornell ISS, as indicated in the acknowledgments.

Notes

Author Biographies

Michael Manville is Assistant Professor of Urban Planning at UCLA Luskin School of Public Affairs. He studies transportation, land use, and public finance.

Emily Goldman has a PhD in City and Regional Planning from Cornell and is currently teaching GIS and open data analysis for the Civic Innovation Fellowship Program in New York City.

References

Alesina

Alberto

Glaeser

Edward

. 2006. Fighting Poverty in the US and Europe: A World of Difference. Oxford, UK: Oxford University Press.

Anas

Alex

Lindsey

Robin

. 2011. “Reducing Urban Road Transportation Externalities: Road Pricing in Theory and Practice.” Review of Environmental Economics and Policy 5 (1): 66–88.

Arnott

Richard

de Palma

Andre

Lindsey

Robin

. “The Welfare Effects of Congestion Tolls with Heterogeneous Commuters.” Journal of Transport Economics and Policy 28:139–61.

Atkinson

Anthony

Stiglitz

Joseph

. 1980. Lectures in Public Economics. London: McGraw-Hill.

Banister

2008. “The Big Smoke: Congestion Charging and the Environment.” In Road Congestion Pricing in Europe: Implications for the United States, edited by Richardson

Harry

Bae

Christine

, 176–97. Cheltenham, UK: Edward Elgar.

Baron

Jonathan

Ritov

Ilana

. 1995. “Reluctance to Vaccinate: Omission Bias and Ambiguity.” Journal of Behavioral Decision Making 3:263–77.

Barth

Matthew

Boriboonsomsin

Kanok

. 2008. “Real World Carbon Dioxide Impacts of Traffic Congestion.” Transportation Research Record 2058:163–71.

Beevers

Sean

Carslaw

David

. 2005. “The Impact of Congestion Charging on Vehicle Emissions in London.” Atmospheric Environment 39:1–5.

Blumberg

S. J.

Luke

J. V.

2014. “Wireless Substitution: Early Release of Estimates from the National Health Interview Survey, July–December 2013.” National Health Interview Survey Early Release Program. http://www.cdc.gov/nchs/data/nhis/earlyrelease/wireless201412.pdf.

10.

Blumenberg

Evelyn

Manville

Michael

. 2005. “Beyond the Spatial Mismatch.” Journal of Planning Literature 19 (2): 182–205.

11.

Brand

Anna

. 2015. “The Politics of Defining and Building Equity in the 21st Century.” Journal of Planning Education and Research 35 (3): 249–64.

12.

Brueckner

Jan

Helsey

Robert

. 2011. “Sprawl and Blight.” Journal of Urban Economics 69 (2): 205–13.

13.

Brugge

Durant

J. L.

Rioux

2007. “Near-Highway Pollutants in Motor Vehicle Exhaust: A Review of Epidemiologic Evidence of Cardiac and Pulmonary Health Risks.” Environmental Health 6 (1): 1–12.

14.

Cain

Alasdair

Jones

Peter

. 2008. “Does Urban Road Pricing Cause Hardship to Low-income Car Drivers?” Transportation Research Record 2607:47–55.

15.

Case

Anne

Fertig

Angele

Paxson

Christina

. 2005. “The Lasting Impact of Childhood Health Circumstances.” Journal of Health Economics 24: 365–89.

16.

Chen

Chao

Varaiya

Pravin

. 2002. “The Freeway Congestion Paradox.” Access 20 (Spring): 41–42.

17.

Christainsen

Gregory B.

2006. “Road Pricing in Singapore after 30 Years.” Cato Journal 26 (1): 71–88.

18.

Currie

Janet

. 2009. “Healthy, Wealthy and Wise.” Journal of Economic Literature 47 (1): 87–122.

19.

Currie

Janet

. 2011. “Inequality at Birth.” American Economic Review 101 (3): 1–22.

20.

Currie

Janet

Hyson

Rosemary

. 1999. “Is the Impact of Health Shocks Cushioned by Socioeconomic Status? The Case of Birth Weight.” American Economic Review 89 (2): 245–50.

21.

Currie

Janet

Walker

Reed

. 2011. “Traffic Congestion and Infant Health: Evidence from E-Z Pass.” American Economic Journal: Applied Economics 3 (January): 65–90.

22.

Deakin

Elizabeth

Harvey

Greg

. 1996. Transportation Pricing Strategies for California: An Assessment of Congestion, Emissions, Energy and Equity Impacts: Final Report. Sacramento, CA: California Air Resource Board.

23.

Downs

Anthony

. 2004. Still Stuck in Traffic. Washington, DC: Brookings Institution.

24.

Ecola

Liisa

Light

Thomas

. 2009. Equity and Congestion Pricing. Santa Monica, CA: RAND Corporation.

25.

Edlin

Aaron

Karaca-Mandic

Pinar

. 2006. “The Accident Externality from Driving.” Journal of Political Economy 114 (5): 931–55.

26.

EPA (Environmental Protection Agency). 2014. “Cars, Trucks, Buses, and ‘Nonroad’ Equipment.” http://www.epa.gov/airquality/peg_caa/carstrucks.html.

27.

Eto

Joseph

LaCommare

Kristina

. 2008. “Tracking the Reliability of the US Electric Power System.” Lawrence Berkeley National Laboratories. https://emp.lbl.gov/sites/all/files/lbnl-1092e.pdf.

28.

Gyourko

Joseph

. 1998. “Place-based Aid vs People-based Aid and the Role of an Urban Audit in a New Urban Strategy.” Cityscape 3 (3): 205–29.

29.

Houston

Douglas

Jun

Ong

Paul

Winer

Arthur

. 2004. “Structural Disparities of Urban Traffic in Southern California.” Journal of Urban Affairs 26 (5): 565–92.

30.

Howard-Snyder

Frances

. 2011. “Doing vs. Allowing Harm.” In Stanford Encyclopedia of Philosophy, edited by Zalta

Edward N.

http://plato.stanford.edu/archives/win2011/entries/doing-allowing/.

31.

Fruin

Kozawa

Mara

Paulson

S. E.

Winer

A. M.

2009. “A Wide Area of Air Pollutant Impact Downwind of a Freeway during Pre-sunrise Hours.” Atmospheric Environment 43 (16): 2541–49.

32.

IPCC (Intergovernmental Panel on Climate Change). 2014. Summary for Policymakers. Climate Change 2014: Impacts, Adaptation, and Vulnerability. New York: Cambridge University Press.

33.

Karner

A. A.

Eisinger

D. S.

Niemeir

2010. “Near-Roadway Air Quality: Synthesizing the Findings from Real-world Data.” Environmental Science and Technology 44 (14): 5334–44.

34.

King

David

Manville

Michael

Shoup

Donald

. 2007. The Political Calculus of Congestion Pricing. Transport Policy 14 (2) (March): 111–123.

35.

Knittel

Christopher

Miller

Douglas

Sanders

Nicholas

. 2011. “Caution, Drivers! Children Present: Traffic, Pollution and Infant Health.” NBER Working Paper 17222. National Bureau of Economic Research, Cambridge, MA.

36.

Levinson

David

. 2010. “Equity Effects of Road Pricing.” Transport Reviews 30 (1): 33–57.

37.

Levy

Jonathan

Buonocore

Jonathan

von Stackelberg

Katherine

. 2010. “Evaluation of the Public Health Impacts of Traffic Congestion: A Health Risk Assessment.” Environmental Health 9:65.

38.

Lin

2008. “Traffic-related Air Quality Assessment for Open Road Tolling Highway Facility.” Journal of Environmental Management 88:962–69.

39.

Lindert

Peter

. 2004. Growing Public. New York: Cambridge University Press.

40.

Manville

Michael

. 2014. “Parking Pricing.” In Parking: Issues and Policies, edited by Ison

Stephen

Mulley

Corrine

, 137–55. Bingley, UK: Emerald Press.

41.

Manville

Michael

Beata

Alex

Shoup

Donald

. 2013. “Turning Housing into Driving.” Housing Policy Debate 23 (2): 350–75.

42.

Medina-Ramon

Mercedes

Schwartz

Joel

. 2008. “Who Is More Vulnerable to Die from Ozone Pollution?” Epidemiology 19 (5): 672–79.

43.

Mitchell

Gordon

. 2005. “Forecasting Environmental Equity: Air Quality Responses to Road User Charging in Leeds, UK.” Journal of Environmental Management 77 (3): 212–26.

44.

Mitchell

Gordon

Namdeo

Anil

Milne

David

. 2005. “The Air Quality Impact of Cordon and Distance Based Road User Charging: An Empirical Study of Leeds, UK.” Atmospheric Environment 39 (33): 6231–42.

45.

NHTS (National Household Travel Survey). 2009. “U.S. Department of Transportation.” Federal Highway Administration. http://nhts.ornl.gov.

46.

Parry

Ian

. 2008. Pricing Urban Congestion. Washington, DC: Resources for the Future.

47.

Percoco

Marco

. 2013. “Is Road Pricing Effective in Abating Pollution? Evidence from Milan.” Transportation Research Part D: Transport and Environment 25 (December): 112–18.

48.

Piketty

Thomas

Saez

Emmanuel

. 2003. “Income Inequality in the United States, 1913–1998.” Quarterly Journal of Economics 118 (1): 1–39.

49.

Ponce

Ninez

Hoggatt

Katherine

Wilhelm

Michelle

Ritz

Beatte

. 2005. “Preterm Birth: The Interaction of Traffic-Related Air Pollution with Economic Hardship in Los Angeles Neighborhoods.” American Journal of Epidemiology 162 (2): 140–48.

50.

Pucher

John

Renne

John

. 2003. “Socioeconomics of Urban Travel: Evidence from the 2001 NHTS.” Transportation Quarterly 57 (3): 49–77.

51.

Rioux

Christine

Gute

David

Brugge

Doug

Peterson

Scott

Parmenter

Barbara

. 2010. “Characterizing Urban Traffic Exposure Using Transportation Planning Tools.” Journal of Urban Health 87 (2): 167–88.

52.

Ritov

Ilana

Baron

Jonathan

. 1995. “Outcome Knowledge, Regret and Omission Bias.” Organizational Behavior and Human Decision Process 64 (2): 119–27.

53.

Ritov

Ilana

Baron

Jonathan

. 2004. “Omission Bias, Individual Differences and Normality.” Organizational Behavior and Human Decision Processes 94:74–85.

54.

Rosenthal

Stuart

. 2014. “Are Private Markets and Filtering a Viable Source of Low-Income Housing?” American Economic Review 104 (2): 687–706.

55.

Ruggles

Steven

Alexander

J. Trent

Genadek

Katie

Goeken

Ronald

Schroeder

Matthew

Sobek

Matthew

. 2011. Integrated Public Use Microdata Series: Version 5.0. Minneapolis, MN: University of Minnesota.

56.

Rutten

Timothy

. 2009. “Congestion Pricing: A Slippery Slope to Toll Roads.” Los Angeles Times, June 10.

57.

Santos

Georgina

. 2008. “London Congestion Charging.” In Brookings-Wharton Papers on Urban Affairs, edited by Burtless

Gary

Pack

Janet Rothenberg

, 177–234. Washington, DC: Brookings Institution Press.

58.

Schelling

Thomas

. 1995. “Intergenerational Discounting.” Energy Policy 23 (4/5): 395–401.

59.

Schrank

David L.

Lomax

Timothy

. 2014. Urban Mobility Reports. College Station, TX: Texas Transportation Institute.

60.

Schweitzer

Lisa

Taylor

Brian

. 2008. “Just Pricing: The Distributional Effects of Congestion Pricing and Sales Taxes.” Transportation 35 (6): 797–812.

61.

Small

Kenneth

. 1992. “Using the Revenues from Congestion Pricing.” Transportation 19 (4): 359–82.

62.

Small

Kenneth

. 2005. “Unnoticed Lessons from London: Road Pricing and Public Transit.” Access 26 (Spring): 10–15.

63.

Taylor

Brian

Morris

Eric

. 2015. “Public Transportation Objectives and Rider Demographics: Are Transit’s Priorities Poor Public Policy?” Transportation 42 (2): 347–67.

64.

Tobar

Hector

. 2015. “How Los Angeles Is Becoming a Third-World City.” New York Times July 6.

65.

Union of Concerned Scientists. 2014. “Why Clean Cars?” http://www.ucsusa.org/clean_vehicles/why-clean-cars/air-pollution-and-health/cars-trucks-air-pollution.html#.VKnzC3tkYqg.

66.

US Census Bureau. 2011. “Current Housing Reports, Series H150/09.” In American Housing Survey for the United States: 2009. Washington, DC: US Government Printing Office.

67.

US Census Bureau. 2014, November. “Summary File.” In 2008–2012 American Community Survey. Washington, DC: US Census Bureau’s American Community Survey Office. http://ftp2.census.gov/.

68.

US Department of Transportation. 2011. 2009 NHTS User’s Guide. Washington, DC. http://nhts.ornl.gov/2009/pub/usersguidev2.pdf.

69.

US EIA (US Energy Information Agency). 2016. “Average Consumer Prices and Expenditures for Heating Fuels during the Winter.” http://www.eia.gov/forecasts/steo/tables/pdf/wf-table.pdf.

70.

Varaiya

Pravin

. 2005. “What We’ve Learned about Highway Congestion.” Access (Spring): 2–9.

71.

Wang

Jon

Jeong

C.-H.

Zimmerman

Healy

R. M.

Wang

D. K.

Evans

G. J.

2015. “Plume-based Analysis of Vehicle Fleet Air Pollutant Emissions and the Contribution from High Emitters.” Atmospheric Measurement Techniques 8:3263–75.

72.

Winston

Clifford

Shirley

Chad

. 1998. Alternate Route: Toward Efficient Urban Transportation. Washington, DC: Brookings Institution.

73.

Zhou

Levy

J. I.

2007. “Factors Influencing the Spatial Extent of Mobile Source Air Pollution Impacts: A Meta-analysis.” BMC Public Health 7:89.

74.

Zhu

Hinds

W. C.

Kim

Shen

Sioutas

2002. “Study of Ultrafine Particles Near a Major Highway with Heavy-Duty Diesel Traffic.” Atmospheric Environment 36:4323–35.

75.

Zint

Bradley

. 2013. “Plans for Tolls on 405 in O.C., Dismissed as ‘Lexus Lanes,’ Draw Heat.” Los Angeles Times, November 1.