Places with the Worst Pain at the Pump in the US by Zip Code

Johannes Moenius Analysis, Archive, Economic Risks, Methodology

Places with the Worst Pain at the Pump in the US by Zip Code

We calculate and map the average burden of gasoline prices on consumer’s budgets for commutes to work by zip‐code for the United States, including Hawaii and Alaska. We find that, for the entire United States, average drive time to work location explains most of the variation in the burden of oil on consumer’s disposable income, followed by the choice of carpooling behavior and income differences. Nonetheless, areas with the higher oil‐burden are characterized by lower average disposable income, longer commute‐times, higher shares of carpooling commuters, but also lower shares of commuters utilizing alternative modes of transportation, such as buses or streetcars. Thus, choice ‐ or availability ‐ of locations to live far from workplace, availability of alternative modes of transportation, and income distribution all seem to contribute to the burden on gasoline prices on consumers’ budgets.

Almost exactly three years ago, Forbes magazine published a report on “Cities with the worst pain at the pump” (http://www.forbes.com/2008/05/07/pain-pump-cities-forbeslife-cx_mw_0506realestate.html). Then, on May first 2008, retail gasoline prices for regular unleaded gasoline hovered around four dollars in Southern California ($3.91, to be exact). And the Inland Empire took the top spot in daily expenses per commuter in the country. Now gas prices are even higher: the same source used in the Forbes study – Gasbuddy.com – reports an average gas price of $4.22 on May 9th 2011 for Southern California. We therefore revisited the Forbes findings. Forbes only compared the major Metropolitan Statistical Areas (MSA) with each other, and calculated the expense per day and commuter. We improved on their method in three substantive ways: we calculated commute costs at the zip-code level for almost 30,000 zip-codes to better account for within MSA variation as well as areas outside of MSAs. Second, we related these expenses for commuters relative to the disposable income in each household, taking variations in workforce participation into account, a variable which we call the average oil burden per household in each zip-code. We have displayed the results in an interactive map, which is accessible at http://isea.redlands.edu/analysis/2011/05/10/places-with-the-worst-pain-at-the-pump-in-southern-california/. Finally, we identified drivers that explain some of the patterns in the data. In order to do so, we had to implement some important simplifications relative to the earlier Forbes method. Thus calculated variable values for particular zip-codes have to be taken with a grain of salt. Please read the “caveats” section below for details. Nonetheless, we expect our analysis to reflect the true patterns reasonably well.

Overall results: A look at the US map quickly reveals that oil-burden is concentrated in certain areas: While some states seem to be – at least in relative terms – affected only a little, others carry a higher burden – and areas being closer to the coasts (but not right along the coastline) and those along the Mississippi and Missouri seem to be harder hit, and where the two coincide – namely the state of Louisiana and some neighboring regions – these areas seem to be especially strongly affected. In contrast, little affected states seem to be located where Mountain regions border the Midwest, such as Montana, Wyoming, North Dakota, and South Dakota. Closer inspection reveals that even in harder hit states there exist areas that show lighter burdens. Specifically centers and more densely populated areas seem to fare better: In the 33% zip-codes with the lowest oil-burden, population density is more than four times higher than in the 33% highest oil-burden zip-codes. Moreover, comparing those two groups again, average drive-time to work is almost 50% higher, income noticeably lower, carpooling higher, and use of alternative (public) transportation substantially lower for the highest oil-burden group of zip-codes. At the same time, home-ownership is higher, the share of the population without graduate or postgraduate degrees is higher, and the share of rural population is also higher for those zip-codes with the highest oil- burden. The following table summarizes the calculated averages:

[table]
Tercile Lowest 33% Middle 33% Highest 33%
Oil-burden per Household as share of disposable income 2.8% 4.4% 6.2%
Gasoline expense for commuting to work per commuter per day $4.30 $5.38 $6.72
Share of drivers driving alone 72% 77% 77%
Share carpooling 11% 13% 15%
Share using other means of transportation 17% 10% 8%
Average time to work 22 min 25 min 31 min
Share using other means of transportation 17% 10% 8%
Average disposable income 57,000 48,000 42,000
Population density: population per sqm 2350 940 500
Share of homeowners 55% 63% 66%
Share without graduate degrees 69% 80% 85%
Share of rural population 40% 61% 83%
[/table] Unweighted averages of zip-code-level data based on 29877 zip codes with valid data across the US; 369 zip-codes were excluded from the analysis due to insufficient data.

Implications: The burden of oil-prices rests quite unequally on American consumers’ shoulders. Cities that offer alternative means of transportation provide their residents with the possibility to evade this burden, while in locations outside of cities, carpooling seems to be the dominant alternative to reduce the weight of gas prices on consumers’ budgets. Adjustments in purchasing behavior, both at the pump as well as at retail stores (see Jim Hamilton’s excellent contribution on econbrowser, http://www.econbrowser.com/archives/2011/05/will_high_oil_p.html), are more likely to happen in areas where the oil-burden is high. One should, however, not forget that individual choices in terms of education, living conditions such as home-ownership, distance of residence relative to workplace, as well as political choices such as income distribution, zoning, and public transportation availability all jointly affect America’s oil consumption and should be viewed as a system, partly designed and constructed under the assumption of availability of affordable oil. Once this assumption is no longer valid, the system in its entirety should be reviewed.

Data-Sources: All community data by zip-code comes from year 2009 estimates generated and published by ESRI based on the American Community Survey. Gas data was retrieved from www.gasbuddy.com, accessed 5/9/2011. Freely available averages at the state and metro levels were used for this report, with metro-level data replacing state averages for whichever metro area they were available. This was true for 3,654 out of the 30246 zip-codes total.

Method: The variable oil-burden was calculated as a proxy for average gasoline expenses for daily commutes to work as a share of disposable income per household. This takes into account the share of household-members going to work, their commute choice, and technical data, in particular average gasoline consumption of vehicles in the US and shares driven on highway versus city based on EPA test standards. However, choices and behavior have been assumed to be unchanged since the ACS 2009 data release. In particular, no adjustments have been made for changes of location or behavior due to high gas prices, such as commuting behavior, choice of residence, or changes in employment status.

Caveats: The strongest caveat is the impossibility to adjust for behavioral changes when gas-prices rise. These range from different commuting choices, labor supply / demand decisions, vehicle choice, and so on. All of those are well documented in the literature, but not available as data for inclusion in our calculation. Moreover, the commute data as published by ESRI is based on their estimates and dates back to 2009. Thus, changes that have happened thereafter are not accounted for, that is, we need to assume that changes in commuting behavior, disposable income, and household size have been at least small enough not to interfere with the displayed patterns. We concede that these assumptions are strong, but likely somewhat less stringent than the 2008 Forbes data, which appear to be based on year 2000 census data. To our defense, we exclusively investigate commute choices to work, that is, gasoline expenditure for getting to work and returning. Demand for commute to work is likely much less elastic than for leisurely travel. Nonetheless, we regard our measures more reliable as indicators of relative oil dependence (which areas are more dependent than others) rather than of absolute oil dependence (how much does the average household in a particular region spend on gasoline).