Pain at the Pump: Favorite National Parks of a Gas-Conscious Region


by Sarah Dasher


Abstract
This study has several aims:  a) to investigate the strength of a previously-proposed correlation between National Park attendance and changing gasoline prices and b) to determine whether an individual park's proximity to population centers has any effect on that correlation. 

My results confirm the supposition that gas prices correlate with overall National Park attendance. Furthermore, they shows that certain parks experience more attendance fluctuation than others, when gas prices rapidly climb or fall.  Finally, they suggests that the parks nearest to population centers (50 miles or less) experience the greatest attendance fluctuations in response to gas price changes.  One possible explanation for this surprising result is that people cut back on weekend trips to nearby parks when gas prices are high, but continue to go on long-distance National Park vacations.

Introduction
As with any other vacation destination, National Park visitation varies with economic fluctuations.  For example, park attendance
plummeted in 2008, ostensibly because the soaring price of gasoline in summer of 2008 kept many people from taking long road trips to the parks.  This paper aims to determine how National Park visitation has varied with the summer price of gasoline over the past twenty years. Several relationships will be investigated:  1) whether gasoline price has historically correlated with overall park visitation,  2) if proximity to a population center influences that correlation.  All of the park usage statistics and are provided by the National Park Service for public use.  The gasoline price data and maps of population centers are assembled from other Internet sources.

Background

Previous work has linked declining National Park attendance to a variety of factors, including gasoline prices (Johnson and Suits, 1983; Morgan, 1986), and an increasingly digital culture (Pergams and Zaradic, 2006). None of the studies I encountered, however, took a spatial approach to analyzing park attendance,m and at least three were particular to individual parks or visitor centers (Morgan, 1986; Geurts, 1982) Furthermore, the gas price/park attendance papers were all over twenty years old and firmly rooted in the energy crisis of the late 1970s and early 1980s, which also coincides with the end of the first "backpacking boom" in America. No studies existed linking low gas prices with the relatively high park attendance in the 1990s. Although I found newspaper articles speculating about the effect that last summer's gas hike would have on park attendance, I could find no formal statistical research.


Methods
Data for this project were culled from three sources: the National Park Service (Department of the Interior), the Department of Energy, and the Bureau of Labor Statistics.  All data were provided for free on the agencies' websites.

Data Acquisition.

Park Usage Statistics
Park Usage statistics are provided by the National Park Service in either park-by-park or year-by-year form, at this site.  Data was collected for every parcel (park, national monument, national seashore, et cetera) controlled by the Department of the Interior, dating back to 1979, and compiled into a database.

Gasoline Prices
Gasoline price data is available from the US Department of Energy.  Data was collected for regular grade gasoline in each month dating back to 1980, using this table.  Gas prices were then adjusted for inflation using the Consumer Price Index (CPI) for that month, available from the US Bureau of Labor Statistics here.

Population Centers
Population data was obtained in layer form from the National Atlas, a cooperative project of 20 federal agencies.  The population layer gives spatial and population data for over 30,000 US cities and towns.

Park Locations
NPS administrative regions and park boundaries were obtained as two separate layers from the NPS Data Store: "Current Administrative Boundaries of National Park System Units 05/11/2009", and "Regions of the National Park Service, USA".


Editing the Dataset.

Data scarcity was not a problem for this project--editing it was.

The scope of this project changed considerably in the face of information overload:  For example, the NPS oversees 391 distinct units, including:

It took me a while to compile visitation data for all these parks, going back to the earliest available time (1980).  Because parks come, go, and change designations over time, there wasn't a simple way to copy and paste information into a data table, and I devoted a fair amount of editing time to getting all the years and parks lined up. Eventually, I decided to limit the scope of the study to National Parks lying within the Pacific West administrative area.  This narrowed a list of 391 candidates to 10.

Similarly, I began with population data for 30,000 towns.  That list shrank somewhat when I excluded  cities with a population of less than 100 people:  because they were encoded as population -99999, leaving them in would wreak havoc down the road, when it came time to sum them.

Numerous other errors in the dataset cropped up as the project proceeded. For one, there is an obvious problem with underreporting in North Cascades National Park, resulting in a 2000% reduction in attendance over the 30 year study interval. Olympic National Park, which lies within 50 miles of Seattle/Tacoma, had corrupted data that for some reason prevented it from being joined to buffer cities, despite repeated attempts. Eventually, I eliminated Olympic National Park from the study.

Putting the Pieces Together.

Three layers were loaded into ArcGIS: a base layer of National Park regions (Northeast, Midwest, Pacific West), a layer showing specific NPS parcel boundaries, and a "cities" layer containing population data.  In order to limit the scope of the project, the parks of interest were limited to National Parks (no monuments,seashores, recreation areas) falling within the Pacific West administrative region, but excluding Alaska and Hawaii.

A buffer was applied to the parks of interst at a radius of 50 miles.  Cities within those buffers were then joined to the buffer layer using a point-in-polygon spatial join.  Total population of cities within each buffer were summed using the summarize feature. 

The buffering, joining, and summarizing was repeated for radii of 100 and 200 miles.


Results

The data, as it exists, clearly shows the following relationships:

1.  There is an inverse relationship between system-wide park attendance and gas prices averaged both annually, and averaged over the summer months (June, July August), with winter prices omitted. (Figure B1)


2.  While most parks did lose visitors when gas prices soared, Crater Lake, Death Valley, North Cascades and Joshua Tree (CRLA, DEVA, NOCA and JOTR) did not abide by this pattern. (Figure B2)

3. Generally speaking, parks that have the most people living withing a 50-mile radius are more susceptible to gas price fluctuations than the other parks. (Figure B3, Figure B4, Figure B5)


4.  Variations in attendance over time may be for spatial or non-spatial reasons. 

Analysis

Crater Lake, Death Valley, North Cascades and Joshua Tree are oddball parks: they do not appear to lose visitors when gas prices soar.

This assertion is based on finding the mean attendance for each park over the interval 1980-2008, and measuring the deviation from this mean attendace in each year.  Of particular interest are the interval 1980-1986, when gas prices dropped radically (320-172 cents per gallon), and 2001-2008, when gas prices rose radically (170-320 cents per gallon). Deviation from each park's mean attendance was found over these intervals, as a way of measuring whether park visitation had increased or decreased. Population growth over this period (~1% each year +/ .13) was also considered, but not included, because I assumed that there would be some sort of delay before that population growth showed up in park attendance.  In any case, "deviation from mean attendance over 30 years" is the part of my analysis that I am least happy with.  Sadly, it is also the cornerstone upon which my spatial analysis is based.

Population within an x mile radius is also a crude tool.  For one, it is not possible to ascertain from this crude analysis where park visitors are actually coming from--although Mount Rainier is very near to Seattle, its visitors may well be coming from Vancouver, or much further away.  Also, some parks, while close by crow's-flight standards, actually require a lot of driving miles to access.  Buffering does not account for either of these subtleties.

Apart from these known weaknesses, the unusual nature of CRLA, DEVA, NOCA and JOTR attendance patterns may have spatial and non-spatial explanations. 

Non-spatial
Joshua Tree (JOTR) and Death Valley (DEVA) aren't as vulnerable to summer gas price fluctuations as most parks, because they receive many of their visitors during the early spring, when gas prices are more stable.  (Figure M1) For example, in 2008, gas prices shot up 96 cents between February and July.  This means that while 2008 was an expensive year to visit Yosemite--which receives five times as many visitors in summer as it does in winter--it was a comparatively cheap year to visit Joshua Tree or Death Valley, which receive many early-spring visits.

As for North Cascades (NOCA), there appears to be a problem with under-reporting.  As can be seen from the graph, park visitation at NOCA today is a mere 5% of what it was back in the 1980s.  A closer inspection of the data reveal that only 7 visits were recorded during March of 2008.  Most likely, this shows some sort of omission, or radical change in reporting criteria that may skew the validity of this data.

Spatial
What CRLA, DEVA, NOCA and JOTR have in common is that they all have comparatively few people living within 50 miles (Figure C1).When the radius expands to 100 (Figure C2) or 200 (Figure C3) miles, their population bases are much like the other parks.  Therefore, it may be true that outlying parks are less susceptible to attendance fluctuations than parks in more settled areas.  One possible explanation for this pattern is that when gas is expensive, people cut back on weekend trips to close-by parks, but continue to make big trips to far away parks.

Conclusion

There is an inverse relationship between system-wide park attendance and gas prices, and that pattern holds true for most individual parks.  However, in the Pacific West region, Crater Lake, Death Valley, North Cascades and Joshua Tree (CRLA, DEVA, NOCA and JOTR) did not abide by this pattern.  Variations in these parks' attendance over time may be for spatial or non-spatial reasons.  One possible explanation for this surprising result is that people cut back on weekend trips to nearby parks when gas prices are high, but continue to go on long-distance National Park vacations.

Sources

Geurts, Michael. Forecasting the Hawaiian Tourist Market. Journal of Travel Research, Vol. 21, No. 1, 18-21 (1982)

Johnson, Rebecca and Suits, Daniel. A Statistical Analysis of the Demand for Visits to U.S. National Parks: Travel Costs and Seasonality. Journal of Travel Research, Vol. 22, No. 2, 21-24 (1983)

Morgan, James. The Impact of Travel Costs on Visits to U.S. National Parks: Intermodal Shifting Among Grand Canyon Visitors. Journal of Travel Research, Vol. 24, No. 3, 23-28 (1986)

Oh, Kyushik and Jeong, Seunghyun. Assessing the spatial distribution of urban parks using GIS. Landscape and Urban Planning, Volume 82, Issue 1-2, 15 August 2007, Pages 25-32

Pergams, Oliver and Zaradic, Patricia. Is love of nature in the US becoming love of electronic media? 16-year downtrend in national park visits explained by watching movies, playing video games, internet use, and oil prices. Journal of Environmental Management, Volume 80, Issue 4, September 2006, Pages 387-393

United States Bureau of Labor Statistics. Historical Petroleum Prices ftp://ftp.bls.gov/pub/special.requests/cpi/cpiai.txt

United States Dept. of the Interior. National Park Usage Statistics. http://www.nature.nps.gov/stats/park.cfm

United States Dept. of the Interior. National Atlas. http://www.nationalatlas.gov/

.