Marketing,congestion,and demarketing in Utah’s National Parks

Abstract

Utah’s Mighty 5 ad campaign was designed to attract out-of-state visitors to southern Utah’s five National Parks (NPs). Using the synthetic control method, we find the campaign to have contributed to rapid visitation growth at Arches, Canyonlands, and Capitol Reef NPs. No ad campaign effect was found for Bryce Canyon and Zion NPs, which implies that recent increased visitation at these parks has been driven by the national trends and not the Mighty 5 promotion. Arches, Bryce Canyon, and Zion currently suffer from excess demand (congestion), and the US National Park Service is actively engaged in visitor management efforts to mitigate the environmental pressures associated with overvisitation. The state has responded to congestion by crafting a new “demarketing” campaign to provide a more tailored tourism experience aimed at increasing the value of a visit and diverting potential tourists to high-quality alternative sites selected to match visitor interests.

Keywords

demarketing marketing national park congestion national park visitation synthetic control method

JEL codes: C23, Q41

Introduction

Following two decades of an overall decline in per capita outdoor recreation-based tourism (Pergams and Zaradic, 2008), National Parks (NPs), World Heritage Sites, and other outdoor tourism destinations around the world have experienced excess demand for at least the past decade. Overcrowding of nature tourism sites has been documented on every continent. A small sampling of the literature related to overcrowding at NPs and heritage sites—and efforts to control overfull demand—includes Ferreira and Harmse (2014) in Africa; Prakash et al. (2019) and Sim et al. (2018) in Asia; Armstrong and Kern (2011) and Beeton and Benfield (2002) in Australia; León et al. (2015) and Mejía and Brandt (2015) in South America; Oklevik et al. (2019) and Medway et al. (2011) in Europe; and Upchurch (2015) and Lawson et al. (2011) in North America.

In addition to lowering the perceived quality of recreational visits at protected areas, congestion at outdoor recreation sites can lead to a wide variety of ecological impacts. Greater pedestrian or vehicle traffic can affect species abundance and affect the normal behavior of animals (Prakash et al., 2019). Parking lots that are full lead to private vehicles parked along roadways, compacting and damaging adjacent soils, and increasing off-trail use of park resources (US National Park Service (USNPS), 2017a). Transportation systems designed to relieve congestion by distributing visitors to more sites within a protected area or more evenly across seasons can cause impacts as visitor use spreads to new areas or to more ecologically sensitive times of the year (Monz et al., 2016). The five US NPs stretching across southern Utah (Figure 1) have not been immune to issues associated with overuse and congestion, and the USNPS has struggled to find the appropriate balance between its obligation to protect ecologically sensitive areas while simultaneously keeping the parks open to tourists (USNPS, 2014, 2017a,b).

Figure 1.

Location of five NPs in the study. NP: national park.

Sustainable resource management in the face of visitor pressure has become an important line of inquiry in tourism research, but this is necessarily linked to product management (marketing) efforts, as classified by Leask (2016). Park visitation in Utah has been rising since 2008, yet, in 2013, the Utah Governor’s Office of Economic Development (Utah GOED, 2013a) initiated The Mighty Five marketing effort, an aggressive multimedia campaign to promote visitation to Utah’s five NPs (Arches NP, Bryce Canyon NP, Canyonlands NP, Capitol Reef NP, and Zion NP).¹ In 2017, the USNPS cited the promotion campaign (hereafter, “Mighty 5”) as contributing to excess demand at Arches NP (USNPS, 2017a: 1).

We begin with a brief review of the marketing and demarketing literature as applied to nature-based tourism attractions. We then introduce the method of synthetic control as a means to measure the effect of the Mighty 5 marketing on each of Utah’s NPs against the backdrop of increasing visitation throughout the NP system. To our knowledge, this is the first application of synthetic control in a marketing context in tourism economics. Our empirical results are provocative: we find no statistically significant effect of the campaign at two parks with well-known congestion problems, but it significantly contributed to rapidly increasing visitation at three other parks. That is, the marketing campaign had uneven effects across the five NPs, although each park was treated equally in the campaign. As a result, the campaign increased congestion at previously uncongested parks. Our discussion focuses on how federal and state entities responded to the success of the Mighty 5 by initiating a demarketing campaign with the multiple goals of (1) reducing visitor congestion at crowded NPs, (2) more closely tailoring nature-based tourism sites to specific interests of potential visitors, and (3) spreading the economic benefits of nature-based tourism more evenly across the region.

Our second primary contribution is to describe how the very success of the Mighty 5 marketing campaign later forced federal and state entities to investigate and/or initiate efforts to “demarket” the NPs. The USNPS and the Utah Office of Tourism adopted different demarketing approaches. The USNPS focused on developing rules to restrict visitation across time and space. The state of Utah, which enjoys a wealth of visually stunning landscapes adjacent to NPs, used a fundamentally different approach. Its demarketing efforts advertised nature-based tourism sites more closely tailored to specific interests of potential visitors. This also satisfied its goal of spreading the economic benefits of nature-based tourism more evenly across the region.

Background and literature review

Promotion of NPs by private and public entities in the United States

Since their inception in the late 19th century, NPs have been the subject of intense promotion by both private and public entities. The first director of the USNPS, Stephen Mather, enlisted support from the transportation and tourism industries as part of his overall strategy to expand the park system (Wilson, 2014: 82–83). Encouraging railroads, the emerging automobile industry, and local chambers of commerce to promote NPs was hardly difficult. Railroads proved to be especially enthusiastic promoters of NP tourism (Runte, 1991); in 1918, 12 western railroads spent the equivalent of US$ 5.6 million (constant US$2013) on booklets, brochures, and other promotional materials encouraging visits to NPs located along their main and branch lines (Blodgett, 2007). Railroads also profited from luxury lodging they had constructed within the parks; these grand hotels remain popular to this day. The National Park Service (NPS) did not leave promotion solely to the private sector; visitation was also encouraged by constructing an expansive road system to accommodate the growing use of the automobile. The USNPS’ Mission 66 program was “…an aggressive system-wide construction program designed to expand park visitor accommodations and facilities,” providing a boost to park tourism (Keiter, 2010: 86). More recently, the USNPS celebrated its centenary in 2016 with a multimedia promotion called Find Your Park—a coordinated public–private partnership to boost awareness of and visitation to NPs (USNPS, 2018b).

Beginning in March 2013, Utah’s Mighty 5 ad campaign initially targeted residents of six major cities in the western United States in an effort to attract out-of-state visitors to its five NPs. The original US$3.1 million campaign consisted of dramatic NP images on television programs, websites, digital billboards, magazines, and, in Los Angeles and Denver, wallscapes on large buildings. The television advertisement depicted one family’s adventure to Utah’s five NPs, thus drawing directly on the increasing trend in adventure vacations by families (Schänzel and Yoeman, 2015).² The prime-time television ad portrayed and mentioned, by name, each of Utah’s five NPs. Further, the ad illustrated the parks as part of a multiday holiday excursion, stretching from west to east across the southern portion of the state, with the tagline, “Five iconic parks, one epic experience.” Thus, the ad provided potential visitors with immediate help in solving the first stage of the tourist trip design problem (see, e.g. Kotiloglu et al., 2017). After 9 months, the campaign was judged to be hugely successful: analysis by Strategic Marketing & Research, Inc reported that 5.75 million people had been exposed to the campaign, with 6.5% more visits from ad-aware people than from those who were unaware of the campaign. The study concluded that there were 375,000 more trips to Utah NPs in 2013 as a result of the promotional effort (Businesswire, 2014). The campaign’s impact was expected to increase over time as the promotion continued and potential tourists had more time to plan a visit in the future.

Demarketing campaigns

The Mighty 5 was a traditional marketing campaign designed to increase visitation but (Sharpley and Pearce, 2007) argue more generally that marketing is a demand management tool that can be used to either increase or decrease demand for a good or service. “Demarketing” campaigns can be used to reduce overall demand for a product or to shift demand in time and space. For example, the highly successful anti-smoking and anti-drunk driving publicity campaigns influenced how much, and where, people chose to smoke and drink. Such efforts are easily extended to manage excess demand at tourism sites.

Armstrong and Kern (2011) fit a broad range of NP visitor management strategies within the context of demarketing campaigns. Demarketing may take the form of a change in the product (e.g. activity or length-of-stay restrictions at a park or site), its distribution among visitors (visitor capacity constraints or a reservation system), the price of access (entrance fees), and/or how the natural area is promoted (emphasis on activity restrictions or nonpromotion of certain areas or sites). The most common demarketing dimension used by (or recommended to) policymakers is nonprice mechanisms designed to allocate visitors over space and time. Visitor capacity constraints and reservation systems can be used to both limit overall congestion as well as shift visitation to nonpeak seasons (e.g. Ferreira and Harmse, 2014; Prakash et al., 2019). Restrictions on private vehicles in favor of shuttle systems will reduce vehicular congestion but can result in long wait times for shuttles if total visitor constraints are not also implemented (Upchurch, 2015). Congestion pressure can also be limited at specific sites within a park, particularly for individual activities or access to ecologically sensitive sites (Armstrong and Kern, 2011; León et al., 2015).

When excess demand is present, economists will usually tout the role of price as a means to efficiently allocate a scarce resource. Walls (2016) has advocated a differentiated pricing structure based on season and location in an effort to shift visitor use to other times of year and other parks. As Park et al. (2010) note, though, NPs, heritage sites, and other public lands are managed as public goods to benefit society as a whole, so that social equity issues associated with price increases are often of great concern to managers. Consequently, demarketing efforts based solely on price are rare; if prices are raised for congested areas, this action is often taken in combination with other demarketing actions. Another form of demarketing is “diversion demarketing” (Medway et al., 2011). Rather than accentuating the negative aspects of demarketing efforts for a congested site, promotion efforts can instead highlight the positives of alternative places to visit, while ceasing promotion of congested or ecologically sensitive sites altogether (Armstrong and Kern, 2011). This form of demarketing can be effective for regions such as southern Utah, where close substitutes for congested NPs are readily accessible to tourists.

Methods and data

Methods

Our goal is to examine the differential effects of the Mighty 5 at five different NPs. Thus, the selected statistical approach must be able to measure promotion effects for each park. Further, the promotion effect must be disentangled from the overall increase in visitation experienced by the NP system as a whole in recent years. A traditional difference-in-difference (DD) approach could be used (e.g. see Ashenfelter and Card, 1985), where the Mighty 5 marketing campaign is a “treatment” and parks enjoying this promotion are called treatment units. All other NPs not subject to the Mighty 5 campaign would serve as control units against which the treated units could be compared. However, a DD approach does not satisfy our goal. First, the measured treatment effect would be summarized in a single parameter; in effect, DD measures an average effect across all treated units and not the desired park-level effect. Second, DD assumes that treatment and control units are similar in every way except for the variables used in the statistical modeling. Finally, each control unit—even if it is a poor comparison to treatment units—receives the same weight in the overall analysis. DD is thus heavily reliant upon analyst judgment in choosing an appropriate set of controls.

The synthetic control method (SCM) relaxes the restrictive assumptions of DD by using data-driven process to weight control units differentially (Abadie and Gardeazabal, 2003; Abadie et al., 2010, 2014). Intuitively, one can think of synthetic control in two stages. First, the characteristics of all units are used to identify those characteristics that are most important in predicting the outcome-of-interest (annual visitation at a NP) during the pre-intervention period (before the promotion campaign). Second, control units with important characteristics that most closely match those of the treatment unit receive a greater weight than units who did not match well. Control unit weights (which sum to one) are used to create a synthetic control unit. The time path of the outcome of interest for a good synthetic control will closely match the time path of the treatment unit during the pre-intervention time period. The difference between the actual outcome and the synthetic outcome during the post-intervention period is the measured treatment effect. Synthetic control has been used to evaluate numerous public policies including, among others, California’s tobacco control program (Abadie et al., 2010). nuclear power facilities (Ando, 2015), and greenhouse gas policies (Kim and Kim, 2015). Within the context of our application, synthetic control has been used to analyze the effect of tourism development policies (Castillo et al., 2017) and designations of large national monuments (NMs) (Jakus and Akhundjanov, 2019). The key advantage of SCM over DD is that a weighted combination of NPs control units provides a better comparison for the treatment NPs in Utah than an arbitrarily selected set of control NPs. See the Online Appendix for more discussion of the SCM.

Let $y_{i t}$ measures the number of visitors to NP i ( $i = 0$ for a treatment NP and $i > 0$ for control NPs) at time t $(t = 1, \dots, t_{0}, \dots, T)$ . Let t₀ denotes the time of the intervention so that $t = 1, \dots, t_{0} - 1$ are the pre-intervention periods and $t_{0}, \dots, T$ are the post-intervention periods; in our case, the Mighty 5 campaign began in 2013. Let $y_{i t}^{N}$ be the number of visitors observed for NP i at time t, where superscript N indicates no Mighty 5 campaign. Let $y_{i t}^{I}$ be observed visitation for NP i at time t if NP i is treated with the Mighty 5 during periods t₀ to T. Further, let $α_{i t} = y_{i t}^{I} - y_{i t}^{N}$ measures the effect of the marketing campaign for unit i at time t and let $d_{i t}$ be an indicator that takes value one if NP i is treated with the Mighty 5 at time t and zero if not. Observed visitation for any NP i at time t is

\begin{array}{l} y_{i t} = y_{i t}^{N} + α_{i t} d_{i t} \end{array}

Because only the first NP $(i = 0)$ is exposed to the intervention at period t₀, we have that

\begin{array}{l} d_{i t} = \{\begin{array}{l} 1 if i = 0 and t \geq t_{0} \\ 0 otherwise \end{array} \end{array}

We aim to estimate $α_{0} = (α_{0, t_{0}}, α_{0, t_{0} + 1}, \dots, α_{0, T})^{'}$ . For $t \geq t_{0}$

\begin{array}{l} α_{0 t} = y_{0 t}^{I} - y_{0 t}^{N} = y_{0 t} - y_{0 t}^{N} \end{array}

Here, $y_{0 t}^{N}$ , the visitation at a treatment unit had the treatment not occurred, is unobserved and must be estimated. Suppose that $y_{i t}^{N}$ is a function of predictor variables,

\begin{array}{l} y_{i t}^{N} = δ_{t} + θ_{t} z_{i} + λ_{t} μ_{i} + ε_{i t} \end{array}

where $δ_{t}$ is a time trend common to all units, $z_{i}$ is a $r \times 1$ vector of observed covariates (not affected by the intervention), $μ_{i}$ are permanent unobserved variables, $θ_{t}$ and $λ_{t}$ are unknown parameters, and $ε_{i t}$ are an unobserved error term with zero mean.

A synthetic control (synthetic NP) could be generated as a weighted average of NP units in the control group. Consider a vector of weights, $w$ , each element of which is less than one and whose sum equals one; any value chosen for $w$ is equivalent to choosing a synthetic control (Abadie et al., 2014). Following Abadie et al. (2014), let $x_{1}$ be a $k \times 1$ vector of pre-intervention characteristics for a treated unit, and our goal is to find $w$ that matches the vector as closely as possible. If $X_{0}$ is a $k \times N$ matrix of the same variables for the units in the control group, then the goal is to choose $w$ so as to minimize the pre-intervention difference the treatment unit and its synthetic control, $x_{1} - X_{0} w$ , that is, we choose $w^{*}$ and $V$ to minimize

\begin{array}{l} ∥ x_{1} - X_{0} w ∥ = \sqrt{{(x_{1} - X_{0} w)}^{'} V (x_{1} - X_{0} w)} \end{array}

where $V$ is a $k \times k$ matrix used to weight the most important unit characteristics associated with visitation. Optimal values for $V$ and $w$ minimize the mean square error prediction of the synthetic control estimator,

\begin{array}{l} MSPE = (y_{1} - y_{0} w^{*})^{'} (y_{1} - y_{0} w^{*}) \end{array}

The synthetic control estimator of the effect of the Mighty 5 is given by

\begin{array}{l} {\hat{α}}_{0, t} = y_{0, t} - \sum_{i = 1}^{N} w_{i}^{*} y_{i, t} for t = t_{0}, t_{0} + 1, \dots, T \end{array}

Data

Our outcome of interest is annual visitation to 54 NPs with continuous data over the period 1993–2015³ (Table 1). Collected from the NPS Visitor Statistics portal, the data set consists of five treatment units (Arches NP, Bryce Canyon NP, Canyonlands NP, Capitol Reef NP, and Zion NP) and 49 control units (USNPS, 2018a).⁴ The Mighty 5 campaign started in March 2013, giving us 20 years of observations during the pretreatment period (1993–2012) and 3 years during the posttreatment phase (2013–2015). Our posttreatment period ends in 2015 for two reasons. First, in December 2015, the USNPS began its centenary celebration Find Your Park (USNPS, 2018b). Second, the Utah Office of Tourism de-emphasized the Mighty 5 in favor of a new campaign, Road to the Mighty 5, promoting other outdoor destinations (Utah GOED, 2016). We cannot control for all state, local, and private marketing efforts at all NPS units over our 23-year time period.

Table 1.

Treatment and control NPs.

Treatment NPs	Control NPs
Arches	Acadia	Grand Canyon	Mesa Verde
Bryce Canyon	Badlands	Grand Teton	Mount Rainier
Canyonlands	Big Bend	Great Basin	North Cascades
Capitol Reef	Biscayne	Great Sand Dunes	Olympic
Zion	Black Canyon of the Gunnison	Great Smoky Mountains	Petrified Forest
	Carlsbad Caverns	Guadalupe Mountains	Redwood
	Channel Islands	Haleakala	Rocky Mountain
	Congaree	Hawaii Volcanoes	Saguaro
	Crater Lake	Hot Springs	Sequoia
	Cuyahoga Valley	Isle Royale	Shenandoah
	Death Valley	Joshua Tree	Theodore Roosevelt
	Denali	Kenai Fjords	Voyageurs
	Dry Tortugas	Kings Canyon	Wind Cave
	Everglades	Lake Clark	Wrangell-St. Elias
	Gates of the Arctic	Lassen Volcanic	Yellowstone
	Glacier Bay	Mammoth Cave	Yosemite
	Glacier

NP: national park.

Figure 2 provides an initial comparison of visitation at treatment units relative to combined visitation at the control sites. Visitation to Utah parks stagnated or fell through about 2006, after which visitation began to increase for the remainder of the pre-intervention period. In contrast, aggregate visitation at all other NPs did not begin to rebound until 2013, the start of our post-intervention period. A visual comparison of the slopes for visitation at Utah’s NPs versus the other NPs suggests faster growth in Utah during the post-intervention period.

Figure 2.

Logged number of visitors in NPs in Utah. NP: national park.

Synthetic control relies upon a set of unit characteristics to identify which unit attributes best predict the outcome of interest during the pre-intervention period. Our set of predictor variables was based on the studies of Henrickson and Johnson (2013) and McIntosh and Wilmot (2011), who investigated the factors affecting visitation at US NPs. Park visitation was hypothesized to be a function of the budget of the NP in the previous year⁵ (a proxy for park popularity); park age; a dummy variable equal to 1 if the NP is located in the contiguous United States and 0 otherwise; distance to, population, and per capita income in the nearest population base (urban areas) with population over 1 million⁶; the price of gasoline⁷; cooling degree days (CDDs)⁸; and number of NPs within 1000 miles. For NPs that had previously been designated as a NM, park age was defined as the date at which it was designated as a NP (see Cline et al. (2011) for a discussion of the NP designation effect). A logarithmic transformation was applied to variables measuring park visitation, NP budget, distance, population, per capita income, and gas price. Following Kaul et al. (2015), we use only one observation of the outcome variable as a unit characteristic; we have chosen the lagged visitation in 2012, immediately preceding the Mighty 5. Mean values for the Utah parks and its set of controls are shown in the upper portion of Table 2.

Table 2.

Visitation predictor means.

Mean for treatment and control NPs	Arches	Bryce Canyon	Canyonlands	Capitol Reef	Zion	49 Control NPs
Park budget (million US$)	1.13	2.54	4.82	1.75	5.45	6.03
Park age (years)	31.50	74.50	38.50	31.50	83.50	53.84
Located in continental United States (yes = 1)	1	1	1	1	1	0.84
Distance (miles)	230	260	238	221	159	581
Population (millions)	1.63	1.58	1.63	1.63	1.58	4.09
Per capita income (000 dollars)	27.4	31.1	27.4	27.4	31.1	33.9
Gas price (US$)	2.01	2.02	2.01	2.02	2.00	2.03
CDDs	237.03	27.81	205.29	144.72	283.92	129.71
Number of parks	30	28	29	27	27	14
Lagged visitation in 2012 (millions)	1.04	1.30	0.47	0.67	2.83	1.14
	Synthetic NPs
Means for synthetic NPs	Arches	Bryce Canyon	Canyon-lands	Capitol Reef	Zion
Park budget	4.67	5.45	7.93	5.72	12.31
Park age	37.94	34.60	60.22	49.99	80.94
Located in continental United States	1.00	1.00	1.00	1.00	1.00
Distance	222	242	280	266	210
Population	8.53	5.97	1.66	1.66	2.43
Per capita income	30.7	32.1	27.9	28.0	28.6
Gas price	2.05	2.04	2.07	2.07	2.08
CDDs	196.90	63.92	134.84	127.58	238.64
Number of parks	25	24	26	26	23
Lagged visitation in 2012	1.07	1.57	1.02	1.21	2.98

Note: NP: national park; CCD: cooling degree days. All variables are averaged for the pre-intervention period (1993–2012).

Results

Construction of synthetic controls

Synthetic controls were estimated using the Synth package in R. The bottom portion of Table 2 describes the characteristics of the synthetic NPs, while the control unit weights used to construct each synthetic control are reported in Table 3. Each synthetic control was composed of a weighted average from between four control units (Bryce Canyon NP) and nine control units (Canyonlands NP). Figure 3 shows the actual time path of visitation to each treatment unit (solid line) relative to the time path associated with its synthetic counterpart (dashed line) during the entire 1993–2015 time period. The synthetic NPs appear to follow the trajectory of the actual NPs relatively closely during the pre-intervention period (to the left of the vertical line in each graph), although actual visitation to Arches, Bryce Canyon, and Capitol Reef NPs begin to exceed that of the synthetic control during the Great Recession period of the late 2000s. During the post-intervention period, the plots in Figure 3 show that, with the exception of Zion NP, visitation to treatment NPs appears to have increased faster than their synthetic counterparts.

Table 3.

Weights for Synthetic NP Unit.

NP	Arches	Bryce Canyon	Canyonlands	Capitol Reef	Zion
Black Canyon of the Gunnison	0.003	0.009		0.001
Death Valley	0.145		0.133	0.160	0.075
Grand Canyon			0.003		0.487
Grand Teton	0.048	0.432	0.021	0.354	0.238
Great Basin	0.048		0.374	0.293
Great Sand Dunes		0.295
Guadalupe Mountains	0.008
Joshua Tree	0.456	0.264	0.002	0.002
Mesa Verde	0.291		0.280	0.181
Petrified Forest			0.001	0.003
Rocky Mountain			0.001		0.003
Saguaro				0.004
Sequoia					0.193
Yellowstone			0.183

NP: national park.

Figure 3.

Number of visitors: Treated NPs versus synthetic NPs in Utah. NP: national park.

Figure 4 distills the information included in Figure 3 by plotting the differences in visitation between the actual unit and its synthetic control. Also called the Actual–Synthetic Gap, a good match between a treatment unit and its synthetic control will show small gaps (near zero) during the pre-intervention period, with larger gaps occurring only after the intervention. The plots in Figure 4 suggest that the overall effect of the Mighty 5 campaign on the visitation in Utah’s NPs has been positive. However, the analysis to this point has relied entirely on visual inspections; we must still determine if the observed gaps are significantly different from zero.

Figure 4.

Gap in number of visitors between treated NP and synthetic NP in Utah. NP: national park.

Placebo and MSPE tests

Following the procedures of Abadie et al. (2010), we perform placebo (falsification) tests on all control NPs, that is, those NPs not exposed to the Mighty 5 campaign, to assess the significance of the results presented in the previous section. We run the SCM for these NPs. Here, we treat control NP units as if they had been exposed to the Mighty 5 campaign when in fact they had not; that is, we are acting as if these units had been given the Mighty 5 campaign placebo treatment. If the synthetic control of the placebo NP creates gaps of magnitude observed for units receiving the placebo are found to be of similar magnitude to the gaps in visitation for treated NPs, our SCM analysis does not provide strong evidence of a Mighty 5 effect; those observed for treatment units, then any actual–synthetic differences reported in the “Construction of synthetic controls” section were obtained by chance and observed effects are statistically insignificant. If the gaps for placebo NP units are smaller or close to zero than those for the treatment unit, then we may conclude that the Mighty 5 campaign has been effective in attracting more visitors to the treatment units. We have more robust evidence for the effectiveness of the Mighty 5 campaign.

We sequentially treat each control with the placebo and then measure the gaps between it and its synthetic control. We also identify those placebo units whose synthetic controls were poorly fitted during the pre-intervention period. Poorly fitted units were defined as those whose synthetic pretreatment mean square prediction error (MSPE) statistic (see the Online Appendix) was 20 times, 10 times, or 2 times larger than the pretreatment MSPE calculated for the true treatment unit. For any possible placebo unit, a large pre-intervention MSPE value implies that its true time path of visitation cannot be well-replicated using a combination of other control units, and it, therefore, represents a poor comparison unit.

Figure 5 presents the placebo test results using different levels of tolerance; the dark line is the actual–synthetic gap for the treatment unit, whereas the lighter shaded lines show the actual–synthetic gaps for each placebo unit. Poor synthetic controls are eliminated as our pre-intervention tolerance becomes more stringent; this causes the pre-intervention dispersion across units to tighten around the expected pre-intervention gap of zero. For example, out of 49 possible comparison units, 44 placebo comparisons are allowed for Arches NP under the 20 × MSPE criterion but only 33 are permitted using the 2 × MSPE criterion.

Figure 5.

Placebo gaps in number of visitors with MSPE below selected cutoff point. Note: solid black lines: gaps in number of visitors in NPs in Utah and gray lines: gaps in number of visitors in control NP. MSPE: mean square prediction error; NP: national park.

During the pre-intervention period, the time path for the actual–synthetic gap should be clustered around zero for both placebo and treatment units in Figure 5. A potentially significant treatment effect would be evident if, during the post-intervention period, the black line for the true treatment unit was significantly above the gray lines for the placebo units. That is, the placebo should have had no effect on a control unit, and its actual–synthetic gap remains near zero in the post-intervention period. With the exception of Zion NP, the actual–synthetic gaps are largest, or almost the largest, for the true treatment units. Tightening the cutoff point further to 10 times larger MSPE (10 × MSPE) or two times as large (2 × MSPE) does not change the results very much; the gaps for Arches, Bryce Canyon, Canyonlands, and Capitol Reef remain larger than nearly all of the comparison placebo units, especially under the strictest interpretation of “good” comparison units.

Another way to evaluate the statistical significance of the measured treatment effect is to compare the ratio of the post-intervention MSPE to the pre-intervention MSPE. Control units should have a pre-intervention MSPE value that is similar to its post-intervention value, so that the post-/pre-ratio has a value near one. In contrast, strong treatment effects should result in large post-intervention MSPE so that the post-/pre-ratio should have a relatively large positive value. Large ratios for treatment units relative to small ratios for placebo units mean that any gap observed in the post-intervention period is likely not a coincidence (Abadie et al., 2010). We rank the post-/pre-ratios from largest to smallest and use the position in the distribution to determine a p-value testing the null hypothesis that the intervention had no effect on the outcome of interest.

Table 4 presents the rank and p-value of the post-/pre-intervention MSPE ratio for each NP in Utah relative to the set of controls under various choices concerning the quality of control units. When using all 49 possible control units, Canyonlands NP and Capitol Reef NP have the highest post-/pre-campaign MSPE ratio relative to the remaining 49 control NPs. Given a total of 50 NP units, the probability of observing this outcome is 2% (1/50). That is, the probability that the Mighty 5 campaign had been assigned randomly to one of these 50 parks is 2%, which is a highly unlikely proposition. Similarly, the probability that the post-/pre-intervention ratio for Arches NP ranked third out of 50 units was 6%; that is, the probability that the Mighty 5 campaign had randomly assigned is 6%, another somewhat unlikely outcome. The 2% and 6% probabilities are in the range of the 5–10% probability levels typically used in conventional tests of statistical significance, which supports the effectiveness of the Mighty 5 ad campaign. In contrast, the post-/pre-MSPE ratio for Bryce Canyon NP and Zion NP ranked 6th and 14th, respectively, out of 50, and the probability of random treatment assignment is 12% (=6/50) and 28% (=14/50). Thus, the impact of the Mighty 5 campaign on visitation at Bryce Canyon NP and Zion NP is statistically insignificant. Similar calculations for more stringent synthetic controls (better quality) yield similar results. Statistically significant effects are found consistently for Arches NP, Capitol Reef NP, and Canyonlands NP.

Table 4.

Rank of post-/pre-MSPE ratio and p-value for H₀: Mighty 5 had no effect on annual visitation.

Park	All possible comparisons	20 × MSPE	10 × MSPE	2 × MSPE
Arches	3/50	3/45	3/43	2/34
	0.06	0.07	0.07	0.06
	(49)	(44)	(42)	(33)
Bryce Canyon	6/50	6/45	6/43	3/33
	0.12	0.13	0.14	0.09
	(49)	(44)	(42)	(32)
Canyonlands	1/50	1/42	1/34	1/15
	0.02	0.02	0.03	0.07
	(49)	(41)	(33)	(14)
Capitol Reef	1/50	1/42	1/37	1/17
	0.02	0.02	0.03	0.06
	(49)	(41)	(36)	(16)
Zion	14/50	14/43	14/42	8/23
	0.28	0.33	0.33	0.35
	(49)	(42)	(41)	(22)

Note: Number of comparison units in parentheses.

Discussion

Our synthetic control analysis has revealed a statistically significant response to the Mighty 5 campaign in three of Utah’s five NPs. The synthetic control gaps can be used to calculate the visitation response associated with the campaign (Table 5). In 2013, our modeling predicts just under 346,000 additional visits due to the campaign, which is remarkably close to the number predicted by the Utah Office of Tourism’s own marketing analysis (Businesswire, 2014). Over the 3-year post-intervention period, our models predict an annual average of roughly 500,000 additional visits to Utah’s NPs as a direct result of the Mighty 5 promotional effort. A more interesting result is the response of tourists to the Mighty 5 campaign: two of Utah’s most congested NPs (Bryce Canyon and Zion NPs) did not have a statistically significant response, whereas two of Utah’s more remote and lightly visited NPs did (Canyonlands and Capitol Reef NPs).

Table 5.

Gap in number of visitors between treated NP and synthetic NP in Utah (visits).

	2013	2014	2015	Three-year average
Arches
Actual	1,082,866	1,284,767	1,399,247	1,255,627
Gap	242,425	337,509	295,616	291,850
Percentage of actual	22.4	26.3	21.1	23.2
Canyonlands
Actual	462,242	542,431	634,607	546,427
Gap	45,691	73,878	121,258	80,276
Percentage of actual	9.9	13.6	19.1	14.7
Capitol Reef
Actual	663,670	786,514	941,029	797,071
Gap	56,539	118,159	210,043	128,247
Percentage of actual	8.5	15.0	22.3	16.1
Sum of three parks
Actual	2,208,778	2,613,712	2,974,883	2,599,124
Gap	344,655	529,546	626,917	500,373
Percentage of actual	15.6	20.3	22.1	19.0

The results suggest that Utah’s potential tourists are aware of congestion issues at Bryce Canyon and Zion NPs. Over the course of our study time frame, Bryce Canyon consistently attracted more than 1 million visitors annually, and Zion annually drew 2.2 million visitors or more. Given its narrow canyon and thusly limited transportation and parking infrastructure, Zion NP has had a shuttle system in place since 2000. In recent years, parking lots are routinely filled to capacity by 9:00 a.m. on a typical summer day, and peak season wait-times for shuttles can exceed 60 min at popular trailheads (USNPS, 2017b). Zion NP is actively investigating a reservation system to allocate visitors by time-of-day and by season in an effort to reduce congestion pressure. Bryce Canyon NP, located atop the relatively narrow Paunsaugunt Plateau, is also similarly limited in its ability to handle vehicles. Bryce Canyon instituted its own shuttle system in 2000 and recently revisited its transportation management plan in response to increasingly inadequate shuttle capacity and congestion at parking areas (USNPS, 2014).

Arches NP did not break the 1 million annual visitors mark until 2010, but it has drawn more than 1 million visitors every year since. Arches NP has a single-paved entrance, and its transportation infrastructure is not currently amenable to a shuttle bus system (USNPS, 2017a). Given these constraints, park officials are considering a reservation system similar to that under review by officials at Zion NP and used at other congested NPs (e.g. Muir Woods NP). In 2018, wait times to enter Arches NP during the peak season were in excess of 30 min. It is likely that potential tourists will soon become aware of congestion at Arches. Similar to what has been observed at Bryce Canyon and Zion NPs, future promotional efforts may have little effect on Arches NP visitation.

In addition to Arches NP, the Mighty 5 boosted tourism to less well-known, more remote NPs whose primary entrances are located some distance from an interstate highway or a major population center. Canyonlands and Capitol Reef NPs each averaged less than 630,000 visitors per year and had never attracted as many as 1 million visitors during our study period. Evaluated at average visitation over the 3-year post-intervention period, our modeling suggests the Mighty 5 caused a 15–16% annual increase in visitation at Canyonlands and Capitol Reef NPs (a time during which visitation grew by 40% in both parks). Both parks have continued to see increasing visitation in the post-2015 period, with 2018 visitation numbers that are 17% and 30% higher than 2015, for Canyonlands and Capitol Reef NPs, respectively.

Does growing visitation, at even remote parks, mean that Utah’s outdoor recreation-based tourism has reached its maximum capacity? Are sensitive scenic and ecological treasures preserved in the parks at risk of being overwhelmed by excess demand? Perhaps not. The broad swath of southern Utah where the five NPs are located is also home to an expansive area of protected public lands. Though Bears Ears NM and the Grand Staircase-Escalante NM (originally totaling 12,100 km²) are experiencing a legal dispute regarding their size, their smallest protected area (5300 km²) would still be in excess of the total area of the five NPs (3400 km²). Further, Glen Canyon National Recreation Area (4400 km²), the recently designated San Rafael Swell National Recreation Area (4000 km²), numerous state parks, and abundant public lands without formal protective designations are readily accessible as substitute sites for congested NPs. Given the abundant possibilities to redirect tourists to high-quality substitute sites, the elements that made Mighty 5 a successful tool at boosting overall demand for Utah’s NPs between 2013 and 2015 can be restructured as a diversion marketing campaign.

In 2017, the state of Utah began its Red Emerald Initiative, a promotional effort that does not directly acknowledge congestion issues but contains strong demarketing elements (Utah GOED, 2016). First, the initiative’s many television and web advertisements explicitly name and depict outstanding landscapes for numerous alternatives to NPs. Second, one of Red Emerald’s goals is to “prioritize quality, not quantity of visitors.” If a person’s website or television viewing habits reveal, for example, an interest in niche activities such as fishing, hiking, motorized recreation or paleontology, niche advertisements are then targeted at that specific viewer or audience. As such, the Red Emerald campaign has much in common with the activity clustering of tourists recently examined by Derek et al. (2019). This campaign also achieves another goal anticipated by Derek et al. (2019) in that tailored tourism results in a higher valued experience, so that the net economic benefits of tourism are increased. Thus, the state’s new campaign could reduce congestion pressures, improve ecological outcomes at NPs, and increase social welfare. Further, by directing tourist traffic to lightly visited sites, visitor expenditures and the benefits of public lands tourism would be distributed more widely throughout rural regions of southern Utah.

Conclusions

The Mighty 5 ad campaign was designed to attract out-of-state visitors to southern Utah’s five NPs. We find the campaign to have been effective in raising visitation at Arches (p-value < 0.10), Canyonlands (p-value < 0.05), and Capitol Reef (p-value < 0.05) NPs. No treatment effect was found for Bryce Canyon NP and Zion NP. While Bryce Canyon NP and Zion NP have also experienced more visitors in recent years, any increased visitation due to Mighty 5 cannot be separated from the broad upward trend in visitation experienced by the NP system as a whole.

Arches, Bryce Canyon, and Zion NPs currently suffer from excess demand, and the NPS is actively engaged in visitor management efforts to mitigate the environmental pressures associated with overvisitation. The primary means by which Arches and Zion NP officials are approaching the congestion problem is to restrict visitor use by time-of-day and by season through the introduction of a reservation system. Zion and Bryce Canyon NPs restrict the use of private vehicles in the park (essentially raising the cost of access without formally increasing the entrance fee). These proposed solutions, of course, have proved to be controversial among the general public. Utah’s Mighty 5 campaign appears to have been working at cross-purposes with federal efforts to mitigate the unintended effects of congestion. The state of Utah’s new Red Emerald campaign allows the state tourism promotion agency to craft a demarketing campaign that relieves congestion at overcrowded sites, promote tailored tourism to high-quality alternative sites, and spread the benefits of nature-based tourism more evenly throughout the state.

Footnotes

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) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Man-Keun Kim

Notes

Author biographies

Tatiana Drugova who is a recent PhD graduate from Department of Applied Economics at Utah State University and is now the postdoctoral researcher in the same department. Her main research area is Agribusiness. For the manuscript, she reviewed the previous studies, collected data, and ran the statistical analyses as well as discussed the results and contributed to the final manuscript.

Man-Keun Kim is an associate professor in Department of Applied Economics at Utah State University. He is an applied economist specializing in the field of regional economics with subspecialties in natural resources and environmental economics. His current work focuses on tourism demand vulnerability and resilience to political issues and natural disaster such as wildfire. Regarding the manuscript, he conceived of the idea, explored, and verified the methodologies. He reviewed the previous studies, discussed the results, and wrote the first draft of the manuscript as well as contributed to the final manuscript.

Paul Jakus is a professor in Department of Applied Economics and Institute of Outdoor Recreation and Tourism (IORT) at Utah State University. He is an applied economist specializing in the field of environmental and ecological economics with subspecialties in regional economics. The bulk of his current work focuses on the economics of public land in the United States; recent topics have included the possible fiscal consequences of a land transfer to western states and the regional economic effects of landscape-scale national monuments. Regarding the manuscript, he reviewed the previous studies, discussed the results, revised, and contributed to the final manuscript.

References

Abadie

Diamond

Hainmueller

(2010) Synthetic control methods for comparative case studies: estimating the effect of California’s Tobacco control program. Journal of the American Statistical Association 105(490): 493–505.

Abadie

Diamond

Hainmueller

(2014) Comparative politics and the synthetic control method. American Journal of Political Science 59(2): 495–510.

Abadie

Gardeazabal

(2003) The economic costs of conflict: a case study of the Basque country. American Economic Review 93(1): 113–132.

Ando

(2015) Dreams of urbanization: quantitative case studies on the local impacts of nuclear power facilities using the synthetic control method. Journal of Urban Economics 85: 68–85.

Armstrong

Kern

(2011) Demarketing manages visitor demand in the blue mountains national park. Journal of Ecotourism 10(1): 21–37.

Ashenfelter

Card

(1985) Using the longitudinal structure of earnings to estimate the effect of training programs. The Review of Economics and Statistics 67(4): 648–660.

Bai

Ouyang

(2014) Property taxes and home prices: a tale of two cities. Journal of Econometrics 180: 1–15.

Beeton

Benfield

(2002) Demand control: the case for demarketing as a visitor and environmental management tool. Journal of Sustainable Tourism 10(6): 497–513.

Blodgett

(2007) Defining uncle Sam’s playgrounds: railroad advertising and the national parks, 1917-1941. Historical Geography 35: 80–113.

10.

Businesswire (2014) Utah office of tourism summer travel campaign created by struck yields impressive results. Available at: https://utahpulse.com/index.php/features/marketing/708-utah-office-of-tourism-summer-travel-campaign-created-by-struck-yields-impressive-results (accessed 31 January 2020).

11.

Castillo

Figal Garone

Maffioli

, et al. (2017) The causal effects of regional industrial policies on employment: a synthetic control approach. Regional Science and Urban Economics 67: 25–41.

12.

Cline

Weiler

Aydin

(2011) The value of a name: estimating the economic impact of public land designation. Social Science Journal 48: 681–692.

13.

Derek

Woźniak

Kolczyk

(2019) Clustering nature-based tourists by activity: social, economic, and spatial dimensions. Tourism Management 75: 509–521.

14.

Ferreira

Harmse

(2014) Kruger national park: tourism development and issues around the management of large numbers of tourists. Journal of Ecotourism 13(1): 16–34.

15.

Henrickson

Johnson

(2013) The demand for spatially complementary national parks. Land Economics 89(2): 330–345.

16.

K-S

Pesaran

Shin

(2003) Testing for unit roots in heterogeneous panels. Journal of Econometrics 115: 53–74.

17.

Jakus

Akhundjanov

(2019) The antiquities act, national monuments, and the regional economy. Journal of Environmental Economics and Management 95: 102–117.

18.

Kaul

Klößner

Pfeifer

, et al. (2015) Synthetic control methods: never use all pre-intervention outcomes together with covariates. MPRA Paper 83790, University Library of Munich, Germany.

19.

Keiter

(2010) The national park system: visions for tomorrow. Natural Resources Journal 50: 71–110.

20.

Kim

M-K

Kim

(2015) Estimating impact of regional greenhouse gas initiative on coal to gas switching using synthetic control methods. Energy Economics 59: 328–335.

21.

Kotiloglu

Lappas

Pelechrinis

, et al. (2017) Personalized multi-period tour recommendations. Tourism Management 62: 76–88.

22.

Lawson

Chamberlin

Choi

, et al. (2011) Modeling the effects of shuttle service on transportation system performance and quality of visitor experience in rocky mountain national park. Transportation Research Record 2244: 97–106.

23.

Leask

(2016) Visitor attraction management: a critical review of research 2009-2014. Tourism Management 57: 334–361.

24.

León

Araña

, et al. (2015) Tourists’ preferences for congestion, residents’ welfare, and the ecosystems in a national park. Ecological Economics 118: 21–29.

25.

Levin

Lin

C-F

Chu

C-S

(2002) Unit root tests in panel data: asymptotic and finite-sample properties. Journal of Econometrics 108: 1–24.

26.

Masini

Medeiros

(2019) Counterfactual analysis with artificial controls: inference, high dimensions and nonstationarity. SSRN Paper 3303308. Available at: https://ssrn.com/abstract=3303308 or https://dx-doi-org-s.web.bisu.edu.cn/10.2139/ssrn.3303308 (accessed 6 July 2020).

27.

McIntosh

Wilmot

(2011) An empirical study of the influences of recreational park visitation: the case of US National Park Service sites. Tourism Economics 17(2): 425–435.

28.

Medway

Warnaby

Dharni

(2011) Demarketing places: rationales and strategies. Journal of Marketing Management 27(1–2): 124–142.

29.

Mejía

Brandt

(2015) Managing tourism in the Galapagos islands through price incentives: a choice experiment approach. Ecological Economics 117: 1–11.

30.

Monz

D’Antonio

Lawson

, et al. (2016) The ecological implications of visitor transportation in parks and protected areas: examples of research in US National Parks. Journal of Transport Geography 51: 27–35.

31.

Oklevik

Gössling

Hall

, et al. (2019) Overtourism, optimisation, and destination performance indicators: a case study of activities in Fjord Norway. Journal of Sustainable Tourism 27(12): 1804–1824.

32.

Park

Ellis

Kim

, et al. (2010) An investigation of perceptions of social equity and price acceptability judgements for campers in the U.S. national forest. Tourism Management 31: 202–212.

33.

Pergams

Zaradic

(2008) Evidence for a fundamental and pervasive shift away from nature-based recreation. Proceedings of the National Academy of Science 105(7): 2295–2300.

34.

Prakash

Perera

Newsome

, et al. (2019) Reasons for visitor dissatisfaction with wildlife tourism experiences at highly visited national parks in Sri Lanka. Journal of Outdoor Recreation and Tourism 25: 102–112.

35.

Runte

(1991) Promoting the golden west: advertising and the railroad. California History 70(1): 62–75.

36.

Schänzel

Yoeman

(2015) Trends in family tourism. Journal of Tourism Futures 1(2): 141–147.

37.

Sharpley

Pearce

(2007) Tourism, marketing, and sustainable development in the english national parks: the role of national park authorities. Journal of Sustainable Tourism 15(5): 557–573.

38.

Sim

Koo

C-D

Koo

, et al. (2018) An analysis of perceived crowding level reported by domestic visitors of south korean national parks: a multilevel ordered logit approach. Asia Pacific Journal of Tourism Research 23(3): 281–296.

39.

Upchurch

(2015) Zion national park, Utah: enhancing visitor experience through improved transportation. Transportation Research Record 2499: 40–44.

40.

US National Park Service (2014) Bryce Canyon National Park multimodal transportation plan EA. Available at: https://parkplanning.nps.gov/document.cfm?parkID=34projectID=37876documentID=57631 (accessed 31 January 2020).

41.

US National Park Service (2017a) Arches national park traffic congestion management plan. Available at: https://parkplanning.nps.gov/document.cfm?parkID=25projectID=59437documentID=83785 (accessed 31 January 2020).

42.

U.S. National Park Service (2017b) Zion national park visitor use management plan and EA: preliminary alternative concepts. Available at: https://parkplanning.nps.gov/document.cfm?parkID=113projectID=58542documentID=81614 (accessed 31 January 2020).

43.

US National Park Service (2018a) Find your park. Available at: https://www.nps.gov/subjects/centennial/findyourpark.htm (accessed 31 January 2020).

44.

US National Park Service (USNPS) (2018b) Visitor use statistics. Available at: https://irma.nps.gov/Stats/ (accessed 30 March 2019).

45.

Utah Governor’s Office of Economic Development (Utah GOED) (2013a). Mighty 5 advertisement. Available at: https://www.youtube.com/watch?v=QlKOcyf915M (accessed 30 March 2019).

46.

Utah Governor’s Office of Economic Development (Utah GOED) (2013b) Utah launches the Mighty Five™ spring/summer advertising campaign. Available at: https://business.utah.gov/news/utah-launches-mighty-five-springsummer-advertising-campaign/ (accessed 30 March 2019).

47.

Utah Governor’s Office of Economic Development (Utah GOED) (2016) Road to the Mighty 5 advertisement. Available at: https://www.visitutah.com/road-to-mighty/ (accessed 30 March 2019).

48.

Walls

(2016) Protecting our national parks: new entrance fees can help. Common Resources (18 April 2016) Available at: https://www.youtube.com/watch?v=QlKOcyf915M (accessed 31 January 2020).

49.

Wan

S-K

Xie

Hsiao

(2018) Panel data approach vs synthetic control method. Economic Letters 164: 121–123.

50.

Wilson

(2014) America’s Public Lands: From Yellowstone to Smokey Bear and Beyond, Lanham: Rowman & Littlefield.