As many American states have recently considered policies that make it more difficult for residents to vote, some political elites, commentators, and business leaders have warned these laws would negatively impact states’ economies by discouraging tourism. Some such harms have manifested through organizers relocating events, such as Major League Baseball moving its 2021 All-Star Game from Atlanta, Georgia, to Denver, Colorado, following Georgia’s passage of new voting restrictions – a move estimated to have cost Georgia over $100 million in economic activity.Footnote 1 However, observers also warn that these and other laws perceived to undermine democracy motivate individuals to avoid travel to states adopting them; for instance, a report from The Perryman Group on restrictive voting legislation Texas passed in 2021 warned the $6.6 billion it projected the state would lose in tourism revenue by 2025 was partially attributable to “socially conscious consumers” who would vacation elsewhere.Footnote 2 Thus, individual revulsion at backsliding and decisions to steer tourism dollars elsewhere has been posited as a mechanism by which the public can constrain anti-democratic impulses.
This paper examines whether states’ adoption of democratic backsliding policies affects leisure travel preferences consistent with this mechanism. While recent work indicates individuals are less willing to accept jobs in states that adopt backsliding policies (Nelson and Witko Reference Nelson and Witko2022, Reference Nelson and Witkon.d.), tourists are not personally affected by restrictions on democratic rights like voting; therefore, tourists’ proclivity to boycott backsliding states must rely on affinity for democracy rather than self-interest in living under a democratic government. That consumer behavior can be influenced by corporations’ partisan political activities offers some hope that support for democracy might drive vacation preferences (e.g., Kam and Deichert Reference Kam and Deichert2020; Panagopoulos et al. Reference Panagopoulos, Green, Krasno, Schwam-Baird and Endres2020), but it is unclear if pro-democracy inclinations are as powerful as partisanship. Consequently, it is unknown if states face consequences through individual-level tourism preferences that might discourage backsliding.
We evaluate this mechanism through two pre-registered survey experiments assessing how a state’s backsliding conditions respondents’ vacation preferences. We find that backsliding reduces respondents’ interest in vacationing in affected states; for instance, in our conjoint experiment, the negative effect of backsliding on respondents’ destination choices (1–3 percentage points) is similar to or larger than the effects of other important leisure travel considerations, such as increasing travel time to the destination by ≈2 hours or decreasing the destination’s average temperature by 3–6°F. Further, we find that this effect is driven largely by Democratic respondents.
Our findings suggest individuals’ aversion to backsliding not only influences decisions where their self-interest is at stake but also where condoning backsliding entails no personal costs. States should be conscious that backsliding not only discourages organizations from holding events in their jurisdictions but also deters individuals from traveling there. However, as we discuss in our Conclusion, the effects of these individual-level sanctions are of modest size and may cost states, in aggregate, less than 1% of their annual gross domestic product – a cost which may not, on its own, deter policymakers from backsliding.
Backsliding as a deterrent to tourism
Scholars and political observers have expressed alarm at the recent rise in democratic backsliding, or policies or actions that erode fundamental democratic institutions, in the United States. While many of these concerns stem from events during Donald Trump’s presidency, such as interfering in government investigations and stoking political violence (e.g., Carey et al. Reference Carey, Helmke, Nyhan, Sanders and Stokes2019; Levitsky and Ziblatt Reference Levitsky and Ziblatt2018), backsliding has manifested in many other contexts. Grumbach (Reference Grumbach2022) notes marked declines between 2000 and 2018 in the quality of democracy in states under unified Republican control relative to those under unified Democratic control or divided government. Even after Trump’s departure from office, many state and local governments have continued to threaten basic democratic principles by considering or passing legislation making it more difficult to vote,Footnote 3 criminalizing protests,Footnote 4 and drawing electoral districts to dilute the voting power of certain partisan or racial groups.Footnote 5
As backsliding has become salient in the United States, recent studies have endeavored to not only explain these trends but also understand backsliding’s behavioral consequences for Americans. For instance, Nelson and Witko (Reference Nelson and Witko2022, Reference Nelson and Witkon.d.) find states’ adoption of backsliding policies makes people less willing to accept jobs in those states. Differently, Simonson et al. (Reference Simonson, Lacombe, Green and Druckmann.d.) argue individuals who perceive backsliding and instability increase their gun-buying activity. Additionally, Schneider (Reference Schneider2022) demonstrates that when out-partisans in control of the government commit backsliding actions, Americans are more supportive of copartisan candidates who pledge to retaliate with their own backsliding actions.
While these studies help illuminate the implications of backsliding on American political behavior, their ability to identify the mechanisms underlying these behaviors is limited. Principally, in these studies, individuals are assumed to both have predispositions toward democratic principles and be subject to the consequences of the backsliding policies in hypothetical or real ways. Put differently, individuals’ responses to backsliding may be influenced by personal beliefs and values or by self-interest (Chong Reference Chong2000). For instance, in Nelson and Witko (Reference Nelson and Witko2022), respondents may express aversion to accepting a job in a state that adopted backsliding policies because those policies clash with their beliefs and values or because they do not want to be subjected to them, respectively. Distinguishing between these mechanisms is important because many posited deterrents to backsliding across the United States, such as boycotting backsliding jurisdictions, depend on persons not directly affected by backsliding; for instance, tourists who spend a few days in a backsliding state are not impacted by its voting laws. Thus, the effectiveness of those deterrents relies on the degree to which unaffected persons’ democratic predispositions prompt them to sanction backsliding.
Findings from recent studies on political consumerism provide optimism that backsliding can influence tourism. Like basing vacation choices on backsliding, consumers whose purchasing decisions are influenced by companies’ political activities do not bear the costs of those activities; rather, individuals engage in political consumerism to align purchasing habits with political predispositions (Newman and Bartels Reference Newman and Bartels2011). For instance, Kam and Deichert (Reference Kam and Deichert2020) find consumers are more likely to boycott or patronize businesses that treat workers poorly or well, respectively, and that the effect of poor treatment on boycotting overshadows that of positive treatment on patronizing. Relatedly, Panagopoulos et al. (Reference Panagopoulos, Green, Krasno, Schwam-Baird and Endres2020) demonstrate consumers are less (more) likely to patronize companies who donate to non-copartisan (copartisan) candidates of the consumer. Similarly, we expect individuals are less likely to vacation in states that adopt backsliding policies.
In an era of heightened polarization where Americans increasingly view all objects through a partisan lens, it is important to consider how partisanship might condition reactions to backsliding. While members of both the Democratic and Republican parties embrace basic tenets of democracy, recent studies indicate Democrats react more negatively to backsliding. For instance, Nelson and Witko (Reference Nelson and Witko2022, Reference Nelson and Witkon.d.) find that a state’s adoption of a backsliding policy makes Democrats, but not Republicans, less willing to relocate there (see also Carey et al. Reference Carey, Helmke, Nyhan, Sanders and Stokes2019; Simonovits et al. Reference Simonovits, McCoy and Littvay2022). Thus, we expect that the negative effect of backsliding on travel preferences is stronger among Democrats than Republicans.
Research design and analysis
Assessing the effect of democratic backsliding on tourism preferences is difficult because natural variation in the quality of democracy among US states is correlated with other factors that may also influence vacation choices, such as region and party control (Grumbach Reference Grumbach2022). Consequently, any relationship between tourism states’ experience and their quality of democracy may be spurious.
Alternatively, we conduct two experiments that enable us to isolate the causal effect of backsliding by manipulating the presence or salience of backsliding policies (Miller and Smith Reference Miller and Smith2023). Our first study utilizes a conjoint experiment, which allows us to observe how respondents utilize information about backsliding when embedded in a multidimensional decision-making context alongside other factors relevant to tourism. Our second study alters the salience of a backsliding policy – making it more difficult to vote by mail – recently adopted by Florida, one of the most traveled-to states; in doing so, we assess how emphasizing backsliding in the real-world political milieu affects behavior. As each design involves tradeoffs, we employ both as complements that address potential internal and external validity concerns in each.
Study 1
Our first study utilizes a conjoint experiment, which allows researchers to mimic multidimensional decision-making contexts respondents face in the real world by prompting them to compare profiles consisting of randomly assigned levels of a fixed set of attributes (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014). This design is appropriate for our context because individuals consider a range of destination characteristics when deciding where to vacation, such as the destination’s climate and attractions (Van Nostrand et al. Reference Van Nostrand, Sivaraman and Pinjari2013). Our design can not only identify the causal effect of backsliding on vacation preferences but also determine whether that effect persists in the presence of other considerations and compare its magnitude to those of other factors.
We fielded our experiment in December 2022 using nearly 2,100 respondents recruited through Lucid, which provides researchers samples whose characteristics mirror those of the US population.Footnote 6 Respondents were asked to imagine they had won a 2-week, all-expenses-paid vacation anywhere in the United States in July 2023.Footnote 7 Then, respondents were presented with 10 conjoint tasks, each containing 3 profiles of potential destinations for their vacation. Our profiles included randomized levels of six destination attributes.Footnote 8 Four attributes were apolitical: community type, average July temperature, travel time, and most popular attraction. A fifth attribute, “recent state news,” communicates a recent action by the destination’s state legislature.Footnote 9 Two levels of this attribute are consistent with democratic backsliding – that the legislature limited early voting or the right to protest at the state capitol.Footnote 10 Two more levels are associated with enhancing democracy – that the legislature expanded early voting or the right to protest at the state capitol. The fifth level concerns an action not related to the quality of democracy – that the legislature formed a commission to study economic growth. Our sixth attribute, “state-level 2020 presidential election result,” communicates information about the state’s partisan character. Including these final two attributes allows respondents to distinguish the destination state’s partisanship from any recent backsliding activity, allowing us to isolate the effect of the latter from the former.Footnote 11 In each task, respondents were asked to indicate their interest in each destination and select their most preferred destination.Footnote 12
Figure 1 displays the average marginal component effects (AMCEs) from our choice-based outcome, which reflect the change in the probability a profile with a given attribute-level is selected relative to a randomly generated profile with that attribute’s baseline level (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014).Footnote 13 The left pane presents the AMCEs for all respondents, while the center and right panes show the AMCEs for Democrats and Republicans, respectively. In our full sample, respondents were 3 percentage points less likely to choose destinations with our backsliding “recent state news” attribute-levels – that the destination’s state legislature limited early voting or protest rights – relative to profiles with the attribute’s baseline level that the legislature formed a commission to study economic growth. Additionally, respondents were between 1 and 2 percentage points less likely to choose destinations whose legislatures curtailed early voting or protest rights compared with legislatures that expanded either right.Footnote 14
Effect of democratic backsliding on destination choice. Points and lines represent the average marginal component effects (AMCEs) and 95% confidence intervals, respectively, for each attribute-level on respondents’ destination choice relative to its respective baseline. Left, center, and right panes present AMCEs among all respondents, only Democrats, and only Republicans, respectively.
While these differences may seem substantively small, comparisons with the AMCEs for other attributes traditionally associated with vacation choices reveal backsliding is consequential. For instance, for every ≈2 hour interval increase in travel time (e.g., moving from “less than 2 hours” to “2–4 hours”), respondents are 1–2 percentage points less likely to choose the more distant destination. Similarly, as the destination’s average July temperature decreases each step from 78°F to 72°F, 67°F, and 64°F (the 80th, 60th, 40th, and 20th percentile July temperatures among all states, respectively), respondents are 1–2 percentage points less likely to choose the cooler destination. Thus, the effect of backsliding on respondents’ choices is of similar or larger magnitude than the effect of ≈2 additional hours of travel time or a temperate decrease of 3–6°F. While our design does not allow us to assess whether backsliding would cause tourists to forego their ideal destinations (e.g., whether a nature lover would choose a place with museums that had not backslid vs. a locale with national parks that recently limited early voting; see Graham and Svolik Reference Graham and Svolik2020), these results indicate that, all else equal, tourists prefer destinations that have not recently backslid.
Turning to the partisan-conditional AMCEs, we find that Democrats behave similarly to our full sample, though they exhibit more positive effects for expanding early voting and protest rights. However, Republicans express similar levels of distaste for expansion and curtailment of both rights, as any legislative action decreases the probability of a profile’s selection by 3–4 percentage points. While Republican respondents’ aversion to destinations where rights have expanded is consistent with other recent work on partisan-conditional attitudes toward democratic policies (Nelson and Witko Reference Nelson and Witkon.d.), that they are similarly negative toward laws limiting those rights is unexpected. However, we are reticent to place emphasis on these Republican-conditional effects as they dissipate when using the rating outcome (see Supplemental Information Section SI.B1a).
Our findings from Study 1 support our expectations: respondents are less likely to prefer destinations that recently experienced backsliding relative to destinations that recently enhanced or have not taken action on democratic policies, and this effect manifests among Democrats but not Republicans. However, while our conjoint design allows us to mimic the multidimensional choice context in which individuals make vacation choices, it faces an inherent external validity limitation: because our destinations are abstracted, they do not incorporate the real-world milieu in which people choose destinations and omit details that may obviate the effect of backsliding. Consequently, we conduct a vignette-based study featuring Florida, a popular tourist destination whose recent backsliding policies have garnered national attention. This study constitutes a harder test of our expectations because respondents are likely pre-treated by not only their perceptions of Florida tourism but also news about Florida’s backsliding policies (Gaines et al. Reference Gaines, Kuklinski and Quirk2007).
Study 2
While we randomized the presence of backsliding in Study 1, we are unable to do so for natural stimuli as a destination’s quality of democracy is fixed. Thus, Study 2 instead manipulates the salience of Florida’s backsliding, which enables us to understand how highlighting backsliding influences tourism preferences.
We fielded our experiment in February 2023 using approximately 1,170 respondents recruited through CloudResearch, which provides researchers with samples whose characteristics mirror those of the US population.Footnote 15 Before exposure to our vignette, we asked all respondents to indicate on a five-point scale their interest in vacationing in five states popular for tourism: California, Florida, New York, Nevada, and Illinois. Then, respondents were asked to imagine they are considering vacationing in Florida and that they used an Internet search engine to find information. Below this prompt, respondents saw five search results. Four results were generic websites about Florida tourism, and one result mentioned a new law passed by the state legislature. While control condition respondents saw a result about Florida’s adoption of strawberry shortcake as the state dessert, treatment condition respondents received a result concerning a new law limiting residents’ ability to vote by mail – a backsliding policy that erodes electoral democracy (Grumbach Reference Grumbach2022). Respondents were then asked to again express their interest in vacationing in the same five states and indicate if they wanted more information about vacationing in each state. This first outcome question, together with its pre-treatment analog, facilitates a pre-post estimation of treatment effects on tourism attitudes (Clifford et al. Reference Clifford, Sheagley and Piston2021). Alternatively, the second outcome question, which signals willingness to move beyond merely expressing a preference and expending effort on information search, provides a more costly behavioral indicator of respondents’ preferences.
Figure 2 presents results for our interest and information outcomes in the left and right panels, respectively. Focusing first on the top-most set of points, we see no evidence backsliding is consequential for either outcome in the full sample; while respondents’ interest in vacationing in Florida decreases, as expected, this decrease is substantively small (−0.02 on a five-point scale) and not statistically distinguishable, and respondents’ desire for more information about Florida unexpectedly increases, though by a small, indistinguishable amount (increase in probability of 0.01). Turning to the middle and bottom-most points relating to Democratic and Republican respondents, respectively, we observe effects for our interest outcome consistent with our findings in Study 1 – Democrats are distinguishably less interested in vacationing in Florida when informed of its new voting restrictions (−0.08). The corresponding effect for Republicans is of similar magnitude but positive (0.07), suggesting Republicans are more interested in vacationing in Florida when made aware of its backsliding; however, this effect is not distinguishable at the 95% level (p=0.15) and, unlike the negative effect among Democrats, shrinks considerably when the sample is reweighted to reflect the US population.Footnote 16 Additionally, the effect of backsliding on requesting more information is small and not distinguishable for both Democrats (0.02) and Republicans (−0.02), suggesting the treatment did not impact this behavioral outcome in the same way it changed interest.Footnote 17
Effect of democratic backsliding on attitudes toward vacationing in Florida. Points and lines represent the treatment effects and 95% confidence intervals, respectively, associated with informing respondents that Florida recently restricted early voting relative to the control condition on interest in vacationing in Florida (left pane) and requesting more information about vacationing in Florida (right pane). Top-most points in each pane indicate treatment effects for the full sample, while the middle and bottom-most points indicate effects among Democrats and Republicans, respectively.
Our effects are more modest in Study 2 relative to Study 1 in that we detect a distinguishable effect of backsliding only among Democrats and that effect – a decrease of 0.08 in interest on a five-point scale – is substantively small. However, it is important to note Study 2, which moves away from abstract destinations by incorporating the real-world political milieu, constitutes a harder test of our expectations. Therefore, that we recovered similar effects for destination preferences among the subgroup most averse to backsliding – Democrats – reinforces confidence in the corresponding finding from Study 1.
Conclusion
For many people, vacation destinations are ostensibly apolitical; individuals often choose where to vacation based on what they find enjoyable and relaxing. However, our findings indicate that even in this apolitical context, states’ democratic backsliding decreases tourists’ interest in traveling to those destinations – even though they would be unaffected by those policies during their short trips. Indeed, to the extent that preferences shape ultimate decision-making, when combining our estimated effects of backsliding on destination choice among all respondents in Study 1 – decreases in the share of tourists selecting those destinations of 1–3% – with tourism data from Florida and Georgia – states whose recent backsliding have garnered national attention – we project that adopting backsliding policies could cost those states $981 million to $2.9 billion and $644.6 million to $1.9 billion in annual tourism-related economic activity, respectively.Footnote 18
However, it is important to acknowledge that these economic costs may not, on their own, dissuade policymakers from adopting backsliding policies. Indeed, policymakers may enjoy separate benefits from backsliding, such as higher approval ratings among their core constituents (Schneider Reference Schneider2022), that they weigh against potential harms to their states. Thus, while our findings suggest states face economic costs to backsliding, this accountability mechanism may only temper backsliding to the extent that policymakers place more value on potential tourism activity losses relative to political gains. Given that our negative effects of backsliding are distinguishable but of modest size – our calculated losses in economic activity for both Florida and Georgia represent less than 0.3% of those states’ gross domestic products – some policymakers may accept this tradeoff.
Our analysis also reinforces a normatively troubling finding in other recent studies on attitudes toward democracy in the United States: like many other aspects of American life, individuals’ preferences regarding democracy itself have polarized along partisan lines. Namely, while Democrats distinguish between policies that expand or contract democracy and reward or punish accordingly, Republicans seemingly fail to respond differentially to democracy-enhancing and democracy-eroding policies in ways that hold governments accountable (Carey et al. Reference Carey, Helmke, Nyhan, Sanders and Stokes2019; Nelson and Witko Reference Nelson and Witko2022, Reference Nelson and Witkon.d.; Simonovits et al. Reference Simonovits, McCoy and Littvay2022; but see Graham and Svolik Reference Graham and Svolik2020). Future work should consider the basis for these differences. That Republican elites are responsible for most recent backsliding policies may cue the public to interpret these policies along partisan lines (Grumbach Reference Grumbach2022). Alternatively, differences between Democrats and Republicans may stem from fundamentally different conceptions of democracy (Davis et al. Reference Davis, Goidel and Gaddie2022). Uncovering the sources of these dissimilar responses and discerning how to encourage all Americans to uphold core institutions and norms are essential for preserving democracy.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/XPS.2023.40
Data availability
The experiments included in this paper were pre-registered through the Evidence in Governance and Politics (EGAP) Open Science Framework (OSF) registry (Study 1, https://osf.io/5zt2u; Study 2, https://osf.io/nctze). The data, code, and any additional materials required to replicate all analyses in this article are available in the Journal of Experimental Political Science Dataverse within the Harvard Dataverse Network, at doi: https://doi.org/10.7910/DVN/KA7DLE.
Acknowledgments
We thank Michael Nelson, Erin Rossiter, and Chris Witko for their helpful feedback and guidance on this project.
Competing interests
The authors declare none.
Funding
Support for this research was provided by East Tennessee State University’s Honors College and Office of Research and Sponsored Programs.
Ethics statement
The surveys in which our experiments were embedded received approval from the Institutional Review Board of East Tennessee State University (Study 1 #c0922.4e; Study 2 #c0223.6e). Our research adheres to APSA’s Principles and Guidance for Human Subjects Research (see Supplemental Information Sections SI.A1 and SI.A2 for descriptions of our protocols which address these points).
Open access
