Partisan Conflict, National Security Policy Uncertainty and Tourism

Abstract

This study investigates the spillover effects among partisan conflict, national security policy uncertainty and tourism (i.e., tourist arrivals, exports, and stock) in the U.S. by using the TVP-VAR-based connectedness measures. Specifically, we discuss the association strength, spillover direction and dynamic linkages among the three under this framework. The results show that partisan conflict and national security policy uncertainty are net transmitters of spillovers to tourism, and those effects are more potent for inbound tourism demand than tourism stock performance. Moreover, the magnitude of spillovers among the three is time-varying and increases significantly in times of crisis, especially during the 9/11 attacks, the global financial crisis and the COVID-19 pandemic. Our results have important implications for tourism managers to develop sustainable development strategies to buffer or adapt to the uncertainty impact.

1. Introduction

The diversity of activities in the tourism industry makes tourism a highly connected and interconnected economic sector, making tourism highly vulnerable to uncertainty [1,2]. There are many sources of uncertainty, such as terrorist attacks, natural disasters, political instability, war outbreak, and other negative events that threaten tourism. Partisan conflict is one of the most worrying issues in the near future. The partisan conflict has been shown to increase the probability of crises and thus create uncertainty, as evidenced by the 2018 Sino–US trade war, the 20-year war in Afghanistan, and the recent divergence of views between different political parties during the COVID-19 pandemic [3,4]. The growing polarization creates uncertainty about policy choices and sustainable tourism [5]. Some scholars have documented that partisan conflict is also closely related to tourism exports and reduces returns on travel stocks [6,7]. It is not difficult to find evidence that the influence of partisan conflict has increasingly penetrated every sphere of tourism.

Meanwhile, the tourism industry is vulnerable to security threat shocks. The tremendous effect of the coronavirus pandemic on tourism has proved the fragility of the U.S. tourism industry to security threats. As reported, from March 2020 to October 2021, the impacts of the border closure measures led to a loss of almost 300 billion USD in export earnings and exceeded one million jobs (U.S. Travel Association). Indeed, the choice of destination for tourists relies upon the safety conditions of the goal. A large number of studies have proved that national security threats exert a notable effect on tourism markets [8,9,10,11]. Overall, the influence of national security uncertainty on tourism development has been stressed.

Moreover, some studies have observed the relationship between partisan conflict and national security policy uncertainty. On the one hand, conflicts between different parties create uncertainty about all policy choices [12,13]. For example, partisan conflicts regarding the COVID-19 measures contributed to negative feedback in the U.S., threatening the short-term sustainability of the response with severe consequences [14]. With growing congressional influence, partisan conflicts in Congress increase the likelihood that the national security policy bill will stall [15,16]. On the other hand, national security uncertainty also influences partisan conflict. For example, national security threats can slow partisanship, especially under great management pressure during stressful times since political leaders have low goals and do not pursue partisan interests during this period [17]. Obviously, partisan conflict and national security policy uncertainty interact with each other. On this basis, we find a new question: whether or not there exists a connectedness between partisan conflict, national security policy uncertainty and tourism markets.

To understand the complex hidden correlations, we explore the spillover effects among partisan conflict, national security policy uncertainty and tourism. In economic activities, the spillover effect is usually understood as the development of one aspect of something that drives the growth of other parts of something [18,19]. In short, it is not acquired within the economic activity itself, the effect is acquired outside the economic activity itself. As a highly connected sector, tourism is easily affected by ecological, social, and economic factors. Therefore, some scholars have researched the spillover effect between tourism and its influencing factors. For example, Charbel Bassil (2014) studied the individual impact of terrorism on travel demand in Lebanon, Turkey, and Israel. They show that domestic and transnational terrorism have significant spillover effects on travel demand in all three countries, depending on the intensity of terrorist attacks [20]. Abdelsalam et al. (2021) explored the spillover effects of COVID-19 pandemic uncertainty on the U.S. travel subsector. Through comparison, they found that catering was the most seriously affected industry in the tourism industry, and airlines followed closely [21]. The above studies provide a good inspiration for our study of spillover effects among partisan conflict, national security uncertainty and tourism, but these studies are generally based on regression models, which do not capture dynamic connectivity well. To overcome this shortcoming, we employ a mixed approach introduced by Antonakakis, Chatziantoniou and Gabauer (2020), which is comprised of the DY framework and a time-varying parameter vector autoregressive (TVP-VAR) model [22,23,24,25,26]. This method measures various levels of connectedness from pairs to system-wide connections. It is needless to specify a fixed rolling-window size compared to the original spillover index measures, so the potential variations in parameter values can be determined more accurately. These advantages give us more flexibility to capture possible changes in the connectedness among partisan conflict, national security policy uncertainty and tourism, thereby achieving accurate and robust results.

By examining the combined effects and dynamic transmission paths among partisan conflict, national security policy uncertainty, and tourism, we study for the first time the spillover effects of tourism in the dual context of internal strife and external threat. The contributions are threefold. Firstly, to the best of our knowledge, this is the first research on the connectedness between partisan conflict, national security policy uncertainty, and tourism markets. We provide initial evidence regarding the combined effects of partisan conflict and national security on tourism. Secondly, unlike previous studies that focused on the static impact of partisan conflict on tourism or national security uncertainty on tourism, this paper conducts a dynamic spillover analysis. Thus, the spillover’s time-varying direction and intensity are highlighted. Finally, because of market heterogeneity, considering the impact of partisan conflict and national security policy uncertainty on tourism—tourist arrivals, tourism exports, and the Tourism and Travel Stock Index—can uncover the variance of information spillover to each tourism market at a practical level.

The rest of this paper is organized as follows. Section 2 gives a brief review of the relevant literature. Section 3 presents the method and all data sources. Section 4 presents the empirical results and describes some valuable findings. Finally, Section 5 provides a summary and a discussion of some policy implications.

2. Literature Review

2.1. Impact of Partisan Conflict on Tourism

As a new influencing factor in the tourism market, the partisan conflict has attracted the attention of scholars in recent years. The study found that the impact of partisan conflict on tourism was negative in almost all countries. For example, Mexico’s multilayered partisan conflict led to greater drug violence [27]. The resulting violence is detrimental to the construction and development of tourist sites and affects consumer confidence in the next step. Rother et al. (2016) found that regional conflicts and refugee crises in the Middle East and North Africa have profoundly affected national economies [28]. Such large-scale conflicts are costly, and contagion to the tourism industry is inevitable after economic damage. The academic community has made a series of studies on the impact of party conflict in the United States. For example, Azzimonti (2018) confirmed the hypothesis that political disharmony inhibits private investment by studying the relationship between partisan conflict and private investment in the United States [29]. Apergis et al. (2021) investigated the impact of the uncertainty of party conflict in the United States on international oil prices [30]. They believe that the uncertainty of party conflict in the United States is a systemic risk factor for asset prices. Other scholars have studied the prediction ability of party conflict to income inequality [31]. The above studies have confirmed that the partisan conflict caused by the political polarization in the United States is typical in academic circles. Therefore, based on data availability, we studied the impact of partisan conflict on tourism in America.

Partisan conflict affects tourism in several ways. First, partisanship affects trade policy and, therefore, tourism policy. Left-wing parties prefer policies that increase government spending and induce growth. In the contrast, right-wing parties prefer policies that lead to lower spending, balanced budgets and lower inflation, which inevitably affect tourism [32,33]. Second, party conflict will affect the effectiveness of government policies in preventing adverse economic outcomes, as well as the possibility of tax reform, which will affect investment decisions. The resulting long-term investment changes will cause exchange rate fluctuations and further affect tourism [34]. Third, partisan disputes will also lead to changes in immigration policies. From the control of immigration during the Trump administration and the encouragement of the Biden administration to actively encourage and welcome high-quality immigrants, it can be found that immigration policies are different under different positions. When markets are flawed in the form of information asymmetry, migrant groups can effectively act as intermediaries and as catalysts for international tourism flows [35]. Finally, government interventions resulting from partisan conflicts, such as travel bans (external and internal) and border closures during COVID-19, can also limit tourism development [36].

In the empirical study, Azzimonti and Talbert (2014) proposed an alternative channel for political differences to influence economic decisions according to the standard party model of fiscal policy decisions [37]. They indicated that the intensification of polarization would induce economic policy uncertainty, thus leading to a decline in long-term investment. Through the estimated gravity models, Harb and Bassil (2020) found that the change in immigration ratio caused by immigration policy had a positive impact on the entry of tourists, which would offset the negative effect of terrorism on tourists [38]. In addition, the result of partisan conflict on tourism-related businesses’ financial performance and failure risk is also evident [39]. Intense partisanship reduces travel stock returns and strongly correlates with stock market volatility [40,41,42]. Some other scholars have proved that partisan conflict can predict the stock market, which also shows the close relationship between partisan conflict and tourism demand and the tourism stock market [43,44,45]. In all, the literature above reveals that partisanship has a significant impact on the tourism market. However, a detailed analysis of the different effects of partisan conflict on the physical tourism market and the stock market is lacking.

2.2. Impact of National Security Threats on Tourism

The travel industry is susceptible and prone to panic. Therefore, events such as war, terrorism, tension, ethnic and political violence can affect the behavior of investors and tourists. From the essence of international relations, countries are in an uncertain and Hobbesian world [46]. The concept of security is a more general, deeper and helpful way of studying international relations than power or peace. Tourism as a highly interrelated economic sector is deeply affected by security in international relations. Therefore, the study of the impact of national security on the world tourism market cannot be ignored. For most developing countries, national security is a complex phenomenon. Due to its underdeveloped economy and infrastructure, unstable political system, and ethnic or other social divisions, various problems pose a security threat. It is also reflected in the vulnerability and permeability of most developing societies and economies [47]. In this context, tourism security and development cannot be fully guaranteed. For example, in the Middle East, which is plagued by war, violence and acts of terrorism, as well as threats to crime, corruption, deadly diseases and other elements of national security, tourism is in crisis [48]. Among them are Israel, Iran and Turkey, which have been threatened by terrorism and war in recent years. According to Chisadza et al. (2022), economic instability in Africa has largely limited inbound arrivals. These regions have been hit by political instability and social unrest, which deter tourists [49]. Through the above analysis, we find that national security, whether war, violence or economy, cannot be ignored in developing countries. However, there are few conclusions about developed countries in relevant studies. Developed countries have robust infrastructure and economic development. Will they also be affected by national security in the tourism industry? In fact, large-scale migration and policy changes pose a threat to national security and political structure [50], in particular, the political atmosphere in the United States. Obviously, the polarization of Democrats and Republicans has further threatened America’s national security. Therefore, given America’s position in international relations, and the fact that it is a significant component of the global tourism market, this paper selects the United States as the research background of the spillover effects of the three, and fills this research gap.

National security uncertainty affects tourism in several ways. First, geopolitical risks caused by national security reduce tourism investment and exports and affect tourists’ choice of destinations [51]. International and local visitors mostly enjoy visiting countries or places with a past and present record of social, economic and political security. Second, in the context of countries suffering from armed struggle and terror, military conflict increases the perception of terror risk in the destination. Worse, it may plunge the country into a severe economic recession. It has shaken the country’s population’s sense of security and restricted inbound tourist traffic [52]. Finally, natural disasters threaten national security and affect tourism. The negative impact of disasters on inbound tourism is dominant [53]. Natural disasters affect travel demand by changing tourists’ perception of destination attractiveness and risk [54,55].

Several pieces of literature have empirically discussed the impact of national security threats on tourism. Perles-Ribes et al. (2016) analyzed the impact of economic crisis or threat on different tourist destinations in Spain [56]. They indicated that economic disasters would lead to an increase in the unemployment rate and a negative impact on tourist destinations. What is more, Neumayer and Plümper (2016) pointed out that terrorist attacks will also impact traveler flows to nearby destinations, and even safe areas will be affected by severe security events in neighboring countries [57,58]. Natural disasters and public health crises such as the recent COVID-19 pandemic are widely proven to have a detrimental and statistically significant effect on tourism demand [59,60,61]. In general, national security threats hurt the tourism sector to a large extent and wide range. However, the above literature is limited to case studies. Some aspects of national security cannot fully reflect a country’s tourism safety environment, and the results are not generalized enough. Therefore, the National Security Policy Uncertainty Index (hereinafter NSU) is used to that end in our research. The index was built by Baker, Bloom and Davis (2016), who pointed out that the method based on major newspaper coverage frequency is reliable, accurate, and relatively unbiased [62].

2.3. Relationship between Partisan Conflict and National Security Policy Uncertainty

Whether it is a contradiction caused by artificial factors, such as terrorism, extremism, or different opinions caused by natural disasters, such as hurricanes and epidemics, policy-makers must deal with it when the partisan conflict occurs. There may be a significant change in the degree of conflict, leading to a surge in political uncertainty. Finally, the anticipation of national security is unstable. Several studies have observed a link between partisan conflict and national security policy uncertainty. On the one hand, partisan battles strongly impact national security policy, whose decisions are fraught with uncertainty [63]. For example, Trejo and Ley (2016) found that in Mexico, the government’s political control and intervention failed to work due to the partisan conflict, resulting in a spiral of violent crime and increasing uncertainty about national security [64]. Similarly, partisan conflicts towards the COVID-19 measures contributed to negative feedback in the U.S., threatening the short-term sustainability of the response with serious consequences [14]. The disagreement is also evident over defense spending, whether to wage war and the immigration policy [65,66,67]. With growing congressional influence and partisan conflict in Congress, the likelihood that the national security policy bill will stall has increased.

On the other hand, national security threats also influence partisan conflict. For example, national security threats can slow partisanship, especially under great management pressure, such as dealing with the global epidemic [62]. It could also affect the election, resulting in a short-term increase in partisan conflict [68]. The differences could be due to different empirical periods. Overall, the mutual interactions are confirmed. Therefore, the relationship between partisan conflict and national security policy uncertainty should explain the spillover effect on U.S. tourism.

2.4. Analysis of Different Tourism Performance

The existing tourism literature mainly analyzes the impact of uncertainty on tourism from three perspectives: tourist arrivals, tourism revenues, and the performance of the tourism stock. The change in tourist arrivals and the tourism sector receipts mainly manifest the change in tourism demand. More specifically, the inbound tourists are closely related to the destination’s security environment and institutional quality [69]. Fourie, Rosselló-Nadal and Santana-Gallego (2020) studied how security threats hurt tourism using the number of tourist arrivals [58]. Lanouar and Goaied (2019) addressed the relationship between tourism, terrorist act and political turmoil in Tunisia [70]. Employing tourism exports, Petit and Seetaram (2019) measured the effect of revealed cultural preferences on tourism demand [71]. Hailemariam and Ivanovski (2021) exploited the negative effect of economic policy uncertainty on U.S. tourism net exports [72]. Saha and Yap (2014) measured the moderation effects of political unstableness and terrorist acts on the tourism sector by using two types of tourism demand data: tourist arrivals and tourism revenue [73].

The third angel focuses on the tourism stock market performance. For example, Zopiatis, Savva, Lambertides and McAleer (2019) used the stock indices to study the relationship between the tourism industry and unexpected non-macro incidents [74]. Demiralay (2020) employed the tourism stock index to evaluate the effect of partisan conflict in the tourism industry [7]. Bashir and Kumar (2022) found that tourism stock returns become more sensitive to stockholder awareness and policy-related economic uncertainty during the COVID-19 crisis [75]. In addition, it has been proved that the number of tourists is significant and positively affects tourism revenue. There is a causal relationship between the number of inbound tourists and the performance of U.S. tourism stocks [76,77,78]. In summary, both tourism demand and stock performance can be affected by partisan conflict and national security policy uncertainty, and there are close ties between different tourism markets. Hence, it is more comprehensive and reliable to consider both tourism demand and the different performance of tourism stocks, such as Lee and Chen (2021) and Lee, Chen and Peng (2021) [79,80]. In our study, three proxies: tourist arrivals, tourism exports, and Tourism and Travel Stock, are used to reflect spillovers between the three comprehensively. We will further plot the heterogeneity of the tourism market through a comparative analysis.

2.5. TVP-VAR Based Connectedness Measures

The spillover index introduced by Diebold and Yilmaz in measuring connectedness is widely used [24,25,26]. Combining with the rolling window, the dynamic relationship between the variables can also be obtained. Some papers used the DY model to explore the dynamic information connectedness in financial markets and cross-market information overflow effects [81,82,83]. It is also applied to study the external shocks on tourism demand, such as by Dragouni et al. (2016), who used the spillover index to explore whether U.S. outbound travel needs received a spillover effect from sentiment and mood shocks [84]. Afterwards, several improvement methods are proposed for an extension. Baruník et al. (2016) improved the measurement to quantify asymmetries in volatility spillovers [85]. Gamba-Santamaria et al. (2017) improved the calculation of volatility by using a DCC-GARCH framework [86]. Demirer et al. (2018) proposed the LASSO-VAR model to address the low precision problem resulting from high variable dimensions in the origin method [87]. Despite that, a disadvantage that relies on window width selection remains unsettled. Thus, there is still room for further advancement to overcome some shortcomings regarding connectedness measures.

Antonakakis and Gabauer (2017) first extended the DY framework with the TVP-VAR, enabling variances to differ through a stochastic volatility Kalman filter estimation with forgetting factors [88]. The potential changes in parameter values can be more accurately determined than the original measures. Moreover, Korobilis and Yilmaz (2018) proved that the dynamic connectivity method based on TVP-VAR performs better than the mobile window method in significant crisis moments [89]. Some articles use this method to study the dynamic connectivity of macroeconomic uncertainty indices, economic policy uncertainty indices, the four most traded foreign exchange rates, news sentiment, economic uncertainty and macroeconomic indicators and agricultural products and crude oil futures prices [90,91,92,93]. We believe that applying this method in our research is feasible through their robust results.

To sum up, the existing literature explores the impact of partisan conflict or national security uncertainty on tourism development, including tourism demand and stock performance. The possible existence of the heterogeneity of the tourism market remains under-researched. Therefore, this article aims to address these gaps using our investigation’s TVP-VAR-based measure of connectedness.

3. Methodology and Data

3.1. A Mixed-Method of DY Framework with TVP-VAR Model

To analyze the time-varying transmission mechanism, we apply a mixed methodology introduced by Antonakakis et al. (2020), which is comprised of the DY framework and the TVP-VAR model [22,23,24,25,26]. The model is based on a time-varying variance–covariance structure, which allows possible changes in the underlying structure of the data to be captured more flexibly and robustly. Specifically, due to the improvement brought by the construction, this method does not need the rolling window analysis. Therefore, there is no need for observation loss when calculating dynamic connectivity measures. Since the proposed framework is based on a multivariable Kalman filter, it is less sensitive to outliers [22].

Specifically, the TVP-VAR model can be represented below,

$$y_t={\beta }_t{y}_{t-1}+{\epsilon }_t\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\epsilon }_t|M_{t-1}~N\left(0,E_t\right)$$

(1)

$$vec\left({\beta }_t\right)=vec\left({\beta }_{t-1}\right)+v_t\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{v}_t|M_{t-1}~N\left(0,R_t\right)$$

(2)

where $y_t$ and $y_{t-1}$ are N × 1 and $N_P\times 1$ dimensional vectors, individually.${\beta }_t$ is an $N\times N_P$ dimensional dynamic coefficients matrix. ${\epsilon }_t$ is an N × 1 dimensional error disturbance vector.$vec\left({\beta }_t\right)$, $vec\left({\beta }_{t-1}\right)$ and $v_t$ are $N_{p }^2\times 1$ dimensional vectors. $E_t$ and $R_t$ represent the N × N and $N_{p }^2\times N_{p }^2$ dimensional time-varying variance–covariance matrix, respectively.

Then, to figure the generalized impulse response functions (GIRF) and generalized forecast error variance decomposition (GFEVD), we should alter the VAR to its vector moving average illustration first:

$$y_t={\displaystyle \sum }_{j=0}^{\infty }{F}^{\prime }{G}_t^{j}F{\epsilon }_{t-j}$$

(3)

$$y_t={\displaystyle \sum }_{j=0}^{\infty }{W}_{it}{\epsilon }_{t-j}$$

(4)

The generalized impulse response functions represent all variables’ responses after receiving a shock from variable $i$. There is no structural model, for this reason, we count the differences between a K-step-ahead prediction where once variable $i$ receives a shock and where once variable $i$ receives no shock. The shock on variable $i$ can be explained by this difference, which is counted by

$$GIR{F}_t\left(K,{\tau }_{j,t},M_{t-1}\right)=E\left(Y_{t+K}|{\epsilon }_{j,t}={\tau }_{j,t},M_{t-1}\right)-E\left(Y_{t+K}|M_{t-1}\right)$$

(5)

$${\mathsf{\Psi }}_{j,t}^g\left(K\right)=\frac{W_{K,t}{E}_t{\epsilon }_{j,t}}{\sqrt{S_{jj,t}}}\frac{{\tau }_{j,t}}{\sqrt{S_{jj,t}}}$$

(6)

$${\mathsf{\Psi }}_{j,t}^g\left(K\right)=S_{jj,t}^{-\frac{1}{2}}{W}_{K,t}{E}_t{\epsilon }_{j,t}$$

(7)

With $S_{jj,t}={\tau }_{j,t}^2 , {\tau }_{j,t}$ is the selection variable, which is one when on the $j$ position, otherwise zero. $M_{t-1}$ denotes the information set until $t-1$. ${\mathsf{\Psi }}_{j,t}^g\left(K\right)$ shows the generalized impulse response functions of the variable $j$ and $K$ is the forecast horizon.

Afterwards, we calculate the generalized forecast error variance decomposition which represents the variance share of a variable over others. The calculation is as follows:

$${\tilde{\varphi }}_{ij,t}^g\left(K\right)=\frac{{{\displaystyle \sum }}_{t=1}^{K-1}{ \mathsf{\Psi }}_{ij,t}^{2,g}{ \mathsf{\Psi }}_{ij,t}^{2,g}}{{{\displaystyle \sum }}_{j=1}^N{{\displaystyle \sum }}_{t=1}^{K-1}{ \mathsf{\Psi }}_{ij,t}^{2,g}}$$

(8)

With ${\displaystyle \sum }_{j=1}^N{\tilde{\varphi }}_{ij,t}^g\left(K\right)=1$ and ${\displaystyle \sum }_{i,j=1}^N{\tilde{\varphi }}_{ij,t}^N\left(K\right)=N$. After being normalized, each row adds up to 1, meaning that the forecast error variance of variables $i$ is explained by all variables.

Using the generalized forecast error variance decomposition, we build the total connectedness index (TCI), which represents the common contribution of cross-variable shock spillovers to the variance of total prediction error. In short, it measures the average interdependence between the experimental variables. This index is defined as

$$C_t^g\left(K\right)=\frac{{{\displaystyle \sum }}_{i,j=1,i\ne j}^N{\tilde{\varphi }}_{ij,t}^g\left(K\right)}{N}\times 100$$

(9)

Directed spillovers break down total spillovers into overflows that originate (or flow to) a specific source.

Firstly, we focus on the total directional connectedness to others, which can be described as the impact from variable $i$ to all other variables $j$. The definition is as follows:

$$C_{i\to j,t}^g\left(K\right)=\frac{{{\displaystyle \sum }}_{j=1,i\ne j}^N{\tilde{\varphi }}_{ji,t}^g\left(K\right)}{{{\displaystyle \sum }}_{j=1}^N{\tilde{\varphi }}_{ji,t}^g\left(K\right)}\times 100$$

(10)

Secondly, we compute the total directional connectedness from others, which can be described as that variable $i$ receiving a shock from all other variables $j$. The calculation is as follows:

$$C_{i\leftarrow j,t}^g\left(K\right)=\frac{{{\displaystyle \sum }}_{j=1,i\ne j}^N{\tilde{\varphi }}_{ij,t}^g\left(K\right)}{{{\displaystyle \sum }}_{i=1}^N{\tilde{\varphi }}_{ij,t}^g\left(K\right)}\times 100$$

(11)

Thirdly, we subtract total directional connectedness to others from total directional connectedness from others to obtain the net total directional connectedness, which helps better understand potential dynamics. It is defined as

$$C_{i,t}^g\left(K\right)=C_{i\to j,t}^g\left(K\right)-C_{i\leftarrow j,t}^g\left(K\right)$$

(12)

$C_{i,t}^g>0$ means that variable $i$ drives the network; on the contrary, $C_{i,t}^g<0$ means that variable $i$ is driven by the network.

Finally, we decompose the net total directional connectedness into pairs of connectivity between the specified two variables and obtain the net pairwise directional connectedness:

$$NPD{C}_{ij}\left(K\right)=\frac{{\tilde{\varphi }}_{ji,t}^g\left(K\right)-{\tilde{\varphi }}_{ij,t}^g\left(K\right)}{T}\times 100$$

(13)

3.2. Data

Our indicators for the travel market are Inbound, Exports, and DJUSTT, which represent the inbound tourist departures from all over the world to the U.S., the U.S. travel spending, and Tourism and Travel Stock Index, respectively. We use the PCI index as a proxy for partisan conflict, which is constructed by Azzimonti (2014) [94]. The NSU index is a proxy for national security policy uncertainty, built by Baker et al. (2016) [62]. Both indices are based on the coverage frequency of major newspapers containing related terms. The data on the PCI index comes from the Federal Reserve Bank of Philadelphia (https://www.philadelphiafed.org/research-and-data/real-time-center/partisan-conflict-index (accessed on 18 July 2022)) and the data on the NSU index is obtained from the website of Economic Policy Uncertainty (www.policyuncertainty.com (accessed on 18 July 2022)). The data on all travel indicators are downloaded from Bloomberg.

Figure 1 plots the graph description of raw data trends for all variables. The partisan conflict has fluctuated dramatically since 2008, peaking around 2013 and 2017. The national security policy uncertainty is volatile, and there is a peak in 2001 and 2003 corresponding to the breakout of 9/11 and the war on terror it triggered. Regarding the tourism market, the trends in inbound tourism demand, tourism exports and the Tourism and Travel Stock Index are very similar. The overall trend is increasing until the outbreak of COVID-19.

The sample period for PCI, NSU, and Inbound is January 1996 until December 2021, the Exports are from January 1999 to December 2021, and the DJUSTT is from January 2005 to December 2021. We take the log difference for Exports and DJUSTT. The PCI, NSU, and Inbound are in logarithm. Moreover, we test the stationarity of variables using the augmented Dickey–Fuller and Phillips–Perron models, finding that no sequence has a unit root at a significance level exceeding 1%. The summary statistics and the unit root checks for each data sequence are shown in

Table 1. Descriptive statistics.

Variables	Mean	Max	Min	Std. Dev.	Kurt	Skew	ADF	PP
PCI	4.687	5.603	4.081	0.326	2.745	0.149	−4.553 a	−3.963 a
NSU	4.247	6.386	3.255	0.591	4.092	0.942	−7.376 a	−7.341 a
Inbound	15.214	15.946	13.170	0.519	10.627	−2.205	−3.792 a	−3.911 a
Exports	−0.001	0.310	−0.158	0.064	64.405	−5.972	−10.076 a	−10.102 a
DJUSTT	0.010	0.318	−0.237	0.093	5.189	−0.328	−13.571 a	−13.656 a

Note: The symbol ^a denotes significance at 1% level.

.

4. Empirical Results

4.1. Static Analysis

Following the DY framework, we obtain the connectedness for partisan conflict, national security policy uncertainty, and the tourism market in the entire sample.

4.1.1. Total Connectedness

Table 2. Spillover table (in %): PCI, NSU, and tourism.

Panel A	Inbound	PCI	NSU	FROM
Inbound	69.8	19.5	10.7	30.2
PCI	17.6	79.1	3.3	20.9
NSU	5.9	4.7	89.4	10.6
TO	23.5	24.2	14.0	61.8
NET directional connectedness	−6.7	3.4	3.4	20.6
Panel B	Exports	PCI	NSU	FROM
Exports	84.7	5.8	9.5	15.3
PCI	4.8	91.1	4.1	8.9
NSU	5.6	5.2	89.3	10.7
TO	10.3	10.9	13.6	34.9
NET directional connectedness	−4.9	2.0	2.9	11.6
Panel C	DJUSTT	PCI	NSU	FROM
DJUSTT	89.6	2.2	8.2	10.4
PCI	0.7	91.3	8.0	8.7
NSU	3.9	3.9	92.2	7.8
TO	4.6	6.1	16.2	26.9
NET directional connectedness	−5.8	−2.6	8.4	9.0

Note: The bolded one is the directional spillover index, which represents the total connectivity.

reports the average connectedness of the total sample, offering an outline of the average spillovers and the close “input-output” disintegration of the full spillovers. We break it down into three sub-tables: Panels A, B and C.

From Panel A, the average total connectedness index (TCI) is 20.6%, which indicates strong interconnection between PCI, NSU, and Inbound. From the directional spillover point of view, the total directional spillover index is as high as 61.8% in which Inbound is the largest recipient (30.2%) and PCI is the most significant contributor (24.2%). PCI contributes much more spillover to Inbound (19.5%) than NSU (10.7%), which may be because the uncertainty of security policy stems from partisan conflict. The results show that partisan conflict significantly impacts inbound tourist flows, far greater than the impact of national security uncertainty. Institutional quality is the driving force for tourism volume growth [69]. On the contrary, the disunity of political parties caused by partisan conflict will greatly affect the country’s security and economic decisions, which will further affect tourists’ travel decisions and their perception of destination security, resulting in a decline in the number of inbound tourists.

Panel B in Table 2 shows the average connectedness for PCI, NSU, and Exports. The directional spillover index is still high (34.9%), whereas the average TCI (11.6%) is lower than that of PCI, NSU, and Inbound, given that inbound tourism revenue is less volatile than tourist flows. Exports is the largest recipient (15.3%), followed by NSU (10.7%). Although NSU contributes more spillover to Exports (9.5%), PCI is still the net transmitter. The results indicate that both PCI and NSU affect Exports significantly, while NSU is the net receiver of shock from PCI. It is consistent with our discussion in the theoretical section, that is, partisan conflict can affect national security and thus tourism. This result recognizes the dual role of national security as both producer and mediator of spillover effects. Therefore, part of the spillovers from NSU to Exports comes from the effect of PCI on NSU. Although inbound tourism revenue is more sensitive to changes in national security policy, partisanship has a direct net spillover effect on national security policy.

In Panel C, both the total spillover (9%) and directional spillover (26.9%) are smaller than above, showing that partisan conflict and national security policy uncertainty has a greater impact on the tourism demand than the stock performance. Even though more sensitive, the effect of PCI and NSU on DJUSTT is short-lived. It can be interpreted as the tourism stock market having strong interlinkage with other financial markets in the context of the integration of financial markets. Travel stock performance is affected by more uncertainties than travel demand, so it digests information and risk faster than the actual market. The bilateral impact is the greatest between NSU and DJUSTT. DJUSTT is the largest receiver (10.4%), and NSU is the largest contributor (16.2%). The results show that rather than PCI, DJUSTT is susceptible to NSU. Different from the above results of tourism Inbound and Exports, it is evident that national security has a greater spillover pass-through effect in the tourism stock market. The stock market is highly liquid, while unstable or turbulent countries have a more direct impact on the tourism stock market. After all, partisan conflicts need to play a role in economic and political decision-making, which takes some time to reflect. These findings support and expand the extant studies, showing that policy uncertainty shocks tend to increase volatility in financial markets. In contrast, partisan conflict shocks tend to reduce volatility in financial markets [41,42].

4.1.2. Net Pairwise Connectedness

The net spillovers help us to more intuitively understand how much each market contributes to others. We calculate the net pairwise spillovers based on the table above. Figure 2 exhibits the net directional spillovers of PCI, NSU, and tourism. The number on the arrows represents a specific pairwise overflow, and the arrow sizes show the size of the pairwise directional connectedness. We observe that PCI and NSU are the net transmitters of spillover shocks to both Inbound and Exports, in which the net spillover of PCI to Inbound (1.9%) is greater than that to Exports (1.0%).

Meanwhile, NSU is also the net recipient of PCI spillovers. This finding reveals a transmission mechanism in which part of the spillovers from PCI is transmitted first to NSU, then to Inbound and Exports. In the tourism stock market, the net spillover of NSU to DJUSTT is most significant (4.3%), suggesting that the transmission mechanism of the spillover effect is that NSU first transmits to PCI and then overflows to the tourism stock market. To sum up, shocks from partisan conflict are the main net transmitter and transmit spillovers to the real tourism market directly. Moreover, the partisan conflict also has an indirect impact on the delivery of national security policy uncertainty since partisan conflicts strongly impact national security policy whose decisions are fraught with uncertainty.

4.2. Dynamic Analysis

In this part, following the TVP-VAR-based connectedness measures, we obtain the dynamic spillover effects among partisan conflict, national security policy uncertainty and tourism.

4.2.1. Total Spillovers

Figure 3 shows the results of dynamic time-varying total spillover indices, which provide different and more information than the static spillover effect. There are five main jumps in Figure 3a,b, and four jumps in Figure 3c.

Firstly, the rising total connectivity in 1996–2001(Figure 3a) which likely to be the result of changes in U.S. immigration policy, the impact of the Asian financial crisis, and the Kosovo War. Against the backdrop of rising anti-immigrant sentiment, the U.S. government introduced stricter immigration laws in 1996, hindering the flow of inbound tourism. Because immigrants positively impact the number of inbound tourists, which upholds the Visiting Friends and Relatives (VFR) practical standpoint. Additionally, the shock of the 1998 Asian financial crisis on major source regions (Japan, South Korea, Hong Kong, and Taiwan) where the spending power of residents declined sharply, affected inbound tourism and tourism revenue to a large extent in the U.S. Moreover, the 1999 war in Kosovo significantly increased U.S. military spending and strained internal security, influencing destination selection decisions for inbound tourists.

Secondly, almost every variable has an intense connectivity period around 2001 (Figure 3a,b) because of the 9/11 terrorist attacks. Consequently, U.S. inbound tourism decreased by 12.7%, and tourism revenue decreased by 21.841 billion USD over the next two years (U.S. Travel and Tourism Satellite Accounts for 1998–2003). Meanwhile, it promoted partisanship over the measurement adoption. Republicans adopted a more aggressive military posture, whereas Democrats are more sensitive to information such as the consequences of war and inequality in conscription. Finally, the government revised immigration policy and tightened the 2000-mile border of U.S.–Mexico, and U.S.–Canada. All the policies affected tourism demand. We also can see the high persistent spillover index for 2005–2007; particularly, the fluctuations in Figure 3c are the most significant. During this time, the major natural disasters–the most active Atlantic hurricane season–caused a significant threat to national security and resulted in thousands of casualties and billions of dollars in damage. Therefore, we confirmed a substantial increase between partisan conflict, national security policy uncertainty, and the tourism market when partisan conflict increases and national security is threatened or undermined.

Thirdly, beginning with low values for the system spillover indices, total connectedness climbed back slightly in 2003 amid the most pronounced term of the SARS virus (Figure 3a,b). This crisis was built on the 9/11 terrorist attacks, which plunged U.S. tourism into a severe recession. Similarly, around 2020, all total spillover measures rise substantially and then peak because of the breakout of coronavirus. The global spread of the COVID-19 led many cities to issue travel bans, and lockdowns, with reduced tourism production and demand. At the time, nearly 75 million jobs in tourism were in danger, and the industry dropped more than 2.1 trillion USD in turnover (WTTC, 2020). Additionally, partisan conflicts towards the COVID-19 measures contributed to negative feedback in the U.S., threatening the short-term sustainability of the response with serious consequences. Overall, our findings reveal that the total spillovers between partisan conflict, national security policy uncertainty, and the tourism market increase observably in times of epidemic.

Finally, it is worth noting that there is a spike around 2008 in Figure 3c due to the financial crisis. Bear Stearns declared bankruptcy in March 2008, and Lehman Brothers filed for bankruptcy in September 2008, a financial turmoil that quickly evolved from Wall Street into a global financial tsunami, leading to severe stock market declines and recessions. Furthermore, oil prices climbed in late 2008 after OPEC cut production, increasing volatility in the travel stock market. Another peak of 2016–2017 in total connectedness could be driven by the U.S. presidential election, when partisanship in the United States was severe. In essence, tourism as a flexible commodity was sensitive to the shocks from regime change and presidential policy uncertainty. The latent reason being that this industry is sensitive to the government policy fluctuations and other associated aspects related to safety, social security and stability. Accordingly, we extrapolate that partisan conflict and security policy uncertainties resulting from the financial crisis and regime change enhance the spillover effects on the tourism market.

Overall, total spillover plots illuminate that the spillover effects of shocks among partisan conflict, security policy uncertainties and tourism increase significantly during major events. The demand for tourism is relaxation. Therefore, if tourists perceive danger, or are even just engaged in a tense condition, they will give up that travel.

4.2.2. Net Spillovers and Pairwise Spillovers

Figure 4 presents the results of net spillover effects. The net directional spillovers show prominent time-varying characteristics such as the total connectedness above, with the highest values generally reaching their height during turbulent times. Focusing on the net total connectedness results, it is evident that the tourism market is the primary recipient of spillover effects over the sample period. Partisan conflict and national security policy uncertainty are the main net transmitters while the dominant position of the two is different in different systems. For example, in the interrelationship between the PCI, NSU, and Exports (Figure 4b), PCI is the net transmitter of spillovers during almost the entire sample period except 2001. In the tourism stock market (Figure 4c), NSU is the main net transmitter while changing from information transmitter to information receiver after late 2010. Hence, separating the spillover effects from every indicator is necessary to obtain more specific information about the individual and the combined impact of partisan conflict and national security policy uncertainty on tourism.

We display the net pairwise spillovers for the partisan conflict–national security policy uncertainty, national security policy uncertainty–tourism, and partisan conflict–tourism in Figure 5. As we can see from Figure 5a, shocks to PCI and NSU are net transmitters of spillover effects to inbound tourism demand. The most apparent exclusion is the 1996–2003 period when tourism demand sends spillovers to the NSU. Specifically, the net transmitting role of inbound tourism peaked during 1999–2002. Even though unexpected, it can be interpreted as the truth that a flood of immigrants into the U.S. produced great uncertainty to national security and increased the pressure on government management simultaneously.

When focusing on the remaining periods, the magnitude of these spillover effects varies in partisan conflict and national security uncertainty, indicating that the influence on inbound tourism demand is heterogeneous. In 2013, the U.S. partisan conflict peaked when federal government departments closed. The spillover effect on Inbound reached a peak of about 7%, while the spillover effect of NSU was relatively low. This gap is even more pronounced after the outbreak of COVID-19 in 2020, with net spillover indices of 12% and 4%, individually. Furthermore, in the spillovers between PCI and NSU, PCI is a net transmitter of spillover effects during the entire sample period. Accordingly, we can infer that PCI increases uncertainty about security policies, affecting U.S. inbound tourism demand.

Starting with the interrelationship between the PCI, NSU, and Exports (Figure 5b), it is evident that the spillover effects are nearly zero during the sample period. PCI is also the net transmitter of spillovers to NSU during almost the whole sample period apart from 2001. There is a peak of about 10% from the national security policy uncertainty shock on partisan conflict, which was in tune with the 9/11 attacks and its consequence. Hence, the impact of partisan conflict and national security policy uncertainty shocks on inbound tourism revenues appear insignificant, though this does not remain during significant security threats. Here, again, our findings reflect the importance of national security policy and political stability to tourism.

Regarding the tourism stock market (Figure 5c), PCI is also the main net transmitter, and NSU changes from information transmitter to receiver after late 2010. Indeed, all the events of the United States debt-ceiling crisis in 2011, the European debt crisis in 2012, the Ebola and MERS epidemics during 2014–2015, global stock indexes fell sharply in 2015, and the Dow Jones stock index fell to its largest drop since the 2008 financial crisis in 2018 increased stock market volatility. An interesting discovery exists in the spillovers between PCI and NSU, where NSU is the net transmitter to PCI during 2005–2007, and PCI is the net transmitter to NSU during 2016–2017, which coincides with Atlantic hurricanes and the U.S. presidential election as mentioned earlier, respectively. Therefore, we exploit a path where the bulk of the spillovers is transmitted first to partisan conflict/national security policy uncertainty, then to the tourism stock market.

These findings indicate that the tourism market is the primary net receiver of spillovers during the sample. Within different systems and periods, partisan conflict and national security policy uncertainty are the main net transmitters of shocks while the dominant position of the two is different. Moreover, a spillover transmission mechanism exists in partisan conflict, national security policy uncertainty, and the tourism market.

4.3. Further Discussion

From the static perspective, the results show that the spillover effect initiates partisan conflict for the Inbound and Exports physical tourism market, and the mediator is national security uncertainty. When partisan conflict intensifies, that is, political security faces an uncertain environment, decision-makers face the uncertainty of whether political reform is successful or not. According to Lucas’ paradox, political instability leads to a lack of capital flow from rich to developing countries [95]. Other studies have also verified this by examining different political parameters, such as regime effect, alliance density, etc. [96,97]. That is to say, the political instability caused by the partisan conflict has affected the flow of foreign capital. There is a positive and direct relationship between foreign investment and tourism entry [10]. The spillover effect of the travel stock market is caused by national security uncertainty and mediated by partisan conflict. At the macro level, national security will affect the image of the whole society or tourism destinations, and the resulting negative impact will cause fluctuations in the stock market. At the micro level, threats to national security will impact the behavior and attitude of tourists, which will affect speculators’ views and psychological expectations of the market direction, leading to changes in the tourism stock market. Unlike the physical tourism market, national security uncertainty plays a greater role in the indirect transmission process than partisan conflict. This is due to the high liquidity characteristics of the stock market, which leads to its rapidly accepting shocks and digesting information. The conflict between parties takes time to reflect the polarization of their decisions, which delays the time for the stock market to receive information. Therefore, the negative image of national security is more direct, and the resulting partisan conflict further affects the tourism stock market.

From the dynamic perspective, the spillover effect between partisan conflict, national security uncertainty and the tourism market is time-varying and connected. During frequent global security, economic and political crises, the spillover effect has been significantly enhanced, including the 9/11 attacks of 2001, the global financial crisis of 2007−2008, the presidential election in 2016, and the COVID-19 crisis in 2020. When a particular country is affected by natural disasters or terrorist attacks, it will generally hurt international tourists. This may be due to the destruction of infrastructure, vital scenic spots, and the widespread weakness of the host country’s economy. All these have reduced the ability of the destination to cater to tourism, destroyed the investment in tourism supply, and reduced the attractiveness of the destination [11]. When the world faces the impact of the economic crisis, from a traditional point of view, people may prefer to use their discretionary income for necessities in times of economic difficulties. When the world is in a period of political uncertainty, this can increase uncertainty among households and businesses. Individuals are more inclined to postpone or cancel consumption decisions such as travel plans, leading to a decline in revenue and share price of travel companies [96]. In addition, investors may demand higher returns as compensation for increased political uncertainty, causing share prices to fall.

4.4. Robustness Check

To check the robustness of the results, we use NSP_EMV (National Security Policy Equity Market Volatility Tracker) instead of NSU for investigation. The results are reported in

Table 3. Spillover table (in %): PCI, NSP_EMV, and tourism.

Panel A	Inbound	PCI	NSP_EMV	FROM
Inbound	62.9	25.6	11.5	37.1
PCI	9.1	88.6	2.4	11.4
NSP_EMV	2.9	9.7	87.3	12.7
TO	12.0	35.4	13.8	61.2
NET directional connectedness	−25.1	23.9	1.2	20.4
Panel B	Exports	PCI	NSP_EMV	FROM
Exports	78.9	4.0	17.2	21.1
PCI	1.8	95.3	2.9	4.7
NSP_EMV	9.1	10.2	80.7	19.3
TO	10.9	14.2	20.0	45.1
NET directional connectedness	−10.2	9.5	0.7	15.0
Panel C	DJUSTT	PCI	NSP_EMV	FROM
DJUSTT	84.4	3.0	12.7	15.6
PCI	0.2	98.5	1.3	1.5
NSP_EMV	7.3	4.8	87.8	1.2
TO	7.6	7.8	14.0	29.3
NET directional connectedness	−8.1	6.3	1.8	9.8

Note: The bolded one is the directional spillover index, which represents the total connectivity.

, which is quite similar to the original results.

5. Conclusions and Implications

This paper explores the spillover effects among partisan conflict, national security policy uncertainty and tourism (i.e., tourist arrivals, exports and stock) in the U.S., as well as plotting the heterogeneity of the tourism market through a comparative analysis. To this end, we adopt a mixed connectedness measure introduced by Antonakakis et al. (2020), comprising the DY framework and a TVP-VAR model [22,23,24,25,26]. Using monthly data from January 1996 to December 2021, we obtain the whole sample connectedness and the time-varying dynamic connectedness between partisan conflict, national security policy uncertainty, and tourism. The main empirical findings can be summarized as follows.

First, in the connectedness for all variables, we find that the tourism market is the net receiver of spillover effects sourced by partisan conflict and national security policy uncertainty. Partisan conflict and national security policy uncertainty contribute more to inbound tourism demand than tourism stock. In contrast, the performance of travel stocks is more sensitive to the uncertainty of national security policy than partisan conflict. Second, in terms of the time-varying dynamic total connectedness, changes in dynamic complete volatility connectivity are more dramatic and rapid during periods of security, financial, and political shocks. Third, we obtain a spillover transmission path of the three. When the physical tourism market is the recipient of the spillover effect, the partisan conflict is the transmitter, and national security uncertainty plays a mediating role. When the tourism stock market is the receiver, national security uncertainty is the transmitter and partisan conflict is the intermediary. Dynamic results show that the bulk of the spillovers from national security policy uncertainty is transmitted first to the partisan conflict before 2015, then to the tourism stock market. Afterwards, the transmission path changed, in that most partisan conflict shocks impact the stability of security policy first, then the tourism stock market.

Our findings back up several specific policy-related recommendations. First, our study explores a channel through which improvements in political division and the national security environment benefit tourism sector development. Accordingly, governments in any tourist destination state, most especially in the U.S., should strive to create a stable political and security environment to reap further fruits from the tourism sector. Second, our findings confirmed that the impact of partisan conflict and national security policy uncertainty on tourism is time-varying and event-dependent. Hence, governments should be more cautious about security policy changes in turbulent times. Stable national policies can calm consumer sentiment while reducing the fear of inbound tourists. Finally, the strong interaction effects of partisan conflict and policy-related security instability give essential policy implications. Tourism managers need to concurrently judge the consequences of political instability and policy uncertainty while planning and evaluating robust crisis recovery plans for tourism development.