Democratic transition,political risk,economic instability,and tourist inflows: The case of Tunisia

Abstract

This article assesses the effect of the political risk and economic instability on the tourist arrivals in Tunisia using various wavelet methods. Our findings reveal a substantial effect of political risk over the short and medium terms, while the risk of economic effect is more perceptible over the long run. These outcomes are robust when using standard time series modeling. Terrorist incidents and political uneasiness increase the perception of risk and affect the tourism inflows over the short run. Governments are invited to indorse security and tourism safety because if not, the tourism demand will impede the economic growth over the long run.

Keywords

ARDL model economic risk political risk tourist arrivals wavelets

Introduction

Tourism has become a vital source of revenues and a reliable pillar of the socioeconomic development and economic sustainability, mostly so for emerging countries. It is well recognized that tourist inflows have potential economic benefits in destination areas through several transmission channels. However, these economies are still susceptible to geopolitical risk (GPR) and economic policy uncertainty (EPU) compared to developed economies. GPR shocks are causing substantial business cycle effects in the emerging economies. Indeed, 13–22% of the emerging countries outputs are caused by the global GPR shocks (see, inter alia, Baker et al., 2016; Cheng and Chiu, 2018). The vulnerability of developing economies to changes of macroeconomic policies and innovations in the global and regional GPRs could hinder tourist arrivals (TAs) and subsequently impede local economic growth. Demir and Gozgor (2018) show that the intensification of EPU decreases tourist demand. In tourism literature, several works investigated the causality linkages between the GPR, instability of macroeconomic policies, and the TA. For instance, Muzindutsi and Manaliyo (2016) investigate the short- and long-run connectedness between political risk, TA, and tourism revenues in South Africa. The authors reveal substantial effect of political risk over the long run. No short-run effect is evidenced. Gozgor and Ongan (2017) show that the EPU affects the tourism demand and causes substantial decline in tourism activity. Demir and Gozgor (2018) find that higher levels of EPU result in 3% and 6% reductions of the total and personal travel expenditure, respectively. Using a multiple and partial wavelet analysis, Balli et al. (2018) investigate the impact of global EPU on the tourism demand for five the Organization for Economic Co-operation and Development (OECD) countries. They unveil that the global uncertainties are affecting tourism inflows in various levels. Specifically, the authors point out the substantial and persistent effect of the 9/11 terrorist attacks and the 2008 global financial crisis. Singh et al. (2018) employ the wavelet for the US economy and uncover that the effect of political risk is more pronounced for short-run horizons while the EPU is more pertinent over the medium- and long-run horizons. Quite similar findings are found for emerging countries. For instance, Tiwari et al. (2019) implement the wavelet method and show that, over the long run, the effect of GPR on the tourist flows is greater than the EPU. Balli et al. (2019) reveal the heterogeneous effect of GPR on tourism demand in seven emerging countries. Accordingly, some countries are severely impacted by GPR shocks while others are mostly immune. Using the multiple and partial wavelets, Wu and Wu (2019) reveal substantial coherencies and lead-lag interplays between the EPU and the TA for the Brazil, Russia, India and China (BRIC) countries.

Two foremost reasons are motivating our research. First, we are concerned with one of the renowned tourist destinations in the Mediterranean countries. The World Travel and Tourism Council Report (2018)¹ advocates that the direct contribution of travel and tourism to the gross domestic product was around 14.2%² in 2017 and is expected to grow by 2.5% per year. Moreover, the tourism activity directly supports around 225,000 jobs (6.3% of total employment), and it is expected to rise by 0.6% in the next 10 years.³ Second, over the last decade, Tunisia, as the Arab spring’s birthplace, has encountered several instances of political unrests, regime shifts, terrorist attacks, and economic impediments. Thus, our research contributes to the tourism literature by analyzing the joint and isolated effects of the political and economic risks on tourist inflows within a time–frequency framework for a country in which the tourism sector holds a momentous standpoint in national economic growth.

Data and methods

We use a monthly time series of the TA, the political risk rating (PRR), and the economic risk rating (ERR) for Tunisia. The sample period runs from January 1990 to September 2019, yielding 354 observations. The TA time series is gathered from the CEIC database,⁴ while the PRRs and ERRs are sourced from the Political Risk Services Group.⁵ ERR⁶ ranges from a high of 50 (lowest risk) to a low of 0 (highest risk), while the PRR⁷ alternates from a high of 100 (lowest risk) to a low of 0 (highest risk).

Methodologically, we follow three steps. First, we employ a bivariate wavelet method to analyze the connectedness between the PRR, the ERR, and the TA. Secondly, we resort to the partial wavelet coherence (PWC) and multiple wavelet coherence (MWC) to analyze the interactions of the three variables over time scales and frequencies. The PWC is analogous to a simple correlation, whereas the MWC is similar to multiple correlations (see Ng and Chan, 2012), which allows us to examine the resulting wavelet coherence (WC) of multiple independent variables on a dependent one. Finally, we check the robustness of the wavelet’ outcomes using an autoregressive distributed lag (ARDL) model.

The WC is defined as the localized correlation coefficient between the two signals over time and across frequencies. The WC is given as follows

R^{2} (u, s) = \frac{{| S (s^{- 1} W_{x y} (u, s)) |}^{2}}{S (s^{- 1} {| W_{x} (u, s) |}^{2}) S (s^{- 1} {| W_{y} (u, s) |}^{2})}

The PWC is comparable to a simple correlation. The PWC consists of identifying the WC between two signals $(y)$ and $(x_{1})$ after eliminating the power of a third signal $(x_{2})$ . The PWC can be written as

R P^{2} (y, x_{1}, x_{2}) = \frac{{| R (y, x_{1}) - R (y, x_{2}) \times R (y, x_{1}) \times |}^{2}}{{[1 - R (y, x_{2})]}^{2} {[1 - R (x_{1}, x_{2})]}^{2}}

The MWC assesses the WC of multiple independent variables (x₁ and x₂) on a dependent one ( $y) .$ Formally, the MWC is as follows

R M^{2} (y, x_{2}, x_{1}) = \frac{R^{2} (y, x_{1}) + R^{2} (y, x_{2}) - 2 Re [R (y, x_{1}) \times R (y, x_{2}) \times R (x_{2}, x_{1}]}{1 - R^{2} (x_{2}, x_{1})}

Findings and discussion

The WC between TA, ERR, and PRR is conveyed in Figure 1(a) and (b), respectively. The horizontal axis displays the timeline, while the vertical axis designates the frequencies. The strength of the coherencies is highlighted with red (strong) and blue (weak) colors. The thick black line delimitates the zones of statistical significance at the 5% level. The direction of the arrows is showing the phase difference between variables. Specifically, arrows pointed to the right (left) indicate that the time series are in phase (antiphase). When arrows are pointed to right and up (down), the first variable is leading (lagging). When they are pointed to the left and up (down), the first variable is lagging (leading). From Figure 1(a), we note small red islands over the 2–8-month band showing strong short-run coherencies localized at the beginning and the end of the sample period. Other red islands are detected at up to 16-frequency bands, indicating strong coherencies over the long run. A small red island is identified in 1990, corresponding to the 1990’ Gulf War. Another small region of strong coherency is detected at the end of 2001 and the beginning of 2002 over the 8–16 months’ band, corresponding, respectively, to the New York 9/11 terrorist attacks and the terrorist attack in April of 2002 on the synagogue of the Tunisian island of Djerba, one of the Tunisia’s most important sites.⁸ Other significant coherences are observed in 2011 and 2015. This could be the outcome of the popular uprising in Tunisia which protested regional inequalities, poverty, corruption, and political repression, ultimately forcing the ex-president to step down in January 2011. Tunisia has witnessed several political violence incidents and terrorist attacks. On March of 2015, three extremists attacked the Bardo National Museum and 20 people, mostly European tourists, were killed. Following this terrorist attack, more than 3000 booked holidays were cancelled and reservations dropped by 60%. Three months after, a mass shooting occurred in the tourist resort of Sousse and 38 European tourists were killed. These dramatic incidents have received intense coverage in international media. Subsequently, the TA fell by 25% in 2015 and tourism revenues dropped by 35%, causing mass unemployment and business closures. When inspecting the phase differences, we observe that for most red islands, the arrows are upward and pointed to the left, indicating that the two variables are in an antiphase relationship and the TA is lagging the PRR. Some high coherency regions with arrows pointed to the right and upward are identified mostly during 2018 and 2019. The TA and the PRR are in phase with TA as leading variable.

Figure 1.

(a) The WC: TA and PRR; (b) the WC: TA and ERR. WC: wavelet coherence; TA: tourist arrival; PRR: political risk rating; ERR: economic risk rating.

When looking to the connectedness between TA and ERR (Figure 1(b)), we notice a relatively different situation. Strong coherencies are detected over the 4–32-month frequencies while other relatively high coherencies are perceived over the 64–128-months’ band. This WC plot indicates strong connectedness between the ERR and the TA during the period 2005–2015. Over the long run and for most parts of the red islands, arrows are turned to the right and upward, showing that the ERR and TA are in phase with the ERR as a follower. For the remaining islands, arrows are pointed to the left and down, indicating that TAs are leading the ERR.

Based on the WC’s outcomes, it is clear that the TA is conjointly influenced by political and economic risks and so, it is more useful to isolate the effect of each variable. Figure 2(a) displays the WC between TA and PRR after cancelling out the ERR. We can perceive small islands of strong coherencies at the 4–8-month frequencies, pointing out a strong effect of the PRR on the TA over the short run. Moreover, other significant islands of red color are identified up to a 16-month frequency band, which may be due to some persistent effect and gradual recovery of the tourism activity. Overall, cancelling out the effect of the ERR indicates that the political risk is a significant variable governing the tourism demand in Tunisia. Figure 2(b) displays the WC between TA and ERR after isolating the PRR. We observe that the surface of significance has been reduced substantially and only a few small islands of red color at the 4–8 months’ band could be identified, which means that TA and ERR interact over the short run and the tourists make decision destination depending on the economic strength of a country.

Figure 2.

(a) PWC: TA-PRR cancelling out ERR; (b) PWC: TA-ERR cancelling out PRR. PWC: partial wavelet coherence; TA: tourist arrival; PRR: political risk rating; ERR: economic risk rating.

Figure 3 reports the MWC between PRR, ERR, and TA. The visual inspection reveals the existence of huge islands of strong coherencies at high and low frequencies and covering the whole period, indicating strong joint effect of PRR and ERR on TA. The existence of significant red islands up to 32 months reveals the persistence of political and economic shocks on tourism activity. The violent degradation of the economic and security environments along with political uneasiness will damage the tourist destination reputation, safety, attractiveness, and comfort, negatively affecting the tourist perception of a country, which in turn leads to a depression in the tourism business. Overall, our findings are in line with previous relevant studies focused on emerging countries, including, among others, Muzindutsi and Manaliyo (2016), Balli et al. (2018), Wu and Wu (2019), Tiwari et al. (2019), and Balli et al. (2019), pointing out that the negative impact of political risk and economic uncertainty on TA is varying across time-scales and frequency bands.

Figure 3.

The MWC: TA-PRR and ERR. MWC: multiple wavelet coherence; TA: tourist arrival; PRR: political risk rating; ERR: economic risk rating.

To check the robustness of the wavelet’ outcomes, we estimate an ARDL as a standard time series model.⁹ The ARDL model is written as follows

Δ {TA}_{t} = α_{0} + β_{j} \sum_{j = 1}^{n} {TA}_{t - j} + γ_{j} \sum_{j = 1}^{n} {PRR}_{j - t} + δ_{j} \sum_{j = 1}^{n} {ERR}_{t - j} + φ_{1} {TA}_{t - 1} + φ_{2} {PRR}_{t - 1} + φ_{3} {ERR}_{t - 1} + ε_{t}

where ΔTA _t are logarithmic variation of the TA at time (t). $α_{0}$ is an intercept. The coefficients $β_{j}$ , $γ_{j}$ , and $δ_{j}$ refer to the short-run dynamics, while $φ_{1}, φ_{2}$ , and $φ_{3}$ correspond to the long-run relationship. $ε_{t}$ is the error term. $(n)$ designates the number of the selected lags according to the Akaike information criterion and Schwarz information criterion.¹⁰ Statistically, the rejection of the null hypothesis $H_{0} : φ_{1} = φ_{2} = φ_{3} = 0$ implies the existence of long-run connectedness between the dependent and the two independent variables and allows the estimation of an error correction model (ECM). The estimation of ARDL model shows a no errors serial correlation lagrange multiplier test (LM test), while the cumulative sum control chart (CUSUM) test reveals the stability of the estimated coefficients. Two key comments arise from the estimation results. For the long-run dynamics, the coefficients $φ_{2}$ and $φ_{3}$ are significant and are positively signed (0.913 and 1.54) implying that ERR and PRR are positively associated to the TA, which means that increases of the economic and political ratings (i.e. decrease of risk level) raise TA. When looking to the short-run dynamics, we observe that the lagged risk coefficients are positively signed but insignificant, while the other coefficients relative to the lagged TA are positive and significant. The ARDL model shows no significant short-run connectedness between the TA and the two risk ratings which is inconsistent with the wavelet’ findings. However, this result is in line with the findings of Muzindutsi and Manaliyo (2016) using the ARDL approach for the context of South Africa revealing the inexistence of short-run effect.

We extend the ARDL analysis by examining the ECM. Doing so, the use of the bound test rejects the null hypothesis of no cointegration¹¹ indicating the existence of long-run association between the variables. We estimate the following ECM

Δ {TA}_{t} = α_{0} + β_{j} \sum_{j = 1}^{n} {TA}_{t - j} + γ_{j} \sum_{j = 1}^{n} {PR}_{j - t} + δ_{j} \sum_{j = 1}^{n} {ECR}_{t - 1} + δ {ECT}_{t - 1} + μ_{t}

where ${ECT}_{t - 1}$ is the lagged error correction term from the cointegrated relationship. $μ_{t}$ is an error term. The estimated ECM meets all the statistical requirements in terms of no serial correlation in the errors, stability of the estimated coefficients, and breakpoint tests. The estimated coefficient of the lagged ECT is negative and significant (−0.167), which means that 16.7% of the disequilibrium between the TA, PRR, and ERR is corrected each month. In other words, the changes in PRR and ERR will take around 6 months $(1 / 0.167)$ to have a full effect on TA. The estimated parameters of the lagged PRR and ERR are insignificant, showing no short-run relationship between the two risk variables and the TA. This finding is inconsistent with the WC results. The plausible argument may be that the tourism expenditure may encompass long-run planning which could be insensitive to short-run changes.

Conclusion and policy implications

We find interesting results. First, we uncover substantial joint effects of PRR and ERR on TA. Second, the political risk effect holds over the short run while the economic instability is more relevant over the long run. Unlike other countries in the Middle East and North Africa (MENA) region, the democratic transition has been successful in Tunisia, but the country is still facing high political and economic risks and is operating in a high GPR region. Besides, the recent terrorist incidents and political uneasiness have received intensive attention in international media and has caused a substantial downturn in the tourism business. Hence, Tunisian policy makers are solicited to set prominent strategies and a comprehensive vision of tourism management to draw the tourism sector in a profitable direction. To maintain the economic sustainability, they are also asked to listen to market competition and focus on resources that would enhance the tourism sector.

Table 1.

Results of the bound test for cointegration.

		Pesaran critical values
Model	Estimated	Lower bound	Upper bound
F-value	30.40	6.56	7.3

Footnotes

Acknowledgements

The authors are grateful to the Editor-in-Chief Professor A. Assaf and the anonymous reviewer for his/her helpful comments and suggestions. C. Aloui would like to acknowledge Business, Society & Environment (BSE) Research Lab, Prince Sultan University, Saudi Arabia, for their support. The authors are grateful to Ms. Al Kayed for her professional language editing.

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

Chaker Aloui

Notes

References

Baker

Bloom

Davis

(2016) Measuring economic policy uncertainty. The Quarterly Journal of Economics 131: 1593–1636.

Balli

Shahzad

Uddin

(2018) A tale of two shocks: What do we learn from the impacts of economic policy uncertainties on tourism? Tourism Management 68: 470–475.

Balli

Uddin

Shahzad

(2019) Geopolitical risk and tourism demand in emerging economies. Tourism Economics 25: 997–1005.

Cheng

Chiu

(2018) How important are global geopolitical risks to emerging countries? International Economics 156: 305–325.

Demir

Gozgor

(2018) Does economic policy uncertainty affect tourism? Annals of Tourism Research 69: 15–17.

Gozgor

Ongan

(2017) Economic policy uncertainty and tourism demand: empirical evidence from the USA. International Journal of Tourism Research 19: 99–106.

Muzindutsi

Manaliyo

(2016) Effect of political risk shocks on tourism revenue in South Africa: a time series analysis. International Journal of Business and Management Studies 8: 169–186.

EKW

Chan

JCL

(2012) Geophysical applications of partial wavelet coherence and multiple wavelet coherence. Journal of Atmospheric and Oceanic Technology 29: 1845–1853.

Pesaran

Shin

Smith

(2001). Bounds testing approaches to the analysis oflevel relationship. Journal of Applied Economics 16: 289–326.

10.

Singh

Das

Jana

, et al. (2018) A wavelet analysis for exploring the relationship between economic policy uncertainty and tourist footfalls in the USA. Current Issues in Tourism 22: 1789–1796.

11.

Tiwari

Das

Dutta

(2019) Geopolitical risk, economic policy uncertainty and TA: evidence from a developing country. Tourism Management 75: 323–327.

12.

(2019) A multiple and partial wavelet analysis of the economic policy uncertainty and tourism nexus in BRIC. Current Issues in Tourism. DOI:10.1080/13683500.2019.1566302.