The permanent or transitory nature of shocks to tourism expenditures and receipts: Evidence from new panel stationarity tests with breaks and factors

Abstract

Tourism expenditures and receipts are an important component of many countries’ balance of payments. Indeed, the determination of whether shocks to tourism expenditures and receipts are permanent or transitory in nature is relevant to understanding the appropriate policy response. This study introduces new panel stationarity tests that simultaneously model structural breaks and a common factor structure. The results from several panel stationarity tests including the new tests reject the null hypothesis of stationarity for the panel of 63 countries. Based on tests of stationarity with a common factor structure without structural breaks at the individual country level, the null hypothesis of stationarity is rejected in 24 and 26 countries with respect to per capita international tourism expenditures and tourism receipts, respectively. The inclusion of the Fourier approximation for smooth structural breaks alongside the common factor structure increases the number of rejections of the null hypothesis of stationarity to 48 and 54 countries for per capita international tourism expenditures and receipts, respectively. Hence, shocks to either tourism expenditures or receipts will likely be permanent in nature requiring a policy response in order to return tourism expenditures and receipts to their original trend path.

JEL Codes: Z32, C33.

Keywords

tourism expenditures tourism receipts panel stationarity tests structural breaks PANIC principal component

Introduction

Tourism and the hospitality sector serves a prominent economic role in many countries through its contribution to a country’s foreign exchange earnings, government revenues, employment, and balance of payments. However, with the influence of globalization on the tourism and hospitality industry, this sector is particularly sensitive to exogenous shocks due to economic and political forces, natural disasters, terrorist attacks, and even pandemics, among many others (Hall, 2010; Gil-Alana et al., 2016). As a stark reminder of the severity of such shocks, the lockdowns and travel restrictions associated with the COVID-19 pandemic has resulted in a 74% global decline in international tourist arrivals with an even more dramatic impact on tourist arrivals to developing countries (Vanzetti and Peters, 2012). For policymakers and those in the tourism and hospitality industry the ability to model and predict various tourism indicators is important in order to both anticipate and navigate the economic landscape associated with adverse shocks to this sector.

In this regard, an extensive literature has emerged that evaluates the permanent or transitory nature of shocks to tourism indicators through a variety of unit root and stationarity tests. It is particularly relevant to make the distinction between permanent and transitory shocks in order to craft the appropriate policy response. For instance, if an adverse shock to the tourism and hospitality sector is perceived as transitory (i.e., time series stationary), a policy intervention is less likely as the shock will dissipate relatively quickly as the time series returns to its trend level. On the other hand, if an adverse shock is considered permanent (i.e., time series non-stationary), a policy intervention is more likely to be required to restore the time series to its original trend.

The examination of the unit root and stationarity properties of tourism indicators, more often than not has focused on tourist arrivals with some studies investigating tourism receipts and expenditures. Such studies have employed a wide range of testing approaches from univariate and panel unit root tests with and without structural breaks to second-generation panel unit root and stationarity tests that allow for cross-sectional dependence. For panel data models, the presence of cross-sectional dependence due to unobservable common factors or spatial spillover effects is one of the major concerns in estimation and inference. Ignoring cross-sectional dependence could lead to invalid inference and inconsistent estimators (see Baltagi and Pirotte, 2010), and the recent studies in the panel data literature propose error factor structures to model cross-sectional dependence due to unobservable common factors (Pesaran, 2006; Bai, 2009; among others). Similar to the issue of cross-sectional dependence, the failure to incorporate structural breaks may lead to inconsistent estimation and invalid inference.

The importance of accounting for the effects of cross-correlations and structural breaks have been well-documented in the literature and several studies note that the usual panel estimators can be biased when not correcting for the effects of cross-correlations and structural breaks (Bai, 2010; Baltagi et al., 2016; Baltagi, et al., 2017; among others). The same issue can arise in unit root analysis. Although the presence of both structural breaks and common factors complicates the analysis of the non-stationarity of panel data, allowing both to exist brings the model closer to actual empirical settings. We control for the effects of structural breaks and common factors in the stationarity analysis, and it is flexible enough to model the nature of shocks for a panel of countries subject to structural breaks and common factors. As evident from the results, the incorporation of both structural breaks and cross-sectional dependence are important considerations when undertaking panel data modeling.

Within this line of the tourism research a number of studies have focused on the Asia-Pacific Rim region, specifically in the case of Fiji (Narayan 2005a 2005b, 2005c), Australia (Narayan 2008; Narayan and Prasad, 2008; Valadkhani and O’Mahony 2018), Singapore (Lee 2009, 2010; Tan and Tan 2014), Malaysia (Lean and Smyth 2009; Solarin 2015), Taiwan (Chu et al., 2008; Lee et al., 2014; Yang et al., 2014), India (Bhattacharya and Narayan 2005; Tiwari et al., 2018; Xie et al., 2018), China (Chu et al., 2014), Hong Kong (Au et al., 2005), Bali (Smyth et al., 2009), Cambodia (Tang and Wong 2009), Thailand (Saleh et al., 2011), Laos (Kyophilavong et al., 2019), and Pakistan (Baig and Hussain 2020). In addition to the numerous studies related to Asia-Pacific Rim countries, a number of other country studies have been undertaken: Lorde et al. (2009, Barbados), Bassil et al. (2014, Lebanon), Dedeoglu (2016, Turkey), Solarin (2016, Mauritius), and Charles et al. (2019, France) along with several multi-country studies by Dash et al. (2017, BRIC countries) and Yucel (2021, 20 most visited countries).

In light of the literature to date, this study extends the examination of the permanent versus transitory nature of shocks to tourism indicators on several fronts. First, unlike the majority of the studies, this study investigates the time series behavior of two key variables in a country’s balance of payments: per capita international tourism expenditures and per capita international tourism receipts. Second, this study serves as the largest multi-country study by investigating 63 countries across varying levels of economic development as defined by the World Bank income classification. Third, this study extends the recent work of Dash et al. (2017) and Yucel (2021) in the use of second-generation panel unit root tests with allowance for structural breaks and cross-sectional dependence through the introduction of new panel tests that simultaneously model structural breaks and cross-correlations in panel data.

This study differs from Dash et al. (2017) with respect to handling structural breaks in a factor structure. Specifically, Dash et al. (2017) employ the panel unit root approach of Im et al. (2005) which accounts for structural breaks as a sharp process and does not correct for cross-sectional dependence. Our study not only models structural breaks as a smooth/gradual process rather than abrupt breaks but also considers cross-sectional dependence in a factor framework. Yucel (2021) extends the analysis of shocks to international tourist arrivals by accounting for smooth structural breaks and cross-sectional dependence based on Nazlioglu and Karul (2017) panel stationarity test. As outlined in Nazlioglu et al. (2021), the panel statistic proposed in Nazlioglu and Karul (2017) is asymptotically suitable for panels where the time dimension is larger than the cross-section dimension. Nazlioglu et al. (2021) advance the work of Nazlioglu and Karul (2017) by developing the response surface function which facilitates obtaining critical values for individual test statistics and corresponding p-values for small samples. These new tests based on the combination of p-values are shown to have good size and power properties even in small samples. In addition to the panel stationarity tests of Nazlioglu and Karul (2017), this study also employs new tests proposed in Nazlioglu et al. (2021) that not only ensures the robustness of the empirical results for the whole panel but also to draw robust inferences for cross-sectional units by means of the response surfaces in approximating the rejection probabilities of the individual test statistics.

The remainder of this study is organized such that Section 2 presents the methodology. Section 3 describes the data with a discussion of the results in Section 4. Concluding remarks are given in Section 5.

Methodology

In order to investigate the nature of shocks to international tourism expenditures and receipts, the analysis utilizes recent developments in testing for stationarity in the non-stationary panel data literature. To begin, the data generating process (DGP) is given as follows

y_{i t} = α_{i} + β_{i} t + r_{i t} + ε_{i t}

(1)

r_{i t} = r_{i t - 1} + u_{i t}

(2)

where

y_{i t}

is the tourism variable of interest with

i = 1, \dots, N

the cross-section dimension and

t = 1, \dots, T

the time dimension. Let

σ_{u i}^{2}

be the variance of

u_{i t}

, then the null hypothesis of stationarity is

H_{0} : σ_{u i}^{2} = 0

for all i is tested against the alternative of a unit root given as

H_{1} : σ_{u i}^{2} > 0

for some i’s. The DGP in (1)-(2) does not account for cross-correlations in the panel, and the panel stationarity tests proposed by Hadri (2000) and Yin and Wu (2001) are known as the first-generation panel stationarity tests.

To provide allowance for a factor structure, $ε_{i t}$ is redefined as $ε_{i t} = λ_{i}^{'} F_{t} + e_{i t}$ where $F_{t}$ are unobserved common factors, $λ_{i}$ are factor loading weights, and $e_{i t}$ is an idiosyncratic error term. Hence, equation (1) can re-written in the factor model framework as

y_{i t} = α_{i} + β_{i} t + r_{i t} + λ_{i}' F_{t} + e_{i t} .

(3)

Here, F_t needs to be estimated. By using the method of principal components for estimating common factors, Bai and Ng (2005) suggest the inverse chi-square test (Fisher’s test) of Maddala and Wu (1999) with

P_{P c} = - 2 \sum_{i = 1}^{N} \ln (p_{i}) \sim χ_{2 N}^{2}

(4)

where

p_{i}

is the p-value corresponding to the individual KPSS statistic of Kwiatkowski et al. (1992) obtained from the idiosyncratic components

e_{i t}

of equation (3). PC implies that the test is drawn from the PANIC procedure of Bai and Ng (2004, 2005) using the principal component (PC) approach to estimate unknown common factors.¹ As

T \to \infty

and

N \to \infty

, the

P_{P c}

test will diverge to infinity in probability, and is of little use for large N. In order to deal with this drawback of the

P_{P C}

test, one can use the modified test, in line with Choi (2001), given as

P_{m, P C} = \frac{P_{P C} - 2 N}{\sqrt{4 N}} \sim N (0,1)

(5)

and the inverse normal test,

Z_{P C} = \frac{1}{\sqrt{N}} \sum_{i = 1}^{N} Φ^{- 1} (p_{i}) \sim N (0,1)

(6)

where

Φ (.)

is the standard normal cumulative distribution function (see Bai and Ng, 2005; Nazlioglu et al., 2021).

As documented (Lee et al., 1997), the individual KPSS statistic has an over-size problem if one ignores a structural break in the data. Carrion-i-Silvestre et al. (2005) incorporate structural breaks using dummy variables which require knowledge regarding the number of breaks and their locations. Nazlioglu and Karul (2017) provide an alternative approach through a panel stationarity test with a Fourier approximation which does not require a priori knowledge on the dates, number, and/or form of breaks, but models structural break(s) as a gradual process. Nazlioglu and Karul (2017) define the deterministic term as a function of time with allowance for a common factor. The testing framework draws upon the following regression model

y_{i t} = α_{i} + β_{i} t + γ_{1 i} \sin (\frac{2 π k t}{T}) + γ_{2 i} \cos (\frac{2 π k t}{T}) + λ_{i}^{'} F_{t} + e_{i t}

(7)

where

γ_{1 i}

and

γ_{2 i}

measure the amplitude and displacement of gradual shift, k denotes the Fourier frequency component, and unobserved common factor

F_{t}

is estimated by using the cross-sectional averages approach of Pesaran (2006, 2007). Nazlioglu and Karul (2017) utilize the standardized group-mean statistic

W (k)

given by

W (k) = \frac{\sqrt{N} (L M (k) - m (k))}{v (k)} \sim N (0,1)

(8)

where

L M (k) = N^{- 1} \sum_{i = 1}^{N} F K P S S_{i.}

F K P S S_{i}

is the individual KPSS statistic with a Fourier approximation proposed by Becker et al. (2006),

m (k)

is the mean, and

v^{2} (k)

is the variance of

L M (k)

In a more recent study, Nazlioglu et al. (2021) suggest the combination statistics $P (k), P_{m} (k)$ , and $Z (k)$ by estimating the response surface function for the individual stationarity tests. The response surface estimation provided by Nazlioglu et al. (2021) not only facilitates the implementation of the combination of p-values tests, that is, $P (k), P_{m} (k)$ , and $Z (k)$ but also enables one to obtain critical values of test statistics and corresponding p-values even in small samples.

Data

This study uses annual data from 1995 to 2019 with respect to international tourism expenditures, international tourism receipts, and population, respectively, for 63 countries obtained from the World Bank Development Indicators.³ More specifically, international tourism expenditures are expenditures of international outbound visitors in other countries while international tourism receipts are expenditures by international inbound visitors in the destination country. Both international tourism expenditures and receipts were divided by the country’s population to arrive at per capita international tourism expenditures and per capita international tourism receipts.⁴ Natural logarithm of the respective variables is used in the estimation.⁵

Table 1 displays the summary statistics for the respective variables. The data show the average per capita international tourism expenditures ranges from 1.83 in Malawi to 7.78 in Aruba. With respect to average per capita international tourism receipts ranges from 1.09 again in Malawi to 9.37 in Aruba. The variability of tourism expenditures and receipts is measured by the coefficient of variation. In terms of per capita international tourism expenditures, the coefficient of variation ranges from a low of 0.03 in the Bahamas and Germany to a higher of 0.46 in Cambodia. For per capita international tourism receipts, the coefficient of variation ranges from a low of 0.01 in Cyprus to a high of 0.56 in Angola. The JB normality test of Jarque and Bera (1987) indicates that the null hypothesis of normality is rejected for a few countries, namely Albania, Botswana, and Madagascar for per capita international tourism expenditures; and Albania, Bahamas, Botswana, Ghana, Guatemala, Honduras, and Solomon Islands for per capita international tourism receipts. In general, the normality test indicates support for the normal distributions of per capita international tourism expenditures and receipts, respectively.

Table 1.

Summary statistics.

	Tourism expenditures						Tourism receipts
Country	Mean	Median	Std.	CV	JB	p-val.	Mean	Median	Std.	CV	JB	p-val.
Albania, UMI	5.08	6.11	1.71	0.34	5.75^c	0.06	5.49	6.21	1.31	0.24	5.10^c	0.08
Angola, LMI	2.40	2.47	0.69	0.29	0.51	0.77	2.16	2.42	1.21	0.56	2.50	0.29
Argentina, UMI	5.01	4.99	0.40	0.08	0.32	0.85	4.62	4.70	0.31	0.07	2.71	0.26
Armenia, UMI	4.46	4.99	1.48	0.33	2.22	0.33	4.45	5.18	1.52	0.34	2.59	0.27
Aruba, HI	7.78	7.85	0.33	0.04	2.30	0.32	9.37	9.39	0.30	0.03	1.23	0.54
Australia, HI	6.78	6.95	0.54	0.08	3.03	0.22	6.94	7.09	0.44	0.06	2.59	0.27
Azerbaijan, UMI	4.06	3.79	1.24	0.31	3.09	0.21	3.80	3.61	1.49	0.39	2.89	0.24
Bahamas, HI	7.06	7.01	0.19	0.03	1.38	0.50	8.74	8.71	0.20	0.02	6.91^b	0.03
Belarus, UMI	3.99	4.33	0.85	0.21	3.37	0.19	3.65	3.91	1.19	0.33	3.36	0.19
Bolivia, LMI	3.47	3.67	0.78	0.23	1.98	0.37	3.50	3.52	0.68	0.20	1.93	0.38
Botswana, UMI	4.80	4.85	0.25	0.05	8.32^b	0.02	5.34	5.50	0.43	0.08	4.70^c	0.10
Brazil, UMI	3.94	4.01	0.75	0.19	2.27	0.32	2.94	3.32	0.64	0.22	3.33	0.19
Bulgaria, UMI	5.05	5.23	0.67	0.13	3.34	0.19	5.78	6.21	0.80	0.14	3.58	0.17
Cabo Verda, LMI	5.11	5.29	0.53	0.10	3.50	0.17	6.01	6.58	0.89	0.15	3.08	0.21
Cambodia, LMI	2.45	2.45	1.13	0.46	1.80	0.41	4.22	4.45	1.18	0.28	1.98	0.37
Cameroon, LMI	3.08	3.21	0.46	0.15	2.86	0.24	2.59	2.56	0.49	0.19	1.38	0.50
Chile, HI	4.55	4.42	0.37	0.08	2.70	0.26	4.82	4.90	0.37	0.08	2.44	0.29
Colombia, UMI	4.07	4.05	0.48	0.12	2.76	0.25	4.07	4.09	0.57	0.14	2.39	0.30
Costa Rica, UMI	4.97	4.92	0.30	0.06	2.51	0.29	6.16	6.19	0.41	0.07	1.57	0.46
Cyprus, HI	6.85	7.00	0.39	0.06	2.19	0.33	7.79	7.76	0.11	0.01	0.57	0.75
Ecuador, UMI	3.88	3.94	0.39	0.10	1.09	0.58	3.91	3.78	0.54	0.14	2.69	0.26
Egypt, LMI	3.31	3.33	0.31	0.09	2.14	0.34	4.43	4.45	0.42	0.10	0.94	0.63
Fiji, UMI	4.83	4.84	0.19	0.04	0.66	0.72	6.67	6.76	0.48	0.07	2.19	0.33
Finland, HI	6.67	6.81	0.38	0.06	2.90	0.23	6.51	6.59	0.36	0.06	2.46	0.29
Germany, HI	6.89	6.95	0.22	0.03	2.77	0.25	6.18	6.36	0.36	0.06	3.30	0.19
Ghana, LMI	2.97	3.55	0.92	0.31	2.67	0.26	3.07	3.45	0.98	0.32	19.92^a	0.00
Guatemala, UMI	3.67	3.98	0.58	0.16	3.24	0.20	4.01	4.30	0.44	0.11	5.50^c	0.06
Honduras, LMI	3.68	3.79	0.40	0.11	2.80	0.25	3.94	4.14	0.47	0.12	6.64^b	0.04
Hungary, HI	5.48	5.54	0.30	0.06	2.11	0.35	6.29	6.33	0.37	0.06	1.66	0.44
Israel, HI	6.49	6.41	0.26	0.04	3.36	0.19	6.48	6.60	0.27	0.04	3.01	0.22
Jordan, UMI	4.90	4.95	0.27	0.05	0.80	0.67	5.92	6.09	0.49	0.08	3.03	0.22
Kazakhstan, UMI	4.34	4.41	0.73	0.17	1.22	0.54	4.09	4.30	0.75	0.18	1.85	0.40
Kuwait, HI	7.74	7.79	0.35	0.04	2.52	0.28	5.29	5.31	0.18	0.03	0.65	0.72
Madagascar, LI	2.25	2.32	0.30	0.13	8.58^a	0.01	2.79	3.00	0.60	0.21	2.39	0.30
Malawi, LI	1.83	1.83	0.20	0.11	2.51	0.28	1.09	1.13	0.30	0.28	0.99	0.61
Malaysia, UMI	5.43	5.51	0.54	0.10	2.57	0.28	6.11	6.43	0.52	0.08	2.41	0.30
Mauritius, UMI	5.65	5.73	0.43	0.08	2.02	0.36	6.92	7.02	0.39	0.06	2.71	0.26
Mexico, UMI	4.34	4.43	0.30	0.07	2.78	0.25	4.75	4.71	0.28	0.06	1.28	0.53
Moldova, UMI	4.24	4.61	0.74	0.17	2.96	0.23	4.02	4.35	0.82	0.20	2.22	0.33
Mongolia, LMI	4.25	4.43	1.07	0.25	1.97	0.37	4.06	4.41	0.92	0.23	3.21	0.20
Morocco, LMI	3.62	3.82	0.64	0.18	2.39	0.30	5.08	5.41	0.58	0.11	3.32	0.19
Nepal, LMI	2.48	2.72	0.71	0.29	2.62	0.27	2.51	2.38	0.46	0.18	0.65	0.72
Norway, HI	7.67	7.91	0.51	0.07	3.14	0.21	6.82	6.93	0.36	0.05	2.90	0.24
Pakistan, LMI	2.10	2.29	0.51	0.24	2.45	0.29	1.54	1.55	0.16	0.10	1.36	0.51
Panama, UMI	4.95	4.89	0.58	0.12	2.40	0.30	6.22	6.26	0.92	0.15	2.58	0.28
Paraguay, UMI	3.71	3.63	0.47	0.13	1.67	0.43	3.38	3.48	0.50	0.15	2.44	0.29
Peru, UMI	3.79	3.78	0.55	0.15	1.63	0.44	4.17	4.26	0.63	0.15	2.30	0.32
Philippines, LMI	3.76	3.64	0.76	0.20	0.86	0.65	3.67	3.59	0.50	0.14	0.84	0.66
Poland, HI	5.15	5.33	0.40	0.08	2.74	0.25	5.48	5.56	0.36	0.07	1.50	0.47
Portugal, HI	5.93	6.06	0.31	0.05	1.75	0.42	6.95	7.06	0.49	0.07	1.79	0.41
Romania, UMI	4.18	4.41	0.91	0.22	1.39	0.50	4.06	4.39	0.89	0.22	2.26	0.32
Seychelles, HI	6.46	6.49	0.32	0.05	0.41	0.81	8.25	8.27	0.33	0.04	2.00	0.37
Slovak Republic, HI	5.40	5.83	0.81	0.15	3.12	0.21	5.58	6.07	0.73	0.13	3.03	0.22
Slovenia, HI	6.27	6.44	0.45	0.07	3.02	0.22	6.89	7.11	0.43	0.06	3.08	0.21
Solomon Islands, LMI	4.02	4.28	0.68	0.17	2.25	0.32	3.73	4.16	1.21	0.33	4.57^c	0.10
South Africa, UMI	4.51	4.61	0.37	0.08	1.95	0.38	4.91	5.11	0.43	0.09	3.62	0.16
South Korea, HI	5.78	6.04	0.59	0.10	1.84	0.40	5.48	5.25	0.43	0.08	2.67	0.26
Sri Lanka, LMI	3.65	3.58	0.72	0.20	2.18	0.34	3.97	3.63	0.93	0.23	2.38	0.30
Switzerland, HI	7.43	7.45	0.32	0.04	2.21	0.33	7.54	7.59	0.27	0.04	2.94	0.23
Thailand, UMI	4.52	4.54	0.47	0.10	0.53	0.77	5.72	5.69	0.67	0.12	2.44	0.29
Trinidad Tobago, HI	4.71	4.69	0.31	0.07	1.62	0.44	5.97	6.03	0.34	0.06	0.97	0.62
Tunisia, LMI	3.86	3.89	0.37	0.10	2.33	0.31	5.44	5.40	0.24	0.04	0.73	0.69
United States, HI	5.88	5.93	0.25	0.04	1.13	0.57	6.18	6.10	0.29	0.05	2.74	0.25

Notes: World Bank Income Classification: HI (high-income country), UMI (upper middle-income country), LMI (lower middle-income country), and LI (low-income country). Summary statistics include the mean, median, standard deviation (STD), and the coefficient of variation (CV), JB is Jarque and Bera (1987) statistic for the null hypothesis of normality, p-val. is probability of JB statistic. Statistical significance denoted as follows: 1%(a), 5%(b), and 10%(c), respectively.

Results and discussion

The results from the panel stationarity tests are presented in Table 2. Note that the regression model includes the constant and trend for the tests without breaks, and includes the level and trend shifts for the tests with Fourier breaks. The panel stationarity tests uniformly reject the null hypothesis of stationarity in the panel at the 1% significance level for per capita international tourism expenditures and receipts, respectively. The panel results suggest that shocks to the respective per capita international tourism expenditures and receipts are permanent in nature, which aligns in part with the findings for tourist arrivals by Au et al. (2005), Chu et al. (2008), Lorde et al. (2009), Chu et al. (2014), Lee, (2010), and Tiwari et al. (2018).

Table 2.

Results for panel statistics.

	Tourism expenditures		Tourism receipts
Common factor	Statistic	p-value	Statistic	p-value
P_PC	302.158^a	0.000	399.153^a	0.000
P_m,PC	11.097^a	0.000	17.207^a	0.000
Z_PC	−7.279^a	0.000	−9.308^a	0.000
Common factor and structural breaks
W(k)	149.966^a	0.000	168.283^a	0.000
P(k)	882.395^a	0.000	974.168^a	0.000
P_m(k)	47.648^a	0.000	53.430^a	0.000
Z(k)	−22.879^a	0.000	−25.106^a	0.000

Notes: Statistical significance denoted as follows: 1%(a), 5%(b), and 10%(c), respectively. To construct the panel statistic, the individual statistics are obtained based on the long-run variance estimator with the boundary rule of Sul et al. (2005). The AR model ${\tilde{ε}}_{i t} = ρ_{1 i} {\tilde{ε}}_{i t - 1} + \dots + ρ_{p i} {\tilde{ε}}_{i t - p} + v_{i t}$ is fitted with the optimal lags selected using the Schwarz criterion with the maximum five lags when we estimate ${\tilde{ω}}_{i}^{2} = {\tilde{σ}}_{v i}^{2} / {\tilde{ρ}}_{i} {(1)}^{2}$ . Here, ${\tilde{σ}}_{v i}^{2}$ is estimated using the Bartlett kernel with the bandwidth $1.447 {[4 {\hat{a}}_{i}^{2} T / (1 + {\hat{a}}_{i}^{2}) (1 - {\hat{a}}_{i}^{2})]}^{1 / 3}$ where ${\hat{a}}_{i}$ is the autoregressive parameter based on ${\hat{v}}_{i t} = a_{i} {\hat{v}}_{i t} + e_{i t}$ The p-values for $P_{m}$ and $Z$ tests are based on the standard normal distribution. The p-values for $P$ are based on the chi-square distribution.

However, it is well-known that a strict null hypothesis is imposed on the panel stationarity tests. The null hypothesis implies that stationarity holds in all cross-section units. A rejection of the null hypothesis does not necessarily mean that the data have a unit root process in all countries. It would be informative to examine the individual statistics for further insights into the results. Table 3 reports the individual statistics with a common factor structure without structural breaks (denoted as KPSS_PC) and the individual statistics with a common factor structure and structural breaks (denoted as FKPSS) for per capita international tourism expenditures and receipts, respectively. Note that the p-values for the individual test statistics are based on the response surface function estimates of Nazlioglu et al. (2021) which enables one to draw robust inferences in small samples. The null hypothesis of stationarity of per capita international tourism expenditures based on the KPSS_PC statistic is rejected for 24 of the 63 countries, while the null hypothesis of stationarity of per capita international tourism receipts is rejected for 26 of the 63 countries. Moreover, the null hypothesis of stationarity is rejected for both per capita international tourism expenditures and receipts in only 14 countries based on the KPSS_PC statistics. In those cases in which the null hypothesis of stationarity is rejected, one can infer that shocks will be permanent in altering the trend path of the tourism indicator in question.

Table 3.

Results for the individual countries.

	Tourism expenditures				Tourism receipts
Country	KPSS_PC	p-val.	FKPSS	p-val.	KPSS_PC	p-val.	FKPSS	p-val.
Albania, UMI	0.203^b	0.015	0.721^a	0.000	0.049	0.693	3.870^a	0.000
Angola, LMI	0.274^a	0.002	0.781^a	0.000	0.037	0.871	0.059^b	0.030
Argentina, UMI	0.045	0.746	0.080^a	0.001	0.091	0.239	0.032	0.480
Armenia, UMI	0.105	0.172	0.652^a	0.000	0.737^a	0.000	0.144^a	0.000
Aruba, HI	0.146^c	0.068	0.683^a	0.000	0.440^a	0.000	0.125^a	0.000
Australia, HI	0.044	0.763	0.576^a	0.000	0.111	0.145	0.253^a	0.000
Azerbaijan, UMI	0.268^a	0.002	0.175^a	0.000	0.291^a	0.001	0.136^a	0.000
Bahamas, HI	0.065	0.462	0.076^a	0.002	0.114	0.134	0.200^a	0.000
Belarus, UMI	0.166^b	0.043	0.435^a	0.000	0.519^a	0.000	0.122^a	0.000
Bolivia, LMI	0.045	0.747	0.029	0.609	0.402^a	0.000	0.235^a	0.000
Botswana, UMI	0.119	0.118	0.445^a	0.000	0.096	0.214	0.075^a	0.002
Brazil, UMI	0.436^a	0.000	0.431^a	0.000	0.063	0.481	0.338^a	0.000
Bulgaria, UMI	0.138^c	0.079	0.100^a	0.000	0.038	0.866	1.273^a	0.000
Cabo Verda, LMI	0.095	0.221	0.040	0.230	0.096	0.215	0.538^a	0.000
Cambodia, LMI	0.080	0.311	0.037	0.329	0.024	0.987	0.219^a	0.000
Cameroon, LMI	0.083	0.293	0.044	0.170	0.167^b	0.042	0.472^a	0.000
Chile, HI	0.047	0.729	0.496^a	0.000	0.144^c	0.070	0.197^a	0.000
Colombia, UMI	0.047	0.718	0.347^a	0.000	0.087	0.263	0.136^a	0.000
Costa Rica, UMI	0.039	0.851	0.078^a	0.001	0.090	0.248	0.333^a	0.000
Cyprus, HI	0.044	0.757	0.078^a	0.001	0.037	0.872	0.081^a	0.001
Ecuador, UMI	0.068	0.424	0.049^c	0.092	0.256^a	0.003	0.355^a	0.000
Egypt, LMI	0.072	0.383	1.009^a	0.000	0.047	0.727	0.041	0.227
Fiji, UMI	0.106	0.167	0.132^a	0.000	0.453^a	0.000	0.250^a	0.000
Finland, HI	0.101	0.192	0.046	0.133	0.119	0.118	0.536^a	0.000
Germany, HI	0.204^b	0.014	0.124^a	0.000	0.278^a	0.002	0.049^c	0.099
Ghana, LMI	0.160^b	0.050	1.009^a	0.000	0.055	0.578	0.385^a	0.000
Guatemala, UMI	0.608^a	0.000	0.306^a	0.000	0.234^a	0.005	0.213^a	0.000
Honduras, LMI	0.159^c	0.051	0.037	0.312	0.167^b	0.043	0.468^a	0.000
Hungary, HI	0.075	0.362	0.030	0.564	0.358^a	0.000	0.030	0.584
Israel, HI	0.046	0.740	0.503^a	0.000	0.053	0.615	0.186^a	0.000
Jordan, UMI	0.783^a	0.000	0.033	0.440	0.425^a	0.000	0.048	0.110
Kazakhstan, UMI	0.106	0.166	0.089^a	0.000	0.151^c	0.061	0.124^a	0.000
Kuwait, HI	0.224^a	0.007	0.087^a	0.000	1.317^a	0.000	0.279^a	0.000
Madagascar, LI	0.059	0.523	0.042	0.195	0.084	0.281	0.183^a	0.000
Malawi, LI	0.156^c	0.055	0.272^a	0.000	0.074	0.364	0.248^a	0.000
Malaysia, UMI	0.143^c	0.072	0.297^a	0.000	0.050	0.657	1.623^a	0.000
Mauritius, UMI	0.041	0.816	0.552^a	0.000	0.049	0.693	0.097^a	0.000
Mexico, UMI	0.050	0.657	0.053^c	0.061	0.493^a	0.000	0.385^a	0.000
Moldova, UMI	0.145^c	0.068	0.347^a	0.000	0.342^a	0.000	0.202^a	0.000
Mongolia, LMI	0.197^b	0.017	0.799^a	0.000	0.311^a	0.001	0.058^b	0.036
Morocco, LMI	0.099	0.199	0.084^a	0.000	0.060	0.521	0.250^a	0.000
Nepal, LMI	0.080	0.315	0.042	0.190	0.117	0.126	0.086^a	0.000
Norway, HI	0.098	0.207	0.112^a	0.000	0.105	0.172	0.453^a	0.000
Pakistan, LMI	0.054	0.589	0.107^a	0.000	0.075	0.354	0.287^a	0.000
Panama, UMI	0.108	0.159	0.485^a	0.000	0.073	0.373	0.092^a	0.000
Paraguay, UMI	0.333^a	0.000	0.749^a	0.000	0.081	0.306	0.037	0.320
Peru, UMI	0.137^c	0.082	0.046	0.129	0.174^b	0.036	0.556^a	0.000
Philippines, LMI	0.069	0.409	0.125^a	0.000	0.056	0.565	0.313^a	0.000
Poland, HI	0.066	0.448	0.043	0.172	0.076	0.353	0.146^a	0.000
Portugal, HI	0.154^c	0.057	0.357^a	0.000	0.054	0.598	0.354^a	0.000
Romania, UMI	0.015	1.000	0.138^a	0.000	0.040	0.836	0.198^a	0.000
Seychelles, HI	0.070	0.401	0.526^a	0.000	0.143^c	0.071	0.260^a	0.000
Slovak Republic, HI	0.120	0.117	0.326^a	0.000	0.096	0.216	0.104^a	0.000
Slovenia, HI	0.080	0.309	0.102^a	0.000	0.336^a	0.000	0.168^a	0.000
Solomon Islands, LMI	0.063	0.487	0.045	0.146	0.081	0.303	0.041	0.222
South Africa, UMI	0.063	0.479	0.138^a	0.000	0.046	0.734	0.147^a	0.000
South Korea, HI	0.125	0.105	0.348^a	0.000	0.066	0.446	0.040	0.241
Sri Lanka, LMI	0.191^b	0.020	0.161^a	0.000	0.338^a	0.000	0.048	0.108
Switzerland, HI	0.202^b	0.015	0.402^a	0.000	0.105^a	0.000	0.288^a	0.000
Thailand, UMI	0.135^c	0.084	0.437^a	0.000	0.143^c	0.072	0.321^a	0.000
Trinidad Tobago, HI	0.093	0.232	0.041	0.207	0.026	0.974	0.281^a	0.000
Tunisia, LMI	0.051	0.645	0.048	0.106	0.337^a	0.000	0.160^a	0.000
United States, HI	0.508^a	0.000	0.383^a	0.000	0.062	0.502	0.044	0.160

Notes: See the footnote of Table 1. FKPSS statistic is based on one Fourier frequency. The p-values are based on the response surface function estimates of Nazlioglu et al. (2021). Statistical significance denoted as follows: 1%(a), 5%(b), and 10%(c), respectively.

Table 3 also reports the individual KPSS statistics with a Fourier approximation to account for structural breaks and a common factor structure, denoted as FKPSS. Compared to the KPSS_PC statistic, the KPSS statistic with a Fourier approximation to structural breaks and a common factor structure (FKPSS) yields more rejections of the null hypothesis of stationarity. This result can be considered as empirical support of Lee et al. (1997) which indicates the over-rejection problem of the KPSS test in the presence of a structural break in the data. The null hypothesis of stationarity of per capita international tourism expenditures based on the FKPSS statistic is rejected for 48 of the 63 countries, while the null hypothesis of stationarity of per capita international tourism receipts is rejected for 54 of the 63 countries. Furthermore, the null hypothesis of stationarity is rejected for both per capita international tourism expenditures and receipts in 42 countries based on the FKPSS statistic. As such, the results presented, although focused on per capita international tourism expenditures and receipts, are contrary to a vast majority of the time series and panel studies associated with the stationarity of tourist arrivals.

Concluding remarks

This study extends the existing literature on the permanent or transitory nature of shocks with respect to tourism indicators through the use of stationarity tests along several dimensions. Unlike the majority of previous studies, this study investigates two key variables in a country’s balance of payments: per capita international tourism expenditures and per capita international tourism receipts. Moreover, the analysis serves as the largest multi-country study to date in the examination of 63 countries across varying levels of economic development. Finally, this study introduces new panel stationarity tests that simultaneously model structural breaks and a common factor structure.

Specifically, a transitory shock to the tourism and hospitality sector will dissipate relatively quickly as the time series returns to its trend level, whereas a permanent shock will more likely require a policy intervention to restore the time series to return to its original trend. The findings that the null hypothesis of stationarity with respect to per capita international tourism receipts is rejected slightly more often than in the case of per capita international tourism receipts has important implications for a country’s balance of payments. This suggests that adverse shocks to tourism receipts, and hence foreign exchange earnings will be permanent in nature. As such, policymakers and those in the tourism and hospitality industry need to develop mitigation strategies to minimize the risk associated with the economic repercussions from adverse shocks. Likewise, for those countries heavily reliant on tourism receipts as a major source of foreign exchange, policy discussions to further diversity the country’s economic base and export sector is warranted.

From the point of the empirical modelling approach, the results provide clear evidence on the importance of considering structural breaks and cross-correlation in tourism data. This finding may call attention for future research to re-investigate not only the dynamics of tourism indicators but also the dynamic interrelationships between the macro economy and the tourism and hospitality sector in light of structural changes and cross-sectional dependence. This suggestion in particular may matter more for countries that have a significant share of their income and employment generated from the tourism and hospitality sector. Although this study includes a relatively large panel of countries, additional analysis for those countries absent of this study is warranted especially as sufficient data become available.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

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

ORCID iD

James E Payne

Notes

Author Biographies

James E. Payne, PhD in Economics, serves as the Dean of the College of Business Administration at The University of Texas at El Paso and holds the Paul L. Foster and Alejandra de la Vega Foster Distinguished Chair in International Business. Dr. Payne has authored over 270 peer-reviewed journal articles and serves on the editorial board of a number of academic journals.

Saban Nazlioglu, PhD in Economics, is a Professor in the Department of International Trade and Finance at the Faculty of Economic and Administrative Sciences, Pamukkale University/Denizli, Turkey. His research areas are applied econometrics (in particular panel data), international trade, and commodity prices.

References

AKM

Ramasamy

Yeung

MCH

(2005) The effects of SARS on the Hong Kong tourism industry: An empirical evaluation. Asia Pacific Journal of Tourism Research 10(1): 85–95.

Bai

(2009) Panel data models with interactive fixed effects. Econometrica 77(4): 1229–1279.

Bai

(2010) Common breaks in means and variances for panel data. Journal of Econometrics 157(1): 78–92.

Bai

(2002) Determining the number of factors in approximate factor models. Econometrica 70(1): 191–221.

Bai

(2004) A PANIC attack on unit roots and cointegration. Econometrica 72(4): 1127–1177.

Bai

(2005) A new look at panel testing of stationarity and the PPP hypothesis. In: Andrews

DWK

Stock

(eds), Identification and Inference for Econometric Models. Essays in Honor of Thomas Rothenberg. Cambridge: Cambridge University Press.

Baig

Hussain

(2020) Do shocks have permanent or transitory effects on tourist inflow? An application of stationarity tests with structural breaks: Evidence re-examined for Gilgit-Baltistan, Pakistan. Asia Pacific Journal of Tourism Research 25(2): 120–130.

Baltagi

Feng

Kao

(2016) Estimation of heterogeneous panels with structural breaks. Journal of Econometrics 191(1): 176–195.

Baltagi

Kao

Liu

(2017) Estimation and identification of change points in panel models with nonstationary or stationary regressors and error term. Econometric Reviews 36(1–3): 85–102.

10.

Baltagi

Pirotte

(2010) Panel data inference under spatial dependence. Economic Modelling 27(6): 1368–1381.

11.

Bassil

Saleh

Anwar

(2014) Is tourism in Lebanon subject to permanent or transitory exogenous shocks? Tourism Analysis 19(6): 781–790.

12.

Becker

Enders

Lee

(2006) A stationarity test in the presence of an unknown number of smooth breaks. Journal of Time Series Analysis 27(3): 381–409.

13.

Bhattacharya

Narayan

(2005) Testing for the random walk hypothesis in the case of visitor arrivals: Evidence from Indian tourism. Applied Economics 37(13): 1485–1490.

14.

Carrion-i-Silvestre

Barrio-Castro

Lopez-Bazo

(2005) Breaking the panels: An application to the GDP per capita. The Econometrics Journal 8(2): 159–175.

15.

Cavaliere

(2005) Limited time series with a unit root. Econometric Theory 21(5): 907–945.

16.

Cavaliere

Xue

(2014) Testing for unit roots in bounded time series. Journal of Econometrics 178: 259–272.

17.

Charles

Darne

Hoarau

J-F

(2019) How resilient is La reunion in terms of international tourism attractiveness: An assessment from unit root tests with structural breaks from 1981-2015. Applied Economics 51(24): 2639–2653.

18.

Choi

(2001) Unit root tests for panel data. Journal of International Money and Finance 20(2): 249–272.

19.

Chu

H-P

Chang

H-L

, et al. (2008) Are visitor arrivals to Taiwan stationary? An empirical study based on panel SURADF tests. Empirical Economics Letters 7(10): 1001–1007.

20.

Chu

H-P

Yeh

M-L

Chang

T-Y

(2014) Are visitor arrivals to China stationary? An empirical note. Asia Pacific Journal of Tourism Research 19(2): 248–256.

21.

Dash

Suresh

Tiwari

(2017) Are tourist arrivals stationary? Evidence from BRIC countries. Current Issues in Tourism 20(3): 221–224.

22.

Dedeoglu

(2016) Are shocks to Turkey’s tourist arrivals permanent? Ekonometri ve Istatistik sayi 24: 30–39.

23.

Gil-Alana

Mervar

Payne

(2016) Modeling the degree of persistence in Croatian tourism. Tourism Economics 22(3): 655–664.

24.

Hadri

(2000) Testing for stationarity in heterogeneous panel data. The Econometrics Journal 3(2): 148–161.

25.

Hall

(2010) Crisis events in tourism: Subjects of crisis in tourism. Current Issues in Tourism 13: 401–417.

26.

Hamilton

(1994) Time Series Analysis. Princeton, NJ, USA: Princeton University Press.

27.

Herwartz

(2008) Reviewing the sustainability/stationarity of current account imbalances with tests for bounded integration. The Manchester School 76(3): 267–278.

28.

Lee

Tieslau

(2005) Panel LM unit root tests with level shifts. Oxford Bulletin of Economics and Statistics 67(3): 393–419.

29.

Jarque

Bera

(1987) A test for normality of observations and regression residuals. International Statistical Review 55(2): 163–172.

30.

Kwiatkowski

Phillips

PCB

Schmidt

, et al. (1992) Testing the null hypothesis of stationary against the alternative of a unit root: How sure are we that economic time series have a unit root? Journal of Econometrics 54(1–3): 159–178.

31.

Kyophilavong

Tiwari

Jozwik

, et al. (2019) Are tourist arrivals stationary? Evidence from Laos. Anatolia: An International Journal of Tourism and Hospitality Research 30(4): 622–625.

32.

Lean

Smyth

(2009) Asian financial crisis, avian flu and tourist threats: Are shocks to Malaysian tourist arrivals permanent or transitory? Asia Pacific Journal of Tourism Research 14(3): 301–321.

33.

Lee

(2009) Are tourist arrivals stationary? Evidence from Singapore. International Journal of Tourism Research 11(4): 409–414.

34.

Lee

(2010) Whether a shock has a temporary or permanent effect on visitor arrivals in Singapore. Tourism Review International 14(1): 43–46.

35.

Lee

Ying

Yang

GJA

(2014) Assessing whether tourist arrivals are stationary: Evidence from Taiwan using sequential panel selection method. Journal of China Tourism Research 10(2): 206–221.

36.

Lee

Huang

Shin

(1997) On stationary tests in the presence of structural breaks. Economics Letters 55(2): 165–172.

37.

Lorde

Francis

Skeete

(2009) Are shocks to Barbados long-stay visitor arrivals permanent or temporary: A short empirical note. Journal of Public Sector Policy Analysis 3: 3–19.

38.

Maddala

(1999) A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin of Economics and Statistics 61(S1): 631–652.

39.

Narayan

(2005a) The structure of tourist expenditure in Fiji: Evidence from unit root structural break tests. Applied Economics 37(10): 1157–1161.

40.

Narayan

(2005b) Did Rabuka’s military coups have a permanent effect or a transitory effect on tourist expenditure in Fiji: Evidence from Vogelsang’s structural break test. Tourism Management 26(4): 509–515.

41.

Narayan

(2005c) Testing the unit root hypothesis when the alternative is a trend break stationary process: An application to tourist arrivals in Fiji. Tourism Economics 11(3): 351–364.

42.

Narayan

Prasad

(2008) Examining the behaviour of visitor arrivals to Australia from twenty different countries: An application of panel unit root tests. Economic Papers 27(3): 265–271.

43.

Narayan

(2008) Examining the behavior of visitor arrivals to Australia from 28 different countries. Transportation Research Part A 42(5): 751–761.

44.

Nazlioglu

Karul

(2017) A panel stationarity test with gradual structural shifts: Re-investigate the international commodity price shocks. Economics Modeling 61: 181–192.

45.

Nazlioglu

Payne

Lee

, et al. (2021) Convergence in OPEC carbon dioxide emissions: Evidence from new panel stationarity tests with factors and breaks. Economic Modelling 100: 105498.

46.

Pesaran

(2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74(4): 967–1012.

47.

Pesaran

(2007) A simple panel unit root tests in the presence of cross section dependence. Journal of Applied Econometrics 22(2): 265–312.

48.

Saleh

Verma

Ihalanayake

(2011) Do external shocks have a permanent or a transitory effect on Thailand’s tourism industry? Tourism Analysis 16(4): 483–491.

49.

Solarin

(2015) September 11 attacks, H1N1 influenza, global financial crisis, and tourist arrivals in Sarawak. Anatolia: An International Journal of Tourism and Hospitality Research 26(2): 298–300.

50.

Solarin

(2016) Global financial crisis and stationarity of tourist arrivals: Evidence from Mauritius. Current Issues in Tourism 19(9): 869–875.

51.

Smyth

Nielsen

Mishra

(2009) I’ve been to Bali too’(and I will Be going back): Are terrorist shocks to Bali’s tourist arrivals permanent or transitory? Applied Economics 41(11): 1367–1378.

52.

Sul

Phillips

PCB

Choi

C-Y

(2005) Prewhitening bias in HAC estimation. Oxford Bulletin of Economics and Statistics 67(4): 517–546.

53.

Tan

S-H

Tan

S-K

(2014) Are shocks to Singapore’s tourist arrivals permanent or transitory? An application of stationarity test with structural breaks. Current Issues in Tourism 17(6): 480–486.

54.

Tang

Wong

(2009) The SARS epidemic and international visitor arrivals to Cambodia: Is the impact permanent or transitory? Tourism Economics 15(4): 883–890.

55.

Tiwari

Dash

Narayanan

(2018) Foreign tourist arrivals in India from major source countries: An empirical analysis. Current Issues in Tourism 21(10): 1137–1156.

56.

Valadkhani

O’Mahony

(2018) Identifying structural changes and regime switching in growing and declining inbound tourism markets in Australia. Current Issues in Tourism 21(3): 277–300.

57.

Vanzetti

Peters

(2012) COVID-19 and Tourism: An Update”. United Nations Conference on Trade and Development, pp. 1–24.

58.

Xie

Tiwari

Chang

(2018) Investigating stationarity in tourist arrivals to India using panel KPSS with sharp drifts and smooth breaks. Applied Economics 50(46): 4985–4998.

59.

Yang

GJ-A

Ying

Y-H

Chang

, et al. (2014) Investigating stationarity in tourist arrivals to Taiwan using panel KPSS with sharp drifts and smooth breaks. Tourism Analysis 19(5): 573–580.

60.

Yin

(2001) Stationarity tests in heterogeneous panels. In: Baltagi

Fomby

Hill

(eds), Nonstationary Panels, Panel Cointegration, and Dynamic Panels (Advances in Econometrics, Volume 15). Emerald Group Publishing Limited, pp. 275–296.

61.

Yucel

(2021) Are shocks to tourist arrivals permanent or transitory? A comprehensive analysis on the top 20 most-visited countries. Current Issues in Tourism 24(16): 2294–2311.