Tourism,insularity,and remoteness: A gravity-based approach

Abstract

This study investigates the sensitivity of bilateral tourism flows to distance, relative prices, and cultural and political proximity variables, with a special focus on small island destinations, using a gravity model. We find that these flows are negatively affected by larger distances between origin and destination countries and by lower gross domestic product (GDP) in both countries. There are significant differences between our subset of islands and other nations. On the one hand, small islands have higher elasticities of demand with regard to distance and destination GDP and are at a disadvantage compared to other destinations since they are small and remote. Furthermore, they have a much higher price elasticity of demand. On the other hand, sharing a common language and a common colonial past with the origin country has a greater positive impact on tourism flows to small islands than to other countries.

Keywords

gravity model small islands tourism

Introduction

Many small islands face serious economic handicaps due to their smallness and remoteness. Smallness implies important cost disadvantages resulting from indivisibilities and the impossibility for many businesses to reach minimum efficient scale, small labor market and lack of skilled labor, lag in adopting innovative technologies, markets dominated by monopolies or small oligopolies, and inefficiencies in the provision of public goods. Catering to large open international markets is not an option in these circumstances, especially when these small islands are remote from the larger world consumer markets. Absolute costs handicaps may then preclude the viability of most industries, when the world markets are competitive enough (Winters and Martins, 2004).

The only industries able to survive in this situation are the ones that allow rents due to natural resources and/or imperfect competition in differentiated products. The tourism industry is one of them. Indeed, small remote islands perform relatively well in the international tourism business. Many of them are among the top world destinations. Among the 32 top destinations ranked by tourism receipts per capita, one can find 21 small islands, populated by less than 2 million inhabitants (see Online Appendix). Tourism revenues cover at least 40% of imports of goods and services for 15 small islands countries and over 20% for 10 other small islands (Bertram and Poirine, 2018). Small remote islands have revealed comparative advantages in tourism industries due to unique natural endowments and to other factors linked to culture and history. One can even argue that “islandness” and remoteness can paradoxically be considered as an advantage to attract important categories of international tourists (Sufrauj, 2011).

In this article, we apply a gravity model to investigate further these tensions between the costs of smallness and remoteness which represent resistance factors for international tourism on one hand and attractive forces which generate the competitive advantage of the small islands on the other hand. The gravity model is well adapted for studying these opposing forces, especially when the model is augmented to take into account various historical and cultural factors such as common language, common currency, and past colonial links. A data set from the UNWTO is used to extract the number of arrivals in a country of destination j, from a country of origin i, for 174 countries, which are separated in two groups: 32 small islands and 142 other continental nations. The impacts of the explanatory variables on these tourism flows are estimated for each of these subsamples and compared to determine whether tourism in small remote islands is structurally different. Two alternative techniques of estimation are applied on cross-sectional data for 2015: (i) two-stage least squares (2SLS) and (ii) quasi-maximum likelihood (QML) with exponential count. The first estimation method is also used to compare estimates for various years of the 1995–2015 sample and to test for robustness.

Our empirical results confirm that bilateral tourism flows are negatively related to large distances between countries of origin and destination, to low gross domestic product (GDP) in both countries, but they also highlight significant differences between our subset of small islands and other nations. Small remote islands tend to have higher elasticities of demand with regard to distance and destination GDP, which represent major handicaps in international tourism, but also higher elasticities to a common language and a common colonial past, which can be important assets to overcome these disadvantages.

The article is organized as follows. We first provide a review of the literature, then present the model as well as the methodology, and discuss the empirical findings before offering some conclusions.

Review of the literature

Our article is at the crossroads of three branches of literature regarding economic development in small remote islands. First, there is a well-documented state of knowledge regarding the economic handicap of small remote islands. Second, the revealed comparative advantage of small islands in the tourism industry is also well-known. Third, there is a growing empirical literature using a gravity model to study various aspects of international trade in tourism services.

The economic handicaps of small remote islands are well summarized by the World Bank (2009) as “a three-dimensional challenge”: distance, density, and division. Islands are badly ranked in these three dimensions and, among the islands, the small Pacific Islands are at a particular disadvantage: (i) they are at the end of the distribution in terms of weighted average distance from large markets; (ii) they have very low economic density; and (iii) they suffer from acute geographical divisions from being sea-locked and isolated, as well as artificial divisions from their very often protectionist trade policies, which generate large inefficiencies and low competitiveness.

The various economic weaknesses and vulnerabilities which result from smallness and remoteness of many islands, such as a crucial lack of competitiveness in most industries, due to the lack of economies of scale and the indivisibility of public goods and infrastructure, are now well-known and surveyed (Armstrong and Read, 2006; Bertram and Poirine, 2018; Deidda, 2016, OECD, 2016, 2018).

In a much-celebrated paper, Winters and Martins (2004) show that compared to a median country (Hungary at the time), small remote countries could suffer dramatic cost disadvantages. They demonstrate that, because of this structural wedge between their costs of production and those of bigger and more connected countries, small islands could be able to compete internationally only in a few niche markets, which could include tourism. Nevertheless, in the tourism and hospitality industry, the cost disadvantage could be as high as 57.5% (for micro economies with median population of 12,000) or 28.5% (for very small economies with median population of 200,000). For personal air travel, they find that costs are 115.7% higher in micro economies and 56.8% higher in very small economies. Therefore, if these small island countries wish to rely on tourism to sustain their economic growth, they must be attractive enough, through unique and outstanding sites or activities, in order to overcome the cost disadvantages of smallness and remoteness.

Fortunately, many small and remote islands are endowed with this kind of attractive natural sites and unique historical and cultural resources. Clearly, many small islands have revealed comparative advantage in tourism activities (see Perrottet and Garcia, 2016; Scheyvens and Momsen, 2008; Sufrauj, 2011; UNWTO, 2004, 2012; World Bank, 2017). The sources of this comparative advantage can be found in natural endowments (such as sunny beaches, outstanding landscapes, and beautiful lagoons), reminiscent of the Heckscher–Ohlin theory of trade. In fact, if some of these natural resources are truly unique worldwide, one can explain trade without referring to the Heckscher–Ohlin model, in which different countries have different endowments of the same factors. One can even argue that smallness and remoteness, which generate costs and economic handicaps, may also be a source of attractiveness for international tourists, through natural elements such as biodiversity, “authenticity of natural sites,” and sociocultural assets (see Scheyvens and Momsen, 2008, Sufrauj, 2011).

In addition, it should be noticed that international trade in tourism services is characterized by intra-industry trade flows, which are typical in an industry characterized by product differentiation (natural and historical sites, sociocultural differences between countries) and by a structure of monopolistic competition on the supply side (Hanna et al., 2015; Nowak et al., 2012). Tourism in small islands is indeed the result of natural endowments, in accordance with Heckscher–Ohlin model of trade, and of two-way flows related to the Dixit–Stiglitz–Krugman theory of trade.

Since the Dixit–Stiglitz–Krugman monopolistic competition model of trade predicts that high-income countries will mostly trade with each other, the gravity equation fits well this type of intra-industry trade (see Anderson, 1979; Bergstrand, 1985; Deardorff, 1998; Helpman et al., 2008; and for a textbook treatment, Feenstra and Taylor, 2014: Ch. 6).

In its simplest form, the gravity model predicts that the flow of goods and services between each country pair is positively related to their economic size and negatively related to the distance between them. The gravity model has also been applied to a wide area of issues regarding trade patterns and various sources of trade, including the effects of the Olympics games (Rose and Spiegel, 2011), migrations flows (Karemera et al., 2000; Lewer and Van den Berg, 2008; Massidda et al., 2015), the international flows of ideas (Andersen and Dalgaard, 2011), and the flows of service exports (Ceglowski, 2006; Kimura and Lee, 2006).

In the field of tourism economics and management, gravity equations have been increasingly used, following the footsteps, although in a different framework, of Eilat and Einav (2004), who notice the similarities of their results with a gravity relation. More recently, Morley et al. (2014) have presented microeconomic foundations to justify the use of gravity models in their estimation of tourism flows. They conclude that

The new version of the gravity model can be presented as a valid tool to assess the effects of tourism policies, examining changes in any of the determinants in the equation. Consequently, their use as policy instruments such as evaluation of tourist taxes or promotional expenditure policies should have the same validity than those derived from traditional tourism demand models. Increased understanding of these factors will assist policymakers in developing more effective policies to increase destination competitiveness and attractiveness within a bilateral setting. (Morley et al., 2014: 8)

Culiuc (2014) applies the gravity model to bilateral tourism flows in order to estimate the significance of supply-side determinants of tourism (presence of direct flights and number of hotel rooms in the destination country) and demand-side factors (GDP, distance, remoteness, relative price, and other economic, geographic, cultural, political, and touristic specificities), along both extensive (tourist arrivals) and intensive (duration of stay) margins. Results highlight differences in elasticities to these variables between tourism and trade gravity models, between OECD and other countries, between small island developing states and continental nations, among others. Lorde et al. (2016) use an augmented gravity framework to test Linder’s hypothesis and find similarity in preferences, such as climate conditions in origin and destination countries, which affect tourism flows. Santeramo and Morelli (2015) apply quantile regression to gravity models to study agritourism flows. Saayman et al. (2016) use the gravity equation to study the effects of formal trade agreements and informal economic cooperation influence on international tourism flows. Zhang et al. (2017) explore the impacts of cultural values on tourism flows, using a panel data gravity model. Yang et al. (2019) apply the same model to examine the relationship between cultural distance and international tourism flows, and they find that the inhibiting effect of cultural distance on bilateral tourism flows seems to decrease with time since 2003. Similarly, the effect of cultural distance is also investigated in the study by Bi and Lehto (2018) in the case of Chinese outbound travelers. Harb and Bassil (2020) find that relative accessibility vis-à-vis other countries also affects tourism flows. Petit and Seetaram (2020) construct a proxy for cultural preferences and find it to have a significant effect on bilateral tourism exports in 12 OECD countries.

Our article follows this line of research by applying the gravity model to the study of international flows of tourists, with a special focus on small remote islands. Basically, there seems to be no proper theoretical reason to distinguish small islands from the other countries in a gravity framework. If the gravity equation applies at a general level, there is no reason to think that it may not apply to small countries as well. However, the case of small remote islands is particularly interesting because tourism is one rare sector where insularity enables them to have a strong comparative advantage, whereas smallness and remoteness create in the same time particular handicaps for these countries, for example, the lack of competitiveness and huge factors of resistance to trade in tourism services.

Indeed, one of our objectives is to examine whether the application of the gravity model simply reveals low flows of tourists to small remote islands because they are small and far from the visitors’ countries of origin or for other reasons. First, what is the net effect of remoteness? It could be negative because of the distance factor, but it could also be positive because of opposite factors, such as the effects of “average remoteness,” reminiscent of the “multilateral resistance effects” studied by Anderson and Van Wincoop (2003), or the attractiveness of an exotic far-away destination for certain tourists? Second, what are the effects of other factors in the augmented gravity equation: common language, common currency, and political, cultural, and economic links such as those inherited from former colonization? Finally, are the signs and magnitude of the various coefficients significantly different between the small islands and the other countries of the world?

Methodology

Ideally, we would want to investigate the effects of gravity variables on tourism revenues. But there are no data on bilateral flows of tourism revenues between countries, except for a limited number of OECD countries. The only bilateral data available for a large sample of world countries and small islands are on tourist arrivals, from the UNWTO (2018, 2019). The data set records the number of arrivals in a country of destination j, from a country of origin i, for 174 countries (32 small islands populated by less than 1 million residents, and 142 other continental nations) during the period 1995–2015. Converting a number of foreign visitors into revenues from international tourism is not an easy task since visitors from different countries of origin usually have different habits regarding both the length of stay and the total spending per day. We will therefore use a gravity model to explain these international bilateral flows of tourists, with a special inquiry on the case of small remote islands.

The main explanatory variables are the distance between the origin country and the destination country since distance is correlated with transportation costs, and the economic weights (GDP) of the origin and the destination countries, the latter being interpreted as a potential supply of tourism infrastructures and services. Other variables have been added to the basic gravity model to measure the effects of other factors on tourist flows: economic variables such as the price differential between the origin and the destination countries and a dummy variable for common currency, as well as cultural and institutional links, identified by dummy variables for common language, religion, and colonial links. For example, Okafor et al. (2018) find that a common language has a positive impact on bilateral tourism flows.

We also consider an index of aggregate relative remoteness for each country. This factor may bear some resemblance with the multilateral resistance factors analyzed by Anderson and Van Wincoop (2003) in the sense that tourist flows are not only influenced by the distance between two countries but also by the relative distance of each of these countries with respect to the rest of the world. For example, since Australia is far from all the other countries, compared to the Netherlands, Australian tourists will less hesitate to cover long distances, such as 7831 km to Japan, than Dutch tourists, who face the same distance to Japan. But this index may also represent other determinants of tourism flows, if multilateral average remoteness (which is particularly significant for the more remote small islands) is also a positive factor in the demand function of some groups of international tourists, as Scheyvens and Momsen (2008), and Sufrauj (2011) argue.

Other nontraditional determinants of bilateral tourism flows to small islands, such as the existence of direct flights between the country of origin and the insular destination, or the number of hotel rooms, on the supply side, climate similarity, time difference, touristic attractions, or conflict magnitude, have been proposed by Culiuc (2014). However, the author finds that the first variable abovementioned is subject to reverse causality, flights being added after a rise in tourism flows. Data availability for the second variable is limited, so that these supply-side variables do not appear in our model. In our preliminary empirical research, other demand-side factors, such as the number of UNESCO World Heritage sites, worldwide governance indicators, dummy variables for country independence or for free-trade agreements, among others, were also included in an extended gravity model, but discarded due to lack of significance or econometric issues such as multicollinearity. Thus, we estimate the following model

ln (F_{i j}) = β_{0} + β_{1} ln (D_{i j}) + β_{2} ln ({DM}_{i}) + β_{3} ln ({DM}_{j}) + β_{4} ({GDP}_{i}) + β_{5} ln ({GDP}_{j}) + β_{6} ln (P_{j} / P_{i}) + β_{7} (ComCur) + β_{8} (ComLang) + β_{9} (ComRel) + β_{10} (PastCol) + ε_{i j}

where i denotes the tourist origin country, j denotes the destination country, ln(.) denotes the natural logarithm operator, and the variables are defined as F_ij: number of annual tourist arrivals from i to j; D_ij: distance between i and j; DM _j : multilateral distance of country j (geometrically weighted average distance between country j and the 50 most powerful countries in terms of GDP); GDP: annual gross domestic product converted in US$ at current exchange rates; P_j/P_i: the price level ratio, computed by the World Bank’s 2011 International Comparison Program as the ratio of the PPP conversion factor to market exchange rate (which is a measure of the absolute level of a price ratio whereas the real exchange rate is only a relative index of a price ratio); ComCur: a common currency dummy variable (=1 if same currency in country i and country j); ComLang: a common language dummy variable, proxy of cultural proximity (=1 if same official language); ComRel: a common religion dummy variable (=1 if same religion); PastCol: a dummy variable for institutional proximity (=1 if past colonial links); and ε_ij is an error term.

Another important econometric issue concerned the potential endogeneity of some explanatory variables. In particular, GDP of the destination country can partly depend on tourism, especially in small countries where the tourism industry represents a large share of the economy. To solve this endogeneity problem, we use the following instrumental variables that proxy for GDP: the share of secondary school enrollment (World Development Indicators) which represents the role of education in economic development; an average index of six governance indicators (Worldwide Governance Indicators), to highlight the importance of institutions in economic growth, and latitude (CEPII databases) to take geography into account in economic development (see Poirine and Dropsy, 2019). The definition and source of these variables and instruments can be found in Table 1.

Table 1.

Data description and source.

Variable	Definition	Unit	Source
F_ij	Number of yearly bilateral tourist arrivals from origin country i in destination country j	number	UNWTO
D_ij	Distance between origin country i and destination country j, calculated as the geometrically population-weighted average distance between major cities (or capital) of each country	km	CEPII
DM _i or DM _j	Multilateral distance of country i or j = geometrically GDP-weighted average distance between origin country and the 50 countries with the highest GDP, converted in US$ at current exchange rates	km	CEPII, World Bank (GDP)
GDP _i or GDP _j	GDP of origin country i or j, converted in US$ at current exchange rates	US$	World Bank (WDI #NY.GDP.MKTP.CD)
P_j/P_i	Ratio of price in country j to price in country j (=Price level ratio of PPP conversion factor to market exchange rate)	unit	World Bank (WDI #PA.NUS.PPPC.RF)
ComCur	1 for common currency	0 or 1	CEPII
ComLang	1 for common official or primary language spoken by at least 20% if the population	0 or 1	CEPII
ComRel	1 for common religion	0 or 1	CEPII
PastCol	1 for pair ever in colonial relationship	0 or 1	CEPII
EduSec2	Secondary gross school enrollment	%	World Bank (WDI #SE.SEC.ENRR)
WGI	Worldwide Governance (six-index average) Indicator	unit (−2.5 to +2.5)	World Bank
Latitude	Average latitude of the country’s capital (or economic center if the capital is not populated enough)	degree	CEPII

Note: GDP: gross domestic product; CEPII: http://www.cepii.fr/CEPII/en/bdd_modele/presentation.asp?id=6; UNWTO: http://statistics.unwto.org; World Bank WDI (World Development Indicators): https://databank.worldbank.org.

A data set of bilateral tourist flows between 174 countries, available for the period 1995–2015, is used to perform regression analysis on cross-sectional data in 2015 and then for other years of the sample, rather than on panel data. Indeed, a generalized method of moments’ Arellano–Bond estimation method for dynamic panel data could have been used with the 1995–2015 panel data available, but this technique removes the cross-sectional fixed effects by differencing the dependent variable and would not have permitted to estimate the effect of distance and other time-invariant explanatory variables. On the other hand, cross-sectional analysis based on a single year (here, 2015) does not take into account the dynamic relationships that could explain the evolution of tourism over time, but allow the estimation of time-invariant factors, such as distance, common currency, common language, common religion, past colonization, and on tourism. In order to examine the robustness of our results over time, cross-sectional regressions are also performed for various years of the 1995–2015 sample, when sufficient observations are available, and the coefficient estimates are compared with the 2015 estimates, and then tested for stability.

We use two alternative techniques of estimation for the cross-sectional regression: (i) 2SLS and (ii) QML with exponential count.

The 2SLS estimation enables the instrumental variables described earlier to be included, and an appropriate weighting matrix (generalized least squares for heteroskedasticity, based on heteroskedastic and autocorrelation consistent (HAC) standard errors and covariance with Bartlett kernel, Newey–West fixed bandwidth equal to 11) to provide for robust covariance calculation and additional efficiency.

Since bilateral tourist flows are count data, QML estimation with an exponential distribution is also used to test the coefficients’ sign and significance, given that these nonlinear estimators do not allow to calculate elasticities, or include instruments, as 2SLS is able to do. The possible presence of zeros would normally be treated by an estimation method such as the pseudo-Poisson maximum likelihood. However, the bilateral UNWTO tourism data do not distinguish between possible zeros and missing observations, leaving 10,880 available observations in 2015 for our cross-sectional regression analysis.

Empirical findings

The regressions are run on a bilateral sample made of 32 small islands (with a population inferior to 1 million inhabitants) and 142 other nations of the rest of the world. The empirical results for the gravity model are presented in Table 2. This model appears to fit the data for 2015 quite well, since the adjusted R² is equal to 79% for 2SLS (it is not available for QML). Overall, both estimation methods also yield similar results in terms of signs and significance. Since QML coefficient estimates do not correspond to elasticities due to nonlinearities, only 2SLS elasticity coefficients will be discussed in the following.

Table 2.

Bilateral gravity model for small islands and the rest of the world (2015).

Sample (2015 cross section)	Rest of the world	Small islands	Diff.	Rest of the world	Small islands	Diff.
Number of countries	142	32		142	32
Number of observations	7000			9800
Estimation method	Two-stage least squares			QML (exponential count)
Dependent variable:	Log (number of tourists)			Number of tourists
Log (distance)	−1.73 (50.8)**	−1.90 (31.0)**	(2.5)*	−1.93 (57.2)**	−2.12 (16.0)**	(1.4)
Log (multilateral distance of country i)	0.22 (2.9)**	−0.91 (4.6)**	(5.4)**	0.39 (4.0)**	−0.25 (0.9)	(2.2)*
Log (multilateral distance of country j)	0.63 (7.7)**	1.37 (2.2)*	(1.2)	0.10 (1.0)	0.31 (0.4)	(0.2)
Log (GDP of country i)	0.91 (90.4)**	0.79 (32.7)**	(4.7)**	0.80 (63.0)**	0.77 (20.3)**	(0.8)
Log (GDP of country j)	0.93 (33.9)**	1.32 (16.1)**	(4.5)**	0.68 (43.1)**	0.98 (8.9)**	(2.7)**
Log (price ratio P_j/P_i)	−0.49 (8.5)**	−1.55 (11.6)**	(7.3)**	−0.31 (5.1)**	−1.45 (7.4)**	(5.5)**
Common currency	0.31 (3.1)**	−0.26 (1.1)	(2.2)*	0.01 (0.0)	−1.17 (2.4)**	(2.2)*
Common language	1.06 (18.1)**	1.28 (10.3)**	(1.6)	0.78 (8.6)**	0.29 (1.3)	(2.0)*
Common religion	0.10 (1.2)	0.26 (1.4)	(0.8)	−0.28 (1.9)**	−0.29 (0.9)**	(0.1)
Past colonization	1.13 (7.3)**	1.50 (5.3)**	(1.1)	1.48 (7.0)**	1.21 (2.1)*	(0.4)
Constant	−32.09 (22.8)**	−31.06 (4.4)**	(0.1)	−16.83 (11.4)**	−16.19 (1.7)	(0.1)
R ²	79.3%			Not available
Adjusted R²	79.3%			Not available

Note: QML: quasi-maximum likelihood; GDP: gross domestic product; Z-statistics are presented in parentheses; statistical significance levels of 1 and 5% are represented by “**” and “*,” respectively. The columns entitled “diff.” show the level of significance (based on Z-statistics) of the differences between the estimates for the world sample and the estimates for small islands.

As expected, the bilateral distance between each pair of countries negatively and very significantly affects tourist flows for the two samples (and for the two estimation methods). A column to the right of the coefficients (“Diff.”) indicates the significance (z-statistics) of their difference. For example, this difference between the distance coefficients for the rest of the world sample and for the islands’ sample is significant at a 5% level (z-statistics = 2.5) for the 2SLS method, but insignificant (z-statistics = 1.4) for the QML method.

The coefficients for relative remoteness are positive and significant for the destinations and for the origins of the travelers of the rest of the world, but not for the islands’ subsample. In particular, a significant difference in the marginal impacts of the overall remoteness of the origin countries (i) is observed between the two country samples.

Similarly, the coefficients for GDP, which are all very highly significant and positive, as expected from a gravity model, differ also significantly for the two samples. More precisely, GDP of the origin countries (i) seems to be a less important factor for travelers coming from islands than from continents, whereas GDP of island destinations (j) appears to have more attractive force than for other nations. In other words, raising an island’s income enables it more easily to alleviate its limited capacity to welcome a larger number of tourists and to increase its infrastructure to provide for more tourism services. Fluctuations in GDP can be decomposed into changes in GDP per capita and changes in population size: a rise in either component benefits tourism, since an upsurge in labor productivity is likely to improve the quantity and quality of tourism services, and a demographic increase brings about a rise in the labor force, which can help welcome more tourists: this effect is stronger than in the rest of the world.

There are two possible interpretations for this result, which may be both valid at the same time. The first one would be that the main obstacle to the development of a high-impact tourist industry in small islands is the lack of an adequate labor supply. The second one would be that very small and less populated islands cannot afford the kind of infrastructures needed to develop international tourism due to the lack of significant economies of scale in the cost of building ports, airports, as well as telecommunications and health-care infrastructures. For example, it costs about the same to build an international airport, or a deep seaport able to offer docking space for big container ships and big cruise ships, whether the local population numbers 100,000 or 2 million people. But the cost per capita is much higher in the first case, making it hard to afford such public infrastructure needed to attract international tourists (whether destination or day cruising tourism). Such lack of economies of scale is even a greater challenge for small archipelagos of islands.

It also makes sense to assume that the relative cost of living in the destination country, represented here by the price ratio of the destination country to the origin country (i.e. the real exchange rate), has also an impact on the tourist destination choice. The corresponding coefficients are all negative, in line with the model’s theoretical predictions, and highly significant. According to these estimates, tourism demand also appears to be about three times more price elastic for small islands than for the rest of the world.

The impact of using of a common currency at origin and destination is positive and significant for continental countries, but not for small islands. Once again, such monetary considerations are probably less crucial in selecting a remote luxury destination than a neighboring continental nation, for example, in the European Monetary Union.

In terms of cultural proximity, a common language has a very significant and positive effect in both samples, without significant differences between the two groups of countries. The impact of sharing a common religion appeared to be insignificant. Institutionally, however, having a colonial past together very significantly increases tourism flows, similarly to small islands or to continental nations.

A comparison of the magnitude of the marginal effects of doubling these bilateral factors (distances, GDP weights, or cost differentials) or narrowing ties (currency, language, religion, and colonial past) on tourism flows is now presented in Table 3.

Table 3.

Marginal effects of significant factors on tourism flows (2015).

Effect of:	Rest of the world (%)	Small islands (%)	Significance of difference
Doubling the distance between the origin and destination countries	−70**	−73**	*
Doubling GDP of the origin country	+88**	+73**	**
Doubling GDP of the destination country	+90**	+150**	**
Doubling the relative cost of living in the destination country	−29**	−66**	**
Sharing a common currency	+37**	−23	*
Sharing a common language	+190**	+260**
Sharing a common religion	+10	+30
Sharing a common colonial past	+210**	+347**

Note: GDP: gross domestic product. Statistical significance levels of 1 and 5% are represented by “**” and “*,” respectively, in the last column. Since the estimated model is: $ln (F_{i j}) = β_{0} + β_{1} ln (D_{i j}) + β_{2} ln ({DM}_{i}) + β_{3} ln ({DM}_{j}) + β_{4} ({GDP}_{i}) + β_{5} ln ({GDP}_{j}) + β_{6} ln (P_{j} / P_{i}) + β_{7} (ComCur) + β_{8} (ComLang) + β_{9} (ComRel) + β_{10} (PastCol) + ε_{i j}$ , then doubling the distance leads to the expected marginal effect on tourism flows: $ln (\frac{F_{i j}^{e}}{F_{i j}}) = β_{1} ln (2) thus \frac{F_{i j}^{e}}{F_{i j}} = exp (β_{1} ln (2)) and \frac{F_{i j}^{e}}{F_{i j}} - 1 = 2^{^β 1} - 1$ _. For example, since the cross-section estimate of β₁ = −1.90 for small islands, then the marginal effect of doubling the distance on the number of tourism arrivals is equal to 2^ ^β ¹ − 1 = −73%. On the other hand, sharing a common language has the expected marginal effect on tourism flows: $ln (\frac{F_{i j}^{e}}{F_{i j}}) = β_{7} thus \frac{F_{i j}^{e}}{F_{i j}} = exp (β_{7}) and \frac{F_{i j}^{e}}{F_{i j}} - 1 = exp (β_{8}) - 1$ For example, since the cross-section estimate of β₈ = 1.28 for small islands, then the marginal effect of sharing a common language with some foreign tourists on the number of tourism arrivals is equal to exp(β₈) − 1 = +260%.

Doubling the distance between two countries causes a similar loss of 73% of tourist flows to small islands and 70% to other nations. For example, the distance between Los Angeles, California and Papeete, French Polynesia (6621 km) is 1.64 times greater than the distance between Los Angeles and Honolulu, Hawaii (4028 km): as a result, the model predicts 61% less Angelinos vacationing in Tahiti than in Hawaii (0.61 = 1.64^(−1.90) − 1, given a logarithmic regression coefficient of −1.90, cf. Table 2), ceteris paribus.

Doubling GDP of the country of origin raises tourist arrivals to a small island by 73%, but to another destination by 88%.

Similarly, doubling GDP of an island destination increases the number of tourists by 150%, whereas the same relative rise in GDP in another country causes tourism inflows to grow by 90%. For example, given that French Polynesia’s GDP is about 7% of Hawaii’s GDP in 2015, the total number of tourists visiting the former island should be 97% lower than the latter island, according to the model, ceteris paribus.

Doubling the price level in the destination countries relative to the price level in the origin country reduces the number of tourists visiting small islands by about 66%, but those traveling to other foreign countries by only 29%. For example, Tahiti’s notoriously high cost of living, even by comparison to the Hawaii (we estimate prices to be 46% higher in Papeete than in Honolulu), causes a 44% loss of tourist arrivals in French Polynesia.

Sharing a common currency increases the number of tourist arrivals to continental countries by 37% but has no significant effect on island tourism.

The model also predicts that having a common language can about triple the flows of tourists, whether to a continental nation or a small island. This is one of the reasons why a large proportion of tourists visiting Hawaii are American (61%), and a significant share of tourists traveling to Tahiti are French (20%), in 2015, despite the great distance (17,401 km) and the long flights (about 20 h) separating France and French Polynesia (ISPF, 2014).

Similarly, sharing a common colonial past can more than triple the number of tourists to islands and elsewhere.

The robustness of this model is tested by running cross-sectional regressions for 2010, 2005, and 2000 (since an insufficient number of observations are available in 1995 for the regression to yield estimates) and by comparing the estimates to those previously interpreted for 2015. Table 4 presents the results for 2SLS, while Table 5 displays the results for QML. Almost all 2SLS and QML estimates appear to be stable for the small island sample (except the 2SLS estimates for GDP), but this is not the case for the rest of the world, as demonstrated by the significant differences between the estimates for 2015 and for other years (see * or ** in the columns entitled “2015 vs. 2010,” or “2015 vs. 2005,” or “2015 vs. 2000”). However, it does not appear that the time-varying coefficients follow a linear trend. For example, the 2SLS estimate of the bilateral distance elasticity for non-island nations oscillates between −1.73 in 2015, −1.93 in 2010, −1.99 in 2005, and −1.82 in 2000 (cf. Table 4). Furthermore, QML estimates of the same elasticity do not statistically exhibit the same instability (cf. Table 5).

Table 4.

Bilateral gravity model for small islands and the rest of the world (2015 vs. 2010, 2005, 2000): Two-stage least squares estimation.

Cross section	2015		2010		2015 versus 2010		2005		2015 versus 2005		2000		2015 versus 2000
Number of observations	7000		7032		2015 versus 2010		5921		2015 versus 2005		5201		2015 versus 2000
Sample	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands
Log (distance)	−1.73**	−1.90**	−1.93**	−1.86**	**		−1.99**	−1.99**	**		−1.82**	−1.74**	*
Log (multilateral distance of country i)	0.22**	−0.91**	0.58**	−0.69**	**		1.04**	−0.08	**	**	0.58**	−0.30	**
Log (multilateral distance of country j)	0.63**	1.37*	1.99**	1.12	**		2.29**	1.48*	**		1.63**	0.51	**
Log (GDP of country i)	0.91**	0.79**	0.91**	0.73**			0.92**	0.76**			0.87**	0.65**	**	**
Log (GDP of country j)	0.93**	1.32**	1.30**	0.99**	**	**	1.30**	1.05**	**		1.05**	0.86**	**
Log (price ratio P_j/P_i)	−0.49**	−1.55**	−0.79**	−1.60**	**		−0.66**	−1.15**	**	*	−0.34**	−1.45**
Common currency	0.31**	−0.26	−0.05	0.11	**		−0.38*	−0.08	**		0.19	−0.70
Common language	1.06**	1.28**	0.94**	1.30**			0.74**	1.42**	**		0.92**	0.93**
Common religion	0.10	0.26	−0.35**	−0.15	**		−0.12	0.17	*		−0.33**	0.17	**
Past colonization	1.13**	1.50**	0.99**	1.53**			1.17**	1.34**			1.19**	1.11
Constant	−32.06**	−31.06**	−54.76**	−22.87*	**		−60.63**	−31.97**	**		−43.81**	−15.56	**
R² (%)	79.3		69.3				69.1				73.3
Adjusted R² (%)	79.3		69.2				69.0				73.2

Note: GDP: gross domestic product. Statistical significance levels of 1 and 5% are represented by “**” and “*,” respectively. The columns entitled “2015 versus….” show the level of significance (based on Z-statistics) of the differences between the estimates for 2015 and for the second year.

Table 5.

Bilateral gravity model for small islands and the rest of the world (2015 vs. 2010, 2005, 2000): quasi-maximum likelihood.

Cross section	2015		2010		2015 versus 2010		2005		2015 versus 2005		2000		2015 versus 2000
Number of observations	9800		8591		2015 versus 2010		7671		2015 versus 2005		6783		2015 versus 2000
Sample	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands	Rest of the world	Small islands
Log (distance)	−1.93**	−2.12**	−1.94**	−2.06**			−1.92**	−2.17**			−1.88**	−2.08**	**
Log (multilateral distance of country i)	0.39**	−0.25	0.74**	−0.48	*		0.89**	0.30	**		0.56**	0.48
Log (multilateral distance of country j)	0.10	0.31	0.56**	2.03*	**		0.76**	1.35	**		0.98**	0.95	**
Log (GDP of country i)	0.80**	0.77**	0.77**	0.70**			0.75**	0.71**	*		0.70**	0.73**	**
Log (GDP of country j)	0.68**	0.98**	0.69**	0.76**			0.69**	0.67**			0.75**	0.80**
Log (price ratio P_j/P_i)	−0.31**	−1.45**	−0.49**	−1.44**	*		−0.33**	−1.31**			−0.38**	−1.32**
Common currency	0.01	−1.17*	−0.22	−0.32			−0.17	−0.65			−0.11	−1.01
Common language	0.78**	0.29	0.76**	0.44			0.65**	0.87*			0.59**	0.68	*
Common religion	−0.25	−0.29	−0.73**	−0.71*	**		−0.57**	−0.57			−0.53**	−0.17
Past colonization	1.48**	1.21*	1.59**	1.38*			1.73**	1.17			2.15**	1.38
Constant	−16.83**	−16.19	−23.23**	−24.01*	**		−25.44**	−21.94	**		−24.77**	−23.07	**
Average log likelihood	−9.32		−9.44				−9.49				−9.38

Note: GDP: gross domestic product. Statistical significance levels of 1% and 5% are represented by “**” and “*,” respectively. The columns entitled “2015 versus….” show the level of significance (based on Z-statistics) of the differences between the estimates for 2015 and for the second year.

Similarly, the 2SLS estimate of the bilateral price ratio for rest-of-the-world countries moves between −0.49 in 2015, −0.79 in 2010, −0.66 in 2005, and −0.34 in 2000, whereas the stability hypothesis for this coefficient is not rejected at a 5% level for the QML estimate. The stability over time of other 2SLS estimates for non-insular nations appears erratic. To summarize, the model’s coefficients for islands appear to be quite robust across the sample and for both estimation techniques, while results of stability tests are ambiguous for the rest of the world.

Conclusion

International tourism tends to be the key sector in the development of small island economies. However, small remote islands are characterized by a tension between huge economic handicaps, which tend to harm all the industries open to international competition, including tourism services, and unique attractive factors which provide them with a competitive advantage in international tourism.

Our application of the augmented gravity model, using two different cross-sectional regression methods (2SLS and QML with exponential count) allowed us to obtain new results regarding the opposite forces mentioned earlier and the differences between the coefficients obtained for the sample of small islands and the ones corresponding to the sample of the other countries.

First, we noticed that both regression methods give similar results in terms of signs and significance of the coefficients. The results of the 2SLS show that the main coefficient of the gravity equation has the expected signs both for the islands and for the rest of the world. But there are some significant differences in the level of some coefficients, indicating as we expected that small remote islands have some special characteristics that explain the flows of international tourists. An increase in the GDP of the island destination countries raises the number of international tourists in a significantly larger proportion than in other countries. This seems to confirm the initial handicap of small remote states in terms of infrastructures and hospitality supply.

Furthermore, a doubling of relative prices in the destination countries reduces the number of tourists traveling to small islands by three fifths, that is, twice as much as visitors in other countries. This higher price elasticity for small islands is another major challenge for them to overcome, since their remoteness and smallness tend to raise their cost of living and therefore the relative price of a vacation for nonresidents.

On the other hand, some historical and cultural ties between the origin and the destination countries appear to be a force of attraction for small island tourism. For example, speaking the same language about triples the expected flows of tourists, and having a common colonial past more than triple these flows. These strong effects of cultural and/or political proximity on tourism flows confirm the results of Vietze (2012) and Mc Elroy and Pearce (2006).

Supplemental material

Appendix - Tourism, insularity, and remoteness: A gravity-based approach

Appendix for Tourism, insularity, and remoteness: A gravity-based approach by Vincent Dropsy, Christian Montet and Bernard Poirine in Tourism Economics

Footnotes

Acknowledgements

The authors thank the anonymous referees, the participants of the 2018 QATEM workshop in Tahiti, and the 2018 WEAI conference in Vancouver, as well as Sylvain PETIT at UPF, for their valuable comments. We also thank the UNWTO for providing free access to its Tourism Statistics Database, and Julien VUCHER-VISIN (ISPF) for detailed tourism statistics of French Polynesia.

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

Vincent Dropsy

Supplemental material

Supplemental material for this article is available online.

References

Anderson

(1979) A theoretical foundation for the gravity equation. American Economic Review 69(1): 106–166.

Anderson

Van Wincoop

(2003) Gravity with gravitas: a solution to the border problem. American Economic Review 93(1): 170–192.

Andersen

Dalgaard

(2011) Flows of people, flows of ideas and inequality of nations. Journal of Economic Growth 16(1): 1–32.

Armstrong

Read

(2006) Geographical ‘handicaps’ and small states: Some implications for the Pacific from a global perspective. Asia Pacific Viewpoint 47(1): 79–92.

Bergstrand

(1985) The gravity equation in international trade: some microeconomic foundations and empirical evidence. Review of Economics and Statistics 67(3): 474–481.

Bertram

Poirine

(2018) Economics and development. In: Baldacchino

(ed.) The Routledge International Handbook of Island Studies. Abingdon: Routledge.

Lehto

(2018) Impact of cultural distance on international destination choices: the case of Chinese outbound travelers. International Journal of Tourism Research 20(1): 50–59.

Ceglowski

(2006) Does gravity matter in a service economy? Review of World Economics/Weltwirtschaftliches Archiv 142(2): 307–329.

CEPII databases (2018). Available at: http://www.cepii.fr/cepii/en/bdd_modele/bdd.asp (accessed 1 June 2018).

10.

Culiuc

(2014) Determinants of International Tourism. IMF Working Papers WP/14/82.

11.

Deardorff

(1998) Determinants of bilateral trade: does gravity work in a neoclassical world? In: Frankel

(ed.) Regionalization of the World Economy. Chicago: University of Chicago Press, pp. 7–22.

12.

Deidda

(2016) Insularity and economic development: a survey. International Review of Economics 63(2): 107–128.

13.

Eilat

Einav

(2004) The determinants of international tourism: a three-dimensional panel data analysis. Applied Economics 36(12): 1315–1328.

14.

Feenstra

Taylor

(2014) International Economics. 3rd ed. New York: Worth.

15.

Hanna

Lévi

Petit

(2015) Intra-tourism trade, income distribution and tourism endowment: an econometric investigation. Applied Economics 47(21): 2184–2200.

16.

Harb

Bassil

(2020) Gravity analysis of tourism flows and the ‘multilateral resistance to tourism’. Current Issues in Tourism 23(6(: 666–678.

17.

Helpman

Melitz

Rubinstein

(2008) Estimating trade flows: trading partners and trading volumes. Quarterly Journal of Economics 123(2): 441–487.

18.

ISPF, Institut de la Statistique de Polynésie Française (2014) Les dépenses des touristes internationaux en Polynésie française, Points Forts Etude, N°7.

19.

Karemera

Oguledo

Davis

(2000) A gravity model analysis of international migration to America. Applied Economics 32(13): 1745–1755.

20.

Kimura

Lee

(2006) The gravity equation in international trade in services. Review of World Economics/Weltwirtschaftliches Archiv 142(1): 92–121.

21.

Lewer

Van den Berg

(2008) A gravity model of immigration. Economic Letters 99(1): 164–167.

22.

Lorde

Airey

(2016) Modeling Caribbean tourism demand. Journal of Travel Research 55(7): 946–956.

23.

Massidda

Etzo

Piras

(2015) Migration and inbound tourism: an Italian perspective. Current Issues in Tourism 18(12): 1152–1171.

24.

Mc Elroy

Pearce

(2006) The advantages of political affiliation: dependent and independent small Island profiles. The Round Table 95(386): 529–539.

25.

Morley

Rossello

Santana-Gallego

(2014) Gravity models for tourism demand: theory and use. Annals of Tourism Research 48: 1–10.

26.

Nowak

Petit

Sahli

(2012) Intra-tourism trade in Europe. Tourism Economics 18(6): 1287–1311.

27.

OECD (2018) Making Development Co-operation Work for Small Island Developing States. Paris: OECD.

28.

OECD/The World Bank (2016) Climate and Disaster Resilience Financing in Small Island Developing States. Paris: OECD/The World Bank.

29.

Okafor

Khalid

Then

(2018) Common unofficial language, development and international tourism. Tourism Management 67: 127–138.

30.

Perrottet

Garcia

(2016) Tourism, Pacific Possible Series; Background paper no. 4. Washington: World Bank Group.

31.

Petit

Seetaram

(2020) Measuring the effects of revealed cultural preferences on tourism exports. Journal of Travel Research 58(8): 1262–1273.

32.

Poirine

Dropsy

(2019) Institutions, culture and the tropical development gap: the agro-climatic origins of social norms about thrift and sharing. Journal of Economic Issues 53(3): 677–702.

33.

Rose

Spiegel

(2011) The Olympic effect. The Economic Journal 121(553): 652–677.

34.

Saayman

Figini

Cassella

(2016) The influence of formal trade agreements and informal economic cooperation on international tourism flows. Tourism Economics 22(6): 1274–1300.

35.

Santeramo

Morelli

(2015) Modelling tourism flows through gravity models: a quantile regression approach. Current Issues in Tourism 19(11): 1077–1083.

36.

Scheyvens

Momsen

(2008) Tourism in small Island states: from vulnerability to strengths. Journal of Sustainable Tourism 16(5): 491–510.

37.

Sufrauj

(2011) Islandness and remoteness as resources: evidence from the tourism performance of Small Remote Island Economies (SRIES). European Journal of Tourism, Hospitality and Recreation 2(1): 19–41.

38.

UNWTO (2004) Making Tourism Works for the SIDS (Small Island Developing States). Madrid, Spain: World Tourism Organization.

39.

UNWTO (2012) Challenges and Opportunities for Tourism Development in Small Island Developing States. Madrid, Spain: World Tourism Organization.

40.

UNWTO (2018) Yearbook of Tourism Statistics, Compendium of Tourism Statistics. Madrid, Spain: World Tourism Organization.

41.

UNWTO (2019) Methodological Notes to the Tourism Statistics Database. Madrid, Spain: World Tourism Organization.

42.

Vietze

(2012) Cultural effects on inbound tourism into the USA: a gravity approach. Tourism Economics 18(1): 121–138.

43.

Winters

Martins

PMG

(2004) When comparative advantage is not enough: business costs in small remote economies. World Trade Review 3(3): 347–383.

44.

World Bank (2009) Pacific Islands Development in 3D, in World Development Report: Reshaping Economic Geography. Washington: World Bank Group.

45.

World Bank (2017). Pacific Possible: Review of Key Long-term Economic Opportunities. Washington: World Bank Group.

46.

World Development Indicators (2018). Available at: https://data.worldbank.org/products/wdi (accessed 1 June 2018).

47.

Worldwide Governance Indicators (2018). Available at: https://datacatalog.worldbank.org/dataset/worldwide-governance-indicators (accessed 1 June 2018).

48.

Yang

Liu

(2019) The world is flatter? Examining the relationship between cultural distance and international tourist flows. Journal of Travel Research 58(2): 224–240.

49.

Zhang

(2017) The impacts of cultural values on bilateral international tourist flows: a panel data gravity model. Current Issues in Tourism 22(8): 967–981.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.16 MB