Modeling the International Tourism Expenditures of Individual Travelers

Abstract

This study presents a set of econometric models that measure the response of international tourism expenditures to consumer prices and other country-level factors, while controlling for the demographic characteristics of the individual travelers. We find that the price elasticity of the demand for international tourism services, conditional on traveling overseas, is approximately −0.8. We also find that the level of economic development of the country, its distance from the United States, and the age and income of the individual travelers all had significant positive effects on tourism expenditures.

Keywords

tourism international trade in services

Introduction

Over the last decade, there has been a significant expansion in travel and tourism from the United States to overseas destinations. The number of overseas recreational travelers from the United States increased from 21.1 million in 2000 to 25.9 million in 2009. The travelers’ expenditures outside of the United States, which are counted as U.S. services imports, rose from $25.5 billion in 2000 to $31.8 billion in 2009.¹

These international tourism expenditures can be a significant source of income for the destination countries, contributing revenues to local businesses and local taxing authorities. Therefore, it is important to understand the determinants of international tourism demand, and in particular the sensitivity of this demand to prices. If international tourism demand is not very sensitive to prices, tourism service providers can raise prices, and local authorities can raises taxes on these services, without a substantial loss in revenue.

There is a large and expanding literature that focuses on measuring the price sensitivity of international tourism demand by estimating the price elasticity of demand. The price elasticity summarizes the percentage decline in demand for every 1% increase in prices. If demand is price-inelastic (the price elasticity is less than one in absolute value), then an increase in price will increase total revenue. The literature utilizes econometric models to estimate the price elasticity of demand based on fluctuations in aggregate international travel data. After reviewing these studies of aggregate data, we present a new set of estimates based on individual-level international travel data.

We analyze the tourism expenditures of individual travelers from the United States to 43 different countries. Our individual-level data are from the 2009 Survey of International Air Travelers (SIAT). To our knowledge, this is the first study that uses the individual responses to the 2009 survey to model the overseas expenditures of leisure travelers from the United States.

We estimate a set of econometric models of the variation in expenditures across demographic groups and across countries. The models are based on a microeconomic framework in which the travelers decide how to allocate their expenditures between the domestic market and the overseas markets that they visit. The models satisfy the homogeneity, symmetry, and negative conditions implied by consumer theory. The traveler’s demand in each country varies with the demographic characteristics of the traveler, including the traveler’s income and age, and with characteristics of the destination country. The country-level factors include consumer price levels, distance from the United States, language differences, the level of economic development of the destination country, and the extent of crime and violence in the country. In the next section, we compare our modeling framework to prior studies.

Review of the Literature

Li, Song, and Witt (2005) and Song and Li (2008) survey recent advances in modeling the demand for international tourism. In this section, we summarize the parts of this literature that are closest to our study, and we discuss the advantages and limitations of the different approaches.

Li, Song, and Witt (2005) note that the predominant econometric framework for analyzing international tourism demand through the 1990s was single-equation static demand models based on aggregate international travel data. In the mid-1980s, researchers in the field started to apply the static Almost Ideal Demand System (AIDS) framework pioneered by Deaton and Muellbauer (1980). The static AIDS model involves system estimation using the expenditure shares of each destination country. The advantages of the AIDS framework are that it is well suited for estimating the extent of substitution between destination countries (summarized by the cross-price elasticities of demand), it has a flexible functional form, the econometric specification is derived explicitly from consumer theory, and it is straightforward to test the restrictions on the demand elasticities implied by the theory, including homogeneity, symmetry, and negativity. Fujii, Khaled, and Mak (1985), Syriopoulos and Sinclair (1993), Papatheodorou (1999), Durbarry and Sinclair (2003), and Han, Durbarry, and Sinclair (2006) are important examples of static AIDS models that quantify the price elasticities of international tourism demand based on aggregate international travel data.² The measures of aggregate expenditure shares are often constructed by allocating the receipts of a destination country across the countries of origin based on headcounts of international visitors, as in Papatheodorou (1999) and Han, Durbarry, and Sinclair (2006).

Recent contributions to this literature have focused on extending the static AIDS framework to incorporate dynamics factors like adjustment costs and habit persistence. Li, Song, and Witt (2004), De Mello and Fortuna (2005), and Li, Song, and Witt (2006) are important examples of dynamic AIDS models. The dynamic models are generally more successful at forecasting international tourism demand, but they still rely on aggregate international travel data.

In this study, we estimate the price responsiveness of international tourism based on the international tourism expenditures of individual travelers, rather than the aggregate international travel data analyzed in the static and dynamic AIDS studies. The advantages of the micro data are that they are direct reports of the international tourism expenditures of individual travelers and they allow us to control for the demographic characteristics of the traveler. On the other hand, the micro data have several limitations that shape the design of our econometric models. For example, the micro dataset is a cross-section rather than a time series, and so it is not possible to model dynamic adjustments.

Theoretical Framework

Our theoretical model focuses on individuals who reside in the United States and travel overseas, since these are the individuals that we observe in the survey data. The decision to become an overseas traveler will depend in part on the relative price of consumption in different countries. However, due to data constraints, we do not estimate the price-responsiveness of the number of travelers to a country. Instead, we estimate the price responsiveness of tourism expenditures conditional on the individual visiting the overseas destination.

In the model, the international travelers allocate their expenditures across the domestic market and the overseas markets that they visit. The expenditure function of individual i is E_i (p,Y_i). The vector p represents the prices of consumption in the countries. The variable Y_i is the total consumer expenditure of individual i.³

According to microeconomic theory, specifically Shephard’s lemma, the demand for consumption in country c (q_ic) is equal to the derivative of this expenditure function E_i with respect to the price of consumption in country c (p_c).

q_{i c} = \frac{\partial E_{i} (p, Y_{i})}{\partial p_{c}}

The individual’s demand for consumption in destination country c is equivalent to the individual’s demand for international tourism services in that country. Consumer theory predicts that the quantity demanded will fall as prices rise and will rise as prices fall (since the second derivative of the expenditure function with respect to p_c is negative).

We adopt a specific assumption about consumer preferences in order to derive our econometric specification. We assume that consumer preferences within this stage of expenditure have a Constant Elasticity of Substitution (CES) functional form, with an elasticity of substitution between consumption in the different countries that is equal to σ > 0. The CES demand structure is very common in econometric models of international trade, including Anderson and Van Wincoop (2003), Helpman, Melitz, and Rubinstein (2008), and many related studies. The assumption of CES preferences implies that the demand for tourism services in each location has the log-linear form in equation (2).

q_{i c} = Y_{i} P^{σ - 1} {(p_{c})}^{- σ} {(z_{c})}^{- σ} .

The variable z_c represents characteristics of country c that shift demand toward that country. The variable P is an aggregate price index implied by the CES preferences.

P = {(\sum_{c} {(Z_{c} p_{c})}^{1 - σ})}^{\frac{1}{1 - σ}} .

The demand model in equation (2) satisfies the homogenous, symmetry, and negativity conditions required by consumer theory.

Equation (4) is the econometric specification implied by equation (2).

\ln (q_{i c}) = α + β \ln (p_{c}) + γ \ln (Y_{i}) + ε_{i c} .

The parameter β is the conditional price elasticity of demand. It measures the response of demand to price changes, holding fixed (or conditioning on) the individual’s decision to travel to country c. In other words, it measures the price responsiveness of the intensity of international tourism demand rather than the price responsiveness of the number of travelers. In the context of the CES model, the conditional price elasticity of demand is equal to –σ(1-s_ic), where s_ic represents the share of consumption expenditure in country c in the total expenditures of individual i. Since the individual’s total expenditures include his or her consumption expenditure in the domestic market, s_ic for overseas market c is usually close to zero, and the conditional price elasticity is approximately –σ. The constant term α includes the price index P, because it is a common factor among the travelers, who all travel from the same country of origin in the same year. The variable ϵ_ic represents the error term of the model. We assume that the supply of tourism services in the overseas market is perfectly elastic, and therefore p_c is an exogenous variable in the model. This is a reasonable assumption, since we are modeling the expenditures of individual travelers.

In the econometric analysis, we consider several country characteristics that could shift the travelers’ demands: these include the level of economic development in the overseas destination, its distance from the United States, and whether English is the primary language in the destination country.⁴ We expect that the level of economic development will serve as a proxy for the level of amenities available in the destination country, and therefore we expect that it will have a positive effect on the individual’s demand for tourism services. We expect that international distance traveled is inversely related to the individual’s opportunity cost of time, and therefore we expect that individuals who travel farther will spend more days and dollars overseas.

The CES structural model is well suited for analyzing our data, because the available micro data are sufficient to estimate the parameters in (4) and to calculate the conditional price elasticity of international tourism demand. Absent the CES assumption, the specification in (4) still serves as a log-linear approximation to a more general demand system, and the estimate of β still measures the conditional price elasticity of international tourism demand.⁵

Data

The estimation data set includes international air travelers who responded to the 2009 Survey of International Air Travelers (SIAT) and met the following criteria.⁶ They were U.S. residents who identified leisure, recreation, holidays, or sightseeing as the main purpose of their trip. They traveled to one of 43 overseas countries that together accounted for more than 90% of all overseas U.S. travelers in 2009.⁷ In addition to the travelers’ destination and the main purpose of their trip, the SIAT provides information on the dollar value of the respondent’s expenditures and his or her length of stay outside of the United States, the respondent’s annual household income, and the respondent’s age. Table 1 lists the SIAT response fields that we analyzed. Table 2 lists the 43 overseas countries.

Table 1.

Fields in the Survey Responses

Field Name	Description
AGE	Age
EXPNUM	Number of travelers included in the expenditure estimate
EXPOUT	Amount of money spent outside of the United States, in dollars
INCOME	Eleven different ranges for annual household income
MAINDEST	Main destination of the outbound overseas trip
NTSOUTUS	Number of nights spent outside of the United States
PURPMAIN	Main purpose of the trip

Table 2.

Average Expenditures and Country Characteristics in 2009

Country	Average Expenditure Per Traveler (U.S. Dollars)	Consumer Price Index	Level of Development	Distance (km)
Australia	2,790	121.67	High income	15,958
New Zealand	2,586	102.49	High income	14,098
Singapore	2,433	82.70	High income	15,561
South Africa	2,207	66.72	Middle income	12,723
Thailand	1,938	48.46	Middle income	14,169
Hong Kong	1,905	82.93	High income	13,129
Indonesia	1,711	54.48	Middle income	16,371
Philippines	1,687	51.38	Middle income	13,793
Russia	1,645	54.81	Middle income	7,827
Brazil	1,636	97.72	Middle income	6,799
India	1,613	32.37	Middle income	12,051
Taiwan	1,610	52.37	High income	12,659
Switzerland	1,605	151.81	High income	6,607
Italy	1,556	115.99	High income	7,222
Japan	1,553	123.77	High income	10,910
France	1,523	117.13	High income	6,169
United Kingdom	1,515	96.76	High income	5,904
Poland	1,457	64.55	Middle income	7,183
South Korea	1,432	73.48	High income	11,174
Argentina	1,416	51.53	Middle income	8,402
Spain	1,381	101.65	High income	6,096
Turkey	1,358	73.85	Middle income	8,733
Vietnam	1,355	36.39	Low income	13,704
Germany	1,345	111.94	High income	6,406
Greece	1,295	100.57	High income	8,261
Iceland	1,177	100.92	High income	4,518
Ireland	1,129	134.15	High income	5,448
Austria	1,118	110.20	High income	7,130
Czech Republic	1,096	74.74	High income	6,907
China	1,078	54.81	Middle income	11,154
Colombia	1,018	63.42	Middle income	3,829
Morocco	1,011	84.85	Middle income	6,109
Denmark	1,002	161.78	High income	6,518
Peru	976	58.27	Middle income	5,671
Egypt	928	46.90	Middle income	9,358
Trinidad and Tobago	903	62.11	High income	3,501
Israel	896	100.86	High income	9,452
Costa Rica	859	58.50	Middle income	3,300
Netherlands	815	112.21	High income	6,198
Jordan	712	70.14	Middle income	9,540
Bahamas	712	83.84	High income	1,538
Dominican Republic	551	45.26	Middle income	2,376
Jamaica	509	47.64	Middle income	2,326

Our econometric models also include country-level variables from several sources. The measure of consumer prices in each country is the 2009 value of the price of consumption (PC) series in the Penn World Tables. It is a purchasing power index of consumer prices that is denominated in a common currency. Following Eilat and Einav (2004), we assume that this purchasing power index is a reasonable proxy for the cost of tourism services in the country. We include an indicator variable for the level of economic development in the destination country. It is equal to one if the country is classified by the World Bank as a high-income country and is equal to zero if it is classified as a low- or middle-income country. We measure international distance as the distance from capital to capital. We also include an indicator variable for whether English is the primary language in the destination country.

Table 2 suggests that these country-level variables can explain some of the variation in average expenditure per traveler across the countries. On average, travelers to more distant countries spend more, even though our measure of tourism expenditures does not include airfares. This reflects the longer average duration of trips to more distant countries. For example, Australia ranks at the top of the list, with the highest average expenditure per traveler, while the Netherlands ranks near the bottom, even though consumer price levels in these two countries are similar. On the other hand, when two countries are approximately the same distance from the United States but have different consumer price levels, the one with higher prices usually records higher expenditure per traveler. For example, travelers to China spend much less on average than travelers to Japan, and travelers to France spend much more on average than travelers to Morocco.

In addition, some of the differences in the average expenditure levels across the countries can be explained by differences in the demographic profile of the travelers that visit each country, since overseas travelers who are older and have higher incomes tend to spend more. Table 3 reports average expenditure per traveler for several demographic groups of travelers. Travelers with household income greater than $200,000 spent $1,973 per traveler on average, while travelers with household income less than $100,000 spent only $1,162 per traveler.⁸ Travelers who are younger than 40 spent $1,268 on average, while travelers who are 40 or older spent $1,426 on average. Table 4 reports the differences in the traveler’s demographic profiles across the countries. For example, 61% of U.S. travelers to France reported annual household income greater than or equal to $100,000, and the average age of the travelers was 50 years. In contrast, 15% of the U.S. travelers to Trinidad and Tobago reported income greater than or equal to $100,000, and the average age of the travelers was 40 years.

Table 3.

Average Expenditures by Demographic Group

Demographic Group	Average Expenditure per Traveler (U.S. Dollars)
Household income ($)
<100,000	1,162
100,000-200,000	1,360
≥200,000	1,973
Age
<40	1,268
≥40	1,436

Table 4.

Demographics of U.S Traveler to Each Country

Country	Share with Household Income ≥$100,000	Average Age
Argentina	0.55	48.6
Australia	0.53	49.4
Austria	0.44	53.6
Bahamas	0.69	43.2
Brazil	0.41	38.2
China	0.43	48.4
Colombia	0.27	42.3
Costa Rica	0.47	42.8
Czech Republic	0.64	52.5
Denmark	0.59	53.0
Dominican Republic	0.61	44.4
Egypt	0.42	57.3
France	0.61	50.2
Germany	0.48	48.4
Greece	0.58	44.8
Hong Kong	0.53	41.8
Iceland	0.44	44.0
India	0.58	39.6
Indonesia	0.50	41.8
Ireland	0.48	48.1
Israel	0.52	53.1
Italy	0.56	50.7
Jamaica	0.44	40.4
Japan	0.53	39.6
Jordan	0.37	56.0
Morocco	0.40	49.6
Netherlands	0.47	52.4
New Zealand	0.65	47.6
Peru	0.40	42.7
Philippines	0.44	45.9
Poland	0.40	43.6
Russia	0.49	52.0
Singapore	0.67	47.2
South Africa	0.67	53.7
South Korea	0.39	40.4
Spain	0.53	44.7
Switzerland	0.59	48.4
Taiwan	0.47	39.9
Thailand	0.43	42.7
Trinidad and Tobago	0.15	39.8
Turkey	0.52	50.0
United Kingdom	0.53	45.0
Vietnam	0.52	44.6

Table 5 reports the means and standard deviations of these individual- and country-level variables for the entire estimation sample. The first row in the table is the economic outcome that we are trying to explain, expenditure per traveler in U.S. dollars (not including airfare). The average value of this variable across the 43 countries is $1,373. The remaining variables in Table 5 are economic determinants of these expenditure levels. The purchasing power measure of consumer prices, the level of economic development, the English indicator variable, and international distance are all country characteristics that are common across the individual respondents. The average value of the consumer price index is 85.95, compared to 99.97 for the United States in 2009. In addition, 62% of the travelers in the SIAT visited high-income countries. The average international distance was 8,150 kilometers. The final three rows in Table 5 are individual characteristics of the travelers. As shown, 51% of the travelers had household income greater than or equal to $100,000, and 18% of the travelers had household income greater than or equal to $200,000. The average age of the international travelers was 46.25 years. In the next section, we use multivariate econometric models to estimate the separate contribution of each of these factors to the U.S. demand for tourism services in the 43 countries.

Table 5.

Descriptive Statistics

Variables	Mean	Standard Deviation
Expenditures per traveler (in U.S. dollars)	1,372	1,803
Consumer price index in the country	85.95	30.84
High-income country	0.62	0.49
International distance (in 1000 km)	8.15	3.78
Traveler’s annual household income is ≥$100,000	0.51	0.50
Traveler’s annual household income is ≥$200,000	0.18	0.38
Traveler’s age	46.25	16.10

Econometric Model of International Tourism Expenditures

Table 6 reports the baseline estimates of the parameters in equation (4). The dependent variable of the model is the log of expenditure per traveler, deflated by the consumer price index in the country. The first column of Table 6 reports the ordinary least squares (OLS) estimates of the log-linear model. We find that consumer prices in the destination country have a negative and statistically significant effect of the demand for international tourism services. The level of economic development of the country, its distance from the United States, and the income and age of the individual traveler all had positive and statistically significant effects. An F test rejects the hypothesis that all of the coefficients are equal to zero. A second F test rejects the hypothesis that the coefficients on the individual-level variables are equal to zero.

Table 6.

OLS Models of International Tourism Expenditures (Dependent Variable: Expenditures per Traveler, Deflated by the Consumer Price Index)

Explanatory Variables	OLS	Poisson	Negative Binomial PML
Consumer price index	−0.8354 (0.0670)*	−0.8380 (0.0820)*	−0.8149 (0.0731)*
High-income country	0.1861 (0.0559)*	0.1100 (0.0679)	0.0993 (0.0633)
International distance	0.5718 (0.0272)*	0.5567 (0.0330)*	0.5341 (0.0297)*
Income ≥$100,000	0.1400 (0.0317)*	0.1260 (0.0406)*	0.1509 (0.0368)*
Income ≥$200,000	0.2443 (0.0441)*	0.3720 (0.0534)*	0.3354 (0.0485)*
Age	0.0503 (0.0386)	0.1164 (0.0513)*	0.0846 (0.0492)
Constant	4.4142 (0.3034)*	4.7160 (0.3787)*	4.7831 (0.3614)*
Number of observations	5,134	5,134	5,134
Test of joint significance of all factors	F = 163.05*	Wald χ² = 679.48*	Wald χ² = 815.69*
Test for the significance of the individual characteristics	F = 30.48*	F = 102.64*	F = 112.36*
R²	0.1537	0.1740

Note: The table reports robust standard errors in parentheses.

The coefficient is significantly different from zero, or that the test statistic rejects the null hypothesis, at the 5% level. OLS = ordinary least squares; PML = pseudo maximum likelihood.

Since the model in equation (4) is log-linear, the estimated coefficients can be interpreted as elasticities. For example, the coefficient of −0.8354 on consumer prices means that a 10% increase in the price level in the destination country would, all else equal, decrease the traveler’s demand for tourism services by approximately 8.4% and would increase total revenues by approximately 1.6% (i.e., 10% minus 8.4%). This conditional price elasticity does not include the likely decline in the number of travelers to the country. Our estimate of the conditional elasticity is significantly less than one, indicating that the demand for tourism services (conditional on traveling to the country) is inelastic, but it is still significantly different from zero. The other parameter estimates have similar interpretations. Travelers to a high-income country spent, on average, 19% more. Tourists who traveled 10% farther spent 6% more. Travelers with household income greater than $100,000 spent 14% more, and travelers with household income greater than $200,000 spent 38% more (i.e., the sum of the coefficients on the two household income terms). Travelers who were 10% older spent 0.5% more.⁹

To check the robustness of the baseline estimates, we considered two alternative estimators. Santos Silva and Tenreyro (2006) recommend the use of Poisson models to estimate log-linearized economic models like the specification in equation (4). They demonstrate that OLS estimates of the coefficients of log-linearized models will be biased if there is heteroskedasticity in the error terms, as is often the case in cross-sectional estimation involving many countries. To address this concern, we estimated a Poisson model with the same set of explanatory variables and the same log-linear functional form (but different assumptions about the distribution of the error term). In the Poisson model, the dependent variable is average expenditures per traveler deflated by the consumer price index, in levels rather than logs. We report the Poisson estimates in the second column in Table 6. The Poisson estimate of the price elasticity is −0.8380. The other explanatory variables have the same signs and statistical significance as the OLS estimate.

While the assumptions about the error term in the Poisson model are more appropriate than OLS, they are still restrictive. The Poisson model assumes that the variance of the error term is equal to its mean. The negative binomial pseudo maximum likelihood (NBPML) estimator, on the other hand, generalizes the Poisson model by relaxing this restriction. We report NBPML estimates of the same model in the third column in Table 6. The NBPML estimate of the price elasticity is smaller than the OLS and Poisson estimates in absolute value, though the differences are not statistically significant at the 5% level. We use a likelihood ratio test to compare the NBPML model to the more restricted Poisson model. The test indicates that the dispersion parameter in the NBPML model is significantly different from zero, and therefore it favors the NBPML model over the Poisson model. We conclude that the NBPML estimate (−0.8149) is the best of the three alternative estimates in Table 6.

Additional Sensitivity Analysis

As an additional sensitivity analysis, we estimated the three models without the high-income country variable. In the baseline model in Table 6, the high-income country variable serves as a proxy for the level of amenities available in the country. Since the baseline model includes the high-income country variable, we are estimating the conditional price elasticity based on variation in consumer prices within the set of high-income countries and within the set of developing countries, but not based on variation between the two groups of countries. If the indicator for high-income country is omitted and the demand is positively correlated with the level of amenities available, then the estimate of the conditional price elasticity of demand should be biased toward zero. The estimates in Table 7 support this theoretical prediction. The OLS estimate of the price elasticity is −0.6419, and the NBPML estimate of the price elasticity is −0.7116. The other parameter estimates are very similar to their counterparts in Table 6.

Table 7.

Models without the Country’s Level of Economic Development (Dependent Variable: Expenditures per Traveler, Deflated by the Consumer Price Index)

Explanatory Variables	OLS	Poisson	Negative Binomial PML
Consumer price index	−0.6419 (0.0358)*	−0.7275 (0.0443)*	−0.7116 (0.0404)*
International distance	0.5761 (0.0272)*	0.5628 (0.0328)*	0.5364 (0.0296)*
Income ≥$100,000	0.1443 (0.0316)*	0.1286 (0.0405)*	0.1531 (0.0365)*
Income ≥$200,000	0.2470 (0.0442)*	0.3743 (0.0537)*	0.3380 (0.0487)*
Age	0.0410 (0.0385)	0.1085 (0.0512)*	0.0821 (0.0495)*
Constant	3.7044 (0.2200)*	4.3157 (0.2794)*	4.3956 (0.2464)*
Number of observations	5,134	5,134	5,134
Test of joint significance of all factors	F = 189.83*	Wald χ² = 671.05*	Wald χ² = 790.23*
Test for the significance of the individual characteristics	F = 31.24*	F = 102.95*	F = 114.81*
R²	0.1519	0.1734

Note: The table reports robust standard errors in parentheses.

The coefficient is significantly different from zero, or that the test statistic rejects the null hypothesis, at the 5% level. OLS = ordinary least squares; PML = pseudo maximum likelihood.

Next, we included a control for whether English is the primary language in the country. All of the travelers in our sample are residents of the United States, and we expect that most speak English. If they visit an English-speaking country, they may have access to a wider set of opportunities for excursions. If this were the case, then the English indicator variable would have a positive effect on average expenditures per traveler. Table 8 reports the estimated parameters of this extended model. The English indicator variable does not have a statistically significant effect in the OLS model, but it does have a significant positive effect in the Poisson and NBPML models. Including the English indicator variable lowers the estimated price elasticity to −0.8073 in the OLS model and to −0.7492 in the NBPML model.

Table 8.

Models with English Indicator Variable (Dependent Variable: Expenditures per Traveler, Deflated by the Consumer Price Index)

Explanatory Variables	OLS	Poisson	Negative Binomial PML
English is the primary language	0.0548 (0.0336)	0.1489 (0.0429)*	0.1201 (0.0401)*
Consumer price index	−0.8073 (0.0683)*	−0.7762 (0.0846)*	−0.7492 (0.0745)*
High-income country	0.1631 (0.0574)*	0.0602 (0.0706)	0.0449 (0.0633)
International distance	0.5821 (0.0281)*	0.5746 (0.0328)*	0.5518 (0.0300)*
Income ≥$100,000	0.1393 (0.0317)*	0.1214 (0.0406)*	0.1498 (0.0366)*
Income ≥$200,000	0.2428 (0.0440)*	0.3666 (0.0533)*	0.3298 (0.0482)*
Age	0.0529 (0.0386)*	0.1238 (0.0510)*	0.0916 (0.0482)*
Constant	4.2595 (0.3143)*	4.3711 (0.3831)*	4.4329 (0.3567)*
Number of observations	5,134	5,134	5,134
Test of joint significance of all factors	F = 139.91*	Wald χ² = 699.50*	Wald χ² = 836.26*
Test for the significance of the individual characteristics	F = 30.33*	F = 99.14*	F = 111.60*
R²	0.1542	0.1772

Note: The table reports robust standard errors in parentheses.

The coefficient is significantly different from zero, or that the test statistic rejects the null hypothesis, at the 5% level. OLS = ordinary least squares; PML = pseudo maximum likelihood.

Finally, Table 9 reports the sensitivity of the estimated price elasticity to five additional modeling alternatives. The first alternative includes a measure of crime and violence in the destination country, based on the World Bank’s World Governance Indicator (WGI) of the rule of law in the country. Specifically, we add an indicator variable for whether the country is below the median value of the WGI rule of law measure among the 43 countries.¹⁰ We expect that the level of crime and violence will have a negative effect on spending, reflecting precautionary restrictions on tourism activities and in some cases shorter stays, and this is supported by the data. Table 9 reports that controlling for this rule of law measure slightly increases the estimate of the absolute value of the conditional price elasticity relative to the baseline model in Table 6.

Table 9.

Additional Sensitivity Analysis

Estimated Conditional Own-Price Elasticity of Demand	OLS	Negative Binomial PML
Baseline model (in Table 6)	−0.8354 (0.0670)	−0.8149 (0.0731)
Model that adds the WGI measure of the rule of law	−0.8448 (0.0666)	−0.8214 (0.0731)
Model with alternative measure of economic development	−0.8251 (0.0805)	−0.8232 (0.0890)
Model that adds an indicator for income ≥$40,000	−0.8355 (0.0670)	−0.8149 (0.0731)
Model that omits observations for the richest country and the poorest country	−0.8395 (0.0695)	−0.8290 (0.0758)
Model with length of stay outside of the United States as the dependent variable	−0.0697 (0.0456)	−0.1382 (0.0726)

Notes: The table reports robust standard errors in parentheses. OLS = ordinary least squares; PML = pseudo maximum likelihood.

The second alternative in Table 9 includes an alternative measure of economic development. We replace the indicator variable for high-income countries with an indicator for whether the country’s gross national income per capita for the country is above the median of the 43 countries. Substituting this alternative measure of economic development has no significant effect on the estimated price elasticity of demand.

The third alternative further differentiates the income classes of the travelers by adding an indicator variable for whether their household incomes are greater than or equal to $40,000 per year. (In the baseline model in Table 6, travelers are distinguished by whether their household incomes are greater than or equal to $100,000 per year and whether they are greater than or equal to $200,000 per year.) The additional term is not statistically significant, and its inclusion in the model has no significant effect of the estimated price elasticity of demand.

The fourth alternative in Table 9 examines whether the elasticity estimates are sensitive to extreme observations in the estimation sample. To test this hypothesis, we estimated the model without the individuals who traveled to the richest country (Switzerland) or the poorest country (Vietnam). The elasticity estimates based on this restricted data set are similar to those for the baseline model in Table 6, slightly larger in absolute value.

Finally, we estimated the model using the traveler’s reported length of stay overseas, rather than total expenditures, as the dependent variable. The traveler’s expenditures can be factored into the length of stay (the time dimension) and the expenditures per night (the spending rate). By definition, the traveler’s conditional price elasticity of demand in the overseas destination is the sum of the elasticities along these two adjustment margins. This final variant of the model isolates the part of the conditional price elasticity that reflects the response in the traveler’s number of nights overseas. The elasticity estimate of this variant of the model is reported at the bottom of Table 9. The length of stay is very price-inelastic. Its price elasticity is not statistically different from zero. This implies that most or all of the conditional price elasticity of demand that we report in Table 6 represents adjustments in the traveler’s rate of expenditures overseas, rather than adjustments in the traveler’s length of stay overseas.

Conclusions

The econometric models reported in this study provide reasonable estimates of the price-responsiveness of the U.S. demand for international tourism services in the 43 overseas countries. The estimates imply that a 10% increase in prices should reduce the quantity demanded by each traveler by approximately 8% but should increase the total expenditure of each traveler by approximately 2%. This econometric estimate controls for several important country-level factors and for the demographic characteristics of the individual travelers. The estimates also indicate that the level of economic development of the country, its distance from the United States, and the income and age of the individual travelers all had significant positive effects on international tourism expenditure per traveler. The sensitivity analysis demonstrates that these results are fairly robust to variation in the estimation technique and the set of explanatory variables that are included in the specification.

Studies of international tourism demand have focused on different price elasticity concepts, and this explains the large differences in reported estimates. Eilat and Einav (2004), for example, estimate the price elasticity of the number of travelers to a country. Their estimate of this price elasticity is approximately −1.0. Our model, in contrast, isolates the price elasticity of the traveler’s demand for tourism services, conditional on visiting the country. Our elasticity estimate is approximately −0.8. These two elasticity concepts are complementary: the elasticity of total demand for international tourism services is the sum of our estimate of the elasticity of demand per traveler and the estimate of the price elasticity of the number of travelers in Eilat and Einav (2004). This sum is approximately −1.8. This total elasticity is, in turn, comparable to the estimates in the AIDS studies, which combine the two elasticity components and do not estimate each component separately. When considered in combination, these studies provide a more complete picture of how international tourism demand responds to price changes along these different margins.

These measures of the price elasticity of international tourism demand are crucial inputs into the formulation of pricing. If international tourism demand is not very sensitive to prices, then tourism service providers can raise prices without a substantial loss in revenues. Likewise, local authorities can raises taxes on tourism services without a substantial loss in tax revenues.

Footnotes

Acknowledgements

This research was conducted while the authors were working in the International Trade Administration of the U.S. Department of Commerce. The views expressed are those of the authors.

Declaration of Conflicting Interests

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

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Bios

Maksim Belenkiy is an international economist in the International Trade Administration of the U.S. Department of Commerce. He earned a PhD in economics from the University of California, Santa Cruz.

David Riker is an international economist in the Office of Economics, U.S. International Trade Commission. He earned a PhD in economics from the Massachusetts Institute of Technology.

References

Anderson

James

van Wincoop

Eric

. (2003). “Gravity with Gravitas: A Solution to the Border Puzzle.” American Economic Review, 93: 170-92.

De Mello

Maria

Fortuna

Natercia

. (2005). “Testing Alternative Dynamic Systems for Modeling Tourism Demand.” Tourism Economics, 11: 517-37.

Deaton

Angus

Muellbauer

John

. (1980). “An Almost Ideal Demand System.” American Economic Review, 70: 312-26.

Durbarry

Ramesh

Sinclair

M. Thea

. (2003). “Market Share Analysis: The Case of French Tourism Demand.” Annals of Tourism Research, 30: 927-41.

Eilat

Yair

Einav

Liran

. (2004). “Determinants of International Tourism: A Three-Dimensional Panel Data Analysis.” Applied Economics, 36: 1315-27.

Fujii

Edwin

Khaled

Mohammed

Mak

James

. (1985). “An Almost Ideal Demand System for Visitor Expenditures.” Journal of Transport Economics and Policy, 19: 161-71.

Gil-Pareja

Salvador

Llorca-Vivero

Rafael

Martinez-Serrano

Jose Antonio

. (2007). “The Effects of EMU on Tourism.” Review of International Economics, 15: 302-12.

Han

Zhongwei

Durbarry

Ramesh

Sinclair

M. Thea

. (2006). “Modelling US Tourism Demand for European Destinations.” Tourism Management, 27: 1-10.

Helpman

Elhanan

Melitz

Marc

Rubinstein

Yona

. (2008). “Estimating Trade Flows: Trading Partners and Trading Volumes.” Quarterly Journal of Economics, 123: 441-87.

10.

Gang

Song

Haiyan

Witt

Stephen

. (2004). “Modeling Tourism Demand: A Dynamic Linear AIDS Approach.” Journal of Travel Research, 43: 141-50.

11.

Gang

Song

Haiyan

Witt

Stephen

. (2005). “Recent Developments in Econometric modeling and Forecasting.” Journal of Travel Research, 44: 82-99.

12.

Gang

Song

Haiyan

Witt

Stephen

. (2006). “Time Varying Parameter and Fixed Parameter Linear AIDS: An Application to Tourism Demand Forecasting.” International Journal of Forecasting, 22: 57-71.

13.

Papatheodorou

Andreas

. (1999). “The Demand for International Tourism in the Mediterranean Region.” Applied Economics, 31: 619-30.

14.

Santos Silva

J.M.C.

Tenreyro

Silvana

. (2006). “The Log of Gravity.” Review of Economics and Statistics, 88: 641-658.

15.

Smeral

Egon

. (2010). “Impact of the World Recession and Economic Crisis on Tourism: Forecasts and Potential Risks.” Journal of Travel Research, 49: 31-38.

16.

Song

Haiyan

Gang

. (2008). “Tourism Demand Modelling and Forecasting—A Review of Recent Research.” Tourism Management, 29: 203-20.

17.

Song

Haiyan

Lin

Shanshan

. (2010). “Impacts of the Financial Economic Crisis on Tourism in Asia.” Journal of Travel Research, 49: 16-30.

18.

Song

Haiyan

Romilly

Peter

Liu

Xiaming

(2000). “An Empirical Study of Outbound Tourism Demand in the UK.” Applied Economics, 32: 611-24.

19.

Syriopoulos

Theodore

Sinclair

M. Thea

. (1993). “An Econometric Study of Tourism Demand: The AIDS Model of US and European Tourism in Mediterranean Countries.” Applied Economics, 25: 1541-52.

20.

Office of Travel and Tourism Industries, U.S. Department of Commerce. (2001). 2000 Profile of U.S. Resident Travelers Visiting Overseas Destinations Reported from Survey of International Air Travelers. Washington, D.C.

21.

Office of Travel and Tourism Industries, U.S. Department of Commerce. (2010). Profile of U.S. Resident Travelers Visiting Overseas Destinations: 2009 Outbound. Washington, D.C.