Modeling Tourism Spending Decisions As a Two-Step Process

ABSTRACT - The aim of this study was to investigate the determinants of tourism expenditure patterns of the U.S. households. Using the 1995 Consumer Expenditure Survey, a double-hurdle model was used to separate the decision to spend on tourism from the decision of how much to spend on tourism. The profile of a household most likely to spend on tourism as well as the profile of a household that relatively spends more on tourism are presented.



Citation:

Gong-Soog Hong, Mohamed Abdel-Ghany, and Soo Yeon Kim (1999) ,"Modeling Tourism Spending Decisions As a Two-Step Process", in E - European Advances in Consumer Research Volume 4, eds. Bernard Dubois, Tina M. Lowrey, and L. J. Shrum, Marc Vanhuele, Provo, UT : Association for Consumer Research, Pages: 216-223.

European Advances in Consumer Research Volume 4, 1999      Pages 216-223

MODELING TOURISM SPENDING DECISIONS AS A TWO-STEP PROCESS

Gong-Soog Hong, Purdue University, U.S.A.

Mohamed Abdel-Ghany, University of Alabama, U.S.A.

Soo Yeon Kim, Purdue University, U.S.A.

ABSTRACT -

The aim of this study was to investigate the determinants of tourism expenditure patterns of the U.S. households. Using the 1995 Consumer Expenditure Survey, a double-hurdle model was used to separate the decision to spend on tourism from the decision of how much to spend on tourism. The profile of a household most likely to spend on tourism as well as the profile of a household that relatively spends more on tourism are presented.

BACKGROUND

During the last decade, growth in tourism in the U.S. reached new highs. Tourism expenditures are projected to reach $473 billion in 1998, making it the nation’s largest services export industry. Total travel spending is up 95% from 1986, with domestic travelers contributing 80% (Galper 1998). Due to the increase in leisure time for workers and the rise in their disposable income (Dardis, Derrick, Lehfeld, and Wolfe 1981), it is expected that the trend of growth in tourism will continue.

Most previous studies used cross-section data and applied the ordinary least squares (OLS) method in examining the effects of economic and demographic variables on tourism expenditures. One of the problems associated with the use of such data, however, is that the sample data often contains a significant number of households that report zero expenditure on tourism. Consequently, econometric techniques not accounting for zero expenditure such as OLS lead to bias and inconsistent estimate (Maddala 1983).

So-called double-hurdle models are now established in the literature as being superior to Tobit modeling (Tobin 1958) in dealing with the zero expenditures. In consumption studies double-hurdle models can be used to separate the decision to spend on tourism (participate) from the level of spending (expenditures) and, therefore, provide more meaningful insights into consumers’ behavior than does the Tobit model (Cragg 1971). Moreover, whereas many of the same variables (such as income and demographics) may influence both participation and expenditure, they may have different effects on participation and expenditures. We attempt to address these issues using the double-hurdle model to analyze tourism expenditure in the United States.

REVIEW OF LITERATURE

Travel expenditure research

Using a sample drawn from summer travelers in Michigan’s Upper Peninsular region, Spotts and Mahoney (1991) classified travelers into three groups on the basis of travel expenditures: light, medium, and heavy spenders. The average expenditures per trip were $134.40. The greatest amount was spent for lodging ($39.18), followed by restaurant and bar expenditures ($27.35), vehicle-related spending ($26.71), and expenditures for groceries ($21.52). Heavy spenders were found to have higher incomes than medium or light spenders.

Based on the 1990 Consumer Expenditure Survey (CE) data, Fish and Waggle (1996) examined expenditures on travel and pleasure trips relative to current and permanent income. On average, households traveled almost four times in 1990 and spent $1,234 for trips, which represented 4.4% of total expenditures and 3.8% of after-tax income. Expenditures on travel and pleasure trips tended to increase monotonically as total expenditures increase. The amount of total expenditures was used as an indicator of permanent income. The highest quintile of total expenditures spent 5.3% of their permanent incomes on average for travel and pleasure trips, and the lowest quintile group spent 3.2% for trips. Travel spending as a proportion of before-tax income showed a somewhat different pattern. As before-tax income increased, the share of permanent income spent for travel decreased.

Cai, Hong, and Morrison (1995) also used the 1990 Consumer Expenditure Survey to examine household expenditure patterns for tourism products and services: food, lodging, transportation, and sightseeing and entertainment. Family life cycle, social class, and cultural and geographic factors were studied. Older households headed by those ages 65 and older spent more on food than young households headed by those ages 35 or younger. Married households and households with more children were likely to spend more on travel than non-married households and households with fewer children, respectively. As the education level of the head of the househol increased, the travel expenditure also increased.

Recreation and leisure expenditure research

Dardis, Derrick, Lehfeld, and Wolfe (1981) examined recreation expenditures using the 1972-1973 Consumer Expenditures Survey (CE) data. Households with heads age 65 and older were found to spend less money on recreation than households with heads under age 35. Education, employment status, region, and race were significant factors affecting the total recreation expenditures. As education of the household head increased, recreation spending also increased. Married, non-Black, retired, or unemployed household heads spent more money on recreation than those who were single, black, or employed did. Region variables also were significant. Households residing in the West spent less on recreation than households in other regions. Those living in rural areas spent less on recreation than those in urban areas did. Income was associated positively and significantly with household recreational expenditures.

Fan (1994) computed the mean budget shares of entertainment and transportation using 1980-1990 Consumer Expenditure Survey (CE) data. White households spent 14.2% and 6.3% of household budgets on entertainment and transportation, respectively. These numbers were 3.8% and 12.6% for Black households. About 4.1% of the budget for Hispanic households were spent on entertainment, and 13.9% was spent for transportation. It was shown that White households allocated more financial resources to these spending categories than did Black and Hispanic households.

Using the 1987-1988 Consumer Expenditure Survey (CE) data, Dardis, Soberon-Ferrer, and Patro (1993) investigated leisure expenditures. Their study showed that elderly people (age 55 and older) spent less than younger people on leisure activities did. Consistent with their earlier work (1981), income, education, race, and region variables were found to be significant factors affecting leisure expenditures. However, the region variable showed somewhat inconsistent findings. Compared with those living in the urban Midwest, those living in rural areas or the urban West spent less money on leisure activities. Salary income of households was a significant predictor of leisure expenditures. Income elasticity for salary income ranged from 0.14 to 1.71, suggesting that leisure activities are either a normal or a luxury good.

In a recent study by Dardis, Soberon-Ferrer, and Patro (1994), leisure expenditure was investigated using the 1988-1989 Consumer Expenditure Survey (CE) data. The findings indicated that income, family life cycle, number of adults in the household, number of children, education, race, gender, and region were significantly associated with leisure expenditures. Both the total salary income of all household members (including heads) and unearned income were related positively to leisure expenditures. Both the young-old group (age 55 to 64) and the old group (age 65 and older) spent more on leisure than the middle-aged group (age 35 to 44). As the number of adults or children increased, households spent more money on leisure activities. Households headed by individuals with relatively high levels of education spent more on leisure activities than those with relatively low levels of education. Black households and female-headed households spent less than non-Black or male-headed households. Compared with households in the urban Midwest, households in the urban Northeast or urban West spent less on leisure activities.

METHOD

Data and sample

Data for this study are from the 1995 Consumer Expenditure Interview Survey (U.S. Department of Labor 1995), the most extensive national household expenditure data available in the United States. This survey focuses on consumer units, defined to be all members ofa particular housing unit related by blood, marriage, adoption, or other legal arrangement.

A national sample of consumer units is interviewed once each quarter for five consecutive quarters; the first interview is used for bounding purposes. Using a rotating sample design, one-fifth of the sample is replaced by new units each quarter. For this study, only households that completed interviews for the first quarter during 1995 are included in the sample (N=4,961).

Model

As indicated in the introduction, the double-hurdle model specifies participation (equation 1) and expenditure (consumption) equations (equation 2):

Z* = Xa+m

Z = 1 if Z*>0

Z = 0 if Z* < 0   (1)

E* = Yb+e

E = 0 if E* < 0 and Z=0

E = E* otherwise.   (2)

In equation (1), the probability of traveling (Z*) is modeled as Z*=Xa+m, where X is a vector of explanatory variables, a is a vector of unknown parameter, and m indicates error terms. Z represents a dependent variable for the probit model based on Z*=Xa+m. Travel expenditure model is described in equation (2), where Y is a vector of explanatory variables, and b and e represent a vector of parameters and error terms, respectively. The model distinguishes the likelihood of spending on tourism from the level of travel spending, suggesting that individuals or households face two-step process of decision making. While a tobit model assumes the same set of variables would explain the decisions to travel and the level of travel spending, a double-hurdle model allows different sets of variables to be included in the model of participation and level of spending. A double-hurdle model is therefore preferred to a tobit model.

Variables

The dependent variable, expenditure on tourism, is the amount of expenditure on food, lodging, transportation, and entertainment. The independent variables include permanent income, net worth, number of earners, race of reference person, education of reference person, region of residence, household size, life cycle stage of the household. The number of trips, and types of trips were used as independent variables in the truncated regression model, whereas the number of vehicles owned was used as an independent variable in the probit equation only. The variables and their measurement are presented in Table 1.

The permanent income hypothesis suggests consumption is determined more by permanent than by present income (Friedman 1957). Permanent income is defined as the constant annual income adjusted for expected income and consumption patterns over the lifetime (Bryant 1990). In this study, we used the instrumental variable approach to estimate the permanent income. Measured after-tax income was regressed on a set of socioeconomic variables that represent human and non-human wealth, including race, age, education, and occupation of reference person, geographic region of residence, urban or rural residence, type of household, number of earners, and financial wealth. Wealth is defined as the market value of owned homes. The predicted values of the equation are interpreted as estimated permanent income. [Permanent income was predicted using the following formula: Permanent income = 777.24+6118.81 age (25-34)+8248.49 age (35-44)+7192.81 age (45-54)+3287.76 age (55-64)+6323.69 age (65 over)+2700.21 Northeast urban+4619.55 West urban+3649.19 South urban+490.07 Midwest urban+3153.41 high school graduate+4349.20 some college+11581.00 college-157.23 husband and wife only-7478.87 female single parent-6549.85 one person unit-6425.83 other family unit+1745.57 white-3445.14 female reference person-469.23 blue collar occupation+5552.77 white collar occupation-2540.99 self-employed-5612.04 retired+7585.12 number of earners+.083 market value of house.   Adjusted R2=.37  F=119.79   All explanatory variables were dummy variables except the market value of the house and number of earners. Omitted variables were: rural, less than high school education, age under 25, husband/wife/children family type, non-white, male reference person, other occupation, non self-employed, and non-retired.]

Net worth is the sum of after-tax income, dollar value of saving accounts, checking accounts, bonds, and securities. Number of earners is the actual number of earners in the household Race and education of reference person are used to capture differences of taste and preferences in spending on tourism. Level of education is divided into four categories: less than high school, high school graduate, some college, and college graduate. Household size is the actual number of persons in the household. Climate and cultural differences in each region of the country as well as rural and urban differences influence expenditure patterns (Ketkar and Cho 1982; Ketkar and Ketkar 1987). In this study, region is a categorical variable divided into urban Northeast, urban Midwest, urban South, urban West, and rural. The rural category is not specified further due to data limitations. The life cycle stages developed by Bojanic (1992) were adapted and modified to better suit the observations in data used in this study. Ten life cycle stages were proposed as shown in Table 1.

Due to non-normality of the distribution and to avoid the possibility for heteroskedasticity, expenditure on tourism, permanent income, and net worth were transformed using natural logarithms. The total number of observations in the sample is 4,961, of which 3,305 (66.6%) did not report any expenditure on tourism.

TABLE 1

MEASUREMENT OF VARIABLES

CHARACTERISTICS OF THE SAMPLE

Table 2 gives an overview of household characteristics in the sample. The sample is predominantly white (85%). Thirty one percent of reference persons obtained their high school diploma, whereas an equal percentage (almost 25%) each, either had some college years of education, or obtained their college degrees. Almost 90% of the households lived in urban areas. Slightly over one-fourth of the sample lived in the urban South.

Table 3 presents the mean values and distribution of tourism expenditures for households classified by stages of life cycle, race, and education. Permanent income was higher at $45,333 and $43,600 a year for households in the Full Nest II and Middle-aged Couples without Children life cycle stages respectively, compared with households in other life cycle stages. Expenditures on tourism were higher for households in the Newly married and middle-aged Couples without Children life cycle stages than households in other life cycle stages. Forty seven percent of households in the Newly Married life cycle stage had some expenditure on tourism compared with only 17.7% of the single parent households. The level of permanent income as well as expenditure on tourism was higher for white households compared with non-white households. As can be expected, households whose household head had a college or a higher degree of education had higher permanent income and spent more on tourism.

TABLE 2

DESCRIPTIVE STATISTICS (N=4,961)

TABLE 3

SELECTIVE STATISTICS BY LIFE CYCLE STAGES, RACE, AND EDUCATION

FINDINGS AND DISCUSSION

The parameter estimates of the double-hurdle model are shown in Table 4. As can be seen, net worth, and number of vehicles owned have a statistically significant positive effect on the probability of spending on tourism (i.e., taking a trip). Household size, however, has a negative effect. Neither permanent income nor the number of earners in a household has a significant effect on the probability of spending on tourism among households recording zero expenditure. For households who did not spend on tourism, households with a white reference person would have a higher probability of spending on tourism than their non-white counterparts. The probability of households whose reference person has some college education or a college degree of spending on tourism is higher than households whose reference person has a high school education. On the other hand, other factors equal, households whose reference person has less than high school education, has lesser probability of spending on tourism than households whose reference person has a high school degee.

The probability of households in rural areas, urban Northeast, and urban Midwest, to spend on tourism is lesser than household in urban South. Compared with households in the life cycle stage of Solitary survivor, households in any of the other life stages have a higher probability of spending on tourism, ceteris paribus.

Whereas permanent income has no effect on the probability of tourism spending, it has a highly significant impact on the level of expenditure among those households reporting expending on tourism. Since the natural log of permanent income was regressed on tourism expenditure, then the regression coefficient represents income elasticity. This means a 1% increase in income, ceteris paribus, will result in .19% increase in tourism expenditure. Demand for tourism with respect to permanent income is, therefore, inelastic. Albeit the positive relationship between income and expenditure on tourism is in support of previous studies (Dardis et al. 1981; Dardis et al. 1993; Dardis et al. 1994; Fish and Waggle 1996), the magnitude of the elasticity coefficient is in contradiction of the results obtained by Dardis et al (1981).

As the number of earners increases, the expenditure on tourism decreases. Since the actual number of earners was regressed on the natural log of expenditure on tourism, the marginal propensity to spend is the regression coefficient divided by the mean of expenditure, i.e. (-0.11/186.1). Thus, as the number of earners increases by one, the expenditure on tourism decreases by .06 of a cent. This negative relationship between the number of earners and expenditure on tourism could be attributed to the scarcity of time available to multiple-earner households compared with single-earner households.

Since the dependent variable was in logarithms, the anti-logs of the dummy coefficients are taken, and the resulting values indicate the percentage difference in expenditure for each dummy variable in relation to the omitted category, which has a base of 100 (Carliner 1973; Halvorsen and Palmquist 1980). The transformed coefficients for the dummy variables are shown in Table 5.

Households headed by persons who have college or higher education spent 23% more on tourism than those headed by persons with high school degree. The result is consistent with previous studies (Cai et al. 1995; Dardis et al. 1981; Dardis et al. 1993; Dardis et al. 1994).

Whereas households located in urban Northeast spent 20% more on tourism than their counterparts in the South, households in rural America spent 32% less on tourism than households in urban South. The result is in support of past studies indicating that rural households, holding everything else constant, spend less on tourism than urban households (Dardis et al. 1981; Dardis et al. 1993).

TABLE 4

RESULTS OF PROBIT AND TRUNCATED REGRESSION FOR EXPENDITURE ON TOURISM

The results for life cycle stage indicate that single-parent households spent 57% less on tourism than households headed by a solitary survivor. There are no significant differences, however, between households in the other life cycle stages and households headed by a solitary survivor. The results support the findings by Cai et al. (1995).

Households whose trips were identified as recreation spent 122% more on tourism than households whose trips were intended to visit relatives and/or friends. Similarly, households whose trips were classified as day trips or others spent 95%, and 23% respectively more on tourism than those households whose trips were to visit relatives or friends.

It is interesting to note that whereas the probability of white households to spend more on tourism than non-white households, there is no significant difference in the level of expenditure between the two types of households, ceteris paribus. The result contradicts past studies (Dardis et al. 1994; Fan 1994; Pitts 1990).

TABLE 5

TRANSFORMED REGRESSION COEFFICIENTS FOR DUMMY VARIABLES AS REGRESSORS ON TOURISM EXPENDITURE

CONCLUSIONS

The findings of this study lend credence to the superiority of the double-hurdle mode over the single decision Tobit model. The results clearly illustrate the importance of a two-step approach to tourism spending modeling. Parameter estimates of the double-hurdle model shows that the effects of some variables such as income, net worth, number of earners, life cycle stage, region, household size, education and race of the head of the household were different in each decision step. It provided more information than would have the Tobit model regarding the unique role of each variable in participation and spending decisions. The additional information may be valuable for understanding consumer behavior in the tourism market.

The results of this study can be used to profile the "typical" household as far as tourism spending is concerned. First, the household most likely to travel is a one that is headed by a white household head who is relatively well educated, has a smaller household size, headed by a non-solitary survivor, has higher net worth, owns more vehicles, and locates in the urban South. Second, a profile of a household that spends more on tourism has the following characteristics: its head is better educated and is not a single parent, has fewer earners, has a higher income, takes more trips especially recreational ones, and resides in urban Northeast. These results also provide essential marketing information to segment the pleasure travel market by various economic and socio-demographic characteristics of households.

REFERENCES

Bojanic, David C. (1992), "A look at a modernized family life cycle and overseas travel," Journal of Travel & Tourism Marketing, 1, 61-79.

Bryant, W. Keith (1990), The Economic Organization of the Household, New York; Cambridge University Press.

Cai, Liping. A., Gong-Soog Hong, and Alastair M. Morrison (1995), "Household expenditure patterns for tourism products and services," Journal of Travel & Tourism Marketing, 4, 15-40.

Carliner, Geoffrey (1973), "Income elasticity of housing demand," The Review of Economics and Statistics, 55, 528-532.

Cragg, John G. (1971), "Some statistical models for limited dependent variables with application to the demand for durable goods," Econometrica, 39, 829-844.

Dardis, Rachel, Frederick Derrick, Alane Lehfeld, and K. Eric Wolfe (1981), "Cross-section studies of recreation expenditures in the United States," Journal of Leisure Research, 13, 181-194.

Dardis, Rachel, Horacio Soberon-Ferrer, and Dilip Patro (1993), "Analysis of Leisure Expenditures in the United States," Consumer Interests Annual, 39, 194-200.

Dardis, Rachel, Horacio Soberon-Ferrer, and Dilip Patro (1994), "Analysis of leisure expenditure in the United States," Journal of Leisure Research, 26, 309-321.

Fan, Jessie X. (1994), "Household budget allocation patterns of Asian-Americans: Are they different from other ethnic groups," Consumer Interests Annual, 40, 81-88.

Fish, Mary and Doug Waggle (1996), "Current income versus total expenditure measures in regression models of vacation and pleasure travel," Journal of Travel Research, 35, 70-74.

Friedman, Milton (1957), A theory of the consumption function. Princeton, NJ: Princeton University Press.

Galper, Josh (1998), "Population update for August," American Demographics, 20, 32.

Halvorsen, Robert and Raymond Palmquist (1980), "The interpretation of dummy variables in semilogarithms equations," The American Economic Review, 70, 474-475.

Ketkar, Suhas L. and Whewon Ch (1982), "Demographic factors and patterns of household expenditures in the United States," Atlantic Economic Journal, 10, 16-27.

Ketkar, Kusum W. and Suhas L. Ketkar (1987), "Population dynamics and consumer demand," Applied Economics, 19, 1484-1495.

Maddala, G. S. (1983), Limited Dependent and Qualitative Variables in Econometrics. Cambridge, UK: Cambridge University Press.

Spotts, Daniel M. and Edward M. Mahoney (1991), "Segmenting visitors to a destination region based on the volume of their expenditures," Journal of Travel Research, 29, 24-31.

Tobin, James (1958), "Estimation of relationships for limited dependent variables," Econometrica, 26, 24-36.

U. S. Department of Labor, Bureau of Labor Statistics (1995), Consumer Expenditure Survey: 1995, Interview Survey Public Use Tape and Documentation. BLS, Washington, DC.

----------------------------------------

Authors

Gong-Soog Hong, Purdue University, U.S.A.
Mohamed Abdel-Ghany, University of Alabama, U.S.A.
Soo Yeon Kim, Purdue University, U.S.A.



Volume

E - European Advances in Consumer Research Volume 4 | 1999



Share Proceeding

Featured papers

See More

Featured

Always Trust in Your Friends? Cross-cultural Effects of Review Source and Incentives on Trustworthiness

Dionysius Ang, Leeds University Business School

Read More

Featured

Exploring the Intersection of Digital Virtual Consumption and Family Rituals

Linda Tuncay Zayer, Loyola University Chicago, USA
Jenna Drenten, Loyola University Chicago, USA

Read More

Featured

F3. The Dark Side of Happy Brands: A Case Study of Newport Cigarette Advertising

Timothy Dewhirst, University of Guelph, Canada
Wonkyong Beth Lee, Western University, Canada

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.