Distance Travelled by Tourists-An Econometric Approach

1. Introduction

The summer holiday season is a period of time that most people look forward to, and many will use this opportunity to travel to a tourist location or holiday home. Many goods and services sectors, and tourist locations will rely on this period for the majority of their revenues. Consequently, an understanding of the factors that influence the behavior of holidaymakers will be of great use to the marketing efforts of companies, and tourist locations.

The objective of this assignment is to examine some of the factors that will influence the distance holiday makers travel when vacationing. The OLS model I have established will determine whether a holidaymaker's age, income level and number of children will influence the distance(miles) they travel when going on vacation, and if so to what extent. The data set was obtained from the University of Daytona, and was part of an assignment for an Economics course in 2006 by Prof. Dr. Gustafson. The data consists of 200 observations of individuals from Chicago who went on a summer holiday.

The analysis starts with an investigation of the descriptive statistics for the data set, and is followed by the results of both the initial OLS regression model, and the altered model adjusted to HAC errors. An analysis of the assumptions of a CLRM follows, after which an interpretation is done of the results of the model. The analysis then proceeds with a Wald test to examine the last two regressors of the model, after which the results are concluded.

2. Descriptive Statistics

The data consists of 200 observations of holidaymakers from Chicago, and relates the individuals income (measured in US $), age, and number of children to the miles they travel when going on a summer holiday. Table 1 provides the mean, median, maximum, minimum and standard deviation for the data.

Firstly, the mean income level of $ 63,925 is 44% higher than the average income of $ 44,344 in Chicago for that period. This can be reasoned by the fact that people who are more afluent are more likely to go on holiday. The average age may appear high at first, however this is to be expected as children are not accounted for and the sample represents individuals who are in full-time employment. The mean miles travelled (1054,23) should be expected as Chicago is located in the Northeren part of the USA and so many vacationers will travel South.

The distribution of data as illustrated in Figure 1 clearly shows that the majority of vacationers travel between 400 and 1700 miles on their summer holiday. Although there are some outliers, mostly in the 0 - 100 interval for people who did not vacation outside of Chicago, there are not enough observations to require omission for the OLS model.

The following table is a correlation matrix and will indicate if there is a multicollinear relationship between individual variables.

Upon examination of the correlation matrix there is a 0.237 correlation between age and income level, which is to be expected as people tend to earn more money as they progress in their careers. The correlation between number of children and age is even higher at 0.348. This is rational as younger vacationers will still be planning a family or have just one child whereas older vacationers will be more likely to have children. These correlations do suggest multicollinearity, however this will be ignored in the model with the consequence of having high standard errors for the independent variables, and a high R2. Although there is a possibility of multicollinearity it is unlikely to affect the BLUE properties of the OLS Estimator.

The income, children and age are the independent variables that will be constituents of the OLS regression model. I expect the income coefficient to be positive since people who are more affluent are more likely to go further away on vacation, and can afford the travel expenses. In contrast, people who have children will tend to prefer the convenience of holiday destinations that are closer by. Finally, I expect people who are older to travel further distances as they would like to see new locations, whereas younger holidaymakers will focus more on the experience of the holiday.

3. Regression Model Estimate and Analysis

The results of the regression and regression errors adjusted to HAC errors can be found in Table 3 below. Firstly, to test whether a linear model specification is appropriate the RESET test was used. The results of this test (see Table 3 below) indicate that the null hypothesis of linearity cannot be rejected, and the linear model specification is appropriate. In addition, a regression adjusted to HAC errors was modeled due to evidence of heteroscedasticity.