21
lost because of missing observations on the covariates). In this run, we omit the scenario
dummies, since in this sample they would all be identically equal to zero. For simplicity,
this specification does not include variables about the respondent’s marital status,
education, and perception of the state of conservation of the monument.
The results of this run are reported in table 6. The coefficient on the site dummies
are positive and statistically significant. They imply that, all else the same, people take
fewer trips to Tatev than they do to the other sites. This is reasonable, considering the
difficulty of traveling to Tatev, due to the distance from the capital and the poor quality
of the roads. The coefficients on the Garni, Haghardzin and Khor Virap dummies are not
statistically discernible from one another.
Table 6. Single-site travel cost model. Dependent variable: annual number of trips (actual
visits only). Poisson model with on-site endogenous sampling. Maximum likelihood
estimation. N=468.
|
Variable_______ |
coefficient |
standard error |
t statistic |
|
Tatev__________ |
0.3214 |
0.1214 |
2.6474 |
|
Garni__________ |
1.0231 |
0.0856 |
11.9521 |
|
Haghardzin |
0.9964 |
0.0955 |
10.4335 |
|
Khorvirap_____ |
0.8507 |
0.089 |
9.5584 |
|
Price___________ |
-0.1263 |
0.0162 |
-7.7963 |
|
Pcappinc______ |
-0.002 |
0.0007 |
-2.8571 |
|
Incmiss________ |
0.5253 |
0.2552 |
2.0583 |
|
Log likelihood |
_______________-357.73_______________ | ||
The coefficient on price is negative, as expected, and strongly significant. Its
magnitude, -0.1263, is reasonable, and implies that, starting at 3 trips per year, an
increase in price by 3,000 AMD (a little less than 5 US dollars) reduces the number of
trips by one. Trip frequency is negatively related to income, and tends to be greater
among those people that did not report their income.