24
The coefficients on the site dummies are positive and significant. The magnitude
of the coefficients indicates that people visit Garni more than Haghardzin, that
Haghardzin is visited more than Khor Virap, and that in turn the latter is visited more
than Tatev.
The coefficients on the scenario dummies suggest that people valued most highly
programs that offer conservation and improve services like food and lodging, information
centers, etc. for visitors, followed by programs that enhance the cultural experience of the
visit. In practice, however, the coefficients on the dummies for the SERVICES and
CULTURE scenarios are not statistically different from one another. They are, however,
statistically different from the coefficient on the INFRASTR scenario.
This specification predicts that at Garni, for example, the average welfare change
from the current situation to be 8,871 AMD for the culture-enhancing scenario, 6,458
AMD for the infrastructure-enhancing program, and 10,346 for the service-oriented
program. (Once again, these are the averages of in-sample predictions for the visitors we
interviewed on-site, not for the population of visitors at large.).
In specification II, the coefficients on the scenario dummies are very close to their
counterparts in specification I, whereas somewhat larger differences are observed for the
coefficients on the site dummies.21 We also find that persons with a University degree
visit somewhat less frequently (25% fewer times, all else the same), married individuals
visit 37% more (about one visit a year, all else the same), and that age has a positive
responses (a combination of actual and stated trips) are less price-responsive than post-policy behaviors.
Our interpretation of the literature is that stated and actual trips may or may not exhibit a different degree of
sensitivity to price and to changing other conditions, depending on the study and the context.
21 A likelihood ratio test shows that adding individual characteristics of the respondents, as we do in
specification II, improves the fit of the model significantly. The results of specification II are qualitatively
similar to those of specification I.