UNESCO World Heritage Site label and sustainable tourism in Europe: a user-generated content analysis

ABSTRACT

Officially branding local heritage through recognised labels is a strategy that regions can use to promote economic development. Regions increasingly seek more sustainable tourism development, which can be captured by the quality of local tourist service development. This paper examines whether the UNESCO World Heritage Site (WHS) label is associated with local tourism development of a higher quality and offers the first comparative study across European regions. Using TripAdvisor reviews of over 38,000 European locations, our results reveal a positive correlation between WHS labelling and measures of perceived quality and breadth of local tourist services.

KEYWORDS:

• UNESCO WHS
• label
• brand
• tourism development
• cultural heritage
• UGC

JEL:

• O3
• L1
• L8
• Z3

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author(s).

Notes

1. Briefly, when the χ² is found to be significant, it means that the proportional odds assumption (null hypothesis) is rejected and consequently not satisfied.

2. The Brant test’s results for the outcome ‘overall score’ are: χ² = 50.49, df = 11, and p > χ² = 0.000. Although the results are based on one outcome (overall score), we extended the gologit model to the whole analysis. This must not be a concern since the coefficients produced in gologit should theoretically be similar to the ologit model, as the ‘autofit’ option was employed in the gologit analysis and the proportional odds assumption was not violated for all outcomes. The ‘autofit’ option simplifies the process of identifying partial proportional odds models that fit the data.

3. To strengthen the Brant test and confirm the gologit model as the best fit, we employ the Akaike information criterion (AIC) comparing the goodness of fit for the models that could be used for a categorical model: ordered logit (ologit), multinomial logit (mlogit) and gologit. The reasoning behind selecting models with varying AIC is that the most fitting models are those with the lowest AIC. The results show the gologit model is the most fit where: ologit (AIC = 24,635.77), mlogit (AIC = 24,610.15) and gologit2 (AIC = 24,608.45).

4. We chose the Poisson regression as the outcome of the Diversity Index contains several zeros in the distribution.