Tourism and tax revenue: evidence from stay-over tourists in the Eastern Caribbean

ABSTRACT

This paper investigated the impact of international stay-over arrivals on tax revenue for tourism-dependent Small Island Developing States (SIDS). To this end, we constructed a unique monthly panel data set of stay-over arrivals and tax revenue and its various components for the period 2002–2018 for Eastern Caribbean SIDS. A Bartik-type instrumental variables estimator exploiting differences in the exposure to and the extent of shocks in the origin markets was used to quantify the impact of stay-over arrivals on tax revenue and its different sources. Our results showed a large cumulative positive elasticity to tourism arrivals of 26.98% for total revenue, 26.86% for goods and services revenue, 23.62% for international trade and transactions revenue, 20.5% for income and profits revenue and 5.54% for property revenue. The findings demonstrate the region’s strong dependence on tourism for government income.

KEYWORDS:

• Tax revenue; stay-over tourists; instrumental variables estimator; air travel; developing countries 1; development

Disclosure statement

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

Notes

1 One should note that REVENUE and TOURISTS were logged because there was a number of outliers in these series, and in terms of the former, to ensure that the error term was (log)normal. We also experimented with pure linear and log-linear and linear-log models but their fit was always considerably inferior to the log-log specification.

2 Phillips-Perron non-stationarity tests for the logs of total revenue, goods and services revenue, international trade and transactions revenue, income and profits revenue, property revenue and stay-over tourists produced χ²-statistics of 575.7, 416.9, 359.9, 554.6, 480.0 and 399.8, indicating that the null hypothesis of all panels containing a unit root could be decisively rejected for all variables used throughout the analysis.

3 One may want to also note, that the Born and Breitung (2016) test was designed for contexts where N is larger than T. Unfortunately, as far as we are aware, there is no equivalent tests available for when T is larger than N, as in our case.

4 Prior to this natural disaster tourism constituted over 20% to the economy. One may want to note that recently the Government in Montserrat has outlined an explicit strategy to reinvigorate tourism on the island (Government of Montserrat 2016).

5 One should note, however, that the test was designed for a large N units, so that it may be weak in our context of T > N.

6 For all specifications we explored whether there were effects up to 12 months after the shocks in the IV specification but for neither the total nor for the components was there any impact beyond t − 5.

7 Elasticities were calculated by using the estimated coefficients and means as given in Gnangnon (2020).

8 One should note that another assumption in this approach is that locations are independent, so that there are no spatial spillovers or correlations, and that the data consist of a series of steady states (Goldsmith-Pinkham et al., 2020). With regard to the former, this is arguably the case for stay-over tourists where most visitors mainly visit one island for each trip to the region, unlike for cruise tourism. In terms of the latter, since tourists generally come no more often than once a year, or at least not in subsequent months, it is more natural to think of each as a steady state.

9 The formula given in (6) is for expositional ease given for the case where one does not employ a leave-one-out strategy in terms of defining OTOURISTS. In the actual analysis the necessary adjustment to the weights in order to take account of this, as shown by Goldsmith-Pinkham et al. (2020) was implemented.