Abstract
We examine in this note the impact of COVID-19 on the Spanish tourism sector by using a strong dependence model. Daily data from five equity markets are used and we find that the coronavirus crisis has increased the persistence in the data, moving in some of the series from a mean reverting process to a non-mean reverting one. Thus, shocks that were expected to be transitory have become permanent, implying the need of strong policy measures to come the series back to their long-term projections.
Introduction
In Spain, tourism accounts for 12% of GDP according to the Bank of Spain’s (2020) latest figures (the third major contributor to national accounts), what means that any adverse shock on this sector may have a dramatic impact on the Spanish economy. The recent unexpected perturbation, COVID-19, took place in mid-March 2020 and is stagnating the economic activity in Spain and all over the world. In fact, the most up-to-date forecast of GDP growth, carried out by Funcas’ Consensus (2020), points out that a technical recession (two consecutive quarters of GDP decline) is expected in 2020, deepening the vulnerability of the economic activity. Additionally, the World Tourism Organization (UNWTO, 2020) states that Spain occupies the second position in the ranking of most visited countries in the world after France and the second highest earning destination behind the United States. Thus, it is important for the scientific community to analyse the impact of COVID on this sector in Spain. In order to do so, we examine whether COVID-19 will have a temporary or a permanent effect on the tourist sector in Spain. These results will help practitioners make decisions in the short and in the long run. If we expect a temporary impact, companies could put up with coronavirus taking the appropriate safety measures, but if we expect a persistent effect, managers should radically modify the strategy and stronger measures should then have to be adopted.
At the same time, getting access to daily data is so difficult in the tourism sector (mainly because figures are released monthly by the National Statistics Office), that we have decided to use as proxy variables various measures that look at the evolution of the Spanish equity markets from macroeconomic, mesoeconomic and microeconomic perspectives. To investigate the economic consequences of coronavirus, we analyse the IBEX 35 data; to assess the mesoeconomic impact, we study the Madrid Stock Exchange (SE) Leisure, Tourism and Hotel total index; and to observe the microeconomic impact, we use the Meliá Hotel International and the Amadeus IT Group stock market data, that are the only two tourism companies in the IBEX 35 and therefore the most powerful enterprises in the Spanish tourism sector. Additionally, we have added the NH Hotels equity data to compare and underpin the results (this company does not belong to IBEX 35 but to the Madrid stock market).
This note focuses on the properties of the five aforementioned stock market indices: IBEX 35, Madrid SE Leisure, Tourism and Hotel total index (‘Madrid SE’), Meliá Hotel International (‘Meliá’), Amadeus IT Group (‘Amadeus IT Group’) and NH Hotels (‘NH Hotel’) stock market; as in many other studies on the persistence of shocks (e.g. Gil-Alana and Moreno, 2009; Lovcha and Perez-Laborda, 2018), we use fractional integration methods. The time period considered starts on 14 May 2018 and ends on 14 May 2020, thus making it possible to assess the evolution of the tourism sector prior to and during the pandemic. The disease was first confirmed in Spain on 31 January 2020, when a German tourist tested positive for SARS-CoV-2 in the Canary Islands, and the lockdown was imposed on 14 March 2020.
Data and methodology
We use daily prices data from five equity markets (IBEX 35, Madrid SE Leisure, Tourism and Hotel total index, Meliá Hotel International, Amadeus IT Group and NH Hotels) to assess tourism activity in Spain before and during COVID-19, from 14 May 2018 to 14 May 2020. The number of observations reaches 512 and the data source is Refinitiv Eikon (Thomson Reuters), which is a real-time financial and economic data platform. Meliá Hotel International and Amadeus IT Group belong to IBEX 35 as they are the two biggest touristic companies in Spain. Additionally, we have included the NH Hotels stock market data to compare and underpin the results with the other two companies. Meliá and Amadeus could be correlated with IBEX 35 data, so that the NH variable may clarify the final outcomes.
As mentioned earlier, we use a strong dependence model based on fractional integration (I(d)) to determine the effect of shocks. Thus, if the value of d is below 1, the effect of a shock will be transitory, taking longer to disappear the higher the value of d is; on the contrary, if d is equal to or higher than 1, there is no reversion to the mean and permanency of shocks. Other articles dealing with fractional integration in tourism data include Assaf et al. (2011), Al-Shboul and Anwar (2017) and Gil-Alana et al. (2019).
Results
The model examined is:
where yt is the observed time series (in logs) 1 ; β0 and β1 are unknown parameters corresponding to an intercept and a linear time trend; and xt is I(d), where d is a real value.
Across Tables 1 to 4, the error term ut in equation (1) is white noise; in Tables 5 to 8, autocorrelation is permitted by using Bloomfield (1973), and in Tables 9 to 12, ut is described in terms of a seasonal autoregression. First, we display the results with the data ending on 23 February 2020 which is the time of the first death by COVID-19 in Spain. Then, we enlarge the sample until 14 May 2020 to compare the changes due to the coronavirus crisis.
Differencing parameter in a sample ending at 23 February 2020: White noise.
Note: SE: Stock Exchange. The selected specifications based on the deterministic terms are marked in bold. The values in parenthesis correspond to the 95% bands for the values of d.
Selected coefficients across Table 1.
Note: SE: Stock Exchange.
* Evidence of mean reversion at the 5% level.
Differencing parameter using the whole sample: White noise.
Note: SE: Stock Exchange.
The selected models for each series are in bold.
Selected coefficients across Table 3.
Note: SE: Stock Exchange.
Differencing parameter in a sample until 23 February 2020: Bloomfield case.
Note: SE: Stock Exchange. The selected models for each series are in bold.
Selected coefficients across Table 5.
Note: SE: Stock Exchange.
* Evidence of mean reversion at the 5% level.
Differencing parameter using the whole sample: Bloomfield case.
Note: SE: Stock Exchange. The selected models for each series are in bold.
Selected coefficients across Table 7.
Note: SE: Stock Exchange.
Differencing parameter in a sample until 23 February 2020: Seasonal case.
Note: SE: Stock Exchange. The selected models for each series are in bold.
Selected coefficients across Table 9.
Note: SE: Stock Exchange.
* Evidence of mean reversion at the 5% level.
Differencing parameter using the whole sample: Seasonal case.
Note: SE: Stock Exchange. The selected models for each series are in bold.
Selected coefficients across Table 11.
Note: SE: Stock Exchange. Values in parenthesis in the third and fourth columns are t values.
Starting with the results based on white noise errors and looking at the data ending on 23 February 2020, we notice that the d-estimates are very close to 1 for ‘Amadeus IT Group’, ‘IBEX 35’, ‘Madrid SE’ and ‘NH Hotel’, where the hypothesis of a unit root cannot be rejected. However, for ‘Meliá’, the value of d is substantially smaller (0.92) and the I(1) hypothesis is rejected in favour of reversion to the mean. That means that a shock in the latter series, though persistent, will be a transitory nature, disappearing in the long term. We also observe negative time trend coefficients for ‘Madrid SE’ and ‘Meliá’ series.
If we extend the sample until 14 May 2020, the results are presented in Tables 3 and 4. The time trends are once more significant for ‘Madrid SE’ and ‘Meliá’, and the negative coefficients, as expected, are now higher. Surprisingly, we also observe an important increase in the order of integration in all series, especially for ‘Madrid SE’ and ‘Meliá’ and ‘NH’ (1.11, 1.07 and 1.06, respectively), where the unit root is now rejected in favour of an alternative with d higher than 1. Of particular interest is the case of ‘Meliá’, where d was significantly smaller than 1 prior to the crisis but became significantly higher than 1 when including the data during the coronavirus. This indicates that the effect of the crisis has been particularly serious in this latter series, since the crisis has produced a clear change in the persistence of the series, moving from mean reversion to a lack of it.
As a robustness test of our results, we extend the analysis to other assumptions on the error term. To start with, we consider autocorrelation in the errors. First, with data ending on 23 February 2020, the results appear in Tables 5 and 6. Here, we observe that the I(1) hypothesis cannot be rejected in any of the series, though for ‘Meliá’, the upper value in the interval is precisely 1.00, being therefore close to the mean reversion case. 2 The estimated coefficient for the trend is now only significant for ‘Meliá’. Tables 7 and 8 refer to the complete data and we observe that, as in the previous case, there is a rise in the order of integration in all cases; in fact, the I(1) null is now rejected in the five series against d > 1.
Finally, and based on the monthly frequency used in the data, an Autoregressive of order 1 (AR(1)) seasonal model
is assumed for the errors, and the results are reported in Tables 9 and 10 (with data ending at 23 February 2020), and in Tables 11 and 12 with the whole sample.
The results are similar to those in Tables 1 and 2. Evidence of time trends are observed for ‘Madrid SE’ and ‘Meliá’; the unit root hypothesis is unrejected for ‘Amadeus IT Group’, ‘IBEX 35’, ‘Madrid SE’ and ‘NH’, and this hypothesis is rejected in favour of reversion to the mean for ‘Meliá’. Using the whole sample, there is a rise in d, which is especially remarkable in the cases of ‘Madrid SE’, ‘Meliá’ and ‘NH’. As in the previous cases, including COVID-19 data, there is a change in persistence, and data for ‘Meliá’ moves from mean reversion to a lack of it.
Conclusions
The impact of COVID-19 on the Spanish tourism sector has been examined in this work by using fractional integration. Our results indicate that this sanitary crisis has been particularly serious in the case of companies related to tourism, increasing the level of persistence, and moving from mean reversion (and transitory shocks) before the crisis to lack of mean reversion (with permanent shocks) during it. Thus, strong policy measures should be taken into account by the companies (and authorities) if we want to recover the original levels-trends prior to the crisis. Putting in a different way, if there is now another exogenous shock affecting the tourism series, stronger actions should be adopted to recover the original levels in the series than if that shock would have happened prior to the crisis. One example could be observed in the recent package of measures announced by ‘Meliá’ in the 27th May Press Release, called Stay Safe with Meliá, 3 in which they present a transformation strategy based on four pillars: (1) safety for employees and customers, (2) reduced contact in interactions between customers and employees, (3) optimization of operational processes, simplifying and digitalizing the service, and (4) adaptation of the brand to new paradigms and customer needs. The new policies include extra cleaning and hygiene measures and new personal space: social distance, apart from a new and more flexible cancellation policy.
Footnotes
Acknowledgements
Prof. Luis Alberiko Gil-Alana acknowledges support from the Ministerio de Economía, Industria y Competitividad. A project from the University Francisco de Vitoria is also acknowledged. Comments from the Editor and an anonymous reviewer are gratefully acknowledged.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was financially supported by the Ministerio de Economía, Industria y Competitividad (grant number ECO2017-85503-R).
