Assessing the role of social media in tourism recovery in tsunami-hit coastal areas in Tohoku,Japan

Abstract

Social media have become increasingly popular. However, there has been little investigation on how to effectively mobilize this newly emerging tool to promote tourism in rural areas. The aims of this article are twofold. First, among social media, we explore the role of Twitter for tourism promotion, especially by focusing on the recovery process of tourism in a rural area affected by the huge earthquake and tsunami, Magnitude 9.0, which hit eastern Japan in March 2011. Second, to approach the first purpose, we compare two types of tourism: ordinary tourism and ‘volunteer tourism’. As to the latter, a massive number of volunteers came to these areas to help in the recovery work from the devastation such as removing debris and helping evacuees rehabilitate their lives in Iwate prefecture. We employed a text-mining method to find keywords used in the official Twitter account issued by the Iwate Prefectural government and time-series regression models to identify factors that promote the two types of tourism. Data were collected using official statistics on incoming numbers of ordinary tourists and volunteer tourists. Tweeted information was provided by Iwate Prefecture. The results revealed that, first, tweeted information on cultural and natural heritage had a positive relationship with the number of incoming tourists while information on disaster-related words had negative effects. In contrast, second, tweeted information on tourism resources worked negatively toward the number of volunteer tourists while that on rehabilitation/reconstruction and on volunteers worked positively. Consequently, it is important to design support measures that enable the local tourism sector to attract incoming tourists after a drop-in volunteer tourism as reconstruction of the disaster area progresses. In this context, our results suggest how to effectively utilize Twitter for this purpose.

Keywords

electronic word of mouth (eWOM)natural disaster social media tourism demand Twitter volunteer tourism

Introduction

Social media has played an increasingly popular role in tourism activity both for tourists and operators of tourism services. However, there has been little investigation on how to effectively mobilize this newly emerging tool to promote tourism, especially in rural areas. The aims of this article are twofold. In the first aim, we examine the role of Twitter among social media for tourism promotion in connection with the recovery of tourism after a natural disaster. The second aim focuses on the recovery of tourism in a rural area affected by the huge earthquake and tsunami, Magnitude 9.0, which hit eastern Japan in March 2011.

To achieve the first aim, we explore the role of Twitter in promoting the recovery of tourism after a natural disaster. This is because once a disaster happens, recovery of tourism in rural areas often faces difficulties and proceeds at a much slower pace or even fails to occur at all in comparison with that in urban areas (Fukui and Ohe, 2013). Thus, it is proposed that social media can be an effective tool to overcome the handicaps faced by tourism in rural areas. Since empirical evidence on this topic is lacking so far, we empirically investigated how social media can contribute to the process of recovery of tourism in disaster-hit areas.

The second aim especially focuses on the recovery process of tourism in a rural area affected by the huge earthquake and tsunami that occurred in March 2011. To approach this purpose, we compare two types of tourism: conventional ordinary tourism and volunteer tourism. Thus, this article takes a comparative perspective between tourism in an emergency situation and tourism in an ordinary setting. It was after the Kobe–Awaji great earthquake in 1995 that the significance of volunteer tourism in disaster-hit areas was widely recognized for the first time in Japan (Nakao, 2002). Just after the 2011 earthquake and tsunami, a massive number of volunteers came to the disaster-hit areas to help in the recovery from the devastation such as by removing debris and helping evacuees rehabilitate their lives. Pilling (2014) stated that the large number of volunteers in the 2011 disaster-hit areas showed that people in Japan are not docile nor uniform in behaviour, which is a stereotypic view of Japanese but demonstrated independence and put forth strenuous efforts on their own. These traits of independence and resilience of the Japanese people aid in the recovery from repeated devastations in this country (Pilling, 2014). Such traits are also necessary for recovery of tourism from either natural or man-made disasters in various parts of the world. This aspect, however, has been scarcely studied in the tourism research arena.

Thus, the overall aim is to investigate the role of volunteer tourism and how social media work on this type of tourism in comparison with conventional ordinary tourism. To approach these aims, by focusing on Iwate prefecture, we employed the text-mining method and econometric time-series regression models to explore factors related to these two types of tourism. Finally, policy recommendations were suggested.

Literature review

The topic of this article covers areas overlapping three domains, that is, disasters and tourism, volunteer tourism and social media. Although these three areas have attracted growing interest in tourism research and there has been a large increase in such studies, a topic encompassing the overlapping of the three domains mentioned above has been little studied. Firstly, with respect to literature on disasters and tourism, Ritchie (2009) conducted a system-based approach to crisis and disaster management related to tourism and mentioned that tourism organizations should work with media to ensure provision of consistent and accurate information to the public and stakeholders. Scott et al. (2010) dealt with safety and security in tourism and marketing directed toward recovery after crises; interesting chapters related to the topic of the present article were included. Specifically, Volo (2010) focused on the role of destination marketing organizations’ websites in communicating about the tourism crises caused by the avian flu in 2006 and evaluated these websites as an effective tool to better share information. Albattat and Som (2014) focused on the impact of natural and man-made disasters on tourism business in Thailand and pointed out the role of media in disseminating correct information. Natural disasters have not been a traditional topic in dark tourism research in comparison with well-studied historical topics such as the Holocaust. Studies on natural disasters have been increasing gradually. Wright and Sharpley (2018a, 2018b) focused on the earthquake in L’Aquilla Italy and investigated residents’ perceptions of dark tourists and tourists’ reactions to disaster pictures. Sharpley and Wright (2018) examined the role of media in the same Italian study area. What they studied, however, was conventional local media and international media, not social media. Hepburn (2017) reported on war and earthquakes from the perspective of Nepal’s bus travel based on her long experience in Nepal and concluded that in Nepal bus travel is more dangerous than being in a war or earthquake. Skinner (2018) focused on volcanic eruption events as dark tourism destinations starting from the historical event in Pompei to instances of modern eruptions including the case of a Japanese eruption. These interesting dark tourism studies, however, did not refer to social media or volunteers.

As an econometric evaluation of the impact of earthquakes on tourism inflow, Mazzocchi and Montini (2001) applied event study methodology to tourist arrivals in Central Italy. It is natural that this article did not focus on the roles of social media, which at that time were newly emerging. In Japan, due to frequent natural disasters, it is a matter of great concern how to implement the recovery of tourism from a disaster once it hits. The study of this field of tourism is designated as ‘restoration tourism’ (The Japan Society for Interdisciplinary Tourism Studies, 2013). In this context, the present study broadly examines restoration tourism. The domains examined in the present study have not been studied in the field of restoration tourism.

In short, although in crisis and disaster management of tourism many authors stressed the role of media to disseminate correct information to the public, the roles of social media in tourism recovery in disaster-hit destinations have not been studied.

Secondly, studies on social media in relation to tourism have been published recently at an explosive pace. Leung et al. (2013) conducted a literature review of social media in tourism and hospitality and stressed the importance of social media for competitiveness in tourism. Minazzi (2015) conducted full-fledged research on marketing of tourism using social media and characterized word of mouth (WOM) and electronic WOM (eWOM). Nevertheless, eWOM was not fully referred to as a disaster or crisis communication tool. The main research attention was focused on tourism marketing, for example, national tourism organizations (Hays et al., 2013), airline companies (Dijkmans et al., 2015; Hvass and Munar, 2012), the hotel industry in Hong Kong (Chan and Guillet, 2011) and in the United States (Leung et al., 2015), nature-based tourism (Wood et al., 2013) and recreation and educational institutions (Hajli and Lin, 2014; Zehrer and Grabmüller, 2012). Schroeder et al. (2013) investigated the roles of social media in crisis communications among international tourists and mentioned that social media is increasingly used for communication with and by tourists in times of crisis while stressing further study of social media as a means of crisis communications. This study fills the gap in this respect.

Thirdly, research on volunteer tourism has been extensively conducted since the 2000s. Wearing and McGehee (2013a) conducted a large number of literature reviews. International volunteer tourism often has been studied from pro-poor perspectives (Wearing and McGehee, 2013b); Borland and Adams (2013) for cases in Central America, Conran (2011) and Mostafanezhad (2014a, 2014b) for cases in Thailand, Coren and Gray (2012) in Vietnam and Thailand, Chen and Chen (2011) in China and Crossley (2012) in Kenya. To the authors’ knowledge, no studies have focused on disaster and volunteer tourism. We conducted this study to fill the scarcity gap.

Study area, methods and data

The Great East Japan Earthquake, which included a subsequent large-scale tsunami, hit approximately a 500-km range of coastline and devastated three prefectures in the Tohoku region in northeast Japan in March 2011: Iwate, Miyagi and Fukushima prefectures. This natural disaster took 18,729 lives with 2666 people still missing, and more than 350,000 houses were half or completely destroyed (Table 1). These damages were mostly caused by the tsunami that washed away coastal communities rather than the earthquake itself. The most heavily damaged prefecture was Miyagi, where more than 10,000 people lost their lives, and the least damaged was Fukushima. Nevertheless, Fukushima suffered special problems, that is, radiation issues caused by the crippled Fukushima Daiichi nuclear power plant where the power supply for the cooling system was destroyed by the tsunami. Due to this peculiar situation, the restoration work in Fukushima has progressed the least.

Table 1.

Damage by the Great East Japan earthquake in three prefectures.

Prefecture	Death toll			Number of missing	Number of damaged housing
Prefecture	Direct deaths	Related deaths	Total	Number of missing	Completely destroyed	Half destroyed	Housing collapsed
Iwate	4672	417	5089	1144	–	–	25,023
Iwate	29.5%	14.5%	27.2%	42.9%	–	–	7.0%
Miyagi	9583	873	10,456	1296	82,903	155,090	237,993
Miyagi	60.4%	30.4%	55.8%	48.6%	82,903	155,090	66.6%
Fukushima	1599	1585	3184	226	21,215	73,190	94,405
Fukushima	10.1%	55.1%	17.0%	8.5%	21,215	73,190	26.4%
Three prefectures total	15,854	2875	18,729	2666	104,118	228,280	357,421
Three prefectures total	100.0%	100.0%	100.0%	100.0%	104,118	228,280	100.0%

Source: The surveyed data by each prefecture as of September, 2013.

Although Iwate was the second most heavily devastated prefecture in terms of death toll and housing damage, we herein focus on Iwate for two reasons. The first reason is that the availability of data at the municipal level was better in Iwate than in the most damaged Miyagi where municipal level data were not well collected. The second reason was that Iwate was one of the most active prefectural governments in utilizing public tweeting both in terms of the number of tweets and followers of tweets (Figure 1). Since we employed a text-mining method to find keywords that influenced tourists through public tweets after the disaster, this credible information was crucial for us to conduct our investigation of tourism recovery. For these reasons, Iwate was selected as the most suitable area to study the relationship between social media and tourism recovery. Thus, we used the official Twitter account ‘prfe_iwate’ issued by the Iwate Prefectural government that had been opened in April 2010. We investigated contents of tweets and retweets and extracted representative nouns that well summarized contents of topics using ‘twilog’, which is an online service to obtain tweeted text data. Fortunately, the Iwate official account was registered in twilog so that we could see all previous tweets. Then, we classified tweets by month for text mining. Text mining was conducted by an R-based software for morphemic analysis called RMeCab. Words extracted from this analysis are designated as ‘terms’. The frequency of terms and classified terms by tourism resources were used for the data.

Figure 1.

Ranking of public tweets by prefecture.

The number of tourists coming into Iwate was tabulated by Iwate prefecture. The tourism statistics published by Iwate Prefecture do not show any difference between ordinary tourists, dark tourists and volunteers. The number of volunteers was provided by the Japan National Council of Social Welfare, which is a non-profit national association of social welfare agencies, including those for volunteer work, that was established based on the Social Welfare Law.

We estimated time-series regression models to identify factors that promote the two types of tourism, that is, conventional tourism and volunteer tourism, to statistically test whether the tendency of repeat visits exists by considering the time lag of each tourism demand. Data were collected using official statistics of monthly incoming numbers of ordinary tourists and volunteer tourists from January 2010, before the earthquake, to March 2013, 2 years after the earthquake, thus providing data for 39 months. For model estimations, actually data for 31 months were used due to the treatment of the first stage difference and first lag variables, which we will mention in detail at the model estimations. We used tweeted information provided by Iwate Prefecture.

Tourism trends before and after the disaster in Iwate Prefecture

The trends in tourism in Iwate before and 2 years after the disaster are illustrated in Figure 2, which shows the number of incoming tourists to Iwate in each monthly period in the year before the tsunami to 2 years after. As easily expected, the number of incoming tourists dropped just after the tsunami. Nevertheless, the tsunami hit in March, an off-season for tourism, so that the drop was not large. The wave of seasonality still existed even after the tsunami but on a smaller scale than before. More specifically, we compared the tourist inflow between leisure tourists and business-oriented tourists (Tables 2 and 3).

Figure 2.

Number of inbound tourists to Iwate.

Table 2.

Leisure-tourist inflow in Iwate before and after the tsunami (number of thousand arrivals).

Year	Overnight tourists			Day trippers			Grand total
Year	Total	Outside prefecture	Inside prefecture	Total	Outside prefecture	Inside prefecture	Grand total
2010	2891	1622	1269	9403	3193	6210	12,294
2011	3244	1749	1495	7947	3173	4774	11,191
2012	3013	1885	1128	7899	3198	4701	10,912
2011/2010	1.12	1.08	1.18	0.85	0.99	0.77	0.91
2012/2010	1.04	1.16	0.89	0.84	1.00	0.76	0.89

Source: Tourism Statistics, Iwate Prefecture.

Table 3.

Business-oriented tourist inflow in Iwate before and after the tsunami (number of thousand arrivals).

Year	Overnight tourists			Day trippers			Grand total
Year	Total	Outside prefecture	Inside prefecture	Total	Outside prefecture	Inside prefecture	Grand total
2010	1517	1027	490	1013	808	205	2530
2011	1977	1438	539	248	30	218	2225
2012	2078	1434	644	611	295	316	2689
2011/2010	1.30	1.40	1.10	0.24	0.04	1.06	0.88
2012/2010	1.37	1.40	1.31	0.60	0.37	1.54	1.06

Source: Tourism Statistics, Iwate Prefecture.

In the case of leisure tourists (Table 2), compared with the previous year (2010), which was an ordinary year before the tsunami, the number of overnight tourists increased in 2011 while that of day trippers dropped. In reality, this was a confusing time especially for tourism activity as it was just after the earthquake. In the first 1 or 2 months after the earthquake, people felt sympathy for the victims who had suffered losses and people in general in the disaster-hit areas. So people refrained from making a long trip for purposes of leisure. At the same time, however, people wanted to see what happened in the disaster-hit areas. It is true that people had dichotomous feelings during that time. In 2012, the number of overnight leisure tourists from prefectures other than Iwate increased while that of tourists from within Iwate prefecture decreased compared with 2010. Due to data constraints, we cannot specify from which prefectures the tourists came. This phenomenon was considered to reflect that people came to see what happened in that region, which can be said to be a kind of dark tourism demand. Despite this increase in demand from outside of Iwate, the number of incoming leisure tourists still decreased from the pre-tsunami year.

In contrast, the number of business-oriented tourists who made overnight stays increased sharply in 2011 and 2012 while those of business-oriented day trippers, especially from areas other than Iwate, dropped drastically (Table 3). This is because large-scale public restoration works for the infrastructure such as on roads, ports, railways and river/coast-line banks and new settlements for those local people who had lost their residence had been started. These activities will take years for completion. Up until the present that work continues but with the scale gradually decreasing. Due to this demand for restoration, the total number of business-oriented tourists who stayed overnight sharply increased in 2012 when the public restoration work with heavy machinery went into full swing.

Finally, we look at the trend of incoming volunteers to Iwate before and after the tsunami (Figure 3). The number of volunteers coming from throughout the country sharply increased just after the tsunami to help the disaster victims at the temporary evacuation sites and to clean up houses, roads and farm lands covered with massive amounts of debris and mud. Figure 3 also shows that nearly 50,000 volunteers came at the peak time. Nevertheless, Figure 3 clearly shows that this enthusiasm lasted at most for 6 months, and then declined rapidly.

Figure 3.

Number of incoming volunteers in Iwate.

To summarize, we can say that the tsunami created a peculiar pattern of tourism demand in disaster-hit Iwate just after that disaster; a drop in the total number of leisure tourists and an increase in demand by restoration workers. Another peculiarity observed here was the massive volunteer tourism that emerged just after the disaster. However, this grassroots enthusiasm among people in this country had withered by 6 months after the disaster. This is partly because large-scale professional restoration works using heavy machinery became in full swing replacing the initial hands-on first-aid action by volunteers who came in a human wave. This is an interesting substitution between the two types of special tourism demands, that is, volunteers and restoration workers. This relationship of the substitution of one type of tourism with another has never been studied before. Therefore, we address this relationship empirically.

In this examination, we will use empirical questions to investigate whether leisure tourism and volunteer tourism are completely different from each other. If they are not identical, we will examine to what extent the demand profile differs in terms of Twitter contents. Clarifying these points is crucial to establish tourism restoration plans for future disaster-hit areas in the social media era.

Estimation models

We estimated demand determinant time-series models for conventional ordinary tourism and volunteer tourism, respectively. The explained variables were the numbers of incoming tourists to Iwate in terms of conventional ordinary and volunteer tourism that were obtained from different public data sources described in the preceding sections. Seasonal adjustment was made on the data of ordinary tourism by a seasonal index, which was obtained by centring the moving average. We did not apply a seasonal adjustment to the volunteer tourism because data on volunteer tourism would not show any seasonal pattern. The model TO represents ordinary tourism and TV represents volunteer tourism. To make comparisons with each other, common explanatory variables were, firstly, the monthly counted frequency of tweeted keywords related to local tourism resources and volunteer-related tweeted keywords. Secondly, the repeat-visit effect for these two types of tourism was shown using the first lag of the number of tourists in each category 1 month before the present month’s explained-variable data, indicated by the suffix t − 1. Thirdly, monthly dummy variables were used to control for seasonal fluctuations that might remain for the model TO. Specifically, monthly dummy variables were used (from January to November unity was given for each month and otherwise 0; base month was December). We also considered a temporary shock dummy for the tsunami-hit month (March 2011 = 1, others = 0).

For the model TV, the dummy variables had different meanings from the model TO because volunteer tourism has an off peak in September and another small off peak in December. Therefore, we used the September dummy variable (September in 2011 = 1, other months = 0) and December dummy variable (December in 2011 = 1, other months = 0). We also considered the overall trend of a decreasing number of volunteers since September in 2011 termed as the volunteer off-peak dummy variable (until August in 2011 = 0, since September in 2011 = 1).

With respect to tweeted keywords, Figure 4 shows the classification of terms that were extracted by text mining. Terms were classified into four groups: food, tourism resources, disaster-related issues and others. ‘Shellfish’ was classified as food and ‘oysters’ were put under ‘shellfish’. Oysters are a well-known local delicacy in the coastal Iwate area. ‘Cultural resources’ and ‘festivals’ were placed under tourism resources, with ‘Hiraizumi’ and ‘heritage’ belonging to ‘cultural resources’. The Tohoku region, including Iwate, has an abundant cultural heritage and annual summer festivals attract large numbers of tourists. The Hiraizumi Chusonji Temple was designated as a World Cultural Heritage site by UNESCO in June 2011, just 3 months after the earthquake. This designation greatly cheered up people in Iwate. We considered this cheering-up effect in the estimation model. ‘Restoration/reconstruction’ and ‘volunteers’ are disaster-related terms. ‘Seasons’ were classified under ‘others’. Thus, the monthly frequency of each term was used for the model.

Figure 4.

Words obtained from text mining analysis.

Initially, as a price factor, we considered the seasonal adjusted general consumer price index (CPI) as an economic variable in the form of the first ordered difference. Although we also tried other specific CPIs, such as the transportation CPI, highway CPI, gas CPI and accommodation CPI, none were stably significant due to highly suspected multicollinearity with other variables. Nevertheless, no statistical significance was observed. Although as another important economic variable, we could not use income because the number of tourists of both types was calculated as aggregated data. We also introduced a long-lasting shock dummy for the effect on ordinary tourism that began the month of the disaster. Regarding volunteer tourism, we did not consider any dummy variable that represented an abnormal period because volunteer tourism only takes place when something abnormal happens. Additionally, dummy variables such as the designation of the Hiraizumi World Heritage Site were also selectively used: a temporary dummy variable (June 2011 = 1, before = 0) or a dummy variable was used (June 2011 and after = 1, before = 0). The selective use was to avoid multicollinearity with the tweeted word ‘Hiraizumi’. Again, these Hiraizumi dummy variables did not show any statistical significance. Thus, all of these insignificant variables mentioned above were dropped in the final estimation model.

Before the estimation, we conducted unit root tests, that is, Augmented Dickey–Fuller tests and Phillips–Perron tests, to confirm the stationarity among variables for the models and found that there was non-stationarity. Thus, we took the first ordered difference models (Table 4). We tried the estimation using the seasonally adjusted data that showed the number of ordinary tourists minus the number of volunteer tourists. Nevertheless, the results are the same with the case using the seasonally adjusted data on the number of ordinary tourists in terms of signs and significance levels of parameters except for adjusted R², which were all smaller than the case of the number of ordinary tourists. Thus, we used the data on the number of ordinary tourists, eventually. Consequently, we set up time-series estimation models for the two types of tourism demands, respectively. Two types of estimation models are:

Model T O : Δ T_{t}^{O} = α_{0} + α_{1} Δ T_{t - 1}^{O} + Σ_{k}^{11} ​_{​} α_{2 k} {MTD}_{k} + α_{3} TDD + α_{4} Δ {TW}_{i t} + μ_{t}

Model T V : Δ T_{t}^{V} = β_{0} + β_{1} Δ T_{t - 1}^{v} + Σ_{j}^{2} β_{2 j} {MVD}_{j} + β_{3} Δ {TW}_{i t} + υ_{t}

where Δ represents the first-order difference; $T_{t}^{O} and T_{t}^{V}$ are the number of ordinary tourists to Iwate or number of volunteers; $T_{t - 1}^{O} and T_{t - 1}^{V}$ represent the first lag that represents the customary effect of $T_{t}^{O} or T_{t}^{V}$ ; TW _it represents the frequency of term i extracted from tweeting by Iwate prefecture at period t; MTD _k represents the monthly dummy variables based on December for ordinary tourism (k = 1,…,11); TDD represents the temporary disaster dummy variable (March 2011 = 1, others = 0); MVD _j represents the monthly September or December, dummy variables for volunteer tourism (j = 1, 2); α₀ and β₀ are constant; α_h and β_m are the parameters (h = 1, 2, 3; m = 1, 2, 3); and and μ_t and υ_t are error term.

Table 4.

Results of unit root test on number of tourists and volunteers in Iwate.

Variables	Original data		First ordered difference
	PP test	ADF test	PP test	ADF test
	Z(t)	Z(t)	Z(t)	Z(t)
Number of tourists to Iwate	−3.988***	−3.966***	−8.375***	−8.210***
Number of tourists to Iwate (seasonal adjusted)	−2.102	−2.188	−6.403***	−6.385***
Number of volunteers in Iwate	−1.821	−1.371	−2.642*	−2.659*
Number of tweets by Iwate prefecture	−3.646***	−3.700***	−7.131***	−6.569***
‘Tourism’	−2.505	−2.524	−7.017***	−7.067***
‘Food’	−2.878**	−2.915**	−8.861***	−8.960***
‘Tourism resources’	−3.098**	−3.078**	−6.181***	−6.203***
‘Festival’	−4.454***	−4.542***	−8.494***	−8.739***
‘Shell fish’	−2.836*	−2.856*	−2.852*	−3.876***
‘Cherry Blossom’	−4.664***	−4.639***	−10.711***	−8.960***
‘Cultural heritage’	−4.164***	−4.232***	−10.206***	−8.144***
‘Hiraizumi’	−4.037***	−4.087***	−10.301***	−7.859***
‘Recovery/reconstruction’	−2.366	−2.466	−9.219***	−8.945***
‘Volunteer’	−2.843*	−2.733*	−6.528***	−6.511***

Source: Iwate Prefecture and Japan National Council of Social Welfare: monthly data from January to March 2013.

Note: Null hypothesis of existence of unit root is rejected when statistical significances are observed; PP test: Phillips-Peron test; ADF test: Augmented Dickey-Fuller test.

***, ** and * indicate 1%, 5% and 10% significance level, respectively.

Results of estimation models

The results of the model estimations are shown in Table 5 for ordinary tourists and Table 6 for volunteer tourists. Only models and parameters that had statistical significance are shown since variables with no statistical significance were omitted in the process of model estimation except for volunteer-related words to compare the results between the two types of tourism. Before the interpretation of each parameter, first we looked at model fitness for multicollinearity and heteroscedasticity and found no problem except for the TV-4 model, which had heteroscedasticity. So, a robust estimation was applied to the TV-4 model.

Table 5.

Results of time-series estimation model on number of ordinary tourists.

Explained variable	$Δ T_{t}^{O}$ : First-stage difference of number of ordinary tourists (seasonally adjusted) at period t
Variable/model	TO-1	TO-2	TO-3	TO-4	TO-5
$Δ T_{t - 1}^{O}$	−0.2811*	−0.5271**	−0.3308*	−0.3530**	−0.4299**
Monthly dummy variable (March)	761,522.8000***	568,950.2000***	538,459.6000***	584,134.1000***	499,095.6000***
Disaster dummy variable (March 2011 = 1, others = 0)	−1,369,094.0000***	–	–	–	–
Δ‘Cultural heritage’	–	–	–	–	7316.1310*
Δ‘Tourism resources’	–	–	5793.8200**	5898.6140**	–
Δ‘Festival’	–	10,674.6100	–	–	–
Δ‘Cherry blossom’	10,756.7200*	17,255.8600*	–	–	13,118.8000
Δ‘Volunteer’	−5399.3930	−9425.7010	−7034.7560	–	–
Δ‘Recovery/reconstruction’	–	–	–	−8505.6370*	−10,061.2400*
Constant	−31,461.4000	−35,956.4200	−48,393.9000	−53,493.0800	−40,632.3700
Mean vif	1.60	1.58	1.17	1.16	1.23
Breusch–Pagan test	ns	ns	ns	ns	ns
Durbin alternative test	ns	ns	ns	ns	ns
Sample size	31	31	31	31	31
Adjusted R²	0.7154	0.2470	0.2647	0.3313	0.2762

Source: Iwate Prefecture and Japan National Council of Social Welfare: monthly data from January to March 2013.

***, **, * and ns indicate 1%, 5%, 10% significance and no significance, respectively.

Table 6.

Results of time-series estimation model on number of volunteer tourists.

Explained variable	$Δ T_{t}^{V}$ : First-stage difference in number of volunteers at period t
Variable/model	TV-1	TV-2	TV-3	TV-4	TV-5	TV-6	TV-7
$Δ T_{t - 1}^{v}$	0.5571***	0.5708***	0.5368***	0.6884***	0.6800***	0.7175***	0.6457***
Monthly dummy variable (September)	−10,573.2200***	−11,510.1000***	−10,726.8300***	−10,263.7400***	−7251.5460***	−9481.6780***	−9341.7350***
Monthly dummy variable (December)	−5287.2930**	−5942.4120**	−5414.2550**	−5428.2340***	–	–	–
Δ‘Hiraizumi’	−323.3073**	–	–	–	−707.6674***	–	–
Δ‘Tourism resource’	−132.1368***	−127.3507***	−176.6317***	–	–	–	−163.0945***
Δ‘Heritage’	–	−142.7783*	–	−338.3804***	–	−318.2267***	–
Δ‘Volunteer’	287.9646**	274.2744**	256.2417**	–	–	–	–
Δ‘Recovery/reconstruction’	–	–	–	239.3012***	255.9324***	245.4158***	149.6569**
Constant	1333.5490*	1437.8600*	1299.1690*	1251.7370	530.4837	625.8244	576.0989
Mean vif	1.27	1.43	1.11	1.12	1.07	1.12	1.06
Breusch–Pagan test	ns	ns	ns	–	ns	ns	ns
Durbin alternative test	ns	ns	ns	–	ns	ns	ns
Sample size	31	31	31	31	31	31	31
Adjusted R²	0.7408	0.7277	0.7013	0.7625	0.6759	0.6618	0.6578

Source: Iwate Prefecture and Japan National Council of Social Welfare: monthly data from January to March 2013.

Note: Robust estimation method was employed for the model TV-4, since the original model has heteroscedasticity.

***, **, * and ns indicate 1%, 5%, 10% significance and no significance, respectively.

Now, let us look into the parameters of ordinary tourism in Table 5. First, among the tweeted words related to tourism that had a positive sign with significance were ‘tourism resources’ with 5% significance and ‘cherry blossom’ with 10% significance. These results mean that these words represent both the cultural and natural heritage and induce the effects on ordinary tourists that are generally expected. On the other hand, with regard to volunteer-related tweeted words, that is, ‘volunteer’ had no significance at all and ‘recovery/reconstruction’ was negative. These results showed that the word ‘volunteer’ was neutral and ‘recovery/reconstruction’ worked negatively when related to ordinary tourism. Second, the parameter of the first lag of ordinary tourists was significant with a negative sign, which shows no effect at all on repeat visits. Third, among the monthly dummy variables, that of March had positive signs while that of the temporary disaster dummy variable had negative signs (1% significance). Although the data were seasonally adjusted, the results showed that seasonal factors still worked. Thus, although it is ideal to test seasonal stationarity, we were not able to conduct this test well. In this respect, the results should be carefully interpreted by taking into account the seasonal stationarity issue.

Now turning to the quite contrasting results in Table 6. First, ‘tourism resources’, ‘heritage’ and ‘Hiraizumi’ were all negative with significance (1%, 5%, 10%), which indicates that these words worked in keeping away volunteers. In contrast, volunteer-related tweeted words, ‘volunteer’ and ‘recovery/reconstruction’, were all positive with significance (1%), which means that these words stably worked to attract volunteers. Second, the first lag parameter was positive with significance (1%, 5%), which means that repeat visits by volunteers were confirmed. This result is consistent with the actual behaviour of volunteers from what one of the authors experienced and observed as a volunteer in the disaster area. Third, the monthly dummy variables in September and December and the off-peak dummy variable beginning in September were all negative with significance (1%, 5%). These results are consistent with the fact that the number of volunteers plummeted beginning 6 months after the disaster.

To sum up, the results verified the effectiveness of tweeted information for the promotion of ordinary tourism and attraction of volunteer tourists and indicated that information that is actually effective is quite different depending on the types of tourism. The relationship between tweeted information and volunteer tourism has never been empirically verified before.

Conclusion

This article quantitatively clarified the effectiveness of official tweeted information for the early stage of recovery of tourism in the disaster-hit areas after the Great East Japan Earthquake. For this purpose, we employed the text-mining method and time-series regression models. The results revealed that conventional ordinary tourism was partially substituted by volunteer tourism in the tsunami-hit coastal areas. Although tweeted information was effective for both types of tourism in terms of immediacy, necessary information should properly be provided for those tourists with different orientations; information on the cultural and natural heritage positively worked for ordinary tourists. The disaster-related words worked positively for volunteer tourists but were negative for the ordinary types of tourists. In particular, official tweeting is effective and important in providing creditable information in the confusing time just after a disaster to inspire people to be volunteers to help those suffering from the disaster.

The results of this article also indicated that volunteer tourism for recovery work in disaster-hit areas has various impacts not only on the disaster-hit areas in physical terms, but also on people outside of the disaster-hit areas. The reason is that this type of volunteer tourism shows that people recognize the weakened solidarity among people in every modern society with the result that they decide to take action to reverse this trend even if temporarily. This article clarified that modern information technology can help people take action to connect with each other in the time of emergency. At the same time, we should also recognize that the interest in volunteer tourism for disaster-hit areas diminishes as time goes by. These findings have not been touched upon in previous literature and they add new knowledge on the overlapping area of dark tourism, social media and volunteer tourism related to natural disasters.

Consequently, it is important to design support measures that enable the local tourism sector to attract incoming tourists after a drop-in volunteer tourism as reconstruction of the disaster area progresses. In this context, the restoration process has a long way to go so that further study is necessary based on our results to explore how to effectively utilize Twitter for this purpose.

A limitation of this study is that profiles of volunteer tourists were not touched upon such as by providing information on gender, age, income and where they lived, nor was it examined how those people who behaved as ordinary tourists in normal times became volunteer tourists when the disaster happened. Another limitation is the technical issue of the seasonal unit root, which was not tested exactly in this article although seasonal adjustments were made. These are topics to be addressed in the future.

Footnotes

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 study was funded by J milk and Grants-in Aid for Scientific Research no. 16K14996 and no. 18H03965, Japan Society for the Promotion of Science.

ORCID iD

Yasuo Ohe

References

Albattat

Som

APM

(2014) Thailand’s Tourism Disasters: Natural and Human-Made Disasters. Saarbrücken: Lap Lambert Academic Publishing.

Borland

Adams

(2013) International Volunteer Tourism: Critical Reflections on Good Works in Central America. New York: Palgrave Macmillan.

Chan

Guillet

(2011) Investigation of social media marketing: how does the hotel industry in Hong Kong perform in marketing on social media websites? Journal of Travel & Tourism Marketing 28(4): 345–368.

Chen

(2011) The motivations and expectations of international volunteer tourists: a case study of “Chinese village traditions.” Tourism Management 32: 435–442.

Conran

(2011) They really love me! Intimacy in volunteer tourism. Annals of Tourism Research 38(4): 1454–1473.

Coren

Gray

(2012) Commodification of volunteer tourism: a comparative study of volunteer tourists in Vietnam and in Thailand. International Journal of Tourism Research 14: 222–234.

Crossley

(2012) Poor but happy: volunteer tourists’ encounters with poverty. Tourism Geographies 14(2): 235–253.

Dijkmans

Kerkhof

Beukeboom

(2015) A stage to engage: social media use and corporate reputation. Tourism Management 47: 58–67.

Fukui

Ohe

(2013) Evaluating the tourism recovery process in areas affected by seismic disasters: comparisons among urban, mountainous, and remote island areas (in Japanese). Journal of Rural Problems 49(1): 42–46.

10.

Hajli

Lin

(2014) Developing tourism education trough social media. Tourism Planning & Development 11(4): 405–414.

11.

Hays

Page

Buhalis

(2013) Social media as a destination marketing tool: its use by national tourism organizations. Current Issues in Tourism 16(3): 211–239.

12.

Hepburn

(2017) Everyday darkness and catastrophic events: riding Nepal’s buses through peace, war, and an earthquake. In: Hooper

Lennon

(eds), Dark Tourism: Practice and Interpretation. Abingdon: Routledge, pp. 147–159.

13.

Hvass

Munar

(2012) The takeoff of social media in tourism. Journal of Vacation Marketing 18(2): 93–103.

14.

Leung

Law

van Hoof

, et al. (2013) Social media in tourism and hospitality: a literature review. Journal of Travel & Tourism Marketing 30(1-2): 3–22.

15.

Leung

Bai

Stahura

(2015) The marketing effectiveness of social media in the hotel industry: a comparison of Facebook and Twitter. Journal of Hospitality & Tourism Research 39(2): 147–169.

16.

Mazzocchi

Montini

(2001) Earthquake effects on tourism in Central Italy. Annals of Tourism Research 28(4): 1031–1046.

17.

Minazzi

(2015) Social Media Marketing in Tourism and Hospitality. Heidelberg: Springer.

18.

Mostafanezhad

(2014a) Volunteer tourism and the popular humanitarian gaze. Gioforum 54: 111–118.

19.

Mostafanezhad

(2014b) Locating the tourist in volunteer tourism. Current Issues in Tourism 17(4): 381–384.

20.

Nakao

(2002) A study on “HANSHIN-AWAJI DAISINSAI” (big earthquake disaster) and the reconstruction of “Tourism of Kobe City” (in Japanese). Journal of Tourism Research 40: 108–116.

21.

Pilling

(2014) Bending Adversity: Japan and the Art of Survival. London: Penguin.

22.

Ritchie

(2009) Crisis and Disaster Management for Tourism. Bristol: Channel View Publications.

23.

Schroeder

Pennington-Gray

Donohoe

, et al. (2013) Using social media in times of crisis. Journal of Travel & Tourism Marketing 30(1-2): 126–143.

24.

Scott

Laws

Prideaux

(2010) Safety and Security in Tourism: Recovery Marketing After Crises. Abingdon: Routledge.

25.

Sharpley

Wright

(2018) Disasters and disaster tourism: the role of the media. In: Stone

Hartmann

Seaton

Sharpley

White

(eds), The Palgrave Handbook of Dark Tourism Studies. London: Macmillan, pp. 335–354.

26.

Skinner

(2018) “The smoke of an eruption and the dust of an earthquake”: dark tourism, the subline, and the re-animation of the disaster location. In: Stone

Hartmann

Seaton

Sharpley

White

(eds), The Palgrave Handbook of Dark Tourism Studies. London: Macmillan, pp. 125–150.

27.

The Japan Society for Interdisciplinary Tourism Studies (2013) Fukkou Tourism (Tourism for Restoration from the Disaster) (in Japanese). Tokyo: Dobunkan Shuppan.

28.

Volo

(2010) Communicating tourism crises through destination websites. In: Scott

Laws

Prideaux

(eds), Safety and Security in Tourism: Recovery Marketing After Crises. Abingdon: Routledge, pp. 83–93.

29.

Wearing

McGehee

(2013a) Volunteer tourism: a review. Tourism Management 38: 120–130.

30.

Wearing

McGehee

(2013b) International Volunteer Tourism: Integrating Travellers and Communities. Oxfordshire: CABI.

31.

Wood

Guerry

Silver

, et al. (2013) Using social media to quantify nature-based tourism and recreation. Scientific Reports 3(2976): 1–7.

32.

Wright

Sharpley

(2018a) Local community perceptions of disaster tourism: the case of L’Aquila, Italy. Current Issues in Tourism 21(14): 1569–1585.

33.

Wright

Sharpley

(2018b) The photograph: tourist responses to a visual interpretation of a disaster. Tourism Recreation Research 43(2): 161–174.

34.

Zehrer

Grabmüller

(2012) Social media marketing in tourism education: insights into the development and value of a social network site for a higher educational institution in tourism. Journal of Vacation Marketing 18(3): 221–228.