Secondary Data Analysis
From literature, Haywood (1986) indicated that stage identifications could be relied on the percentage of tourist arrivals and annual growth rate. Therefore, author used tourist numbers as the secondary indicator to cognize tourism development in the four study sites.
35
Firstly, data were collected from Tourism Bureau in Taiwan website to find out tourist numbers as for tourist arrivals. After visiting the database, researcher collected all data about visitors in the principal scenic spots from year 1986 to 2012 (Tourism Bureau, 2013). To run for TALC model, researcher identified the locations for every single spot and sorted them into the four study sites.
The second process is to understand the population shift so as to be another indicator for tourism impacts. In our research, researcher also observe population shift in the four study sites to understand the potential or existing impacts. The population data were from the Ministry of Household (http://www.ris.gov.tw/). All data were grouped by study sites chronically.
Delphi Panel
Delphi method is an organized expert panel for systematic, collaborative
estimating insights toward one specific issue. Norman Dalkey, Olaf Helmer proposed a method in 1963, and the method relied on a panel of experts in two rounds. All experts involved were encouraged to propose their own thoughts without face to face.
Mostly, Delphi method could be used in policy making, education and estimation (Rowe and Wright, 1999, 2000; Green, Armstrong, and Graefe, 2007).
To better understand tourism impacts, researcher formed up an expert panel for professional feedback. All candidates were selected into groups of the four study sites
36
based on their expertise. To construct Delphi Panel, the survey list was selected from National Digital Library of Theses and Dissertations in Taiwan. The participants for Delphi were professionals selected by research which must include at least one related study site. After selecting professionals, researchers acquired 35 people for Little Liuqiu, 58 people for Penghu, 11 people for Green Island and 5 for Orchid Island. The second step was to select randomly of the professionals for each study site. However, the sample for Orchid Island was not enough, researcher took general investigation instead of sampling. As a result, researcher invited 18 people for Little Liuqiu, 19 people for Penghu, 8 people for Green Island and 5 for Orchid Island. Yet, researcher only invited 8 people for Little Liuqiu, 19 people for Penghu, 8 people for Green Island and 5 for Orchid Island successfully. The last step was to generate an online survey system for Delphi panel respondents. The online survey system adopted was on Dosurvey.com (http://www.dosurvey.com.tw). At final, researcher distributed the invitation on 11th February 2014 and finished surveying on 31st March 2014.
In the aspect of tourism development, a given TALC graph led the subjects to answer in which the degree of tourism development was for each study site. The stages were divided into development, stagnation, decline and rejuvenation (Butler 1980; Haywood, 1986; Toh et al., 2001; Kim, Uysal, & Sirgy, 2012). Although theoretical TALC model encompassed 6 stages, researcher simplified into 4 so as to
37
eliminate vague answers. Besides, the Delphi survey conducted by reference of tourism impact studies as noted in previous section. About tourism impacts, all questions mainly contained positive and negative impacts with their details such as environmental, cultural, services and conditions.
However, in this study, there were two rounds survey. For one reason was that all researcher applied Delphi method as “pre-test.” Thus, Delphi panel was just another method to gather feedback in definite issues. Through doing Delphi, researcher could recollect all possible tourism impacts. The second reason was about research scale.
Researcher defined this research as a prior study for cognizing the development and tourism impacts in offshore islands. In addition, researchers involved would list up different tourism impacts happening in different development stages. Therefore, Delphi panel was another approach of investigation in this study.
Online Survey
This research adopted online survey of tourism impacts toward local people and tourists in terms of their perceptions. Through this survey, researcher could better understand the differences and similarities of potential or existing impacts among each group. Finally, all data collected would go under analysis discussed below.
Survey Method
For tourism impacts online survey, researcher collected public contact
38
information and sent various emails for invitations of this survey. The survey was posted on Dosurvey.com, and divided into 4 individual surveys of each study site. In addition, researcher adopted volunteer online survey system for residents and tourists in the four study sites to provide their perceptions toward tourism impacts. To
encourage more participants, researcher also cooperated with Pollster.com.
Pollster.com is an online survey company which runs for market research. The survey was distributed by Pollster.com to its members which would be rewarded by bonus.
At final, researcher also offer lottery for several gift certificate to thank for
participants. Additionally, with the help of online survey company, researcher gained over 1004 samples in total.
While Zikmud (2003) recommended mail survey could be most effective for collecting a larger amount samples by spreading the survey geographical limitations in a short time, it was not possible for the researcher to finish under difficult
circumstances. Instead, online survey system would be less expensive, more effective and time-saving (Davis, 1997; Dommeyer & Moriarty, 2000; Pitkow & Recker, 1995;
Tse, 1998; Witte, Amoroso, & Howard, 2000); also, online survey system could be less offensive for Respondents (Walsh, Kiesler, Proull, & Hesse, 1992).
However, online survey system must went under specific design for improved reliability by eliminating sampling errors. Most researchers argued that online survey
39
system did not meet accessibility to reality. There were common errors such as coverage error, sampling frame and volunteer sample (Li, 2003).