A country's image as tourist destination for external intermediaries

Abstract

Destination image is an important variable of influence for tourist behavior. Previous studies showed that intermediaries . different travel agencies - influence the visiting intentions of potential tourists, being able to form opinions and shape behaviors. Measuring the perception of travel agencies and other intermediaries about the tourist image of a destination country becomes really important. Reliable and valid measurement instruments are necessary. The vast majority of previous studies used instruments developed for individual tourists, and they do not offer details about how the measurement instruments were obtained. Our objective was to show how one can build an appropriate instrument for intermediaries; the main contribution of our study lays in the scrutiny and rigor of the instrument development process. We performed an instrumental research, on five stages: documentary research through literature review, exploratory research for general and specific attributes, initial questionnaire design based on the first two stages, questionnaire testing on experts and reliability checking on a pilot study. Both qualitative (content analysis and expert ranking) and quantitative (reliability tests) methods of analysis were used. The final instrument obtained is presented in the conclusions section and could be used to estimate external intermediaries f attitude towards a destination country. The main implications are methodological . the whole scale development procedure can be used in order to build and test similar research instruments. [PUBLICATION ABSTRACT]

Headnote

Abstract: Destination image is an important variable of influence for tourist behavior. Previous studies showed that intermediaries . different travel agencies - influence the visiting intentions of potential tourists, being able to form opinions and shape behaviors. Measuring the perception of travel agencies and other intermediaries about the tourist image of a destination country becomes really important. Reliable and valid measurement instruments are necessary. The vast majority of previous studies used instruments developed for individual tourists, and they do not offer details about how the measurement instruments were obtained. Our objective was to show how one can build an appropriate instrument for intermediaries; the main contribution of our study lays in the scrutiny and rigor of the instrument development process. We performed an instrumental research, on five stages: documentary research through literature review, exploratory research for general and specific attributes, initial questionnaire design based on the first two stages, questionnaire testing on experts and reliability checking on a pilot study. Both qualitative (content analysis and expert ranking) and quantitative (reliability tests) methods of analysis were used. The final instrument obtained is presented in the conclusions section and could be used to estimate external intermediaries f attitude towards a destination country. The main implications are methodological . the whole scale development procedure can be used in order to build and test similar research instruments.

Keywords: country image, tourist destination, scale validation, instrumental research, travel agencies.

Please cite this article as following: Vicol, O. and Zait, A. (2014), "A country's image as tourist destination for external intermediaries", Management & Marketing. Challenges for the Knowledge Society, Vol. 9, No. 1, pp. 47 ]74.

(ProQuest: ... denotes formulae omitted.)

Introduction

People's visiting decisions are significantly influenced by the tourist destination's image (Echtner and Ritchie, 1991; Chen and Hsu, 2000; Beerli and Martin, 2004; Tasci et al., 2007; Alcaniz et al., 2009). Previous studies in the field of destination image also showed that intermediaries influence the visiting intentions of potential tourists, being able to form opinions and shape behaviors (Gartner, 1993; Lawton and Page, 1997; Baloglu and Mangaloglu, 2001; Martin, 2011). In this light, measuring the perception of travel agencies and other intermediaries on the tourist image of a destination country becomes really important. In order to do this, a reliable and valid measurement instrument is necessary. The vast majority of previous studies used instruments developed for individual tourists and, moreover, they did not provide details about the scale development process. This is why our objective was to perform an instrumental research in order to build an appropriate instrument for intermediaries, the focus being on the methodological details - scrutiny and rigor of the instrument development process. Our instrumental research had five stages: documentary research through literature review, exploratory research, questionnaire design based on the two previous stages, questionnaire testing on experts and reliability checking on a pilot study (see Figure 1).

The literature review process allowed us to identify previous methods for defining and operationalizing the construct "destination" image, as well as different variables and scales used for quantifying the main components of attitude (cognitive, affective and conative). Relevant tourism attributes, both general and country specific, were obtained through an exploratory survey, followed by content analysis. Using the results from the documentary study and the exploratory one, we designed the questionnaire - the first draftof the research instrument. This questionnaire was tested on a panel of experts and we used the results for refining the research instrument. After this qualitative approach, meant to validate the instrument from a conceptual and content point of view, a statistical reliability check was performed, using Cronbach Alpha estimators, and the final research instrument was obtained.

The instrument is designed to measure Romania's image as a destination country, from the perspective of external tourism agencies as intermediaries. The questionnaire was designed in English and could be used either in English or in the native language of the sending country, after a cultural translation process (adaptation of words and back translation checking procedure). The German, French and Italian forms of the questionnaire are available from the authors, on request. Although the instrument is context-specific (Romania as a tourist destination), the procedure through which the instrument was built and tested is generally valid and recommended; the main focus of our study was the methodological process through which the instrument was obtained, through the combined qualitative and quantitative approaches.

Literature review - methods, variables and scales

Several definitions of destination image are used in the literature. From a pragmatic point of view, the destination image represents a fundamental unit on which all the complex dimensions of tourism rely, a central point in developing and providing tourism products (UNWTO, 2004). From a conceptual perspective, most studies use one of the first definitions suggested by Crompton (1979) and Gartner (1989). Crompton (1979) defines destination image as the sum of beliefs, ideas and impressions that an individual holds about a specific destination. Gartner (1989) considers that individuals do not have the cognitive capacity to evaluate objects and attributes, and so they tend to simplify the process and use specific criteria, as a function of their particular situation. We based our study on the definition offered by Crompton and tried to offer on operational (measurable) perspective of the construct, as well.

One of the most used scales for the operationalization of the construct destination image is that suggested by Echtner and Ritchie (1991). They apply both quantitative and qualitative instruments in order to evaluate destination image: a quantitative approach for the identification of common characteristics and attributes and a qualitative approach for determining holistic and psychological impressions associated with destination image. Their scale was also used by other researchers (Chen and Kerstetter, 1999; Court and Lupton, 1997; Joppe et al., 2001; MacKay and Fesenmaier, 1997; Milman and Pizam, 1995; O'Leary and Deegan, 2005; Tapachai and Waryszak, 2000). The main advantage of using both quantitative and qualitative approaches lies in the possibility of obtaining a complete evaluation of the destination image, through the analysis of the cognitive and especially the affective component of attitude, much more personal and subjective. Both quantitative and qualitative methods were previously used by Dann (1996), Echtner and Ritchie (1993), Lubbe (1998), MacKay and Fesenmaier (1997), O'Leary and Deegan (2005), Tapachai and Waryszak (2000), and Walmsley and Young (1998). Beerli and Martin (2004) extracted a list of nine dimensions, containing all the attributes identified in previous scales used for measuring tourist destination image.

The vast majority of previous studies used quantitative approaches, based on structured instruments containing Likert and semantic differential scales. These instruments have normal methodological limitations, because respondents need to evaluate certain characteristics of the analyzed destination which do not correspond to their perceived image (Tapachai and Waryszak, 2000). Qualitative studies are able to focus on the complexity and internal relationships of real life tourist experiences (Ryan and Cave, 2005).

Qualitative research in tourism is quite often associated with storytelling, and previous studies suggest using open questions in order to generate narrative stories (Govers and Go, 2001; Ryan and Cave, 2005). Using a combined approach (quantitative and qualitative) the researcher could balance limits and disadvantages, obtaining a better internal validity through the qualitative approach and a better reliability and external validity through the quantitative approach (Clark and Dawson, 1999; Burns, 2000).

Qualitative methods, although not so present, are more and more recognized as a necessary complement for researches interested in revealing the psychological characteristics and holistic impressions of destination image (Dann, 1996; Echtner and Ritchie, 1993; Mackay and Fesenmaier, 2000; Reilly, 1990). Govers and Kumar (2007) used a qualitative method in order to identify the most appropriate categories of attributes and their components, which they further incorporated into a quantitative research. Their narrative technique, together with the content analysis of the answers, was also used by Martin and Bosque (2008). The study performed by Baloglu and Mangaloglu (2001) on tourism intermediaries suggests that tourist destination image should be measured using both quantitative and qualitative methods, if we want that image to be "real". The methodology in the field is still developing, and comparability among studies is still desiderating. This is why we decided to apply both methods in order to obtain, refine and test our research instrument. In the next sections we present the main results from each stage of our instrumental research. We shortly present the research method with the corresponding justification, then the results in terms of instrument (questionnaire) structure and details (items).

Exploratory research - general and specific attributes

Most of the previous studies (Baloglu and McCleary, 1999a; Beerli and Martin, 2004; Hosany et al., 2006; Lee et al., 2005; Martin and Bosque, 2008; Phillips and Jang, 2008) have evaluated only the general attributes of a destination (beaches, mountains, culture, monasteries etc.), specific attributes (for example painted monasteries from Bucovina) having been considered in just a few studies (Harahsheh, 2009). Results obtained through the analysis of specific attributes could offer important information for the institutions in charge with promoting a country, in general, and Romania, in particular, as a tourist destination.

The first stage of the exploratory research was to identify all the elements, both common and unique, used to promote Romania in the online environment. The approach consisted in the analysis of two types of websites - those of Romanian intermediaries offering packages for foreign tourists and those officially used to promote Romania as tourist destination at national level and in three EU Countries - France, Italy and Germany. In order to identify the most relevant attributes we used an exploratory online survey, on a sample of experts in tourism (from tourism agencies). The questionnaire had three sections, one with open questions, one with multiple choice questions and one for the demographic profile of the respondents. We compared answers to the open questions to those from the multiple choice questions (which used items from previous studies).

The list of websites included in the analysis was obtained by accessing the results from the first page of Google, after the interrogation "travel agency Romania". These links are relevant, considering the fact that the largest number of link clicks is obtained when those links are among the first 10 results (first page) - 94% from the total users (Granka et al., 2004). From the official website of Romania as tourist destination (www.romaniatourism.com) we obtained addresses of similar websites promoting Romania from Germany, Italy and France. Using results from a previous study that realized a content analysis of Romania's image (Vicol and Spiridon, 2010), starting with the interrogation "Romania tourism", we generated a list of 3187 expressions. From this list we selected 115 expressions based on the existence of the key words "travel", "tour", "tourism", "tourists", "tourist attraction", "travel agency", "travel guide", "tourist guide". This short list of 115 key words was used to generate lists of websites - a number of 2186 addresses. From these we selected 9 websites, those that obtained more than 1% from the results offered by Google in the first two pages. From these, 3 cases were eliminated because they did not have an English option, thus resulting the list of 6 analyzed websites (Table 1).

We analyzed all tourism packages and destinations, obtaining two lists of attributes - general and specific ones. Based on these two lists we created the multiple choice items for our exploratory survey questionnaire (Table 2 and Table 3).

Sample and design for the exploratory research

We downloaded a list of approximately 3000 travel intermediaries from the website of the Ministry of Regional Development and Tourism (all licensed companies, active or not) and confronted it with the list of active members of the National Association of Travel Agencies from Romania (ANAT). This is how we obtained a list of 797 intermediaries with the corresponding e ]mail addresses, and this was our sample. We used SurveyGizmo.com in 4 stages, with one initial e ]mail (20.12.2011) and 3 follow ]up e ]mails (12.01.2012, 16.01.2012, 20.01.2012). From 797 sent e ]mails, 141 turned out to be incorrect addresses, which gave us just 656 e ]mails really sent. The total response rate was 17.84% (117 responses, both complete and partial), and the rate of complete responses was of 9.15% (60 responses).

We used two forms of our exploratory questionnaire, one for intermediaries offering tourist packages for Romania (questionnaire A, 12 questions) and one for intermediaries not offering such packages (questionnaire B, 9 questions). The two groups were filtered through an initial question and then the Survey Gizmo program separated their answering path. For group A we had three open questions, concerning the main Romanian attractions promoted, the most visited attractions from their offer and the attractions perceived as less promoted for foreign tourists. For group B, only perceptions about the most promoted attractions were registered. As multiple choice questions we used items/attributes obtained from the analysis of travel websites. The aim was to elicit all potential general and specific attributes for Romania.

Qualitative analysis of data collected through the exploratory research

We used content analysis for interpreting the collected qualitative data (answers to the open questions). For the data coding we used the emergent coding method (Haney et al., 1998). We first examined all the responses, individually (two different investigators), and obtained two separate lists of codes. Then, we obtained a common list, through a process of comparison and unification of previous lists. As a final step, we coded (using the common list of codes) all answers (texts). Responses were classified into general and specific items. The number of entries of every code and the context in which the code appeared was then checked again, in order to validate inferences, and we also checked the inter ]rater coding reliability (Neuman, 2003). We used WordStat 6.1.4 (trial) and obtained a reliability level of 98%. Based on the classification realized by Beerli and Martin (2004), the items we obtained were grouped in general attributes and specific attributes. Group B was eliminated from the analysis because of the small number of respondents (12).

General attributes selection process

A comparative analysis on responses of three open questions (1, 2 and 5 from questionnaire A) was made. From the responses to the fifth question, the following attributes were selected: Historical buildings/Monuments, Mountains, Museums, Culture, Religion, Traditional cuisine, Handmade Art, Folklore, and Architecture. The selection was made based on the fact that this first 9 attributes cumulate 55% of total items, while the rest of the items are included in the last 22 attributes.

From the first and second questions (open ended questions, questionnaire A) we selected the first 6 of each attributes that cumulate 82% from the total items of first question, and 90% from total items of the second question. The 6 attributes obtained out of both questions are the same, meaning that we have a list of 6 attributes from the open questions that is compared with the list of 9 attributes from the closed question. Attributes Cities and Regions were eliminated from the final list because of their lack of specificity, and attributes Religion and Mountains were found in both final lists. The resulting list of attributes is presented in Table 4.

Because of their similarity, the possibility of overlapping meanings for two of the attributes - Castles and Fortresses and Historic buildings/Monuments - was verified in SPSS:

* we identified all respondents that selected Bran Castle and Peles and Pelisor Castle on question 6 and

* the respondents that listed items included in Castles and Fortresses attribute (from open ended question).

* Using cross tabulation analyses we found that:

* 80% selected the attribute Historic buildings/Monuments from the 48 respondents of group A (question 6)

* 100% respondents that selected Historic buildings/Monuments selected also items with Bran Castle code (1 and 2 open ended questions) - specific expression.

* 100% respondents that selected Historic buildings/Monuments selected also items with Peles Castle code (1 and 2 open ended questions) specific expression.

* all 3 items from the general expression code "Castles" were given by respondents which selected Historic buildings/Monuments (open ended question 1)

* all items included in general code "Castles" are given by respondents which selected Historic buildings/Monuments attribute. - general expression (question 1)

* 100% of items from code Fortresses and fortified churches in Transylvania were mentioned by respondents which selected the attribute Historic buildings/Monuments.

Based on this analysis we can say that the 3 specific attributes: Bran Castle, Peles Castle and Fortresses and fortified churches in Transylvania are associated with the general attribute Historic buildings/Monuments, which was recoded in Historic buildings/Castles/Fortresses.

The final list of general attributes selected for the analysis is presented in Table 5.

The selection process of specific attributes

To select specific attributes we used a graphic form of the responses, having as criterion the percentage of occurrence for each attribute in the total number of responses. As a result we obtained three comparable lists of attributes. The analysis included the results of open ended questions 1 and 2 and the closed question 6. The final list of attributes included those that have appeared in all three lists or in the answer options for question six, and in at least one of the lists resulting from open ended questions. The attributes obtained are shown in Table 6. Attributes that refer to historical cities and regions of the country were excluded from the analysis, as in the case of general attributes.

The Medieval Cities attribute was introduced in the encoding items phase, in the general expressions category (code medieval towns and cities), because it includes a number of cities (Bistrita, Brasov, Cluj-Napoca, Medias, Sebes, Sibiu and Sighisoara) promoted under this name (Medieval Cities) on the website www.romaniatourism.com/medieval-towns.html and it is considered a general attribute. This information is relevant because the website appears with a frequency of 98%, following the interrogation of the search engine using 115 keywords realized in the Vicol and Spiridon (2010) study. This classification is very common and, although the attribute is not in the list of specific attributes from the open ended questions (due to the encoding method), because it appears in items of both open ended questions 7 and 3 times, we decided to include the attribute Medieval Cities in the final list of specific attributes.

The Carpathian Mountains attribute is on the 7th place in rank classification and the 4th place in the evaluation based on the percentage of responses of the 6th question. This attribute appears just one time in open ended question 1 and 2 times in the second open ended question. On the other hand, most references were made to the general attribute Mountains, which is included in the final list of general attributes; therefore this attribute was not included in the final list of specific attributes.

The UNESCO Heritage Areas attribute is ranked 6th in both classifications made for question 6, but it was not mentioned in the answers to any of the open ended questions. To check the frequency of occurrence on the websites that promote Romania as a tourist destination, we used the database developed by Vicol and Spiridon (2010). The database contains 548 Word documents that have been processed with the software WordStat. The 3 words phrases were generated and we measured their frequencies. The phrases containing the word "UNESCO" (WORLD HERITAGE UNESCO) was identified in 6.8% of documents with a frequency of occurrence of 62, in 37 cases out of 178 three words phrases identified in total. In comparison with the expressions that are on the first places like: "Hotel in Bucharest" (frequency 706), "Castles and Fortresses" (frequency 487) and "Black Sea Coast" (frequency 355) the attribute that refers to UNESCO's occurs too rarely in order to be included in specific attributes final list.

The attribute House of Parliament/Ceausescu is placed 8th in the ranks top and 7th in the top based on the percentages of answers. Two items were identified in open ended questions, one in each of the two questions. Also, the Palace of Parliament expression was found in 6.8% of documents, with the frequency of occurrence of 69, in 37 cases out of 178 three words phrases. We decided to not include it in the final list of specific attributes.

The attribute Vineyards and Traditional Cuisine is placed 8th (by percent of responses) and 10 (ranks top) and it was not identified as an item in any of the answers of open ended questions. Based on that, we decided not to include it in the final list of specific attributes. The Constantin Brancusi's Monuments attribute ranks 14th (ranks top) and 10 (by percent of responses) and it was not identified in any of the answers of open ended questions, so it was decided not to include it in the final list of specific attributes.

Attributes: Archaeological Sites, Transylvania, Mocanita, George Enescu Festival, Ovidiu's Island and St Andrew's Monastery and Mud Volcanoes are the last placed in responses of closed ended question and they were not mentioned in the answers of open ended questions. As a result, they were excluded from the final list of specific attributes. Although the attribute Archaeological Sites Transylvania was excluded from the final list, in the answers of open ended questions there is another attribute that refers to the region of Transylvania - Cities and Fortified Churches from Transylvania. It occupies 7th place in both classifications of results of first and second open ended questions, and we decided to include it in the list of specific attributes. The last attributes were excluded from the final list: Wooden Churches of Maramures and Moldova's Monasteries.

In conclusion, the final list of attributes of Romania as a tourist destination, obtained from exploratory research includes: 27 general attributes (10 general attributes added to the 17 selected in the process of review of literature) and 8 specific attributes.

The design of the measuring instrument for the final research

Based on the specific and general list of attributes obtained from the exploratory research and after the literature review process, we built the measurement instrument for the tourist destination image. The instrument was developed in English in order to facilitate the translation process into other languages, depending on the tourists' source country (we used German, Italian and French). The instrument was divided into 4 sections (A to E). Section A contains two questions: the first aims to identify visitors and non-visitors and the second question aims to identify the type of visit that respondents had. The open question from Section B was used to identify the first attribute/words coming to the respondents' mind when asked about Romania as a tourist destination. This was adapted from Baloglu and Mangaloglu (2001). In their study, respondents listed the first 3 adjectives or nouns that came to their mind when thinking of a certain tourist destination. Section C is divided into three subsections: Section C1a operationalizes cognitive image, Section C1b assesses respondents' familiarity with specific attributes, Section C2 includes items measuring affective and conative image and section C3 operationalizes the conative component of image "intention to recommend". Section D includes 4 questions that aimed to identify: the main competitors of Romania as a tourist destination, familiarity and attractiveness of the destination and the overall image. The last part of the questionnaire (Section E) includes questions meant to identify the demographic profile of respondents and the type of company in which they work.

Cognitive component of tourist destination image (TDI)

The cognitive component of TDI is measured by the general attributes resulting from the exploratory research and literature review and is operationalized in Section C1a of the questionnaire. In Section C1b we evaluated the intermediaries f familiarity with the specific attributes. We used a total of 27 items to measure the cognitive component and 8 items to measure familiarity with specific attractions. The items selected to measure the cognitive dimension were classified into five categories: infrastructure, socio ]political environment, natural environment, general attractions (culture, history and art), and price ]quality ratio. A number of 3 to 14 items were assigned to each of the five categories, based on previous studies (Obenour, 2005; Calatone et al., 1989; Chen and Hsu, 2000; Alcaniz et al., 2009; Baloglu and Mangaloglu, 2001; Pike and Ryan, 2004; Baloglu and McCleary, 1999a; Pike, 2003; Beerli and Martin, 2004; Lee et al., 2005; Byon and Zhang, 2010; Hui and Wan, 2003; Echtner and Ritchie, 1991). In Table 7 we present the 5 categories with the assigned items.

We used semantic differential scales with values from 1 to 7, where 1 - "offers very little" and 7 - "offers a lot" to measure the 27 items of Section C1. In addition, to avoid errors in measuring non-responses, 2 other options were added: I don't know and I do not want to answer (Aish and Joereskog, 1990). The category Specific attractions / landmarks include items used to measure familiarity, in C1b section of the questionnaire, identified by exploratory research. The eight items are shown in Table 8 and are operationalized using semantic-differential scales with values from 1 to 7, where 1 - "not at all familiar" and 7 - "very familiar" and the 2 non response options.

Affective component of TDI

We used the bipolar semantic differential scale to measure the affective component of image defined in previous studies (Baloglu and Brinberg, 1997; Baloglu and Mangaloglu, 2001; Pike and Ryan, 2004; Martin and Bosque, 2007; Lee et al., 2005; Hosany et al., 2006) as follows: Unpleasant . Pleasant, Sleepy ] Arousing, Distressing . Relaxing, Gloomy ] Exciting. The non ]response options were also available.

Intention to recommend ] conative component of TDI

Previous studies used a range of 2 to 4 items to identify behavioral intentions (Byon and Zhang, 2010; Castro et al., 2007; Chen and Tsai, 2007; Lee, 2009; Chi and Qu, 2008; Alcaniz et al., 2009). In their study, Byon and Zhang (2010) used for this purpose 4 items: intent recommendation to family and friends, intention to (re)visit, saying positive things to other people and willingness to participate in sports competitions. Due to the nature of research, the last item was adapted to our study and transformed into: intention to recommend Romanian travel packages to customers. We used a 1 ]7 semantic differential scale, with 1 . Definitely not and 7 . Definitely yes. The two options of non ]responses were also included. The last non response was transformed into .Not sell".

In section D of the questionnaire we included four questions: an open ended question that aimed to identify the main competing countries in terms of tourism; a question for measuring the overall image of Romania by using a differential scale with values from 1 to 7 (very negative - very positive) (Baloglu and McCleary, 1999a; Beerli and Martin, 2004) and two non-response options. Familiarity and attractiveness (Sonmez and Sirakaya, 2002) were operationalized through questions D3 and D4 that used the same differential scale with 7 steps. The last part of the questionnaire aims to identify the respondents' characteristics and the company that they represent: age, sex, name of the company, company's number of employees, time spent in the company (number of years) and their position (president/ general manager, operations manager, consultant, marketing or other), type of agency (tour operator, travel agent or both), depending on the destination company type sold (incoming, outgoing, both).

The last question segmented the sample into two groups: intermediaries who sell packages to Romania and intermediaries who don't offer such packages. This way, a comparison was possible between the two sub-samples.

Questionnaire testing on experts - content validity

To test the content validity of the instrument that measures TDI, we used a sample of four travel experts: three professors from the University of Al.I. Cuza and one expert in tourism (responsible for promoting Romania as a touristic destination). Content validity was performed on each item and on the overall scale. Relevance and clarity were tested and measured on a 4-point Likert scale (Lynn, 1986; Waltz et al., 1991) presented in Table 9.

Literature provides four kappa indicators for statistical analysis of nominal agreement between more than two experts (Warrens, 2010). Light and Hubert's indicators are multi-expert version of Cohen's indicator for two experts and Fleiss's indicator is an extension of Scott's pi indicator for more than two experts. Finally, S indicator proposed by Randolph is a generalization of Bennett et al. (1954)'s indicator for more than two experts. Assuming that all the experts assigned a minimum proportion of responses for each category, Warrens (2010) compares the four indicators and demonstrates the mathematical superiority of the Randolph S indicator. Specifically, Warrens (2010) shows that Fleiss's indicator is lower than Hubert and Randolph's indicators, and that Randolph's indicator is superior to Hubert and Light's indicators.

The four indicators presented above are classified into two categories: kappa indicators with fixed marginal values and kappa indicators with free marginal values. The difference between them is given by the way items are assigned to categories, in our case, grades assigning. Indicators with free marginal values are used in cases where experts must not assign a fixed number of items/cases to a category. Indicators with fixed marginal values are used in opposite cases.

In our case the agreement between experts was measured using the kappa indicators with free marginal values category. Bennett et al. (1954) proposed S indicator, which adjusts the level of agreement between the two experts to fit it with the agreement expected by chance. S indicator is making the adjustment based on the number of categories included in the analysis. Bennett indicator calculation formula is:

... (1) where:

q = number of categories;

Pa = proportion of agreement between the experts calculated by the formula:

... (2)

n = number of items/cases;

nij = number of responses in each cell of the table resulting from the answers (for item i and category j), and

r = number of experts.

The premise on which Bennett indicator is built is that the probability of agreement by chance is ... when Pa is equal with ... S = 0 which indicates that the observed agreement is greater than chance agreement. Indicator S can take values between .1 and 1. Values below 0 show that observed agreement is smaller than the agreement expected by chance, and values greater than 0 mean that observed agreement is higher than the one expected just by chance.

S indicator depends only on constant values of ... and ... so, the value of S is not influenced by variation in the distribution of responses between groups or experts.

To assess the agreement between more than two experts, Randolph (2005) provides a formula of calculation for S, called multi-expert kappa indicator. The values of this indicator do not vary depending on the symmetry of the marginal distributions, they vary depending on the number of categories included in the analysis; it is used when there are no restrictions on the number of items distributed in each category.

There is no general agreement on the interpretation of S indicator's values. Some lines were drawn by (Landis and Koch, 1977) which mention that:

* values between -1.0 and 0.0 indicate total disagreement among experts;

* values between 0.01 and 0.2 indicate a very low degree of agreement;

* values between 0.2 and 0.4 indicate a fair agreement, values between 0.41 and 0.6 indicate a moderate agreement;

* values between 0.61 and 0.8 indicate a substantial agreement;

* and values between 0.81 and 1.00 indicate a nearly perfect agreement.

This interpretation is not generally regarded as valid because a small number of experts may generate a higher value of the indicator. Randolph (2008) provides an online automatic calculation of his indicator S. This calculation was used to analyze the relevance and clarity of our questionnaire items. For this process all responses were classified by the 4 categories (Clarity: unclear, needs revision, clear with slight modifications, very clear and Relevant: irrelevant, needs review with minor modifications relevant and very relevant). All four categories were included in the calculation module and we got the following results: agreement indicator on the relevance of items obtained a value of 0.599209, approximated to 0.60 and agreement indicator on the clarity of items obtained a value of 0.698416, approximated to 0.70. According to the literature it can be said that the value obtained for relevance of items indicates a moderate degree of agreement among experts, and the value obtained for clarity indicates a substantial degree of agreement.

Content validity of scale and items - content validity index

The next step in verifying the content validity of our instrument was getting the value of the content validity index (CVI) (Lynn, 1986). This value measures the "degree to which an instrument contains a set of items suitable for measuring the construct" (Polit and Beck, 2004). From the methodological point of view, researchers (Mastaglia et al., 2003) consider that the process of content validation of an instrument is based on judgment and is composed of two stages: scale development through analysis of previous studies and relevance estimation of its content through assessments made by experts.

Lynn (1986) uses two types of indicators to measure the content validity of a scale: individual item validation indicator (I-CVI) and overall scale validation indicator (S-CVI). Also, it is recommended that 3-10 experts take part to this process and it should be used a Likert scale with 4 values (to avoid neutral responses) (Lynn, 1986). After the data is collected from all the experts, the 4 values scale is transformed into a dichotomous scale, in our case relevant - irrelevant, clear - unclear (the values 3 and 4 are converted to 1, which means that the item is considered relevant or clear, and the values 1 and 2 are transformed to 0 which means that the item evaluated is irrelevant or unclear).

Indicator I-CVI is used generally in the review, for reformulation and/or deleting items. If the evaluation is made by more than 5 experts, it is recommended to use the value of I-CVI as 1 for every scale item.

There are three methods of calculating the value of S - CVI indicator. The first method involves calculating the average proportions of relevant items obtained from each expert. The result is denoted by S-CVI/Med. A second method of calculation involves gathering all the indicators I- CVI and dividing the sum by the total number of items (S - CVI/I). The third method involves gathering all relevant responses and dividing by the total number of responses (S- CVI/R). All three methods provide the same values of S - CVI indicator. Besides these three methods researchers use the method of universal agreement (S-CVI/UA), considered the strictest method of assessing content validity of an instrument. This means that a sum of all relevant items is made (all items that have the value of 1of CVI-I) and this is divided by the total number of items. In most cases the amount of S-CVI/UA is less than the S-IVC, and therefore are considered acceptable values greater than 0.70 (Waltz et al., 2005). S-CVI indicator is considered acceptable for values greater than 0.80 (Polit and Beck, 2004), but Waltz et al. (2005) recommended the minimum accepted value to be 0.90.

For our instrument, the following indicators were calculated: indicators for overall scale by 4 methods, both for clarity and relevance. The results are shown in Table 10, Table 11, Table 12 and Table 13. Some authors offer information on the acceptance level of the indicator, as follows: a 70% indicates a necessary agreement between the experts, 80% indicates a proper agreement, and 90% can be translated into a good agreement between experts (Hartmann, 1977; House et al., 1981). As shown in Table 10, we obtained CVI index values, calculated by the two methods, of 95% and 84%. Index value of 0.95 S-CVI / Med indicates a good agreement between the experts in terms of relevance and a value of 0.84 S-CVI/AU-indicates a proper agreement between experts in terms of relevance for the assessed scale.

The obtained values after measuring the scale clarity are presented in Table 11. The scale index for clarity was calculated, as in the case of relevance, by both mathematical methods and the results are: S ] CVI/med = 0.97 (97 %) and CVI / AU = 0.89 (89 %).

The value obtained for S - CVI/Med - 0.97 indicates good agreement between the experts in terms of scale clarity the S - CVI/AU - 0.89 indicates adequate agreement among experts in terms of the clarity scale.

According to the methodology proposed by Lynn (1986), when using panels with less than five experts, content validity index for each item must have a value of 1, or obtain perfect agreement. In this case, from all 56 evaluated items we obtained index value of 1 only for 47 for relevance and 50 for clarity. The items with index value of 0 were analyzed separately.

Same items that obtained values less than 1 for relevance obtained less than 1 also for clarity.

Items from Natural Environment category, Congestion and Noise and Environment pollution, with I ]CVI = 0.50 were removed from the instrument: Thus, three items remained in this category with the I ]CVI = 1 of the 5 items originally proposed.

Following the analysis of comments provided by experts, we decided to:

* Remove the following items: personal safety, language barriers, political stability and intention to (re)visit, and

* Review the following items: price level, climate and tourist information.

From the 56 items included in the validation phase, six items were eliminated and three were reviewed (Table 14).

Reliability testing - pilot study

The objective of the pilot study was to estimate the instrument's reliability by checking the internal consistency. Researchers have different views regarding the minimum number of responses required for a pilot study. For example, Hertzog (2008) considers that a sample between 25-40 respondents is sufficient, Johanson and Brooks (2009) argue that a minimum of 30 responses is necessary, and Hill (1998) states that 10 to 30 respondents are sufficient for testing of an instrument.

The pilot study was conducted during 7-13.03.2012 on a sample consisting of Romanian tourism intermediaries who participated in the previous exploratory investigation. The experts' validated instrument was introduced in Survey Gizmo online platform. The generated link with the questionnaire was sent by mail to all respondents who provided complete responses in the previous exploratory investigation. We obtained a 24.6% response rate, 15 complete responses out of 61 sent emails.

The process of estimating instrument reliability by assessing the internal consistency of responses is used for reflective constructs. Internal consistency refers to the correlation between the items, on a test sample, while homogeneity refers to the unidimensionality of the scale. A scale is unidimensional if the items measure a single latent construct. Studies have shown that Cronbach's Alpha estimator is influenced by the number of items used in the scale. If the number of items increases, the value of the estimator will also increase, so it is recommended to report Cronbach Alpha for each instrument construct (Tavakoli and Dennick, 2011).

Cronbach Alpha can range between 0.00 - 1.00 and its value varies according to: the number of items of the scale and inter-item links. The literature offers various ranges of acceptance to its interpretation. For example, Hair et al. (2006) considered the minimum level is obtained for Alpha = 0.70, while Malhotra (2010) suggests a minimum value of 0.60. Most researchers use the range 0.70 - 0.95 according to the interpretation of the estimator, and a value greater than 0.95 is considered redundant and in this case it is recommended to reduce the scale. In Table 15, the Cronbach Alpha values for each construct/image dimension (affective, cognitive and conative image), and for the Specific Attractions category are presented.

The subscales that measure categories: Socio-political environment and natural environment with 2 and 3 items, obtained the Alpha coefficient values lower than 0.6 (0.53 and 0.54). Based on literature (Lee et al., 2005) it was decided to combine the two categories into a single scale titled Socio-political and natural environment. The Alpha coefficient for this new construct is 0.757 and we decided to keep the items in the instrument (see Table 16).

The other three categories that define the cognitive dimension of the scale TDI, Infrastructure, Price- Quality Ratio, General Attractions obtained Alpha's values between 0.74 and 0.935. Cronbach Alpha value for the cognitive dimension is 0.944. These values denote a very good internal consistency of the scale. The constructs for affective and conative dimensions obtained high values for Alpha, 0.928 and 0.948, but not more than 0.95 which would have created suspicion of redundancy of items. The scale for specific attractions obtained a good value Cronbach Alpha (0.87). After analyzing the results of the pilot study the instrument has not suffered any changes. The instrument was developed in English, in order to facilitate the back translation process and avoid errors that could occur by translating the instrument into Romanian and back into English. The final form is presented in the annexes.

Conclusions and future directions

Our study brings an important contribution from the point of view of instrumental research, showing the main stages and procedures necessary for obtaining a reliable and valid research measurement instrument. As a future managerial implication, such a questionnaire could be used to actually measure Romania's destination image from the perspective of external intermediaries from different tourist sending countries, after a cultural translation of the instrument from English to the native language of the analyzed country. Following the same procedure, instruments specific for other countries could be developed - the most important part being the qualitative assessment of the scale.

We started our approach because several studies demonstrated that tourists, especially when they don't have previous direct experiences, trust and turn to professional sources, like travel agencies, for advice (Bitner and Booms, 1982; Snepenger et al., 1990; Baloglu and Managaloglu, 2001). This is why measuring the perceived destination image of intermediaries is important, using a sound instrument that would allow comparisons. Although the literature on destination image is abundant, only a few studies measured this perceived image from the intermediaries' perspective (McLellan and Foushee, 1983; Gartner and Bachri, 1994; Baloglu and Managaloglu, 2001; Xiao and Mair, 2006; Camprubi et al., 2008). As a step forward from these studies, which do not insist on the research instruments used, we present a complex and detailed mixed methodology - qualitative and quantitative - and a thorough description of the whole scale development process, addressing reliability and validity issues. Another plus is the fact that we used real travel intermediaries to test the instrument, while other studies used students. Moreover, such a mixed procedure used on real travel agents leads to a valid instrument for comparisons among countries, despite the large differences in terms of specific tourist attributes.

There are several limits for our study. The main one is related to the small samples used for testing the instrument in various stages: we used four travel experts for testing the content validity and 15 travel agents for testing the internal consistency of the scale. Although some authors suggest that a number of 10 to 30 respondents would be appropriate (Hill, 1998), a bigger number would be more assuring. Also, for the content analysis (in order to get the general and specific attributes), only two coders were used. As future approaches, we intend to retest the instrument using a larger sample of experts for the content validity and a larger sample of travel agents for testing the internal consistency. Applying the instrument in order to obtain Romania's perceived destination image from the external intermediaries' perspective will also allow us to supplementary test the suggested scale.

References

References

Alcaniz, E.B., Garcia, I.S. and Blas, S.S. (2005), "Relationships among residents' image, evaluation of the stay and post-purchase behavior", Journal of Vacation Marketing, Vol. 11, No. 4, pp. 291-302.

Baloglu, S. and Brinberg, D. (1997), "Affective images of tourism destinations", Journal of Travel Research, Vol. 35, No. 4, pp. 11-15.

Baloglu, S. and Mangaloglu, M. (2001), "Tourism destination images of Turkey, Egypt, Greece, and Italy as perceived by US-based tour operators and travel agents", Tourism Management, Vol. 22, No. 1, pp. 1-9.

Baloglu, S. and McCleary, K.W. (1999a), "U.S. international pleasure travelers' images of four Mediterranean destinations: A comparison of visitors and non-visitors", Journal of Travel Research, Vol. 38, No. 2, pp. 144-152.

Baloglu, S. and McCleary, K.W. (1999b), "A model of destination image formation", Annals of Tourism Research, Vol. 26, No. 4, pp. 868-897.

Beerli, A. (2004), "Tourists' characteristics and the perceived image of tourist destinations: a quantitative analysis - a case study of Lanzarote, Spain", Tourism Management, Vol. 25, No. 5, pp. 623-636.

Beerli, A. and Martin, D.J. (2004). "Factors influencing destination image", Annals of Tourism Research, Vol. 31, No. 3, pp. 657-681.

Bitner, M.J. and Booms, B.H. (1982), "Trends in Travel and Tourism Marketing: The Changing Structure of Distribution Channels", Journal of Travel Research, Vol. 20, No. 4, pp. 39-44.

Bosque, I. and Martin, H. (2008), "Tourist satisfaction a cognitive-affective model", Annals of Tourism Research, Vol. 35, No. 2, pp. 551-573.

Calantone, R.J., Benedetto, A.D., Hakam, A. and Bojanic, D.C. (1989), "Multiple multinational tourism positioning using correspondence analysis", Journal of Travel Research, Vol. 28, No. 2, pp. 25-32.

Camprubí, R., Guia, J. and Comas, J. (2008), "Destination networks and induced tourism image", Tourism Review, Vol. 63, No. 2, pp. 47-58.

Chen, C.F. and Tsai, D.C. (2007), "How destination image and evaluative factors affect behavioral intentions?", Tourism Management, Vol. 28, No. 1, pp. 1115-22.

Chen, J.S. and Hsu, C.H.C. (2000), "Measurement of Korean tourists' perceived images of overseas destinations", Journal of Travel Research, Vol. 38, No. 4, pp. 411-416.

Chen, P.J. and Kerstetter, D.L. (1999), "International Students' Image of Rural Pennsylvania as a Travel Destination", Journal of Travel Research, Vol. 37, No. 3, pp. 256-266.

Crompton, J.L. (1977), "A systems model of the tourist's destination selection decision process with particular reference to the role of image and perceived constraints", Crompton.

Dann, G.M.S. (1996), "Tourists' images of a destination - an alternative analysis", Journal of Travel and Tourism Marketing, Vol. 5, No. 1/2, pp. 41-55.

Echtner, C.M. and Ritchie, J.R.B. (1991), "The meaning and measurement of destination image", Journal of Tourism Studies, Vol. 2, No. 2, pp 2-12.

Echtner, C.M. and Ritchie, J.R.B. (1993), "The measurement of destination image: An empirical assessment", Journal of Travel Research, Vol. 31, No. 4, pp. 3-13.

Gartner, W.C. (1993), "Image formation process", Journal of Travel and Tourism Marketing, Vol. 2, No. 2/3, pp. 191-216.

Gartner,W.C. (1989). "Tourism image: Attribute measurement of state tourism products using multidimensional scaling techniques", Journal of Travel Research, Vol. 28, No. 2, pp. 16-20.

Gartner, W.C. and Bachri, T. (1994), "Tour operators' role in the tourism distribution system: An Indonesian case study", Journal of International Consumer Marketing, Vol. 6, No. 3/4, pp. 161-179.

Govers, R., Go, F.M. and Kumar, K. (2007), "Promoting Tourism Destination Image" Journal of Travel Research, Vol. 46, No. 1, pp. 15-23.

Granka, L.A., Joachims, T. and Gay,G. (2004), "Eye-tracking analysis of user behavior", Proceedings of the 27th annual international conference on Research and development in information retrieval SIGIR 04, 478. ACM Press.

Hartmann, D. (1977), "Considerations in the choice of inter-observer reliability estimates", Journal of Applied Behavior Analysis, Vol. 10, No. 1, pp. 103-116.

Hertzog, M.A. (2008), "Considerations in determining sample size for pilot studies", Research in nursing and health, Vol. 31, No. 2, pp. 180-191.

House, A.E., House, B.J. and Campbell, M.B. (1981), "Measures of inter-observer agreement: Calculation formulas and distribution effects", Journal of Behavioral Assessment, Vol. 3, No. 1, pp. 37-57.

Hui, T.K. and Wan, T.W.D. (2003), "Singapore's image as a tourist destination", International Journal of Tourism Research, Vol. 5, No. 4, pp. 305-313.

Johanson, G.A. and Brooks, G.P. (2009), "Initial Scale Development: Sample Size for Pilot Studies", Educational and Psychological Measurement, Vol. 70, No. 3, pp. 394-400.

Joppe, M., Martin, D.W. and Waalen, J. (2001), "Toronto's image as a destination: A comparative importance-satisfaction analysis by origin of visitor", Journal of Travel Research, Vol. 39, No. 3, pp. 252-260.

Lee, C.-ki, Lee, Y.-ki and Lee, B. (2005), "CFA-Korea's destination image formed by the 2002 World Cup", Annals of Tourism Research, Vol. 32, No. 4, pp. 839-858.

Lubbe, B. (1998), "Primary image as a dimension of destination image: An empirical assessment", Journal of Travel and Tourism Marketing, Vol. 7, No. 4, pp. 21-43.

MacKay, K.J. and Fesenmaier, D.R. (2000), "An exploration of cross-cultural destination image assessment", Journal of Travel Research, Vol. 38, No. 4, pp. 417-423.

Malhotra, N.K. (2010), Marketing Research: An Applied Orientation, Pearson: New York.

Malhotra, N., Agarwal, J. and Peterson, M. (1996), "Methodological issues in cross-ultural marketing research: a state-of-the-art review", International Marketing Review, Vol. 13, No. 5,pp. 7-43.

Mastaglia, B., Toye, C. and Kristjanson, L.J. (2003), "Ensuring content validity in instrument development: challenges and innovative approaches", Contemporary nurse a journal for the Australian nursing profession, Vol. 14, No. 3, pp. 281-291.

McLellan R.W. and Foushee, K.D.(1983), "Negative images of the United States as expressed by tour operators from other countries", Journal of Travel Research, Vol. 22, No. 1, pp. 2-5.

Milman, A. and Pizam, A. (1995), "The Role of Awareness and Familiarity with a Destination: The Central Florida Case", Journal of Travel Research, Vol. 3, No. 3, pp. 21-27.

Neuman, W.L. (2003), Social Research Methods: Qualitative and Quantitative Approaches. J. Lasser (ed.) Relevance of social research, Vol. 8, p. 608.

Obenour, W. (2005), "The development of a destination through the image assessment of six geographic markets", Journal of Vacation Marketing, Vol. 11, No. 2, pp. 107-119.

O'Leary, S. and Deegan, J. (2005), "Ireland's Image as a Tourism Destination in France: Attribute Importance and Performance", Journal of Travel Research, 43, pp. 247-256.

Pike, S. (2002), "Destination image analysis - a review of 142 papers from 1973 to 2000", Tourism Management, Vol. 23, No. 5, pp. 541-549.

Pike, S. and Ryan, C. (2004), "Destination positioning analysis through a comparison of cognitive, affective, and conative perceptions", Journal of Travel Research, Vol. 42, No. 4, pp. 333-342.

Polit, D.F. and Beck, C.T. (2006), "The content validity index: are you sure you know what's being reported? Critique and recommendations", Research in nursing health, Vol. 29, No. 5, pp. 489-497.

Reilly, M.D. (1990), "Free Elicitation of Descriptive Adjectives For Tourism Image Assessment", Journal of Travel Research, Vol. 28, No. 4, pp. 21-26.

Snepenger, D., Meged, K., Snelling, M. and Worrall, K. (1990), "Information Search Strategies by destination-naive Tourists", Journal of Travel Research, Vol. 29, No. 1, pp. 13-16.

Sonmez, S. and Sirakaya, E. (2002), "A distorted destination image? The case of Turkey", Journal of Travel Research, Vol. 41, No. 2, pp. 185-196.

Tapachai, N. and Waryszak, R. (2000), "An examination of the role of beneficial image in tourist destination selection", Journal of Travel Research, Vol. 39, No. 1, pp 37-44.

Tasci, A.D.A. and Gartner, W.C. (2007), "Destination Image and Its Functional Relationships", Journal of Travel Research, Vol. 45, No. 4, pp. 413-425.

Tasci, A.D.A., Gartner, W.C. and Tamer Cavusgil, S. (2007), "Conceptualization and Operationalization of Destination Image", Journal of Hospitality and Tourism Research, Vol. 31, No. 2, pp. 194-223.

Tavakol, M. and Dennick, R. (2011), "Making sense of Cronbach's alpha", International Journal of Medical Education, Vol. 2, pp. 53-55.

Vicol, O.C. and Spiridon, S.I. (2010), "Evaluating Online Destination Image By Analyzing Pictures Used in Presenting Romania on Websites", Ovidius University Annals Economic Sciences Serie, Vol. 10, No. 1, pp. 1353-1358.

Xiao, H and Mair, L. (2006), "A paradox of images: representation of China as a tourist destination", Journal of Travel and Tourism Marketing, Vol. 20, No. 2, pp. 1-14.

Walmsley, D.J. and Young, M. (1998), "Evaluative images and tourism: The use of personal constructs to describe the structure of destination images", Journal of Travel Research, Vol. 36, No. 3, pp. 65 ]69.

Waltz, C.F., Strickland, O. and Lenz, E. (1991). Measurement in nursing research. Philadelphia: F. A. Davis.

Warrens, M.J. (2010), "Inequalities between multi ]rater kappas", Advances in Data Analysis and Classification, Vol. 4, No. 4, pp. 271 ]286.

* * * United Nations World Tourism Organization (2004). www2.unwto.org/

Appendix

(ProQuest: ... denotes formulae omitted.)