Instagram influencers attributes and parasocial relationship: A dataset from Qatar

The dataset investigates how social media influencers’ attributes affect followers’ parasocial relationship. It also examines the mediating role of the parasocial relationship between the social media influencers’ attributes and behavioral intentions. A snowballing sampling technique was used to target Instagram users in Qatar. 574 valid responses were analyzed using Partial least squares structural equation modeling (PLS-SEM). The data provides descriptive information about the essential Instagram influencers among users in Qatar. It also gives new insight into the influencers’ characteristics that will impact consumer behavior the most. The dataset could be very helpful for brands and marketers in Qatar in choosing the most effective influencers. The dataset presents a real value for researchers examining social media consumers behavior specifically in GCC countries context or conducting cross-national comparative studies.


a b s t r a c t
The dataset investigates how social media influencers' attributes affect followers' parasocial relationship.It also examines the mediating role of the parasocial relationship between the social media influencers' attributes and behavioral intentions.A snowballing sampling technique was used to target Instagram users in Qatar.574 valid responses were analyzed using Partial least squares structural equation modeling (PLS-SEM).The data provides descriptive information about the essential Instagram influencers among users in Qatar.It also gives new insight into the influencers' characteristics that will impact consumer behavior the most.The dataset could be very helpful for brands and marketers in Qatar in choosing the most effective influencers.The dataset presents a real value for researchers examining social media consumers behavior specifically in GCC countries context or conducting cross-national comparative studies.
© 2024 The Author(s).Published by Elsevier Inc.This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ) Specifications

Value of the Data
• This dataset is interesting as Qatar is the 3rd (96 %) country in the world regarding the percentage of the population using social networks [1] .In addition, Instagram represents after Facebook, the second most-used social network in Qatar [2] .• The dataset contains measures of the main attributes (i.e., Homophily, Popularity, Leverage, Fashionable, and Affinity) of Instagram influencers in Qatar.The dataset also includes measures of the parasocial relationship between Instagram users and influencers, purchase intention (of product recommended by the social media influencer), and word of mouth (intention to recommend or speak positively about the influencer).• The dataset includes several control variables (e.g., demographics and usage of Instagram) that may be used to perform more sophisticated analysis as multigroup analysis MGA or test for moderated mediation.• The dataset could be very helpful for brands and marketers in Qatar in choosing the most effective influencers.Researchers could use the data to identify the fit between the influencers' attributes and their domain of expertise.Data could also be helpful for future cross-national comparative studies that will replicate the data collection in different cultures, countries, and regions.For future research data could be also coupled with other types of data as data obtained from Data mining [3] or Netnographic data [4] .• Participants in our survey were 38 % males and 61 % females living in Qatar.This distribution could be seen as a limitation of our data, as in Qatar, females represent 27.6 % of the population and 34.6 % of the total Instagram users [5] .However, if we report the percentage of females and males to the number of Instagram users in Qatar (40.7 of the population) and the percentage of Instagram users by sex [5] , it is possible to observe a good fit between the distribution of our data and the penetration rate by sex for Instagram users in Qatar: 56 % of females in Qatar use Instagram, and only 35 % of males use Instagram.
• Finally, respondents were 73.3 % Qatari and 26.7 % Non-Qatari Arabic speaking.Qatari citizens represent less than 15 % of the population of Qatar, and Other Arabs represent about 13 % of the total population [6] .However, these segments, specifically the Qatari Citizens, are extremely important from a marketing point of view first because they are one of the wealthiest populations in the world [7] and second because it is difficult to get information from this segment.

Objective
With the continuous increase in social media networks (SMN) use around the world and the rise of time consumers spend on SMN, it became critical for firms and researchers to understand better how SMN could impact consumers' behavior and could be used as effective marketing tools.In this context, social media influencers are becoming the "masterpiece" for any effective social media marketing strategy.This dataset investigates in the context of Instagram users in Qatar, the social media influencers' attributes that effectively impact the parasocial relationship between followers and influencers and lead to behavioral intention: WOM and purchase intentions.It also examines the mediating role of the parasocial relationship between Instagram influencers' attributes and consumers' behavioral intentions.
In addition, as the dataset includes several control variables (e.g., demographics and usage of Instagram), it would be interesting to explore how the proposed model works differently for different groups of consumers (by sex, age, level of usage of SMN, category of product …).

Data Description
In total, 900 people received the link to our final questionnaire.691 responses were collected, indicating a response rate of 76 %.However, only 574 were valid (83 %) and included in our dataset.One hundred seventeen responses (17 %) were eliminated for different reasons, including the age of participants below 18, participants without an Instagram account, or those who do not follow any Instagram influencer.
The dataset associated with this article comes in a raw data table format (.CSV) and an SPSS data file (.sav) that could be used for different analyses.The dataset consists of answers from Instagram users in Qatar asked about their behavior on Instagram, their favorite Instagram influencer, their perceptions of the attributes of their favorite Instagram influencer, their (parasocial) relationship with their favorite Instagram influencer, as well as the possibility of recommending this influencer or purchase goods or services endorsed by their favorite influencer.
A five-point Likert scale survey instrument including 35 items was developed to measure seven different concepts.Considering the importance of the validity and robustness of measurement scales and items, we used only tested and validated scales published in top journals (ranked Q1 CiteScore Best Quartile).Table 1 summarizes the Wording of Measurement Items and their sources.However, after translating the items from English to Arabic and the questionnaire pretest, we slightly changed the wording of some Items to adapt them to the Qatari context and Arabic Language.The changes do not affect the items' meaning nor the scales' content validity.We tested for Construct reliability and convergent validity ( Table 5 ) as well as for discriminant validity ( Tables 6 and 7 ).The results confirm the construct's reliability and validity.
For every item, the responses were scored as 'strongly agree' 5, 'agree' = 4, 'neutral' = 3, 'disagree' = 2, and 'strongly disagree' = 1.First, we adopted four scales from [8] to measure four of the perceived Instagram influencer attributes: Popularity (Pop/ 3 items); Leverage (Lev/ 3 items); Affinity (Aff/ 3 items); and Fashionable (Fash/ 3 items).We measured a fifth Instagram influencer perceived attribute, Homophily (Hom/ 9 items), adapting the scale from [9] .We used an adapted scale from [10] to measure Parasocial relationship (PSI/ 6 items).For behavioral intentions, we used respectively an adapted version of the scale from [11] to measure word of mouth (WOM/ 3 items) and a slightly adjusted version from [12] to measure the intention to purchase (Int/ 5 items).All the scales have reflective Items.Fig. 1 illustrates our conceptual model, and Table 1 represents the wordings of all the items used for the different scales.
The dataset also contains different control variables that could be used for additional analysis.We asked the participants about their age, nationality, revenue, and education.Participants were  Most participants (61.7 %) earn less than QTR 50,0 0 0 annually, as most were students.Table 2 summarizes the profile of our sample.We questioned the participants about their Instagram behavior; we observed that about half of them (49.3 %) spend more than 5 h daily on social media and that the majority of the respondents follow local influencers with substantial diversity in terms of the domain of expertise of their favorite Instagram influencers.Characteristics of respondents related to social media and Instagram behavior are summarized in Table 3 .
Table 4 presents the descriptive statistics for the scales' items.We also used the Common Harman's single-factor to assess possible common method variance problems.The results indicate that the variance accounted for in the first factor is 34.6 % lower than 50 %, indicating that the sample did not contain common method bias.In what follows, we present the PLS-SEM results obtained using SmartPLS.4[13] .eight tables and two figures summarize the measurement model's quality (the instruments' reliability and validity) and the structural model (correlation and hypothesis testing).
Due to low outer loadings, five items were removed from the final measurement model (Hom9, PSI1, PSI5, Lev1, Aff2).The final measurement model is summarized in Fig. 2 .
In general, the measurement model indicates good reliability ( Table 5 ), convergent validity ( Table 5 ), and discriminant validity ( Tables 6 and 7 ).All ρA values exceed the standard threshold of 0.7, all average variance extracted (AVE) values are larger than 0.5, and the values of HTMT criterion exceed the conservative value of 0.85 except for one value that is lower than 0.9 and considered acceptable [14] .Fig. 3 and Table 8 summarize the main results considering the structural model.We also checked that for the Inner model, all the Variance inflation factors (VIF) are lower than 5.00, indicating the absence of high collinearity concerns [12] .Table 9 presents the collinearity test results.
The results for the coefficient analysis R2 are presented in Table 10 and present, in general, a satisfactory level of explanatory power.
We tested the predictive power of our model using CVPAT with PLSpredict.As summarized in Tables 11 and 12 , the model indicates some predictive power to pass the "naïve" IA benchmark but insufficient predictive power to overcome the more conservative LM benchmark [15] .We analyzed the model fit using the SRMR as an indicator.The estimated model has an SRMR of 0.09, higher than the 0.08 threshold but would be accepted if we use the less conservative 0.1 threshold.
Finally, as additional indicators of the model fit we calculated the global goodness-of-fit (GoF) criterion and the effect size.We found a GoF of 0.598 higher than the Threshold of 0.36 confirming the good quality of Fit.Table 13 summarizes the different ƒ² effect size related to the impact of the predictive constructs on the endogenous latent constructs.

Limitations
As indicated in the "Experimental design, materials, and methods" section, we first sent the questionnaire link to over 25,0 0 0 valid emails, but the response rate was lower than 0.2 %.Consequently, we adopted a snowball sampling technique that is effective in situations of difficulty in reaching a specific population or when no population frame is available [17] .Using this technique, we obtained a response rate of 76 % and 574 valid answers, a sample size bigger than the required sample size of 385 (calculated using the Qualtrics sample size calculator [18] ).Our sample is also bigger than the samples of all the articles dealing with Instagram consumer behavior in Qatar and published from 2016 to 2022 (see [19] ).However, the Snowball technique has limitations as the initial participants select the following members for the sample, creating bias and negatively impacting the sample representativeness [20] .In our situation, we obtained more female than male respondents and a majority of Qataris.As discussed earlier, these two elements could be considered strengths for our dataset but must be considered carefully for future usage of our dataset.

Table 1
Wordings of measurement items.
PSI1.I feel close enough to my favorite Instagram influencer to use his(her) Instagram.PSI2.I feel comfortable about my favorite Instagram influencer messages.PSI3.I can rely on the information I get from my favorite Instagram influencer.PSI4.I feel fascinated with my favorite Instagram influencer's Instagram.PSI5.In the past, I pitied my favorite Instagram influencer when he/she made a mistake on his/her Instagram.PSI6.I think that my favorite Instagram influencer is helpful for my interests (in fashion and others).

Table 2
Profile and demographic characteristics of respondents ( n = 574).

Table 5
Construct reliability and convergent validity.