Cross-Cultural Validity of the Psycho-Social Aspects of Facebook Use (PSAFU) Scale

The Psycho-Social Aspects of Facebook Use (PSAFU) scale was developed by
 Bodroza and Jovanovic (2016) to comprehensively measure psychological and
 social aspects of the users? Facebook (FB) experiences i.e. compensatory use
 of FB, self-presentation on FB, socializing, and seeking sexual partners
 through FB, FB addiction, and FB profile as the virtual self. The scale was
 previously validated on two Serbian samples. The aim of this study was to
 examine the validity of the PSAFU scale across multiple cultures, i.e.
 develop a version of the scale that would be suitable for the cross-cultural
 study of FB behaviors. The sample consisted of 1.632 respondents from
 Croatia, Italy, Iran, Serbia, and the UK. Administered questionnaires
 consisted of the PSAFU scale (Bodroza & Jovanovic, 2016), the Big Five
 Inventory (BFI; John et al., 1991), the Narcissistic Personality Inventory
 16 (NPI-16; Ames et al., 2006), socio-demographic, and FB use questions
 (e.g., time spent on FB on a daily basis, frequency of posting, and number
 of FB friends). After some modifications and exclusion of the Virtual Self
 scale, the results of multigroup confirmatory factor analysis show that the
 PSAFU scale is cross-culturally invariant for the cultures included in this
 study and suitable for use in them. The remaining dimensions of revised
 PSAFU scale (PSAFU-R) have statistically significant relationships with
 measured personality traits and with indicators of FB use, although
 considerable cross-cultural variations were obtained.

• Considerable cross-cultural similarities were obtained.
People are mainly motivated to create a Facebook (FB) account because of the possibility to connect and communicate with people they know offlinesocial searching and, to a lesser extent, to explore and meet new, unknown individualssocial browsing (Joinson, 2008;Lampe et al., 2006;Vasalou et al., 2010). More and more, the use of social networking sites (SNS) is reshaping the way people communicate and think (Kross et al., 2013). In 2016, the Psycho-Social Aspects of Facebook Use (PSAFU) scale was developed by Bodroža and Jovanović (2016) to comprehensively measure psychological and social aspects of the users' FB experiences. Having recognized that most of the available instruments at that time were focused on specific characteristics of FB, the goal of creating this scale was to construct an instrument that covers diverse aspects of Facebook use from a psychological standpoint. The authors aimed to measure both objective FB behaviors, as well as deeper psychological processes such as thoughts, feelings, and motives. Previously, this instrument was validated on a population of FB users from Serbia. The goal of this paper was to examine a wide range of FB experiences measured by the PSAFU scale and to explore the scale's properties on a more diverse sample from several countries with different cultural backgrounds: Croatia, Italy, Iran, Serbia and the UK.

Psycho-Social Aspects of Facebook Use (PSAFU) scale
There are numerous studies about FB use and FB behavior, but most of them focus on objective indicators of FB use, such as the number of friends, frequency of posting, number of photos, time spent on FB daily, etc. However, measuring the more objective indicators of FB use leaves us blind to psychological processes and the motives behind these behaviors. Only a small number of studies have explored deeper psychological implications of FB use, such as the feelings evoked during FB interactions, motivations behind certain FB behaviors, and thoughts provoked during FB use (Bodroža & Jovanović, 2016;Ellison et al., 2007;Michikyan et al., 2014;Ross et al., 2009). The PSAFU scale is among the rare measures of FB behaviors that encompass diverse aspects of both FB behaviors and also subtle psychological processes related to FB use. The scale measures five distinct aspects of FB experiences: compensatory use of FB (compensation), self-presentation on FB (self-presentation), socializing and seeking sexual partners through FB (socializing), FB addiction (addiction), and FB profile as the virtual self (virtual self;Bodroža & Jovanović, 2016). The scale 'compensatory use of FB' measures experiences of FB users who find it easier to express themselves and communicate in a FB setting than in an offline setting. When a person carefully chooses their photos and posts in order to create certain impressions on others, they would score highly on the self-presentation scale.
Those who are actively using FB as means to find new friends and sexual partners are representative of 'socializing and seeking sexual partners through FB'. The 'FB addiction' scale includes indicators of the lack of control over the time spent on FB, which in turn affects everyday life and activities. Finally, a high score on the dimension 'virtual self' is descriptive of users who perceive their FB profile as a true and reliable representation of their personality in an online setting.
The structure of the scale was first determined by Exploratory Factor Analysis (EFA) on a general population sample of FB users from Serbia, and then confirmed by Confirmatory Factor Analysis (CFA) on a student sample from Serbia. CFA also led to shortening of the scale in order to achieve a satisfactory fit. Nevertheless, the authors concluded that a longer version, even though less psychometrically sound, may have better content validity (Bodroža & Jovanović, 2016). As for psychometric characteristics, the scale showed satisfactory internal consistency with Cronbach's alphas ranging from .76 (for Virtual self) to .92 (for Compensation). The PSAFU dimensions showed stable patterns of relationships with Big Five personality traits, social anxiety and sensation-seeking. Social anxiety and conscientiousness were shown to be the most consistent determinants of all measured aspects of FB use, suggesting that FB users who are less conscientious and more socially anxious find comfort in an anonymous and controllable FB setting.
In their recent review of the scales for measuring SNS users' engagement, Sigerson and Cheng (2018) stated that the biggest concern regarding the PSAFU scale is the validity of the scale itself. Sigerson and Cheng recommended the authors show that PSAFU dimensions have meaningful relationship with measures of intensity of SNS use (e.g., time spent on SNS or frequency of posting) and measures of bridging and/or bonding social capital. The current study aims to address these concerns by examining the relationship of PSAFU dimensions with time spent online, frequency of posting, and the number of FB friends. Social capital usually refers to the resources an individual can access from their social network (Liu et al., 2016). These resources can be love and affection (bonding social capital) or information and economic benefits (bridging social capital). In this study, the number of FB friends was used as an indirect measure of social capital, which is closer to the notion of bridging than bonding social capital. It is usually assumed that the larger the network of friends on FB one has, the more socially heterogeneous that network becomes and, consequently, the more diverse information one can reach, which is the core concept of bridging social capital (e.g., Granovetter, 1973). Sigerson and Cheng (2018) additionally suggested that objective measures of time spent online and frequency of posting would be superior to self-reported measures. In this study, due to the study design, we did not have objective data on FB use, so we had to opt for self-report measures. There are several studies that explore the relationship between objective and subjective reports of SNS use (Johannes et al., 2021;Scharkow, 2016). Johannes and colleagues (2021) have stated that there is indeed a measurement error of self-reports on SNS use, where persons tend to overestimate their SNS use. According to these authors, this tendency is not random but robust and systematic, meaning that findings that rely on self-reports can still be considered useful and informative.
Finally, one should consider that FB is a worldwide phenomenon. For the questionnaire to be used in different cultures, proof of its cross-cultural validity (i.e., structural invariance) should also be provided.

Facebook Use in a Cross-Cultural Context
Research suggests that culture plays an important role in shaping SNS behavior and should therefore be taken into consideration. There are several studies that have examined the effects of national culture on SNS user behaviors (Chapman & Lahav, 2008;Jovanović et al.,2018;Karl et al., 2010;Krasnova et al., 2012;Vasalou et al., 2010) and all have led to the same conclusion -"culture is the key FB behavioral determinant" (Vasalou et al., 2010, p. 727).
FB is the most used SNS among all countries examined in this study, except Iran, where Telegram and Instagram are currently the most popular SNSs. Although we have witnessed a rise of the popularity of Instagram in the last few years, at the global level Facebook is still the most popular social network with 2.7 billion active users worldwide as compared to 1.3 billion active users of Instagram in April 2021 (Statista, 2021). The UK is currently the 12 th ranking country in the world in the number of FB users, with about 38 million reported users (which is about 56% of the total UK population) in April 2021 (Statista, 2021 However, Iranian sources (Boutorabi, 2012;Janati, 2014), state that there are currently no exact estimates of the number of users, but that previously mentioned numbers are greatly exaggerated and that in 2012 the number of users was somewhere between 4 and 17 million, which is between about 5% and 22% of the total Iranian population at that time.
As an integral part of the validation process, the current study aims to explore the crosscultural validity (i.e., structural invariance) of the PSAFU scale i.e. to come up to the scale version that will be suitable for use in cross-cultural research. In order to do this, the necessity of cross-cultural structural changes to the scale should be considered.

Personality and Facebook Use in Cross-Cultural Setting
Numerous scholars have explored the relationship between personality traits and SNS use (Amichai-Hamburger & Vinitzky, 2010;Bodroža & Jovanović, 2016;Correa et al., 2010;Orchard et al., 2014;Ross, et al., 2009;Skues et al., 2012;Wilson et al., 2010). There are two main hypotheses concerning the relationship between the use of SNS and personality characteristics. The first theory is called "compensation theory" or Poor-Gets-Richer (Kraut et al., 2002), while the other is referred to as Rich-Gets-Richer or "enhancement theory" (Kraut et al., 2002). The premise of the first theory is that those who have difficulties interacting offline (e.g., introverted, shy or socially anxious individuals) will benefit the most from using FB. On the other hand, Rich-Gets-Richer theory claims that those who are sociable and skilled in offline interactions (such as extraverts) will tend to socialize and widen their networks through SNS, because use of such services can additionally enrich their social life. Both theories have been empirically corroborated, leading to the conclusion that the relationship between personality traits and FB use depends on the motives for its use (Moore & McElroy, 2012). Nevertheless, there are no studies exploring these two theories in regards to how national culture and personality traits interact to affect SNS (FB) behavior.
Although the pattern of the relationship between Big Five personality traits and FB related behaviors in different countries can be quite diverse (Caci et al., 2014;Eftekhar et al., 2014;Fallahi & Zahra, 2015;Ivnik, 2014;Jovanović & Bodroža, 2016), some consistencies could be extracted. Most consistent results of the above-mentioned studies suggest that extraversion is almost universally a positive predictor of the number of friends and activity on FB, while higher neuroticism and lower conscientiousness predict increased time spent on FB.
Openness was related to the frequency of posting in few studies, but generally Agreeableness and Openness vary more across studies and therefore seem to be affected more by culture.
Narcissism is also frequently researched in the context of online behaviors (Casale & Fioravanti, 2018;Mamić, 2014) Two meta-studies recently summarized these findings and showed that (grandiose) narcissism is positively related to all aspects of SNS use that were followed: time spent on SNS, number of friends and/or followers, number of photos (especially selfies), frequency of status updates, as well as with more interaction (commenting and liking other's posts; Liu & Baumeister, 2016;McCain & Campbell, 2018). Since numerous studies show that narcissism is an important determinant of online behaviors, we have also included this personality trait it in this study.

The Aims of the Study
The goal of this study was to validate the PSAFU scale in a cross-cultural context, i.e. to develop the version of the scale that would be suitable for the cross-cultural study of FB behaviors. Our aim was to examine the cross-cultural invariance of the PSAFU scale.
Additionally, we tested the predictive validity of the PSAFU scale via its relationship with intensity of FB use, social capital and its discriminant validity via examining its relationship with personality traits. All PSAFU dimensions were expected to be positively related to time spent online with addiction achieving the strongest relationship. Frequency of posting was hypothesized to have the strongest relationship with self-presentation and socialization, since these dimensions encompass aspects of FB use directed at content creation and the active maintenance of social relations that could be achieved through posting different content. On the other hand, addiction, and compensation could be related to the frequency of posts, but to a lesser degree, since time people spend online can also be used passively for browsing othergenerated content. Finally, the number of friends as an indirect measure of (bridging) social capital was expected to be primarily related to the dimension socialization, whose main content refers to establishing new online friendships.
We expected the most cross-culturally stable findings regarding extraversion being related to the indicators of online socializing. Furthermore, higher neuroticism, and lower conscientiousness were expected to be the most consistently related to the indicators of FB addiction (excessive use, inability to control time spent on FB, etc.), and compensation (which covers the other aspect of addictive behaviors). Narcissism was expected to be a relatively consistent predictor of socializing (i.e. initiating online friendships), and self-presentation on FB, since self-presentational concerns are usually achieved through status and photo posting. Also, narcissism was expected to be related to FB addiction, since this PSAFU dimension includes indicators of prolonged time spent on FB.

Participants
Respondents from five countries -Croatia, Italy, Iran, Serbia and the UK, participated in this study. In every country, we aimed to gather a sample of at least 250 participants from the student population. All national subsamples were convenience samples. Gathered subsamples ranged from 251 in the UK to 429 in Italy (see Table 1). Other characteristics of the subsamples in each country are presented in Serbia. The structure of subsamples from Croatia, the UK, and Serbia are very similar regarding gender (primarily female), education (majority with bachelor degree), and the place of living (around half of the sample from urban areas). The UK subsample included more students who are working part-time, while in Croatia and Serbia the majority are unemployed. The Italian subsample is better balanced regarding gender (both genders equally represented) and in place of living. Among these four countries, the structure of the subsamples is very similar regarding marital status of participants with the majority being single or in a dating relationship.

Table 1
Structure of the subsamples from five different countries

INSERT TABLE 1 HERE
The Iranian subsample deserves a more detailed description, since it is slightly different from the subsamples of the other four countries. These differences could be attributed to cultural differences and specific circumstances related to FB use in this country. Since FB use in Iran was illegal at the time this research was conducted, acquiring a sample was challenging. Thus, we allowed non-students to be included in the subsample, which resulted in more full-time employed participants. Also, due to the fact that at the time of data gathering internet infrastructure in this country was mostly limited to urban areas and usually available to welleducated people, this subsample included significantly more participants from big cities and with a higher educational degree, as compared to other countries. Finally, when it comes to marital status, due to cultural specificities, the Iranian sample included more married individuals and significantly fewer individuals in a non-married relationship. The category "common-law union" was omitted from the survey because this kind of partnership is illegal in this country.

Procedure
Since this study did not raise ethical concerns, at the time when it was conducted (in the year of 2018), the ethical approval was not obligatory in any of the countries. However, it was obtained for the UK sample in line with guidance of the British Psychological Study. In Serbia, Croatia and Iran, the questionnaire was administered online while in Italy and the UK pen and paper procedure was utilized. In all the countries, socio-demographic questions were at the beginning of the questionnaire, while the order of the scales followed the order of presentation in the Instrument section. Participants from Croatia, Italy, Serbia and UK were informed about the research topic on their university courses and were invited to take part in exchange for course credit. In Iran, information about the study was shared through academic FB groups, where users were invited to fill out and share the questionnaire. All respondents were informed about the aim of the study, as well as voluntary and anonymous participation. Only individuals who gave their consent participated in the study. 44 items), which has been translated into many languages. In this study, validated, official and/or previously used translations of the scale were used in all countries where they were available (Serbian, Italian, and English;Čolović et al., 2005;Fossati et al., 2011;John et al, 1991). In Iran and Croatia, the scale was back-translated from the English version by the national collaborator.

Psycho-Social Aspects of Facebook Use
Reliability of the BFI scales was satisfactory in most countries (Appendix A) with a few exceptions. The reliability of the Italian version of agreeableness and the UK version of openness fell somewhat below the conventional point of .70 but could still be considered acceptable. The reliability of the Iranian translation of the scales agreeableness and conscientiousness were .56 and .58, respectively. Although these values are not acceptable, we decided to use the data from these scales, but to treat the results with caution. However, in case of future use, these translations need to be refined. In all countries, items were presented along with a 5-point Likert scale (1 = I don't agree at all, to 5 = I completely agree).
Narcissistic Personality Inventory . In this study, the short 16-item version of NPI was used (Ames et al., 2006) to measure narcissism. The scale consists of pairs of alternative items; one of which describes narcissistic behavior, while the other describes non-narcissistic behavior.
Participants answer by choosing one of the two items which describes them best. All five translations of the scale had satisfactory internal consistency (Appendix A). All countries used the existing translation of the scale (Ames et al., 2006;Dinić & Vujić, 2019;Fossati et al., 2013;Jakšić et al., 2014), with the exception of Iran where it was back-translated from English by the national collaborator.
Socio-demographic and FB use questions. Along with the described questionnaires, additional questions regarding socio-demographic variables and a range of questions regarding FB use were administered. To measure time spent online, we asked participants "How many hours a day do you spend on Facebook?" and they responded by choosing between one of four categories: less than one hour, one to three hours, three to five hours, more than five hours. Information on the number of FB friends was gathered via an open-ended question and participants were encouraged to write down the exact number of friends at the time of completion. The question on frequency of posting was formulated as "How often do you post contents (statuses, links, photos, etc.) on your timeline?". Participants answered by choosing one of six options: more than once a day, once a day, once in two to three days, once a week, once in every few weeks, very rarely or never. As for socio-demographic variables, we gathered data on gender, age, education, place of living, employment and relationship status (see Table 1).

Data Analyses
Before starting the analyses, missing data were replaced by EM method. We first employed CFA with Maximum Likelihood estimation procedure (factor loading of the first observed variable set to 1) separately for each country to confirm the structure of the PSAFU scale. When item loadings fell below .40, the item was removed from the model in all countries.
Since our goal was to reach the same model in all countries, the changes were employed in all subsamples until a satisfactory model fit was reached for all. In order to test cross-cultural stability of the PSAFU scale, we conducted a multi-group CFA. To investigate measurement invariance, we analyzed configural invariance (factor loadings and intercepts are allowed to vary between groups), metric invariance (factor loadings are equal across groups) and scalar invariance (item intercepts are equal across groups). The fit of CFA models was established through analyzing the following fit indices (Byrne, 2010): root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR), with values smaller than .08 indicating adequate fit of data to the model; Tucker Lewis Index (TLI), and Bentler's Comparative Fit Index (CFI), with values larger than .90 indicating adequate fit; and χ 2 where values that are not statistically significant indicate adequate fit of data to the tested model.
To test the differences between three nested models in multi-group CFA (configural, metric, and scalar), we calculated ΔCFI, ΔRMSEA and ΔSRMR. The changes in these indicators were used because the difference in the chi-square test has been criticized as inadequate when sample size varies between groups, and when the sample size is large, both of these issues being the case in our study (Chen, 2007;Cheung & Rensvold, 2002). Changes in CFI, RMSEA and SRMR have been shown to be relatively robust and are seen as good indicators (Chen, 2007;Cheung & Rensvold, 2002). Their value should be less than .01 for all three indexes (Cheung & Rensvold, 2002).
After the structure of the final version of the PSAFU questionnaire was determined, scale scores were calculated as means of the belonging items, which gave a theoretical range of the scales scores from 1 to 5. Descriptive statistics (mean score, standard deviation, minimum, maximum, skewness, kurtosis, and K-S test), were also calculated, as well as Cronbach's alpha (internal consistency) coefficients. All of these statistics were done separately for all five countries. Where variable distribution deviated from normal, we applied Blom's formula for normalization. Normalized variables were then used in all further analyses.
Predictive validity of the revised PSAFU scale (PSAFU-R) was explored by regression analyses in which PSAFU-R dimensions were treated as predictors of measures of the intensity of FB use (time spent on FB on daily basis and average number of posts) and of the number of FB friends. Additionally, regression analyses were used to examine the relationship between PSAFU-R dimensions and personality traits. In these analyses, personality traits were the predictors of PSAFU-R dimensions. In all the regression analyses, gender and age were added as control variables. These analyses were carried out separately on subsamples from five countries.

CFA of PSAFU Scale in Five Countries
The first step in the analyses was a series of CFA of the PSAFU scale in five different countries.
As mentioned in the 'Measures' section, the Iranian version of the PSAFU scale did not contain item No. 40. To be able to compare the models in all countries, we removed this item at the very beginning and did not consider it further in any country. We initially tested the structure of the  (Bodroža & Jovanović, 2016). The short version consisted of 26 items that measure the five dimensions as described. This model had a better fit, but was also not adequate to describe the PSAFU structure in the five countries. Starting from the short version of the scale, the next step was to analyze factor loadings and modification indices. This resulted in the removal of the items 9 and 27 from the compensation dimension. Item 39, belonging to the FB addiction scale ("I often spend more than three hours continuously on FB"), had low loading in the Italian sample, so we discarded it from the model in all countries. Since exclusion of this item resulted in FB addiction scale having only 3 items, we retested the model with items from the longer version of this scale. One item (No. 2) showed satisfactory loading in all countries so we decided to retain it. Finally, two out of three items from the scale Virtual Self had low loadings in all subsamples. We retested additional items from the longer version of the scale, but they did not improve the model fit, nor did they have satisfactory factor loading. This indicated that Virtual Self is not sufficiently empirically grounded and that it does not represent a stable construct. Therefore, we decided to completely exclude this scale. The resulting version of the PSAFU scale had 21 items within four scales and it reached a satisfactory level for most fit indices (Supplementary material, Table 1   CFA model diagram for the data from the UK

MG-CFA of PSAFU-R Scale
The final model was achieved in CFA analysesthe model with correlated residuals was tested for multi-group invariance. The difference between the configural and metric model did not exceed the recommended criteria, but between metric and scalar did (see Table 2). This indicated that the fit had deteriorated to an extent in that the scalar model fits worse than the metric model. Muthén and Muthén (2009)   were used in all further analyses.

Predictive and Discriminant Validity of PSAFU-R Dimensions Across Countries
To explore predictive validity of PSAFU-R dimensions, we regressed them on to the measures of the intensity of FB use and social capital, i.e. to the time spent on FB on a daily basis, the frequency of posts, and the number of FB friends. The most consistent predictor of both the time spent daily on FB and the frequency of posting across five countries was addiction.
Socialization also predicted frequency of posting in Italy, Croatia, and Serbia. When it comes to the number of FB friends, none of the PSAFU-R dimensions predicted the criterion in all countries. Compensation was related to the number of FB friends in Italy, Croatia and the UK, and socialization was related to the same criterion in Iran, Italy and Serbia. Other predictors that contributed to the three criterion variables in individual countries could be seen in the Table 3.  Table  4 − by country. They revealed that compensation was uniquely related to introversion (in all countries) and higher neuroticism (in Iran, UK, and Serbia). Self-presentation was predicted by higher neuroticism and narcissism (in all countries except Croatia). Similarly, the most consistent predictor of socialization was narcissism (in all countries except Croatia) and neuroticism (in Croatia, UK and Serbia). Socializing on FB was also more pronounced among males (in all countries except UK) and younger people in Italy, UK and Serbia, but among older people in Iran. Finally, as with self-presentation and socialization, addiction was predicted by narcissism (in all countries except Croatia) and neuroticism (in Croatia, UK, and Serbia). Addiction was also more pronounced among men (in all countries except UK) and younger people (in Italy, UK and Serbia). Other predictors of PSAFU-R dimensions which were less consistent across countries can be seen in Table 4.

Table 4
Regression analyses predicting PSAFU-R dimensions based on the personality traits across five countries

Discussion
The main focus of this study was to examine the applicability of the Psycho-Social However, scalar invariance was not obtained, suggesting that intercepts of the scales could vary in different countries and thus, the scale should not be used for mean-level cross-cultural comparisons of FB users' behaviors. The lack of scalar invariance could be attributed to few possible causes. First, it might be due to the differences in the samples from the different countries, primarily with regard to gender and age structure, which varied significantly. Creating a uniform sample is an endeavor for the future studies of cross-cultural invariance of PSAFU-R.  Table  4 (https://www.pcb-3d.com/wordpress/wp-content/uploads/Supplementary%20materials.pdf). Therefore, caution is needed when interpreting these findings.
Our final step was to examine discriminant validity i.e. which psycho-social aspects of FB use are universally related to personality characteristics and which of these relations could be interpreted in terms of cross-cultural similarities vs. specificities. In line with the compensation theory and with our hypothesis, compensation is most consistently related to introversion and neuroticism in all countries. Therefore, the typical compensating user profile is a less sociable individual who has fewer social contacts and high emotional reactivity, which greatly confirms the profile described by Bodroža & Jovanović (2016).
Self-presentation, socialization and FB addiction revealed a similar pattern of relationships with personality traits. Higher narcissism and higher neuroticism seem to motivate self-presentational concerns and a need for establishing social contacts in a FB setting, as well as prolonged and addictive FB use. These results are consistent with the findings of several previous studies (Mehdizadeh, 2010;Michikyan et al., 2014;Ong et al., 2011;Seidman, 2013).
Since (grandiose) narcissism and neuroticism are not usually associated (e.g., Miller et al, 2011), it is more probable that these results refer to two different FB personality profiles. Narcissistic personality profile suggests that these individuals use online settings to acquire more social interactions i.e. they are getting socially richer from using the Internet (Rich-Gets-Richer theory; Kraut et al., 2002). However, it seems that getting "rich" doesn't necessarily have to be interpreted in terms of social contacts, but can also be in terms of general self-image enhancement. Making new FB friends means a greater number of connections and these connections may empower a person and make them feel worthier, especially when effective selfpresentation tactics are enacted. On the other hand, individuals characterized by high neuroticism seem to use FB friendships and opportunities for impression management to compensate personal insecurities, which is in line with Poor-Gets-Richer i.e. compensation theory (Kraut et al., 2002 Table  6, https://www.pcb-3d.com/wordpress/wpcontent/uploads/Supplementary%20materials.pdf) revealed that gender differences in FB addiction in most countries did not exist and where they did, their direction was not consistent.
There are a number of limitations of this study that need to be addressed. In order to control the sample variability, but also due to its availability, we opted for student samples in all countries. Thus, conclusions from this study are not entirely generalizable to the wider population. Also, although in Iran students were invited to participate in the study, this sample is somewhat different in structure and characteristics. Data obtained on the Iranian sample showed a weaker fit to the scale structure, but it is hard to conclude whether this was the consequence of the specificity of the sample, of cultural factors or of something else. Further studies should investigate these concerns. The distribution of gender was also not controlled, which resulted in a predominantly female samples in Croatia, Serbia, and the UK. In our study, we relied on the selfreported data on features of FB use, which were shown to be only moderately valid in comparison to objective data from client log files (Johannes et al., 2021;Scharkow, 2016). Therefore, whenever it is possible, objective data should be used. In the future, it would be interesting to examine the psycho-social aspects of FB use on samples from a more general and more representative population. As previously mentioned, one possible reason for the lack of scalar invariance of PSAFU-R scale might be due to inadequate rigor around translations of the scale. It is possible that some items might have been interpreted differently in different cultural contexts. Thus, translations used in this study should be additionally analyzed and, if needed, refined.
To conclude, this study shows that the PSAFU-R scale can be used in Croatian, Italian, Iranian, Serbian, and the UK cultures for studying cross-cultural similarities and differences in Facebook use. Authors working in this field of study could take this instrument into consideration when designing their studies.   Appendix C

New version of the scale Psycho-Social Aspects of Facebook Use
The list of PSAFU items that were retained in the final model (C-compensation, SP-selfpresentation, SOC-socialization, A-addiction). Numbers of the items correspond to the item numbers in the long version of the scale (Bodroža & Jovanović, 2016) and also to the numbers on the Figure 1   CFA model diagram for the data from the UK