The current study received ethical approval from the ethics committee for the Faculty of Psychology (ref: 315) Ruhr University Bochum, Germany. The data for this current study were taken from Bochum Optimism and Mental Health program (BOOM), aimed to explore the positive and negative factors of mental health across different countries. Pakistani University students fluent in English, aged 18 to 50 were recruited from different universities in Pakistan. Participants gave written informed consent for contribution in the study after being assured about anonymity and confidentiality issues. German students took part in the study from October 2011 to December 2015, whereas Pakistani students took part between May 2016 and March 2017. A total of 1841 Pakistani and 7890 German students completed the survey. Demographic information included age, gender, discipline, faculty, and monthly income. Participant characteristics are presented in Table 1.

Valid versions of the following scales already exist in English, which is the official language in Pakistan and is the medium of instruction in schools. Therefore, we applied the English version of these scales for Pakistani students. The internal consistency of all scales in both samples are presented in Table 2.

Positive Mental Health Scale (PMH-scale; German original version: Lukat, Margraf, Lutz, van der Veld, & Becker, 2016). The PMH-scale was used to measure psychological and subjective aspects of well-being. This instrument includes nine items rated on a 4-point Likert scale ranging from 0 (I disagree) to 3 (I agree). Previous research revealed good psychometric properties of the PMH-Scale such as Cronbach’s alpha α = .93. The English version of the PMH-scale was translated from the original language version by using the translation-backtranslation-modification procedure (Berry, 1989).

Satisfaction with Life Scale (SWLS; English original version: Diener, Emmons, Larsen, & Griffin, 1985; German version: Glaesmer, Grande, Braehler, & Roth, 2011). Life satisfaction was assessed with the well-established SWLS that includes five items rated on a 7-point Likert scale ranging from 1 (Strongly disagree) to 7 (Strongly agree). Previous research revealed good psychometric properties in both languages (German version: Cronbach’s alpha α = .92, English version: Cronbach’s alpha α = .87).

Resilience Scale (RS-11; German original version: Schumacher, Leppert, Gunzelmann, Strauss, & Brähler, 2005; English version: see Brailovskaia, Schönfeld, Kochetkov, & Margraf, 2017). The RS-11 was used to measure resilience. This instrument includes eleven items rated on a 7-point Likert scale ranging from 1 (Disagree) to 7 (Agree). Previous research reported good psychometric properties for the RS-11 in both investigated languages: (German version: Cronbach’s alpha α = .91and English Version Cronbach’s alpha α = .80).

Brief Form of Perceived Social Support Questionnaire (F-SozU K-14; German original version: Fydrich, Sommer, Tydecks, & Brähler, 2009; English version: see Brailovskaia et al., 2017). The F-SozU K-14 assessed perceived anticipated and subjective support from one’s social network. This instrument consists of 14 items rated on a 5-point Likert scale ranging from 1 (Not true) to 5 (True). Previous research reported good psychometric properties for the F-SozU K-14 in both investigated languages (German version: Cronbach’s alpha α = .94 and English version: Cronbach’s alpha α = .87).

Depression Anxiety Stress Scales (DASS; English original version: Lovibond & Lovibond, 1995; German version: Henry & Crawford, 2005). The well-established DASS was used to assess depression, anxiety, and stress symptoms. This instrument consists of three subscales (Depression, Anxiety, and Stress). Each subscale includes seven items. The in total 21 items are rated on a 4-point Likert scale ranging from 0 (Did not apply to me at all) to 3 (Applied to me very much or most of the time). Previous research reported good psychometric properties for the DASS in both investigated languages (German version: Cronbach’s alpha of Depression = .88, Anxiety = .76, and Stress = .86, English version: Cronbach’s alpha of Depression = .88, Anxiety = .82, and Stress = .90).

Family affluence scale (FAS-II). The FAS-II measures family wealth with four questions asking the number of vehicles, computers, child’s own bedroom, and holidays of the family. The FAS-II was developed by WHO Health Behavior in School-aged Children Study and validated across 35 countries.

Data were checked for missing values and response sets. Psychometric properties including mean, standard deviation, skewness, and kurtosis were analyzed. Internal consistency was measured by Cronbach’s alpha. Correlation between PMH, RS-10, F-SozU k-14, SWL scales and negative mental health scales (Depression, anxiety, and stress) were calculated to measure construct validity. The SPSS (version 25.0) was used for the above analyses.

Single-group confirmatory analysis (CFA) was tested (Hirschfeld & Von Brachel, 2014) to examine the proposed factor structure of PMH, RS-10, F-SozU K-14, and SWL scales across both countries. Within each sample, one-factor model was defined for testing. Maximum likelihood estimators were used for model testing. Given that we have a large sample size (>5000), we used a combination of goodness of fit as recommended by Marsh, Hau, and Wen (2004) and by Hu and Bentler (1999). A model was accepted if Standardized Root Mean Square (SRMR) <.08, Comparative Fit Index >.90, Tucker Lewis index (TLI) > .95 and Root Mean Squared Error of Approximation (RMSEA) < .05. Maximum likelihood estimators were used for model testing.

Furthermore, Jöreskog (1971) multigroup factor analysis approach was used to test measurement invariance of these scales using R-package (Lavaan) (for more detail see: Hirschfeld & Von Brachel, 2014). Configural invariance is analyzed with no equality constraints across groups and confirms whether factor structures are identical across countries. In testing for metric invariance, factor loadings were constrained to be equal across both countries. All intercepts were constrained to be equal across both countries. For the reference country, factor mean was set to 0 and scaling factor to 1. If intercept measurement was established, then scalar measurement could be expected (Millsap & Yun-Tein, 2004). Each model was accepted when (a) CFI, RMSEA and SRMR values showed good model fit, and (b) ΔCFI in association to the previous model was not greater than .01 (Cheung & Rensvold, 2002). If full invariance was not established, partial weak and partial strong invariance was accepted after estimating freely variant item’s factor loading or intercept (Byrne, Shavelson, & Muthén, 1989).

Psychometric properties of each scale such as mean, standard deviation, skewness, kurtosis and internal consistency are demonstrated in Table 2. The values of skewness and kurtosis for both Pakistani and German samples are near to 0 (e.g., SK < 2 and Ku < 5), thus we treat all the scales as continuous (Muthén & Muthén, 2009). All variables demonstrated acceptable normal univariate distributions (George & Mallery, 2010) and had slight right-tailed distribution on PMH, and RS-11.

The results of the single group CFAs for PMH, RS-10, F-SozU k-14, and SWL scales of each sample are presented in Table 3–6, respectively. The model fit of the CFA of PMH and SWLS for both samples indicated that the one-factor model was a goot fit of the data from both countries. Meanwhile, the model fit of the one-factor model of RS-11 and F-SozU K-14 was modest to poor for both countries, although all factor loadings in both samples were above 0.35.

For RS-11, modification indices showed that there were many correlated error terms between Item 3 (Keeping interested in things is important to me) and Item 7 (I keep interested in things), Item 10 (Sometimes I make myself do things whether I want to or not) and Item 11 (I have enough energy to do what I have to do), as well as Item 8 (I can usually find something to laugh about) and Item 9 (I can usually look at a situation in a number of ways). All these items have similar wording and deal with topics of interest, ability to do tasks and seeing things from different perspectives. After correlating terms model fit improved.

For SWLS, modification indices showed a correlated error term between item 2 (The conditions of my life are excellent.) and item 3 (I am satisfied with my life.). Both items have similar wording and theme of satisfaction with living conditions.

Correlation matrixes were investigated between positive constructs (i.e., PMH, RS-11, F-SozU K-14, and SWLS) and negative mental health factors (Depression, Anxiety, and Stress). Within each sample, the mean, SDs, and correlations of these constructs with the total score of each instrument respectively are presented in Table 2. All correlations were statistically significant and in the anticipated direction; positive mental health constructs were negatively associated with depression, anxiety and stress, whereas positively associated with each other.

The results of the measurement invariance using multi-group CFAs for PMH indicated good invariance across both countries (Table 3). Structural (configural), metric (weak) and scalar (strong) invariance showed a good model fit as well as CFI, which suggests that the PMH has same structure, factor loadings and intercepts in both countries, therefore each item contribute equally in measuring the latent variable.

Table 4 shows model testing results of measurement invariance for resilience. The configural model and weak invariance model both showed an acceptable global fit. Next, the test of strong measurement invariance indicated poor model fit, as CFI exceeded .01. Expected parameter change suggested higher parameter changes when Item1 and Item 2 were estimated freely. By setting free the constraint of Item 1 (When I make plans, I follow through with them) and Item 2 (I usually manage one way or another), the CFI value was less than 0.01. Thus, partial strong measurement invariance was established.

Table 5 presents the model testing results of measurement invariance of F-SozU K-14. The global fit of the configural model and weak invariance model were both acceptable. The drop of the CFI from the full strong invariance model to weak invariance model exceeded 0.01 compared to the weak model. Expected parameter change statistics suggested that factor loadings of Item 1 (I can easily find someone who will look after my apartment when I am not there), Item 2 (There are people who take me as I am, without limitations), Item 11 (There are people who share joy and sorrow with me) and Item 12 (With some friends and family I can really let loose) were not equal across both countries (Kaplan, 1989). After releasing constraints of these items, partial strong measurement invariance was established.

Table 6 displays the model testing results of measurement invariancefor the SWLS. In testing measurement invariance, configural, weak and strong invariance assumption for the SWLS was supported, and the global fit of this model remained good.

After fulfilling the precondition of latent means, latent mean comparisons of PMH, RS-10, F-SozU k-14, and SWLS were investigated between Pakistani and German university students. Pakistani sample was used as reference group. All mean differences between Pakistan and Germany were significant apart from RS-11 (z = −1.66, p = .096, d = −0.77). The German sample had significantly higher latent means of PMH (z = 3.94, p < .001, d = 0.058), F-SozU K-14 (z = 25.50, p < .001, d = 0.38) and SWLS (z = 0.74, p < .001, d = 0.99) than the Pakistani sample.

The measurement invariance testing for PMH demonstrated full strong measurement invariance, suggesting that the score of PMH can be used in direct mean comparison across cultures. Moreover, latent mean comparison indicates that Pakistani students experience less positive mental health. Due to poor educational system and financial crises, Pakistani students experience emotional and educational challenges such as relationship problems, competition for good grades, uncertainty regarding the future, which badly influences their psychological well-being. These findings are consistent with the previous studies of Bibi, Blackwell, and Margraf (2019) and Irfan (2016), who reported that Pakistani students face many challenges during their university life and experience various psychological issues.

The Resilience scale showed a modest fit with the original one-factor model in both countries. One possible reason for the poor model fit of RS-11 in the German sample could be that the item selection process for RS-11 was established according to statistical criteria rather than a theoretical concern. Furthermore, partial strong measurement was established by releasing the intercept of Item 1 (follow through with plan) and Item 2 (manage one way or another). Both items were more strongly associated with the trait resilience in the German sample as compared to the Pakistani sample. A possible explanation could be that in Pakistan people usually do not plan their work and routines as much as people in Germany, therefore may fail to manage their life in one way or another. Therefore, these items do not fit the circumstances in Pakistan. Our findings of measurement invariance of resilience are in contrast with Yang, Li, and Xia (2012), who found only four of 14 items invariant across U.S. Chinese and Taiwanese students. Therefore, cross-cultural comparisons should be interpreted with caution. In addition, no significant differences were found for resilience latent mean comparison, indicating that both German and Pakistani students have an equal ability to cope with adverse situations of life.

For social support, partial strong measurement invariance was established with the freely estimating of the intercepts of Item 1 (easily find someone to look after apartment), Item 2 (people who take me as I am) Item 11 (people to share joy and sorrow with), and Item 12 (let loose with people). The intercepts of all these items were higher in the German sample apart from Item11. This could be because F-SozU K-14 is developed in German cultural framework and put emphasis on instrumental social support, therefore, these items may not fit the culture of Pakistan. However, in Pakistan people have strong social bonds and extended family members usually live together. They have less instrumental social support and selflessly support each other (Jibeen, 2016). Results also indicated that the mean scores of German students were higher than Pakistani students. However, effect sizes were small, therefore these results should not be over-interpreted.

For life satisfaction, strong measurement invariance was established as well. Results of the latent mean comparison of the four variant items indicated that German students scored higher on life satisfaction compared to Pakistani students. It could be because Pakistani students experience more financial, academic, family and political issues, which influence their quality of life and thus lowers their life satisfaction. Results are in line with previous studies which found that western students experience less academic stress than Asian students, who as a result experience less life satisfaction (Nguyen, Le, & Meirmanov, 2019; Xiong, 2018).

In spite of the strengths of the present study such as the large sample size and the cross-cultural setup, the following limitations should also be considered. First, a convenient purposive sampling technique was used to collect the data, which may not be truly representative of the populations. Second, the justification for nonequivalent intercepts of resilience, social support, and life satisfaction scales are still unclear, since traditional or methodological factors might have contributed to the items’ non-invariance (Van de Vijver, 2007). Future studies may use focused group research to explore the potential sources of biases. Third, though latent means comparison was based on only non-biased items which could underrepresent the construct. There are many differences between the two countries that may explain the variances we found in the current studies, such as individualistic and collectivistic cultural, developing and developed status, economic status, and religious. Future studies may further investigate whether any of those factors can explain the variances between countries.