Researchers typically conceive of impulsivity as a multidimensional construct. Numerous conceptualizations of impulsivity have been proposed, which are rooted in different personality theories (Dickman, 1990; Eysenck & Eysenck, 1977; Zuckerman et al., 1993; for an overview, see Whiteside & Lynam, 2001). However, the present section will focus only on the conceptualization of impulsivity proposed by Whiteside and Lynam (2001), as the authors comprehensively integrated existing conceptualizations of the construct and based their model of impulsivity on the most prominent model of personality, the Big Five (McCrae & Costa, 1990).

Following Whiteside and Lynam (2001), four distinct personality facets that lead to impulsive behavior can be distinguished: (1) urgency, that is “a tendency to commit rash or regrettable actions as a result of intense negative affect” (p. 677); (2) lack of premeditation, that is, acting without careful thinking and planning and without reflecting on the consequences of an act; (3) lack of perseverance, that is, the inability to remain focused on boring or difficult tasks or to ignore distracting stimuli; and (4) sensation seeking, that is, a tendency to engage in and enjoy exciting and risky activities.

These four facets are related to some of the Big Five personality traits (Goldberg, 1992): Urgency encompasses neurotic aspects. Negative emotions, as part of the domain of Neuroticism, often result from impulsive actions that ignore long-term negative consequences. Individuals high in premeditation are thoughtful and rational; individuals high in perseverance have great self-discipline—dispositions that characterize the Big Five domain of Conscientiousness. Moreover, individuals who are high in sensation seeking engage in risky and exciting activities—a disposition that characterizes the domain of Extraversion (Keye et al., 2009; Whiteside & Lynam, 2001).

The four personality facets of impulsive behavior (i.e., urgency, lack of premeditation, lack of perseverance, and sensation seeking) can have several adverse consequences. For example, individuals high in urgency, lacking premeditation and perseverance, and seeking sensation are more likely to engage in negative and harmful behaviors such as problematic gambling (Nower et al., 2004). Notably, certain personality facets specifically drive different kinds of (impulsive) behavior: A high lack of premeditation is strongly related to high delinquency (Miller et al., 2003; White et al., 1994) and drug use (Miller et al., 2003). High urgency is strongly associated with self-injury behavior (Glenn & Klonsky, 2010), and high sensation seeking is associated with alcohol abuse (albeit only in individuals who are high in antisocial personality traits; Whiteside & Lynam, 2003; Whiteside et al., 2005).

The four personality facets associated with impulsive behavior are also closely linked to several (sub)clinical and pathological cognitions and behaviors. For instance, because high urgency is an expression of the inability to regulate and cope with negative emotions, it is strongly associated with forms of psychopathology such as borderline personality disorder or pathological gambling (Whiteside et al., 2005). Persons with attention deficit hyperactivity disorder (ADHD) typically have a high lack of premeditation (Miller et al., 2003).

To date, researchers have mainly investigated the associations between (sub)clinical and pathological cognitions/behaviors and urgency, lack of premeditation, lack of perseverance, and sensation seeking. By contrast, non-clinical associations have largely been ignored so far and thus must be investigated in more detail.

Assessing impulsivity: The four-factor model

To assess the four personality facets related to impulsive behavior, Whiteside and Lynam (2001) developed the 45-item Urgency, Premeditation, Perseverance, and Sensation Seeking (UPPS) Impulsive Behavior Scale. The UPPS scale includes items from various other impulsivity scales (e.g., Dickman’s Functional and Dysfunctional Impulsivity Scales; Dickman, 1990; see also Zuckerman, 1994) and from personality inventories (e.g., Revised NEO Personality Inventory; NEO-PI-R; Costa & McCrae, 1992). A series of exploratory factor analyses using data from the NEO-PI-R, eight widely used impulsivity measures, and additional “impulsiveness” items created by the investigators identified four personality facets that lead to impulsive behavior: urgency, lack of premeditation, lack of perseverance, and sensation seeking (Whiteside & Lynam, 2001). These four facets, which had previously been lumped together under the umbrella term “impulsivity,” were found to be theoretically distinct (Whiteside & Lynam, 2001).

Several studies that tested the four-factor model of impulsivity provided evidence that the scale does indeed capture four distinct personal facets, and that these facets have different intercorrelations (Schmidt et al., 2008; Whiteside & Lynam, 2003; Whiteside et al., 2005). Whereas lack of premeditation and lack of perseverance were found to be substantially correlated (r = .45), all other intercorrelations were substantially lower (.00 ≤ |r| ≤ .29; d’Acremont & Van der Linden, 2005; Whiteside & Lynam, 2001). The rank order of the intercorrelations did not hold in every sample (which is unsurprising given the many low intercorrelations). For example, in a sample of 20-year-olds, Miller et al. (2003) replicated a strong correlation between lack of premeditation and lack of perseverance (r = .50), and they found similarly large correlations between these two facets and urgency (r = .56 and r = .41, respectively). Keye et al. (2009) replicated Miller et al.’s (2003) findings in a German student sample. Correlations between lack of premeditation and lack of perseverance seemed to be generally strong across all of the aforementioned studies (d’Acremont & Van der Linden, 2005; Keye et al., 2009; Miller et al., 2003; Whiteside & Lynam, 2001).

Although the 45-item UPPS scale (Whiteside & Lynam, 2001) is frequently used in research on impulsivity, it is too long for many surveys—especially for multi-thematic surveys (e.g., the German Socio-Economic Panel, SOEP) in which questionnaire space is very limited.

Item generation and selection

To enable researchers to survey impulsive behavior in contexts with limited time and resources, Kovaleva et al. (2014a) developed the German-language Skala Impulsives-Verhalten-8 (I-8). The I-8 scale captures all four factors of the UPPS model with just two items each. The construction of I-8 was based mainly on two German-language versions of the UPPS scale developed by Keye et al. (2009) and Schmidt et al. (2008). To construct the subscales urgency, lack of premeditation, and lack of perseverance, Kovaleva et al. (2014a) used the results of Keye et al.’s (2009) and Schmidt et al.’s (2008) factor analyses. They selected two items per subscale based on the level and stability of factor loadings across studies and theoretical aspects. To construct the subscale sensation seeking, they first created six items based on the UPPS items (Whiteside & Lynam, 2001). These six items were later reduced to two based on factor loading patterns and item total correlations in an initial construction sample (Kovaleva et al., 2014a). The preliminary items of I-8 were cognitively pretested and subsequently modified based on the pretest results. Kovaleva et al. (2014a) psychometrically validated the resulting I-8 in three large independent quota and random samples representing the adult population in Germany (aged 18 and older). The four-factor structure was corroborated, and the German-language I-8 was shown to be a psychometrically valid and reliable scale to measure traits that lead to impulsive behavior.

To enable I-8 to be used in English-language research, Kovaleva et al. (2014a) adapted the scale to English. When doing so, they followed the two-step procedure recommended by the International Test Commission (2010; see Kovaleva et al., 2014a), which aligns with the TRAPD (Translation, Review, Adjudication, Pretesting, and Documentation) approach (Harkness, 2003): First, two professional translators (English native speakers) translated the items independently. One translation was done in British English, the other in American English. In the second phase of the adaptation process, a reconciliation meeting was held during which the proposed adaptations were discussed and revised in a group comprising experts on the psychological characteristics in question, the translators, and an expert on questionnaire adaptation.

However, the testing of the psychometric quality of the English-language adaptation of I-8 has remained a desideratum until now. The present study aims to fill this gap and, when doing so, to compare the English-language version of I-8 with the German source version and test the measurement invariance of the scale across the two languages. Measurement invariance is an important prerequisite for valid cross-cultural comparisons (Chen, 2008).

Samples

To establish the psychometric properties of the English-language adaptation of I-8, to compare them with those of the German-language source instrument, and to test the measurement invariance of the scale across the two languages, the respective versions were administered to respondents in a web-based survey conducted in parallel in the UK and Germany in January 2018 by the panel provider respondi AG using computer-assisted self-administered interviewing (CASI).

For both nations, quota samples were drawn that represented the heterogeneity of the adult population in terms of age, sex, and educational attainment. The basis for the quotas for the UK and Germany was the latest 2011 German census (https://ergebnisse.zensus2011.de). The purpose of the research¾to investigate the quality of several questionnaires¾was explained to all respondents. Only native speakers of the respective languages were recruited. All respondents consented to participation in an anonymous online survey and were financially rewarded for their participation. A subsample was reassessed after approximately 3 to 4 weeks (median time intervals: 28 days in the UK and 20 days in Germany). According to the local legislation and requirements of the institution, our study on human participants did not require review and approval by an ethics committee as we collected data without any reference to the participants’ identity. The data collection was completely anonymous. We adhered to ethical standards comparable to the 1964 Declaration of Helsinki.

Table 2

Sample Characteristics by Nation

Note. The educational attainment levels were as follows: Low (UK) = Never went to school/Skills for Life/1–4 GCSEs A*–C or equivalent; Low (Germany) = No educational qualifications/lower secondary leaving certificate (ohne Bildungsabschluss/Hauptschulabschluss); Intermediate (UK) = Five or more GCSEs A*–C/vocational GCSE/GNVQ intermediate or equivalent; Intermediate (Germany) = Intermediate school leaving certificate (mittlerer Schulabschluss); High (UK) = Two or more A-levels or equivalent; High (Germany) = Higher education entrance qualification (Fachhochschulreife/allgemeine Hochschulreife).

Only those respondents who did not abort the survey prematurely were included in our analysis. This resulted in a gross sample size of N_UK = 508 (retest: N_UK = 117) for the UK and N_DE = 513 (retest: N_DE = 125) for Germany. To ensure high data quality, we applied three criteria simultaneously. The first criterion was the ipsatized variance, that is, the within-person variance across items (Kemper & Menold, 2014). We excluded respondents with an ipsatized variance below 5%. The second criterion was the Mahalanobis distance of the respondent’s response vector from the average sample response vector (Meade & Craig, 2012). Respondents within the upper 2.5% of the sample distribution of the Mahalanobis distance were excluded. The third criterion was the response time. Respondents who answered all items in less than 1 s per item, on average, were excluded. We chose relatively liberal heuristics for exclusion criteria to avoid erroneously excluding valid cases and thereby creating systematic bias in our data. These quality checks resulted in the exclusion of 8% of cases in both the UK and the German subsamples. The final net sample sizes were N_UK = 468 (retest: N_UK = 111) and N_DE = 474 (retest: N_DE = 117). Table 2 details the sample characteristics and their distribution. The target and real sample sizes per quota are listed in Table 3.

Table 3

Quotas: Target and Actual Sample Sizes for the UK and Germany

Note. The educational attainment levels were as follows: Low (UK) = Never went to school/Skills for Life/1–4 GCSEs A*–C or equivalent; Low (Germany) = No educational qualifications/lower secondary leaving certificate (ohne Bildungsabschluss/Hauptschulabschluss); Intermediate (UK) = Five or more GCSEs A*–C/vocational GCSE/GNVQ intermediate or equivalent; Intermediate Germany) = Intermediate school leaving certificate (mittlerer Schulabschluss); High (UK) = Two or more A-levels or equivalent; High (Germany) = Higher education entrance qualification (Fachhochschulreife/allgemeine Hochschulreife). The target n per quota was calculated based on the latest German census (2011; https://ergebnisse.zensus2011.de).

The real n per quota was calculated after data cleansing.

Material

We validated I-8 by evaluating its nomological network, which consists of a broad range of non-clinical correlates. Whereas the associations between the I-8 factors and (a) the Big Five domains of personality were clearly established within the framework of the development of the German-language scale (Whiteside & Lynam, 2003), associations with other individual-difference constructs were less clear and had not been established to date. Based on the definition of the four factors of the I-8 scale, we tentatively assumed the following: (b) Self-esteem is independent of the four factors; (c) higher self-efficacy is associated with a higher level of premeditation, perseverance, and sensation seeking; (d) higher (internal) locus of control is related to lower urgency, lack of premeditation, lack of perseverance, and sensation seeking; (e) higher life satisfaction is associated with lower urgency and with a higher level of premeditation (Ravert & Donnellan, 2020); (f) high risk proneness is related to high sensation seeking. We assumed correlations with (g) general health to be minor, acknowledging that higher urgency, lack of premeditation, lack of perseverance, and sensation seeking are related to many psychological disorders but not to physiological disorders. We further assumed that (h) socially desirable responding as a form of response bias is present when answering the I-8 items. Individuals commonly perceive impulsive behavior—and thus also personality facets that lead to such behavior—to be socially undesirable (as evidenced indirectly by the fact that the facets are associated with harmful behaviors). Accordingly, the respective language versions of the following short scales were also administered as part of the survey:

1. The extra-short form of the Big Five Inventory–2 (BFI-2-XS; English version: Soto & John, 2017; German version: Rammstedt et al., 2020)
2. The Rosenberg Self-Esteem Scale (RSES; English version: Rosenberg, 2014; German version: von Collani & Herzberg, 2003)
3. The General Self-Efficacy Short Scale–3 (GSE-3; Doll et al., 2021)/Allgemeine Selbstwirksamkeit Kurzskala (ASKU; Beierlein et al., 2014)
4. The Internal–External Locus of Control Short Scale–4 (IE-4; Nießen et al., 2022)/Internale–Externale-Kontrollüberzeugung–4 (Kovaleva et al., 2014b)
5. The General Life Satisfaction Short Scale (L-1; Nießen, Groskurth, et al., 2020a)/Kurzskala zur Erfassung der Allgemeinen Lebenszufriedenheit (Beierlein, Kovaleva, László, et al., 2015)
6. The Risk Proneness Short Scale (R-1; Nießen, Groskurth, et al., 2020b)/Kurzskala zur Erfassung der Risikobereitschaft (Beierlein, Kovaleva, Kemper, & Rammstedt, 2015)
7. The single-item European Social Survey (ESS) question measuring self-reported general health (ESS, 2016)
8. The Social Desirability–Gamma Short Scale (KSE-G; Nießen et al., 2019)/Kurzskala Soziale Erwünschtheit–Gamma (Kemper et al., 2014)

And finally, we also investigated the associations between respondents’ sociodemographic characteristics and different values on the four factors of I-8. Thus, the survey included several sociodemographic variables: employment status, income, educational attainment, age, and sex. Employment status was surveyed with the following nominal categories: (1) employed, (2) self-employed, (3) out of work and looking for work, (4) out of work but not currently looking for work, (5) doing housework, (6) pupil/student, (7) apprentice/intern, (8) retired, and (9) none of the above. We recoded the variable into (1) unemployed (i.e., out of work and looking for work or out of work but not currently looking for work), and (2) employed (i.e., employed or self-employed). All remaining categories were regarded as missing values.

Item analyses

I-8 conceptualizes four distinctive personality facets related to impulsive behavior—namely, urgency, lack of premeditation, lack of perseverance, and sensation seeking. I-8 should reflect these four facets with two items each. We tested this assumption by fitting a four-dimensional confirmatory factor analysis (CFA) model. The factors were allowed to covary. We assessed model fit via heuristics for fit indices. In line with common guidelines, we judged model fit as adequate if the confirmatory fit index (CFI) was .950 or higher, the root mean square error of approximation (RMSEA) was .060 or lower, and the standardized root mean residual (SRMR) was .080 or lower (Hu & Bentler, 1999). We fixed the first loading of each factor to one and the first intercept to zero to identify the model. We estimated our model with robust maximum likelihood estimation (MLR).

The proposed four-dimensional model fit well in both nations—UK/Germany: χ²(14) = 13.29/23.49, p = .504/053, CFI = 1.000/.992, RMSEA = .000/.038, SRMR = .016/.022, Bayesian information criterion (BIC) = 9,695/8,881. Brosseau-Liard et al. (2012) and Brosseau-Liard and Savalei (2014) found that applied robust corrections of the fit indices using MLR were not theoretically justified. Thus, they supported new corrections, which were implemented as so-called robust CFI and robust RMSEA values in R/lavaan (UK/Germany: robust CFI = 1.000/.993, robust RMSEA = .000/.039). As the proposed four-dimensional model fit exceptionally well, we tested a more restrictive model in which we imposed equality constraints on the factor loadings of the two items of each subscale/latent variable. The proposed four-dimensional model, which has freely estimated factor loadings, is called “tau-congeneric.” In contrast, the more restrictive model is called “essentially tau-equivalent” because it has identical factor loadings for all indicators (Graham, 2006). The essentially tau-equivalent model fit as well as the congeneric one; thus, we accepted the former instead of the latter—UK/Germany: χ²(18) = 15.11/25.77, p = .654/.105, CFI = 1.000/.993, robust CFI = 1.000/.994, RMSEA = .000/.030, robust RMSEA = .000/.032, SRMR = .016/.026, BIC = 9,672/8,859. Figure 1 shows the final essentially tau-equivalent measurement model.

Figure 1. Four-factor essentially tau-equivalent measurement model of I-8 with standardized coefficients. We omitted residual terms for clarity. NUK = 468; NDE = 474.

In the essentially tau-equivalent model shown in Figure 1, the correlations of the subscales were as follows: We observed the strongest correlation between lack of premeditation and lack of perseverance (UK/DE: r = .66/.61), which is in line with earlier research (d’Acremont & Van der Linden, 2005; Keye et al., 2009; Miller et al., 2003). Other correlations between subscales were lower. Urgency correlated positively with lack of premeditation (UK/DE: r = .26/.44) and lack of perseverance (UK/DE: r = .08/.33). Urgency also correlated positively with sensation seeking (UK/DE: r = .50/.25). The correlation between lack of premeditation and sensation seeking was comparatively small (UK/DE: r = −.07/.07). Lack of perseverance correlated negatively with sensation seeking (UK/DE: r = −.26/−.16). Because theoretically (Whiteside & Lynam, 2001) and empirically there is no general construct of impulsivity, as evidenced by the substantially varying correlations across subscales, we do not recommend using a total scale score (index score) across all four factors. Rather, we recommend analyzing the scores of the subscales separately. Unit-weighted mean scores should be computed only for the individual subscales of impulsive behavior. Individual answers should be aggregated to the subscale level only if both items of each subscale were completed. If missing values occur, researchers should use appropriate methods for handling missing data, such as full information maximum likelihood estimation (FIML; Arbuckle, 1996), the pseudo-indicator method (PIM; Rose et al., 2019), or multiple imputation (Rubin, 1987).

Item parameters

Table 4 shows descriptive statistics¾the mean, standard deviation, skewness, and excess kurtosis of each of the eight items of I-8—for the English-language adaptation and the German-language source version. We found no substantial divergence of descriptive statistics across nations.

Table 4

Descriptive Statistics of the I-8 Items by Nation

0.45

0.23

−0.49

−0.21

0.59

−0.81

−0.20

Note. Response options ranged from 1 (does not apply at all) to 5 (applies completely). UK = United Kingdom (N = 468); DE = Germany (N = 474). (R) denotes items that were recoded.

Table 6

Overall, the correlations across nations were highly similar, as evidenced by the strong metacorrelations (i.e., the correlation of the correlations from the UK and Germany), which ranged from .79 to .95 (see the bottom row in Table 6). This implies that the nomological networks of all four dimensions were largely equivalent in both countries

Descriptive statistics

Table 7

Reference Ranges of the I-8 Scale Scores for the Total Population and Separately for Sex and Age Cohorts

−0.25

−0.02

−0.90

Note. Answer options ranged from 1 (does not apply at all) to 5 (applies completely). UK = United Kingdom (N = 468); DE = Germany (N = 474).

Table 7 provides detailed information on reference ranges and accompanying descriptive statistics of   I-8 for the total population and clustered by sex and age.

Further quality criteria

I-8 is an economical scale with a completion time of less than 1 min (estimated value).

We evaluated the comparability of I-8 across the UK and Germany via measurement invariance tests by means of multiple-group CFA (Chen, 2008; Steinmetz, 2013; Vandenberg & Lance, 2000; Widaman & Reise, 1997). We examined measurement invariance in a sequential fashion: First, we investigated configural invariance (same measurement model), then metric invariance (same factor loadings; required for comparing latent (co)variances) and scalar invariance (same item intercepts; required for comparing latent and observed means). We also investigated the invariance of residual variances (equal precision). We further investigated the structural invariance of latent variances (equal latent variances), latent covariances (equal latent covariances), and latent means (equal latent means). Cutoffs for fit indices helped us to evaluate the levels of measurement invariance reached. We applied the cutoffs proposed by Hu and Bentler (1999) to assess the absolute magnitude of fit indices. According to Hu and Bentler (1999), CFI should be below .950, RMSEA should be below .060, and SRMR should be below .080 to indicate a well-fitting model. We applied the cutoffs Chen (2007) recommended to evaluate the changes in fit indices, comparing a more restricted invariance level to a less restricted one. According to Chen (2007), metric non-invariance is given when CFI changes ≤ −.010 supplemented by changes of ≥ .015 in RMSEA or changes of ≥ .030 in SRMR, comparing the metric to the configural model. Scalar non-invariance (or non-invariance of residual variances) is given when CFI changes ≤ −.010 supplemented by changes of ≥ .015 in RMSEA or changes of ≥ .010 in SRMR, comparing the scalar to the metric model (or the invariance model of residual variances to the scalar model). We apply the latter cutoffs also for evaluating the structural non-invariance of latent variances, latent covariances, and latent means.

Invariance tests were based on the essentially tau-equivalent model identified by setting the loadings of each factor to one and first intercept to zero. Thus, the configural and metric models were equivalent.

Table 8 shows the fit of the different models. The fit indices suggested that the metric model had an acceptable fit to the data, indicating the comparability of the latent variances and covariances across the UK and Germany (Bluemke et al., 2016). However, when the scalar model was tested, the misfit induced by constraining the intercepts was not acceptable relative to the less restricted model.

When the full scalar model cannot be accepted, a so-called partial scalar model can be tested, thereby allowing non-invariant parameters to be freely estimated (Byrne et al., 1989; Steenkamp & Baumgartner, 1998). Thus, the partial scalar model has a common-group and a group-specific share of parameters. Non-invariant intercepts can be identified via modification indices (MI), showing the change in the χ² test statistic if the corresponding parameter were released (Yoon & Kim, 2014). The intercept with the largest modification index belonged to the first item of the urgency subscale (MI_Urgency1 = 24.40). Releasing that intercept resulted in a well-fitting and acceptable partial scalar model in both absolute and relative terms. More specifically, all subscales were scalar invariant, except for the urgency subscale, which was only metric invariant. The urgency subscale cannot even be deemed to be partial scalar invariant, as partial scalar invariance implies that at least two intercepts are invariant across groups (Byrne et al., 1989). The urgency subscale had only one invariant intercept. However, the intercept difference on the urgency subscale was relatively small (τ_{Item 1, UK} ‒ τ_{Item 1, DE} = ‒0.252; standardized difference). We thus tentatively conclude that not only the latent (co)variances, latent means, and observed means for lack of perseverance, lack of premeditation, and sensation seeking but also the ones for urgency can be compared across nations without incurring major bias, although care should be exerted when comparing observed means for urgency across nations (see also Steinmetz, 2013).

Table 8

Fit of Different Models Testing for Invariance.

Note. RMSEA = root-mean-square error of approximation; CFI = comparative fit index; SRMR = stan-dardized root mean residual; BIC = Bayesian information criterion. The configural invariance model is equivalent to the metric invariance model because the measurement invariance tests are based on the essentially tau-equivalent model.

^aThe intercept with the largest modification index belonged to the first item of the urgency subscale (MI_Urgency1 = 24.40).

Further, we tested the invariance of residual variances. The model with equal residual variances for the UK and Germany fit slightly worse than the partial scalar invariance model according to most fit indices (i.e., χ², CFI, RMSEA, SRMR, but not BIC). As only the ΔCFI cutoff was slightly exceeded (i.e., ΔCFI = ‒.012, which should be ΔCFI > ‒.010) and all other cutoffs were passed, we accepted the model with equal residual variances. When a model with equal residual variances for the UK and Germany fit, the precision across nations is the same (Steenkamp & Baumgartner, 1998). Observed (co)variances can be compared across the UK and Germany without inducing bias (Millsap & Olivera-Aguilar, 2012).

Additionally, we evaluated the structural invariance (Vandenberg & Lance, 2003) by setting the latent variances, latent covariances, and latent means equal across the UK and Germany. All these parameters can be compared across groups (as evaluated by the previous measurement invariance tests). We hereby evaluate a substantial question: Do the UK and Germany have equal latent variances, latent covariances, and latent means of the I-8 subscales urgency, lack of premeditation, lack of perseverance, and sensation seeking? To evaluate the substantial question, we restricted either the latent variances, latent covariances, or latent means of the four subscales to be equal across nations (in addition to the residual variances, intercepts, and loadings). Then, we compared the values of fit indices to those of the invariance model up to the level of residual variances. According to Chen’s (2007) cutoffs, the increase in misfit was substantial for all of these models. Thus, latent variances, latent covariances, and latent means of the I-8 subscales differ across the UK and Germany.

Further literature

The scale was first published in the journal PLOS ONE.

Acknowledgement

We would like to thank Melanie Partsch from GESIS – Leibniz Institute for the Social Sciences for supporting us in preparing the data. We further thank Constanze Beierlein from the University of Applied Sciences Hamm-Lippstadt for helping us to understand the development process of the I-8 scale in detail.

The study was funded by an internal grant provided by GESIS – Leibniz Institute for the Social Sciences.

Katharina Groskurth, GESIS – Leibniz Institute for the Social Sciences, P.O. Box 12 21 55, 68072 Mannheim, Germany; e-mail: katharina.groskurth@gesis.org

The dataset supporting the conclusions of this article is available in the GESIS SowiDataNet | datorium repository, https://doi.org/10.7802/2097.