General perceived self-efficacy pertains to optimistic beliefs about being able to cope with a large variety of stressors. In contrast to other constructs of optimism, perceived self-efficacy explicitly refers to one’s competence to deal with challenging encounters. It is measured with a parsimonious ten-item scale that was developed for use across cultures. The research question aims at the cross-cultural equivalence of multiple adaptations of this instrument. The present paper compares the language-specific adaptations that were examined in 14 cultures from all over the world. A total of 12,840 individuals responded to the instrument. The unidimensional nature of the scale was replicated in all samples, using reliability analyses as well as exploratory and confirmatory factor analyses. Self-efficacy differences between the 14 cultures are discussed.

The present article introduces the theoretical construct of self-efficacy and describes a brief scale designed to measure this construct at the level of a general personality disposition. The original German instrument has been proven reliable and valid in various field studies (Schwarzer, 1993; Schwarzer & Jerusalem, 1995). So far, the German, Chinese, and Spanish versions are well validated (Schwarzer, Bäßler, Kwiatek, Schröder, & Zhang, 1997; Zhang & Schwarzer, 1995). This paper deals with 14 cultures, comparing the psychometric properties for the German, English, Dutch, Spanish, Russian, Greek, Arabian, Hungarian, Polish, Chinese, Indonesian, Japanese, and Korean versions. The purpose of the study is twofold: to examine whether the theoretical construct of perceived self-efficacy is universal, and to attain psychometrically sound adaptations of the inventory that can be used with the above populations.

Perceived Self-Efficacy: Theory and Measurement

The construct of self-efficacy, which was introduced by Bandura, represents one core aspect of his social-cognitive theory (Bandura, 1977, 1997). While outcome expectancies refer to the perception of the possible consequences of one’s action, self-efficacy expectancies refer to personal action control or agency. A person who believes in being able to cause an event can conduct a more active and self-determined life course. This "can do"-cognition mirrors a sense of control over one’s environment. It reflects the belief of being able to control challenging environmental demands by means of taking adaptive action. It can be regarded as a self-confident view of one’s capability to deal with certain life stressors.

According to theory and research (Bandura, 1995), self-efficacy makes a difference in how people feel, think and act. In terms of feeling, a low sense of self-efficacy is associated with depression, anxiety, and helplessness. Such individuals also have low self-esteem and harbor pessimistic thoughts about their accomplishments and personal development. In terms of thinking, a strong sense of competence facilitates cognitive processes and performance in a variety of settings, including quality of decision-making and academic achievement. When it comes to preparing action, self-related cognitions are a major ingredient of the motivation process. Self-efficacy levels can enhance or impede motivation. People with high self-efficacy choose to perform more challenging tasks (Bandura, 1995). They set themselves higher goals and stick to them. Actions are preshaped in thought, and people anticipate either optimistic or pessimistic scenarios in line with their level of self-efficacy. Once an action has been taken, high self-efficacious persons invest more effort and persist longer than those who are low in self-efficacy. When setbacks occur, they recover more quickly and maintain the commitment to their goals. Self-efficacy also allows people to select challenging settings, explore their environments, or create new environments.

Self-referent thought has become an issue that pervades psychological research in many domains. It has been found that a strong sense of personal efficacy is related to better health, higher achievement, and more social integration. This concept has been applied to such diverse areas as school achievement, emotional disorders, mental and physical health, career choice, and sociopolitical change. It has become a key variable in clinical, educational, social, developmental, health, and personality psychology (Bandura, 1997; Maddux, 1995; Schwarzer, 1992, 1994).

Self-efficacy is commonly understood as being domain-specific; that is, one can have more or less firm self-beliefs in different domains or particular situations of functioning. But some researchers have also conceptualized a generalized sense of self-efficacy. It refers to a global confidence in one’s coping ability across a wide range of demanding or novel situations. General self-efficacy aims at a broad and stable sense of personal competence to deal effectively with a variety of stressful situations (Schwarzer, 1994).

Research Question

The present study aims at examining the psychometric properties of 13 versions of the General Self-Efficacy Scale (see Appendix). This is done with the assumption that self-efficacy is a universal construct that applies to different cultures and that can be measured by inventories in different languages. The purpose is to confirm this assumption and provide measures that can be adopted in other countries for collecting further evidence. Two steps are taken to obtain the necessary psychometric data: First, the internal structure of the instrument is scrutinized, which includes item analyses, principal component analyses, and a confirmatory factor analysis to test the equivalence across languages. Second, mean differences between languages and gender are analyzed. It has to be kept in mind that this is mainly a psychometric study, not a truly cross-cultural one. Psychometric equivalence across languages can be seen as a prerequisite for subsequent cross-cultural studies that also take indigenous characteristics of the specific cultures into account.

Measure

The German version of this scale was originally developed by Jerusalem and Schwarzer in 1981, first as a 20-item version and later as a reduced 10-item version (Jerusalem & Schwarzer, 1992). Typical items are "Thanks to my resourcefulness, I know how to handle unforeseen situations," and "When I am confronted with a problem, I can usually find several solutions." It has been used in numerous research projects, where it typically yielded internal consistencies between alpha = .75 and .91. The scale is not only parsimonious and reliable, it has also proven valid in terms of convergent and discriminant validity. For example, it correlates positively with self-esteem and optimism, and negatively with anxiety, depression and physical symptoms. Previous studies are described in the manual (Schwarzer, 1993; Schwarzer & Jerusalem, 1995), which includes not only the scale in English, German, Spanish, French, Hebrew, Hungarian, Turkish, Czech, and Slovak, but also presents the results of five studies conducted to examine the psychometric properties of the German version. 

The ten self-efficacy items were adapted to the 13 languages by bilingual native speakers based on the German and English versions of the instrument. The English version is enclosed in the Appendix.

Samples

Except for infants, a life-span coverage of age-groups is represented in the collated data (see Figure 1).

Figure 1: Age distribution

Also, different levels of education and socioeconomic status are represented. The entire sample includes an amazing array of diverging, potentially confounding variables which is one of the unique features of studies that comprise many cultures and groups within the global community.

The first German sample consisted of 425 university students who studied different subjects at two universities in Berlin, and 238 students from the University of Düsseldorf. There were also 255 other adults from Berlin, and 220 senior citizens from Düsseldorf, giving a total of 2,115 individuals with a mean age of 33 years (SD = 17 years). There were 977 men with an average age of 30 years (SD = 12) and 1,138 women with an average age of 36 years (SD = 20.3). The age difference was statistically significant, F(1,2113) = 59.6, p < .01.

The second German sample consisted of German teachers. As part of a school innovation project, 269 teachers from ten schools in ten states of Germany responded to the Perceived Self-Efficacy questionnaire. There were 112 men and 150 women who taught in classes from grade 1 to 13, but most of them taught in grades 7 to 10. Age was assessed in categories with 6% being 30 years old or younger, 29% being 31 to 40 years old, 43% being 41 to 50 years old, and 22% were 51 years or beyond.

The third German sample consisted of German high school students Within the same research project, data from 3,077 high school students are available. age

There were 52% boys and 48% girls. They attended grade 7 (22%), grade 8 (24%), grade 9 (22%) and grade 10 or higher (24%).

The Spanish-speaking sample consisted of 959 university students from Costa Rica, including 605 women with an average age of 21.3 years (SD = 6.9) and 354 men with an average age of 21.0 years (SD = 6.3).

There were two Chinese samples, one consisting of 294 first-year undergraduate students and the other of 774 high-school students in Hong Kong. The university students attended three introductory classes on general psychology at the Chinese University of Hong Kong, but most of the students did not choose psychology as their major. There were 94 male students with an average age of 19.7 years (SD = 1.4) and 200 female students with an average age of 19.5 years (SD = 1.4). Among the high-school students, there were 248 boys with an average age of 15.0 years (SD = 1.8) and 525 girls with an average age of 15.1 years (SD = 1.9).

The Indonesian version of the scale was given to 536 students of tourism or educational sciences in Bandung, a large city on the Indonesian island of Java. There were 260 female students with an average age of 20.2 years (SD = 3.00) and 276 male students with an average age of 20.97 years (SD = 3.03).

The Japanese sample consisted of 430 undergraduate students from three universities in the metropolitan area of Tokyo. The 236 women had an average age of 18.48 years (SD = 0.87), and their male fellow students had an average age of 19.13 years (SD = 1.08).

The Korean sample included 35 men with an average age of 68.4 years (SD = 4.8) and 111 women with an average age of 66.4 years (SD = 5.0). The sample was mainly recruited in meeting places for the elderly in Seoul.

The Arabian sample consisted of 264 persons from different regions of Syria. Their mean age was 26 years (SD = 6.6 years). There were 115 men and 149 women.

The Russian sample was made up of 495 individuals from a southern Russian city, among them 264 university students, with a mean age of 27 years (SD = 11.2 years). There were 205 males and 290 females.

The Polish sample consisted of 570 persons from Lodz, including 100 police novices. Their mean age was 26 years (SD = 3 years), with more women (n = 415) than men (n = 155).

The Hungarians were 158 medical students from Budapest with a mean age of 24 years (SD = 3.8 years). Unfortunately, 95 of them did not indicate their gender (39 women, 25 men).

The Greek were 100 persons (50 women, 50 men) with a mean age of 40 years (SD = 10.3 years).

The Dutch sample (n = 697) was composed of a group of benzodiazepin users and a control group, with a mean age of 61 years (SD = 14.7 years). There were 519 women and 178 men.

The first English sample (Great Britain) consisted of 219 arthritis patients in the U.K., aged 60 (SD = 12.5 years), with 26 men and 193 women.

The second English sample (Canada) consisted of Canadian university students. There were 290 students, 104 women and 185 men, who were requested to fill out a questionnaire on students’ reactions to a faculty strike at York University in Spring 1997. This part of the study was conducted by Esther Greenglass. Among other instruments, the students responded to the Perceived Self-Efficacy questionnaire. Age was not asssessed.

The third English sample (Internet data) was composed of computer users worldwide who were responding to a website survey. Out of N=1,437 useable respondents (as of December 17, 1997), half were 25 years of age or younger. The age categories 15-20 and 21-25 were each filled with 24% of participants. Two percent were younger than 15, and 40% were between 26 and 50 years of age. There were 762 men (53%) and 583 women (41%) while six percent did not disclose their gender. The number of men was lower than might have been projected given stereotypes of gender differences in computer use. The majority came from North America (78%), with about 9% from Europe and negligible numbers from other part of the world. The most common education level was "some college," (34 %) but education level ranged from 11% with "some high school" to 19% with a Masters degree or beyond. Grade point average (GPA) was between 3.51 and 4.00 for 47%, and between 3.01 and 3.50 for 31% of the sample. The most frequently reported income level was the under $15,000 category (20%), but there was a wide range: 13% fell into the $15,000 to $25,000 category, 14% into the $25,000 to $35,000 category, 17% fell into the $35,000 to $50,000 category, and 38% were above $50,000. These demographics compare well to the Georgia Tech survey of on-line users, and other demographic surveys about WWW usage (e.g., http://ernie.bgsu.edu/departments/tcom/survey.html or http://www.teleport.com/~tbchad/stats1.html). In other words, the sample isn't just the freshman male college student, and it is more gender balanced than most stereotypes allow.

This section focusses on mean differences among the samples and on the psychometric properties for each of the 13 language adaptations. Table 1 displays descriptive statistics for each sample broken down by gender.

Table 1

Self-efficacy scale means and standard deviations broken down by sample and gender
 

Sample

Gender

Mean

SD

N

 

Internet	Women	28.57	5.21	577
	Men	29.84	5.15	755
Canada	Women	29.87	4.98	185
	Men	31.64	4.47	104
German Teachers	Women	28.36	5.25	110
	Men	28.25	4.86	149
German Students	Girls	28.93	4.27	1477
	Boys	29.65	3.99	1587
Germany (others)	Women	28.73	5.40	1138
	Men	29.91	4.51	977
Hong Kong	Women	22.27	4.96	724
	Men	24.68	4.86	342
Indonesia	Women	30.04	5.12	260
	Men	30.17	5.08	276
Japan	Women	20.17	6.00	236
	Men	20.28	6.43	194
Korea	Women	28.08	6.09	111
	Men	27.13	6.70	35
Syria	Women	28.95	4.73	149
	Men	29.05	4.08	115
Russia	Women	31.38	4.60	290
	Men	32.70	4.84	205
Poland	Women	27.13	5.25	415
	Men	29.03	4.70	155
Hungaria	Women	28.18	5.09	39
	Men	28.30	3.76	25
Greece	Women	29.60	3.89	50
	Men	31.94	3.84	50
Netherlands	Women	30.66	5.53	519
	Men	31.86	4.89	178
Great Britain	Women	28.97	5.43	193
	Men	31.15	4.06	26
Costa Rica	Women	33.04	4.60	602
	Men	33.50	4.16	351

--------------------------------------------------------------------------------------------------------

Note: ** = p < .001

Figure 2 displays the frequency distribution of self-efficacy sum scores. The mean was 28.63 (SD = 6.18, N =7,767 ). The ten-item sum score had a theoretical range from 10 to 40, due to the 1-to-4 response format. There is a slight "ceiling effect," that is, extremely self-efficacious individuals are less well identified than low self-efficacious individuals. The skewness of the distribution was -.42 and the kurtosis -.19.

Item Characteristics and Internal Consistency

Item analyses were carried out separately for each scale adaptation. Each item had a response range from 1 to 4. Item means and corrected item-total correlations are given in Table 2. All coefficients, except Item 2 in the Indonesian, Greek, and Spanish samples, and Item 3 in the Greek sample, turned out to be satisfactory. No overall improvement was possible by eliminating items. The internal consistency of Cronbach’s alpha = .91 was best for the Japanese version, and least satisfactory for the Greek adaptation (alpha = .78). In sum, the internal consistency was very satisfactory, considering that the scale contained only ten items.

Table 2

Item Means and Corrected Item-Total Correlations for Ten Self-Efficacy Items in 13 Languages
 

	Chinese (Hong Kong)   (n = 1,068)		Indonesian   (n = 536)					Japanese   (n = 430)					Korean   (n = 147)
Item	Mean	Correla-  tion	Mean	Correla-tion		Mean			Correla-tion		Mean			Correla-tion
1	2.76	.45	3.49	.37		2.45			.60		2.86			.64
2	2.35	.41	3.67	.25		2.13			.57		2.60			.57
3	1.78	.54	2.11	.33		1.64			.59		2.59			.60
4	2.09	.65	2.83	.51		1.86			.71		2.80			.59
5	2.04	.65	2.94	.55		1.73			.69		2.88			.70
6	2.82	.51	3.00	.52		2.11			.66		2.90			.53
7	2.41	.62	2.75	.54		1.80			.73		2.68			.53
8	2.35	.52	2.95	.60		2.16			.74		2.82			.64
9	2.53	.60	2.84	.55		2.19			.77		2.78			.65
10	1.94	.62	3.54	.43		2.16			.71		2.94			.55
alpha	.85		.80				.91					.88
-----------------------------------------------------------------------------------------------------------------
	Arabian (Syria)  (n = 264)		Russian   (n = 495)					Polish  (n = 570)					Hungarian   (n = 158)
Item	Mean	Correla-tion	Mean		Correla-tion		Mean			Correla-tion		Mean			Correla-tion
1	3.16	.42	3.38		.58		2.91			.50		2.98			.61
2	3.03	.47	3.23		.56		2.54			.46		2.72			.58
3	2.67	.42	3.09		.60		2.70			.51		2.98			.55
4	2.87	.40	2.85		.52		2.57			.60		2.64			.61
5	3.02	.61	3.31		.65		2.75			.66		2.48			.65
6	2.84	.56	3.40		.60		3.09			.58		2.14			.61
7	3.02	.38	3.06		.52		2.77			.56		2.90			.61
8	3.04	.56	3.38		.36		2.85			.55		3.02			.46
9	2.84	.51	3.11		.56		2.70			.63		2.86			.70
10	2.48	.34	3.11		.53		2.77			.61		2.88			.73
alpha	.79		.85				.86					.88

 

In previous studies (Schwarzer et al., 1997), the scale turned out to be homogeneous. To study the scale characteristics of the 13 adaptations, principal components analyses were computed separately for each language version. In most instances, the first eigenvalue was clearly higher than the others, and the second eigenvalue was below unity. The scree tests suggested one-factor solutions. These findings, in conjunction with the high internal consistencies, suggest unidimensionality of each scale adaptation. The overall analysis, collapsing all samples, yielded a one-factor solution with eigenvalues of 4.9, 0.81, 0.72, etc. Half of the variance is accounted for by this component, whereas a second component would only account for 8% of the variance.

Multigroup confirmatory factor analysis (CFA) was applied in previous analyses with several adaptations (Schwarzer et al., 1997), but for 13 groups this seems to be less appropriate. Instead, the correlation matrix of the total sample was chosen as the starting point for a CFA. The one-factor model was tested with the LISREL 8 program (Jöreskog & Sörbom, 1993). Input was a correlation matrix of the ten observed variables, and the parameters were estimated by the unweighted least squares method. The model fit was evaluated in terms of chi-square, root mean square residuals (RMR), and various goodness of fit indices. The chi-square divided by the degrees of freedom can be seen as a less biased fit estimate (chi²/df) than the chi-square itself because it is dependent on sample size. This ratio should be small, and values below three are considered to be satisfactory (Bollen & Long, 1993). The root mean square residual should be very small, with values below .05 being desirable. The goodness of fit index (GFI) should be above .92. The same applies to the adjusted GFI (AGFI; adjusted for degrees of freedom). These are rough fit indicators only. A more comprehensive assessment includes further fit indices as well as a careful inspection of the parameter estimates, accounted variance, and modification indices. The one-dimension confirmatory factor analysis yielded the factor loadings in the last column of Table 3.

Table 3

Discriminative Power and Factor Loadings of the Ten Self-Efficacy Items
 

Item	Wilk´s  lambda	Factor Loading  (PCA)	lambda X  (LISREL)
1	.83	.66	.61
2	.81	.56	.51
3	.74	.67	.62
4	.79	.75	.71
5	.77	.77	.74
6	.78	.65	.61
7	.84	.75	.71
8	.83	.72	.68
9	.87	.74	.70
10	.70	.73	.69

The fit indices turned out to be excellent, which confirmed the unidimensionality of the instrument. The fit was chi² = 164.34 (35 df, p < .01), chi²/df = 4.70, RMR = .053, GFI = .99, and AGFI = .98.

Language and Gender Differences

Figure 2 illustrates that students in Hong Kong and Japan obtained the lowest self-efficacy scores in this investigation, whereas the Costa Ricans and the Russians obtained the highest values. As can be seen in Table 1, this applies to men and women in the same manner. Thus, there might be cultural differences in perceived self-efficacy.

 

There are obviously mean differences between these 14 data sets in terms of culture and gender. A two-way factorial analysis of variance (13 levels of language and two levels of gender) was computed to determine these effects. Due to some missing values, the analysis was based on 7,655 persons. Table 1 contains the cell means, standard deviations, and cell sizes.
There was a significant main effect for language, F(12, 7629) = 282.91, p <.001, partial eta² = .30, a significant main effect for gender, F(1, 7629] =24.26, p <.001, partial eta² = .03, and a significant interaction effect, F(12, 7629) =  3.47, p <.001, partial eta² = .01. However, the latter two effects are negligible since the sample sizes are large, and only less than 3% of variance was accounted for.

Differences between the 13 adaptations were further explored by computing a discriminant analysis with all ten items as discriminators (see second column in Table 3). All ten items discriminated significantly between groups, with Item 5 ("Thanks to my resourcefulness, I know how to handle unforeseen situations") being the one with the highest power to separate the groups.

The present study focussed on a comparison of 13 language adaptations for 14 countries. It was found that in all languages the psychometric properties that were investigated in this study were satisfactory. Internal consistencies, item-total correlations, and factor loadings indicated that the General Self-Efficacy Scale can be seen as homogeneous and unidimensional. By achieving these characteristics it has been suggested that the self-efficacy construct tends to be a universal one, claiming construct validity across very different cultures. However, this is mainly a psychometric study, not a truly cross-cultural one. Psychometric equivalence across languages can be seen as a prerequisite for subsequent cross-cultural studies that also take indigenous characteristics of the specific cultures into account.

A problem arises by inspection of the mean differences between groups. Why do the Costa Ricans (Spanish version) have the highest self-efficacy scores and the Japanese the lowest? Group differences could also be an undesired side effect of the instrument development. Perhaps the Spanish items are "easier" than the Japanese items. This is a general problem of all cross-cultural studies using multilanguage versions of the same inventory. The endorsement of items can be seen as being multiply determined. Among the factors that influence the endorsements are characteristics of the cultural context, those of item wording, and numerous biases, such as situational circumstances of test administration. Attempts can be made to control some of these factors, but essentially the problem is unsoluble.

In spite of the limitations that are typical for cross-cultural studies, some psychometric properties of the General Self-Efficacy Scale are now established for several languages. It is suggested that large-scale field studies may also include these ten items for enrichment purposes and that further psychometric properties be investigated cross-culturally. For example, this inventory could replace other scales that are often included in an exploratory manner. Future research should, in particular, focus on validity and intervention. How can self-beliefs be changed, and how can this instrument reflect such changes? Are more specific measures more appropriate for this purpose? Perceived self-efficacy is now a well-established construct, based on social-cognitive theory that has high explanatory and operative power (Bandura, 1995). That is, perceived self-efficacy not only explains human functioning quite well, it is also easily alterable by interventions. It remains unclear at which levels of specificity of item wording the hypothesized effects can be made most visible. The General Self-Efficacy Scale can be used tentatively for screening people at risk for coping deficiencies, which can set the stage for subsequent prevention programs. Collaborative efforts of psychologists from the global community are needed to foster cross-cultural validity and intervention studies in this field.

Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191-215.

Bandura, A. (Ed.) (1995), Self-efficacy in changing societies. New York: Cambridge University Press.

Bandura, A. (1997). Self-Efficacy: The Exercise of Control. New York: Freeman.

Bollen, K. A., & Long, J. S. (1993). Testing structural equation models. Newbury Park, CA: Sage.

Jerusalem, M., & Schwarzer, R. (1992). Self-efficacy as a resource factor in stress appraisal processes. In R. Schwarzer (Ed.), Self-efficacy: Thought control of action (pp. 195-213). Washington, DC: Hemisphere.

Jöreskog, K. G., & Sörbom, D. (1993). LISREL 8.03. Chicago: Scientific Software International.

Maddux, J. (Ed.) (1995). Self-efficacy, adaptation, and adjustment: Theory, research, and application. New York: Plenum.

Schwarzer, R. (Ed.) (1992). Self-efficacy: Thought control of action. Washington, DC: Hemisphere.

Schwarzer, R. (1993). Measurement of perceived self-efficacy. Psychometric scales for cross-cultural research. Berlin, Germany: Freie Universität Berlin.

Schwarzer, R. (1994). Optimism, vulnerability, and self-beliefs as health-related cognitions: A systematic overview. Psychology and Health: An International Journal, 9, 161-180.

Schwarzer, R. & Born, A. (1998). Optimistic self-beliefs: Assessment of general perceived self-efficacy in 13 cultures. World Psychology, (in press).

Schwarzer, R., Bäßler, J., Kwiatek, P., Schröder, K., & Zhang, J. X. (1997). The assessment of optimistic self-beliefs: Comparison of the German, Spanish, and Chinese versions of the General Self-Efficacy Scale". Applied Psychology: An International Review, 46 (1), 69-88.

Schwarzer, R., & Jerusalem, M. (1995). Generalized Self-Efficacy scale. In J. Weinman, S. Wright, & M. Johnston, Measures in health psychology: A user’s portfolio. Causal and control beliefs (pp. 35-37). Windsor, UK: NFER-NELSON.

Zhang, J. X., & Schwarzer, R. (1995). Measuring optimistic self-beliefs: A Chinese adaptation of the General Self-Efficacy Scale. Psychologia, 38(3), 174-181.

Appendix

1) I can always manage to solve difficult problems if I try hard enough.

2) If someone opposes me, I can find means and ways to get what I want.

3) It is easy for me to stick to my aims and accomplish my goals.

4) I am confident that I could deal efficiently with unexpected events.

5) Thanks to my resourcefulness, I know how to handle unforeseen situations.

6) I can solve most problems if I invest the necessary effort.

7) I can remain calm when facing difficulties because I can rely on my coping abilities.

8) When I am confronted with a problem, I can usually find several solutions.

9) If I am in trouble, I can usually think of something to do.

10) No matter what comes my way, I'm usually able to handle it.

AUTHOR NOTES

All 21 translations and further information are available on the World Wide Web http://www.yorku.ca/academics/schwarze/selfscal.htm.

.