http://orcid.org/0000-0002-8485-1616
Figures
Abstract
The Environmental Attitudes Inventory (EAI) was developed to evaluate the multidimensional nature of environmental attitudes; however, it is based on a dataset from outside the Arab context. This study reinvestigated the construct validity of the EAI with a new dataset and confirmed the feasibility of applying it in the Arab context. One hundred and forty-eight subjects in Study 1 and 130 in Study 2 provided valid responses. An exploratory factor analysis (EFA) was used to extract a new factor structure in Study 1, and confirmatory factor analysis (CFA) was performed in Study 2. Both studies generated a seven-factor model, and the model fit was discussed for both the studies. Study 2 exhibited satisfactory model fit indices compared to Study 1. Factor loading values of a few items in Study 1 affected the reliability values and average variance extracted values, which demonstrated low discriminant validity. Based on the results of the EFA and CFA, this study showed sufficient model fit and suggested the feasibility of applying the EAI in the Arab context with a good construct validity and internal consistency.
Citation: AlMenhali EA, Khalid K, Iyanna S (2018) Testing the psychometric properties of the Environmental Attitudes Inventory on undergraduate students in the Arab context: A test-retest approach. PLoS ONE 13(5): e0195250. https://doi.org/10.1371/journal.pone.0195250
Editor: Mohammad Shahid, Aligarh Muslim University, INDIA
Received: May 22, 2017; Accepted: March 19, 2018; Published: May 14, 2018
Copyright: © 2018 AlMenhali et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
In environmental psychology studies, several models on the substantial influence of human behaviors on the natural world have included environmental attitudes as an indicator of behavior that either improves or worsens the environmental quality. Attitude is defined as an enduring evaluation of people, objects, or ideas [1], a combination of motivational, emotional, perceptual, and cognitive processes with respect to some aspect of the environment [2], and the learned predisposition to consistently respond to a given object [3]. Environmental attitudes are more of an individual concern for the physical environment, which is related to the degree of cognitive, affective, and behavioral concerns toward the environmental problems [4].
Environmental attitudes have been considered a relevant part of environmental psychology models and therefore, are often included in measures of environmental behavior, sometimes considering school students [5–7], university students [8–10], and the general community together [11–13].
Studies on environmental attitudes heavily rely on the subjective measures of environmental cognitions and environmental affection in comparison to objective measures [14–16]. Environmental attitudes are a latent construct and thus unlikely to be observed directly [15]. Traditionally, attitude measurement is designed as direct self-report methods [17–18]. Conventionally, environmental measurement is designed as self-administered methods [19–20] based on the perceived ecological issues and problems. Although some studies have measured environmental attitudes, only two measurements have received much attention due to their validity and reliability assessed [21–22] across cultures.
To avoid the dated issues of the general environmental topics, Dunlap and Van Liere [23] developed the new ecological paradigm (NEP) scale to measure the overall relationship between humans and the environment. Later, the revised version of the NEP (NEP-R) [24] was developed as a better alternative with stronger psychometric properties using sound, grounded theories. Over 49 studies worldwide have used this scale to examine and better comprehend the environmental attitudes driven by values and beliefs about the natural environment and its management. The NEP scales measure ecocentric systems of beliefs as opposed to the anthropocentric systems of beliefs [21]. The 15 items tabulated in the five-factor model of the NEP-R is the most widely-used measure to investigate environmental attitudes. Dunlap, Van Liere, Mertig, and Jones [24] presented a five-factor model that emphasized the awareness of consequences to reflect environmental concerns and values to explain the underlying beliefs from a global perspective. The five-factor model, which encompasses balance of nature, eco-crisis, anti-exemptionalism, limits of growth, and human domination, consists of 15 items. This scale investigates multiple environmental phenomena related to beliefs, attitudes, intentions, and behaviors. It measures multiple environmental issues including pollution, waste management, recycling, and natural resources. Thus, this measure can be considered a multiple-topic assessment method [21].
Much attention was given by a previous study to treat the NEP-R as a measure of environmental beliefs rather than ecological worldview as proposed by Dunlap et al. [24]. Dunlap explains that environmental beliefs provide a more accurate interpretation due to the degree to which respondents view and perceive the world from an ecological viewpoint [21]. Turaga, Howarth, and Borsuk [25] posited that the importance of the NEP/NEP-R for measuring the environmental concerns in the world has been emphasized both from theoretical and practical viewpoints, and used successfully in cross-cultural research [21]. Recent studies have revealed that the NEP-R yields a 2-factor model [26], 4-factor model [27], and 5-factor model [28] instead of unidimensionality of the 15 items. Due to extensive use of the NEP-R for classifying people’s ecological worldview and monitoring changes in worldview over time, the NEP was problematic because of its overestimation [22, 28–32].
The NEP-R has been argued for being overly simplistic and outdated [33–34]. These studies have argued that the NEP-R scale has omitted critical components of the pro-ecological worldview, and the ecocentric perspective resulting from the late twentieth-century environmental ethics literature was excluded. The NEP-R has also been criticized for being a poor predictor of environmental behavior. Earlier studies showed that the relation between the NEP-R and attitude in predicting behavior is weak; thus, it failed to measure worldview precisely [22, 27, 35]. Psychometric analyses conducted on the 15 items showed that the NEP-R is best represented by its five correlated dimensions [28, 36]. These studies have shown that testing the unidimensionality of the NEP/R’s 15 items entails a statistical bias when used as a predictive score or outcome variable. A meta-analysis of 69 studies using the NEP/NEP-R showed that differences in sample type and scale length have a significant effect on NEP scores, making comparison of scores across studies difficult [37].
Milfont and Schultz (2016) have argued that environmental beliefs should be supported by environmental attitudes to define the environmental responsibilities that are reflected from the evaluation of environmental consequences. Bronfman et al. [38] posited that culture is the key source of social influence because of its proximity to, permanent contact with, and authority over environmental attitudes. The environmental attitudes inventory (EAI) [22] was established to address the limitations of the NEP-R. The EAI was developed as an alternative to the NEP and suggests a complex multidimensional structure of environmental attitudes to measure individual’s beliefs regarding the natural environment across 12 dimensions [22, 39]. The multifaceted structure of environmental attitudes was developed from a pool of 200 scale items drawn from established pro-environmental behavior measures: the NEP [21, 24], the Ecological WorldView Scale [40], and the Environmental Perception Scale [22, 41]. The psychometric criteria for the 200 items were tested unidimensionally and multidimensionally for reliability, validity, and intercorrelation evidence; they were revised as a short version of the 12-factor model with six items per scale [22].
Recent studies have shown that the EAI yields a 6-factor model [42], 9-factor model [39], 10-factor model, and 11-factor model [34], the one-dimensional environmental movement activism [35], an 11-factor model [22], 7-factor model [43], and a 6-factor model [44]. Even though the EAI is considered an appropriate tool to measure the influence of ecological initiatives, factors such as socio-demographic effects on cultural biases [45], social-desirability bias [46], and non-response bias [47] may affect the internal consistency of the data from a different population. Like other replication studies, the results of the psychometric analysis of the EAI results show that the extracted factor structures are distinct from the original constructs. It is argued that there is a possibility of the existing studies not being able to assemble enough evidence to draw a definite conclusion about the dimensionality of environmental attitudes in cross-cultural research [48–50]. Thus, it is argued that the generalization of environmental dimensions to underlie the evaluation of the ecological belief, values, and attitudes across cultures is not conclusive. More studies should be conducted to include other distinctive cultures when measuring environmental attitudes to capture the differences in perceived ecological responsibilities and ecological consequences.
Therefore, it is necessary to measure the level of ecological information available within a specific culture, as well as how people perceive ecological issues that shape their environmental attitudes. Thus, it is necessary to utilize a suitable instrument to carry out this function. In psychology-ecology research not many studies have been designed to establish what is essential in the adaptation process of a questionnaire in a culture-specific context [51–52]. Careful translation is not enough to validate a survey because linguistic terms must be relevant to the social and cultural conditions of the population studied [53]. Psychometric measures within a specific context are needed because each society has its beliefs, habits, and attitudes that reflect a country’s culture. López and Amerigo [54] concur that careful identification of individual norms and values for each ecological issue is required to establish meaningful results. A fundamental aspect of developing a useful and meaningful tool is the number of questions. Furthermore, the time taken to answer the questions is another essential factor in selecting the questionnaire to be used. This is consistent with recent concerns related to response burden or response fatigue due to a large number of questionnaire items that affect willingness to participate [47]. The use of a questionnaire to assess young adults’ environmental attitudes when facing ecological issues is an appropriate source for investigating attitudes that influence responsibilities toward ecology, based on the perceived consequences of ecological problems. Attitudes are a latent construct and thus must be inferred from overt responses [55–56]. The explicit measurement technique of a direct self-report method using a questionnaire is the most popular method for measuring environmental attitudes [21–22].
This study aimed to investigate the reliability, construct validity, and factor structure of Milfont and Duckitt’s EAI [22] owing to the multidimensional nature of environmental attitudes. The EAI assesses broad perceptions regarding ecology and the factors affecting its quality. Based on the arguments discussed, the 12-factor EAI seems to cover the most widely examined environmental attitudes factors sufficiently. The psychometric properties of the EAI were measured for two primary purposes: (1) to determine whether the 72-item EAI is suitable for use in the Arab context; and (2) to establish the baseline framework for environmental attitudes.
Study 1
Materials and methods
Hagel (2014) advocates that a psychometric theory requires unidimensionality, while Ziegler and Hagemann (2015) posited that this technique is parallel to testing the dimensionality of items. To determine which scale items of the EAI should represent a common, latent construct, an exploratory factor analysis (EFA) was employed. This varimax rotation method is employed to develop a smaller number of variables that will account for variance of the latent constructs. Varimax rotation is used to maximize the variance of the factor loadings on all constructs in a factor matrix. This technique has the effect of differentiating the original variables by extracted factors [57] due to factor intercorrelations. As a result, it reduces many items and latent constructs into fewer numbers of factors. This analysis is important to identify critical dimensions that would compromise the structure and criterion variables in subsequent analyses. Furthermore, it enables understanding of the structure of the variables and determination of the correlation among them in the data set [58]. In the second stage of the EAI testing, confirmatory factor analysis (CFA) is used to confirm the factor structure and specify how the construct is measured [57].(For this study, the analyses were based on an item pool of 72 items derived from Milfont and Duckitt’s model [22].
Participants
An anonymous and consented questionnaire was administered to final year undergraduates at Abu Dhabi University, UAE, and 59.2% (148) of the students voluntarily agreed to participate. Of the subjects, 52.0% (130) eligible for the EAI measurement provided complete and consistent responses. Patterns of non-response were shown by 102 subjects (40.8%). Eighteen incomplete and inconsistent responses were identified, withdrawn, and not analyzed as part of the valid responses. A total of 130 participants (92 females and 38 males) with ages ranging from 21 to 35 years (M = 28, SD = 0.745) sharing very similar socio-demographic profiles (males and higher age group) from social sciences and science and technology colleges completed the questionnaire.
Instruments
The twelve dimensions of the 72-item questionnaire developed by Milfont and Duckitt [22] were used to test the validity of the EAI in the Arab context. No ethical issues were present in this study as permission was sought from the original author to use the measure (Table 1 and Table 2). The study also was approved by the Institutional Review Board for Human Subjects Research (IRB) of Abu Dhabi University. Responses to all measures were made on a 7-point scale from 1 = strongly disagree to 7 = strongly agree. The following measures along with demographic information (e.g., age, gender, and program) were included in the questionnaire. None of the 72 items were translated into the Arabic language because English is the communication medium for teaching and learning, and all students are expected to meet the specified minimum score on the English proficiency of International English Language Testing System (IELTS) at the time of admission. The EAI consisted of 72 items that assess the 12 dimensions of environmental attitudes.
Results
All data analyses were performed using SPSS version 18.0 [59]. The unidimensionality of the EAI was checked, and a principal component analysis (PCA) using the extraction method of principal axis factoring with oblique rotation was performed. The EAI factors were proven to be statistically important according to the Kaiser-Meyer-Olkin measure of sampling adequacy of 0.891 and Bartlett's test of Sphericity (χ2 = 4458.963, df = 666, p<0.000); this indicates that the non-existence of multicollinearity and correlations between items were sufficient to perform a factor analysis. The eigenvalues for each scale were more than 1.0, and the cumulative percentage variance as 75.564. The following scales were excluded from the EAI dimensions: Scale 1-Enjoyment of nature, Scale 6-Environmental fragility, Scale 8-Personal conservation behavior, Scale 9-Human dominance over nature, Scale 12-Support for population growth policies. The reduction of items was due to cross-loadings and factor loading values of less than 0.5 as suggested by Hair et al. [60]. The EFA was performed using varimax rotation to examine the dimensionality of the items on the 7-factor constructs of the 35 items. The EFA demonstrated that the 35-item EAI was rotated into 7-factor models of the EAI (Table 3) without losing information and internal consistency. The data were considered normal because skewness and kurtosis were between the acceptable limits of ±2.0 [61] and ±7.0, respectively [62].
To examine the adequacy of convergent and discriminant validity, a confirmatory factor analysis (CFA) was performed using SPSS AMOS version 23.0. The measurement model of Study 1 produced an adequate convergent and discriminant validity, met the threshold suggested for factor loadings (≥0.5) [63], AVEs (>0.5) [63], the Cronbach's Alpha (CA) (>0.60) [64], and composite reliability (CR) (≥0.60) [64] (Table 4). A cut-off value of less than 0.80 for RMSEA [62], 0.80 for the NFI [65], and 0.90 for the TLI and CFI [66] were necessary due to the sensitivity to some variables and insensitivity to the sample size. Thus, it indicated acceptable goodness-of-fit indices (CMIN = 899.236; CMIN/df = 1.716; CFI = 0.907; NFI = 0.805, TLI = 0.894 and RMSEA = 0.074).
Study 2
One step towards establishing validity and reliability is the use of the repeatability approach, in which measurements are done under the same conditions and within a short period of time using the same 72-item 12-dimension EAI instrument [67]. This method was used to test the stability and reliability of the 35 items of the 7-dimension EAI (confirmed in Study 1) over time. AMOS version 23.0 [68] was used to test Anderson and Gerbing's [69] two-step modeling method to examine the measurement model. A CFA was performed for the individual construct and measurement model.
Participants and Instruments
A month after conducting Study 1, 400 anonymous and consented questionnaires were administered to the same sample of final year undergraduates at Abu Dhabi University, UAE. Only 32.5% (130) of the subjects were qualified for the study based on the criterion of giving complete and consistent feedback. The non-response rate was 65.0% (260), and five incomplete responses were withdrawn. The withdrawn responses had similar socio-demographic profiles (males, and higher age group). The participants with valid responses were 83 females and 47 males, ranging in age from 21 to 35 years (M = 25, SD = 0.785), and social sciences and science and technology college students. Study 2 adopted the procedure of Study 1 to test Milfont and Duckitt’s [22] EAI items.
Results
A CFA was performed on the 72 items of the 12-dimension EAI and measurement model; the final model confirmed 21 items of the 7-factor model of the EAI (Table 5). The acceptable limits for skewness and kurtosis were ±3.0, corresponding to a normal distribution of the data [70]. The factor loadings (Table 6) and AVEs were greater than 0.50 [71]. The CRs and CAs were well above 0.70, suggesting good convergent validity of the studied variables [62]. All seven constructs were between 0.347 (r2 = 0.120) and 0.858 (r2 = 0.736) at p<0.01, with AVEs greater than r2, exhibiting sufficient discriminant validity) (Table 7). In Study 2, all measurement models yielded satisfactory goodness-of-fit. However, five constructs, Scale 1-Enjoyment of nature, Scale 6-Environmental fragility, Scale 8-Personal conservation behavior, Scale 9-Human dominance over nature, and Scale 12-Support for population growth, were withdrawn because the r2 values were higher than AVEs, thereby indicating insufficient discriminant validity among the constructs [62]. As a result, the number of items was reduced from 72 to 21 for the 7-factor model. The final 7-factor model of the EAI yielded satisfactory goodness-of-fit indices (CMIN/df = 1.648; CFI = 0.948; NFI = 0.808, TLI = 0.935 and RMSEA = 0.071). The final structural model of the 7-factor model is illustrated in Fig 1. A cut-off of less than 0.80 for RMSEA [62], 0.80 for the NFI [65] and 0.90 for the TLI and CFI [66] are necessary due to the sensitivity to some variables and insensitivity to the sample size. A CFI greater than 0.90 indicates adequate fit [72] and can be reproduced for future study.
Discussion
This paper reports on the findings of two studies that investigated whether the 12-factor model identified in previous validation studies is suitable for a new dataset from student subjects [22, 34, 47]. Both studies were conducted on the same target sample with a different timeline to understand the extent of environmental attitudes using the test-retest approach. These approaches developed different models to arrive at an appropriate instrument for use in future research. The EFA and CFA were conducted in Study 1, and the model fits were discussed. Study 1 examined the EAI unidimensionality and dimensionality of each construct following the ABC of test construction [67]. EAI as the construct being measured is clearly discussed. The test-retest analysis was used to justify the intended use of the EAI measure and its use among undergraduate students in the Arab context. This development strategy has provided the basis for the correct use of the test for future study and can be used by other researchers when building on these findings. In Study 2, the CFA was conducted to evaluate parameter estimates, fit indices, and potential modification indices to facilitate theory-testing. This is consistent with the suggestion of earlier studies, which concur that psychometric analysis is applicable for construction and validation of assessment instruments concerned with attitudes [6, 8, 11, 28, 42–43]. The CFA was performed to test the new factor structure from the dataset and compare it with the 7-factor model in Study 1. The data were obtained from educated subjects, consistent with earlier environmental attitude-related studies, which posited that a well-educated sample was expected to be more exposed to information about ecological issues and be more capable of understanding the ecological perspective implied in the measurement [24, 34].
In Study 1, the EFA was conducted to extract the new factor structure of the dataset and identified the 7-factor structural model (Table 3). The 35 items of the EAI were shown to have better unidimensionality, which indicated that the non-existence of multicollinearity and correlations between items was sufficient for a factor analysis of the 7-factor model. The eigenvalues for each scale were more than 1.0, and the cumulative percentage variance was 75.564. These findings are consistent with earlier studies, which concur that the EAI shows evidence of unidimensionality [8, 22, 47]. The unidimensionality was shown in this study because the items of the 7-factor model with 35 items had greater loadings on their corresponding factors and almost all items could be explained by one factor. The unidimensionality of each dimension of the EAI was further examined in the EFA for each factor using a test-retest approach to ensure balanced models. The criteria for convergent validity and discriminant validity of the measurement model and item balance were also confirmed (Table 3). In comparison with the previous studies, this study’s dataset had 7 factors that were more well-separated [39, 42–44].
In general, factor loadings should be equal to or greater than 0.50 for good convergent validity [63]. All 35 items had factor loadings greater than 0.50, indicating high convergent validity. The CA and CR values met the threshold value of at least 0.70 per dimension [64]. All 7 factors of the model had high convergent validity except for Scale 7-Altering nature (AN) (CA = 0.57). Thus, these models may influence the independence of the final structural model. The low convergent validity of Scale 7-Altering nature (AN) was due to low factor loadings of all 5 items; thus, items from this factor may overlap with other factors [73]. Cho [74] argues that CA tends to systematically underestimate reliability due to the assumption that all factor loadings are equal. Other scholars have posited that CR is a reliable measure of reliability [64, 74]. These recommendations support the acceptance of Scale 7-Altering nature (AN) into the construct (CR = 0.71). In this study, the average variance extracted (AVE) for the 5-factor model was satisfactory, meeting the threshold of 0.5 [71], except for Scale 4-Conservation motivated by anthropocentric concern (CM) (AVE = 0.49) and Scale 7-Altering nature (AN) (AVE = 0.33). Fornell and Larcker [71] have suggested that CR must be higher than 0.6 if the AVE is less than 0.5, to support the acceptance of a factor model.
Bivariate-correlation (r) values for each scale were significant with p<0.05, between 0.198 and 0.628 and r2 ranging from 0.052 to 0.394. AVEs also were found to have a value greater than the r2 values, indicating sufficient discriminant validity [75–76]. The shortened version of the 35 items was also found to be similar to that of Milford and Duckitt [22] in that it had a mixture of positive and negative worded items in each scale to prevent response bias. This result indicates that the 7-factor model with 35 items of the EAI can be causally interpreted [66]. These test constructions align with earlier studies [72] that have suggested that a construct should be measured for unidimensionality before a series of model tests for multidimensionality can be conducted. The unidimensionality of a construct is important as the nomological items should be defined and interpreted before results can be obtained regarding the structural model and ideally replicated. Therefore, in this model and dataset, the factors were associated with one another. There are two possible explanations. First, the latent factors constituting the environmental attitudes cannot be independent. Additionally, since perceptions of the consequences of an individual’s actions may reflect their responsibilities, the factors are closely associated and expressed together. Second, since the subjects were highly educated, it is possible that they were more exposed to ecological issues due to the abundance of information provided to them. As a result, they could relate the value of ecology to concerns about ecology.
Study 1 considered the criteria for various model fit indices to discuss the model fit of the CFA. It has been suggested that RMSEA values less than 0.08 are acceptable [77]. Therefore, the RMSEA value of 0.074 in this sample indicates an acceptable fit. In this sample, the CFI value of 0.907, NFI value of 0.805, and TLI value of 0.894 are considered an acceptable fit. Even though it has been argued that the NFI and TLI values should be over 0.90 to indicate good fit, a value between 0.80 and 0.90 (0.80<NFI<0.90) is generally considered an acceptable limit [62–63, 75, 78]. Based on these indices, this sample had an acceptable fit for the 7-factor model. It has provided a measure for the assessment of the EAI in a specific cultural context, and for reproduction in future studies. Furthermore, this illustrates ways to assess the usefulness of a model in a new setting.
Although the EAI was shown to be the new culture-general and fully-balanced tool for measuring environmental attitudes [34, 47], the validity and test-retest reliability in the Arab context has yet to be measured. In this study, the factor structure of the EAI was investigated by both EFA and CFA in both studies based on the suggestion of Altman and Bland [79] that repeatability will evaluate the accuracy of responses. Thus, applying the CFA as a step forward to the EFA to test the EAI is reasonable for this study because it measures the stability and reliability of the original instrument in a different context using the same sample size, repeated over a short period of time, using the same 72-item 12-dimension EAI. Study 2 is a step forward for EAI testing. Study 2 tested alternative factor models to identify the best fitting factor model that sufficiently met the goodness-of-fit indices thresholds to explain the EAI better. Even though the EAI was tested in multiple countries across continents using undergraduate students and was found to have reasonable internal consistency and homogeneity across samples, this study showed a different outcome. Fifteen models were tested, and fit indices were generated. Of these 15 models, the 7-factor model (21 items) showed a better model fit (CMIN/df = 1.648; CFI = 0.948; NFI = 0.808, TLI = 0.935, and RMSEA = 0.071) (Table 4). The rest of the constructs were withdrawn due to factors being insignificant (p>0.05). The 7-factor model with 21 items showed an improved model fit in comparison to Study 1 (CMIN/df = 1.716; CFI = 0.907; NFI = 0.805, TLI = 0.894, and RMSEA = 0.074). This result indicated that the 21-item EAI is a reliable and valid tool for investigating multidimensional environmental attitudes due to its sufficient convergent validity and discriminant validity to support the development of the measurement model (Table 6). In line with previous studies of NEP/R [21, 23] and EAI [22], psychometric testing yielded different numbers of dimensions—between two and five for NEP/R [26, 27, 28] and between six and eleven for the EAI short version [34, 39, 42, 44]. Even though studies have reported that the 12-factor model offers empirical evidence [22, 39], the findings of these studies showed an insignificant correlation value in the 12-factor model; however, dimensions were not withdrawn. The factor that was not withdrawn due to its insignificance led to misleading results because the coefficient of correlation measures the strength associated with the dimensions and determines how much the variables have been changed due to insufficient discriminant validity between the dimensions (AVEs < r2).
These reported results support the validity and reliability of Milford and Duckitt's [22] EAI. Overall, the 7-factor model was shown to have unidimensional factors with high internal consistency, homogeneity, and test-retest reliability; however, it is not free from cultural bias. This is consistent with earlier studies, which showed that socio-demographic factors affect social-desirability bias and the internal consistency of the data of environmental attitudes studies [21, 45–46]. The replication of the EAI version in the Arab context in Study 1 and Study 2 was shown to extract factor structures that were different from the original constructs. It was found that Scale 1-Enjoyment of nature (EN), Scale 6-Environmental fragility (EF), Scale 8-Personal conservation behavior (PC), Scale 9-Human dominance over nature (HD), and Scale 12-Support for population growth policies (SP) were not considered and perceived necessary by students in the Arab context. Differences in the factor structures affected by the psychological constructs of the respondents of the current study and previous studies may be due to the cultural context that contributed to social-desirability bias.
This is supported by scholars and previous studies who stated that dismissing this factor may result in a statistical bias [22, 34, 42–44]. Wu [80] argued that results from previous research were not adequate for diverse populations and not sustained across cultures. Even though the characteristics and practices of the target population are common, cultural differences exist between norms, beliefs, and values; this may represent the social-desirability bias in the current study. Study 1 and Study 2 have established a baseline framework for environmental attitudes in the Arab context. The shortened version of the EAI was found to be reliable, valid, and sufficiently detailed in all aspects of environmental traits. In both studies, more than 40% of the sample did not provide usable data; this may have led to non-response bias affecting the results of the current study. Even though Howcroft and Milfont [37] have criticized the utility of the NEP/R for comparing scores across studies with different sample types and scale lengths, the usability of the EAI was found to be challenging due to a large number of items and redundancy in the positive and negative items. Feedback received from participants suggested that many students found the questionnaire to be excessively long and many of the items overly repetitive. The long-winded questionnaire of the original EAI has contributed to fatigue and a high percentage of unusable data.
Despite acknowledging this feedback, this study maintained the 72 items of the EAI to measure its dimensionality, reliability, and validity and to establish the utility of the environmental attitudes measure in the Arab context due to its psychometric stability across cultures [34, 47]. Since no study has been conducted in the Arab context to measure the reliability and validity of an environmental attitudes tool for investigating multidimensional environmental attitudes, it was rational and practical to adopt the EAI construct for the current study. The results from Study 1 and Study 2 revealed that the extraction of the items into a 7-factor model from the 12-factor model with between three and eight items for each dimension in Study 1, and three-item scales of all seven dimensions in Study 2 did not compromise the scale's internal consistency and reliability; thus, the content validity of the environmental attitude is intact.
Furthermore, the test-retest method is known for providing the opportunity for respondents to recall their feedback that would most likely impact their subsequent responses. Even though Altman and Bland [79] have suggested that certain criteria such as the same sample, same location, and a short interval between the tests must be fulfilled when establishing the repeatability approach, the test-retest approach entails the risk of recall bias due to the short time gap between the first and second studies [81]. Study 2 showed a significant reduction in the amount of complete and consistent feedback at 32.5% in comparison to Study 1 with 52.0%. A long-winded questionnaire and fatigue might have contributed to the signification reduction in the number of responses in these two studies. Since the responses from both studies were retrieved from the same sample, the negative experiences from the first study may have influenced the undergraduate students to provide void responses [82].
The usage of EFA should be strictly considered due to its usability as a data-driven method [76, 81]. On the other hand, CFA should be chosen to balance the aims of the study and concerns on theory-driven method [69, 76]. Even though some argue that EFA can explore the latent variable structure of a dataset [83], Byrne [57] contended that the CFA is the most robust method as it requires an a priori hypothesis to test whether the obtained dataset is suitable for the model. In this study, the PCA and EFA in Study 1 were used to verify the unidimensionality of the construct and were followed by the examination of a new factor structure extraction. The method was test-retesting, allowing for CFA to measure the model fit of the 7-factor model obtained from the same dataset in a different timeline. Unlike earlier studies that measured cross-cultural EAI on an educated sample, this study offers a different approach to administering the test on a well-educated sample, given the aim of determining the suitability of the EAI in the Arab context to establish the baseline framework. This may have resulted in a small difference in the factor structure. Despite confirmation of a similar structure in both studies, the relatively low reliability and validity of a few factors in Study 1 may also have contributed to the relative difference between factor structures.
Consistent with earlier studies, this study considered undergraduate students as educated (e.g.; [10, 34, 36, 38, 39, 43, 71] because they have learned about and experienced sustainability through curriculum, shared practice, and collaborative learning via co-curricular activities, experiences through sustainability research, and interdisciplinary collaboration [83]. Prior research has also proven that undergraduate students have a reasonable level of environmental awareness [10, 36, 34] to justify their attitudes toward pro-environmental behavior.
To overcome these limitations, and to acquire predictability and validity, prospective studies should include the gold standard of socio-demographic factors to further compare the environmental attitudes toward ecology issues. Zsoka et al. [84] clarified that previous studies have focused on the importance of attitude modeling. Sutton and Gyuris [47] have urged for future studies to explicitly address the need to measure change in people’s affective environmental attributes or actual behavior as a consequence of their ecological experiences. Although many studies have associated environmental attitudes with education and measures for a significant duration of the university experience, Chen et al. [85] argued that the effort to study environmental attitudes could be extended to less educated people, those with low employment status, and inhabitants of smaller cities. The uncommon function of the psychosocial aspect in environmental psychology studies in measuring environmental attitudes is consistent with suggestions that the social cost of environmental attitudes varies by social and cultural contexts. Thus, educational level may be helpful but not necessarily sufficient [85–89].
Conclusions
Collectively, this study provides a baseline for the environmental attitudes inventory in the Arab context. The results indicate that the 21-item scale of the EAI in Study 2 is a reliable and valid tool for investigating multidimensional environmental attitudes. Potential response bias due to response burden can be reduced. A reduction in some items of the scale will improve the response rate and completion rate compared to the longer version of the EAI. Furthermore, potential pitfalls due to cultural bias and social-desirability bias can be controlled for and minimized because items that are not considered important in relation to the environmental issues can be excluded from the survey. This paper established that the 72-item EAI with 12 dimensions by Milford and Duckitt [22] is a potential gold standard to measure environmental attitudes in the Arab context. It can be presented as a 21-item instrument to reduce the response burden, increase the response rate, and improve the completion rate while preserving the balance, dimensionality, reliability, and validity of the original version.
References
- 1.
Petty RE, Wegener DT. Attitude change: Multiple roles for persuasion variables. In: Gilbert D, Fiske S, Lindzey G. (Eds.). The Handbook of Social Psychology (4th ed.). New York, NY: McGraw-Hill; 1998. pp. 323–389.
- 2.
Krech D, Crutchfield RS. Theory and problems of social psychology. New York, NY: McGraw-Hill; 1948.
- 3.
Fishbein M, Ajzen I. Belief, attitude, intention, and behavior. Reading, MA: Addison-Wesley; 1975.
- 4. Schultz PW, Shriver C, Tabanico JJ, Khazian AM. Implicit connections of nature. J Environ Psychol. 2004; 24: 31–42.
- 5. Biasutti M, Frate S. A validity and reliability study of the attitudes toward sustainable development scale. Environ Education Res. 2017; 23: 214–230.
- 6. Sousa E, Quintino V, Palhas J, Rodrigues AM, Teixeira J. Can environmental education actions change public attitudes? An example using the pond habitat and associated biodiversity. PLoS ONE. 2016; 11(5): e0154440. pmid:27148879
- 7. Tsoi KH, Chan SY, Lee YC, Ip BY, Cheang CC. Shark conservation: an educational approach based on children’s knowledge and perceptions toward sharks. PLoS ONE. 2016; 11(9): e0163406. pmid:27684706
- 8. Geng L, Xu J, Ye L, Zhou W, Zhou K. Connections with nature and environmental behaviors. PLoS ONE. 2015; 10(5): e0127247. pmid:25985075
- 9. Sammalisto K, Sundström A, von Haartman R, Holm T, Yao Z. Learning about sustainability—What influences students’ self-perceived sustainability actions after undergraduate education? Sustainability. 2016; 8(6): 510.
- 10. DiBenedetto CA, Lamm KW, Lamm AJ, Myers BE. Examining undergraduate student attitude towards interdisciplinary education. J Agr Educ. 2016; 57: 167–178.
- 11. Dean AJ, Fielding KS, Newton FJ. Community knowledge about water: who has better knowledge and is this associated with water-related behaviors and support for water-related policies? PLoS ONE. 2016; 11(7): e0159063. pmid:27428372
- 12. Mamady K. Factors Influencing attitude, safety behavior, and knowledge regarding household waste management in Guinea: a cross-sectional study. J Environ Public Health. 2016; 9305768. pmid:27092183
- 13. Pooley JA, O’Connor MP. Environmental education and attitudes: emotions and beliefs are what is needed. Environ Behav. 2016: 32: 711–723.
- 14. de Leeuw A, Valois P, Ajzen I, Schmidt P. Using the theory of planned behavior to identify key beliefs underlying pro-environmental behavior in high-school students: Implication for educational interventions. J Environ Psychol. 42: 128–138.
- 15. Kormos C, Gifford R, Brown E. The influence of descriptive social norm information on sustainable transportation behavior: a field experiment. Environ Behav. 2015; 47: 479–501.
- 16. Tyrväinen L, Ojala A, Korpela K, Lanki T, Tsunetsugu Y. The influence of urban green environments on stress relief measures: a field study. J Environ Psychol. 2014; 38: 1–9.
- 17. Lönnqvist JE, Verkasalo M, Walkowitz G, Wichardt PC. Measuring individual risk attitudes in the lab: Task or ask? An empirical comparison. J Econ Behav Organization. 2015; 119: 254–266.
- 18. Lu J, Kim KH. Understanding self-report multidimensional attitudes scale toward people with disabilities: an exploratory analysis. Rehabil Psychol. 2017; 62(2): 110. pmid:28383946
- 19. Collins A, Cooper C. Measuring and managing the environmental impact of festivals: the contribution of the ecological footprint. J Sustain Tourism. 2017; 25: 148–162.
- 20. Hiramatsu A, Kurisu K, Hanaki K. Environmental consciousness in daily activities measured by negative prompts. Sustainability. 2016; 8(1): 24.
- 21. Dunlap RE. The New Environmental Paradigm Scale: from marginality to worldwide use. J Environ Educ. 2008; 40: 3–18.
- 22. Milfont TL, Duckitt J. The Environmental Attitudes Inventory: a valid and reliable measure to assess the structure of enviornmental attitudes. J Environ Psychol. 2010; 30: 80–94.
- 23. Dunlap RE, Van Liere KD. A proposed measuring instrument and preliminary results: the new environmental paradigm. J Environ Educ. 1978; 9: 10–19.
- 24. Dunlap RE, Van Liere KD, Mertig AG, Jones RE. Measuring endorsement of new ecological paradigm: A revised NEP scale. J Soc Issues. 2000; 56: 425–442.
- 25. Turaga RR, Howarth RB, Borsuk ME. Pro-environmental behavior: Rational choice meets moral motivation. Ann Ny Acad Sci. 2010; 1185: 211–224. pmid:20146771
- 26. Wen X, Shouying Z. The Environmental worldviews and climate change mitigation behaviors: Testing the New Ecological Scale in the smallest space analysis for Chinese samples. International J Environ Sci Dev. 2015; 6: 547–550.
- 27. Shephard K. e is for exploration: assessing hard-to-measure learning outcomes. Brit J Educ Technol. 2009; 40: 386–398.
- 28. Amburgey JW, Thoman DB. Dimensionality of the new ecological paradigm: issues of factor structure and measurement. Environ Behav. 2012; 44: 235–256.
- 29. Harraway J, Broughton-Ansin F, Deaker L, Jowett T, Shephard K. Exploring the use of the revised New Ecological Paradigm Scale (NEP) to monitor the development of students' ecological worldviews. J Environ Educ. 2012; 43: 177–191.
- 30. Khan A, Khan MN, Adil M. Exploring the New Ecological Paradigm (NEP) scale in India: item analysis, factor structure and refinement. Asia-Pacific J Manage Res Innov. 2012; 8: 389–397.
- 31. Lundmark C. The new ecological paradigm revisited: anchoring the NEP scale in environmental ethics. Environ Educ Res. 2007; 13: 329–347.
- 32. Olli E, Grendstad G, Wolleback D. Correlates of environmental behavior: bringing back social context. Environ Behav. 2001; 33: 181–208.
- 33.
Anderson MW. (2012). New ecological paradigm (NEP) scale. In: The Berkshire Encyclopedia of Sustainability: Measurements, Indicators and Research for Sustainability. Berkshire, UK: Berkshire Publishing Group; 2012. pp. 260–262.
- 34. Milfont TL, Duckitt J, Wagner C. A cross-cultural test of the value-attitude-behavior hierarchy. J Appl Soc Psychol. 2010; 40: 2791–2813.
- 35. Tan KP. Anthropomorphism of nature and efficacy in coping with the environmental crisis. Soc Cognition. 2014; 32: 276–296.
- 36. Erdogan N. Testing the new ecological paradigm scale: Turkish case. African J Agr Res. 2009; 4: 1023–1031.
- 37. Howcroft LJ, Milfont TL. The use (and abuse) of the new enviornmental paradigm scale over the least 30 years: a meta analysis. J Environ Psychol. 2010; 30: 143–158.
- 38. Bronfman NC, Cisternas PC, López-Vázquez E, de la Maza C, Oyanedel JC. Understanding attitudes and pro-environmental behaviors in a Chilean community. Sustainability. 2015; 7: 14133–14152.
- 39. Milfont TL, Duckitt J. Preservation and utilization: understanding the structure of enviornmental attitude. Medio Ambiente y Comportamiento Humano. 2006; 7: 29–50.
- 40. Blaikie NH. The nature and origins of ecological world view: an Australian study. Soc Sci Quart. 1992; 73: 144–165.
- 41. Bogner FX, Wiseman M. Toward measuring adolescent environmental perception. Eur Psychologist. 1999; 4: 139–151.
- 42. Koc I, Kuvaca M. Preservice science teachers' attitudes toward environment. Int J Environ Sci Educ. 2016; 11: 5925–5941.
- 43. Ferreira MC, Milfont TL, Corrêa e Silva AP, Fernandes HA, Almeida SP, Mendonça H. Evaluation of psychosocial stressors in the labor context scale: development and psychometric evidence. Psicologia: Reflexão e Crítica. 2015; 28: 340–349.
- 44. Yu JK, Yokohari M, Tanaka T. Public perceptions of nature and landscape preference in Singapore. Hum Ecol. 2014; 42: 979–988.
- 45. Grendstand G, Sundback S. Socio-demographic effects on cultural biases: a Nordic study of grid-group theory. Acta Sociol. 2003; 46: 289–306.
- 46. Milfont TL. The effects of social desirability on self-reported environmental attitudes and ecological behavior. Environmentalist. 2009; 29: 263–269.
- 47. Sutton SG, Gyuris E. Optimizing the enviornmental attitudes inventory: Establishing a baseline of change in students' attitudes. Inter J Sustain High Educ. 2015; 16: 16–33.
- 48. Boyer P, Lienard P, Xu J. Cultural differences in investing in others and in the future: why measuring trust is not enough. PLoS ONE. 2012; 7(7): e40750. pmid:22848399
- 49. Kågesten A, Gibbs S, Blum RW, Moreau C, Chandra-Mouli V, Herbert A, Amin A. Understanding factors that shape gender attitudes in early adolescence globally: a mixed-methods systematic review. PLoS ONE, 2016; 11(6): e0157805. pmid:27341206
- 50. Sircova A, Karimi F, Osin EN, Holme P, Strömbom D. Simulating irrational human behavior to prevent resource depletion. PLoS ONE, 2015; 10(3): e0117612. pmid:25760635
- 51. Hope AL, Jones CR. The impact of religious faith on attitudes to environmental issues and carbon capture and storage (CCS) technologies: a mixed methods study. Technol Soc. 2014; 38: 48–59.
- 52. Sayles JS, Mulrennan ME. Securing a future: Cree hunters’ resistance and flexibility to environmental changes, Wemindji, James Bay. Ecol Soc. 2010; 15(4): Available from http://www.ecologyandsociety.org/vol15/iss4/art22/.
- 53. Kennedy EH, Beckley TM, McFarlane BL, Nadeau S. Why we don’t “Walk the Talk”: understanding the environmental values/behaviour gap in Canada. Res Hum Ecol. 2009; 16: 151–160.
- 54. López AG, Amerigo M. Relationship among values, beliefs, norms and ecological behaviour. Psicothema. 2008; 20: 623–629. pmid:18940060
- 55. Schwarz N. Attitude construction: evaluation in context. Soc Cogn. 2007; 25: 638–656.
- 56. Wicker AW. Attitudes versus actions: the relationship of verbal and overt behavioral responses to attitude objects. J Soc Issues. 1969; XXV: 41–78.
- 57.
Byrne BM. Structural equation modeling with AMOS: Basic concepts, applications and programming (2nd ed.). Mahwah, NJ: Erlbaum; 2010.
- 58.
Field A. Discovering Statistics Using IBM SPSS Statistics (4th ed). London: Sage Publication; 2013.
- 59.
SPSS Inc. Released 2009. PASW Statistics for Windows, Version 18.0. Chicago, IL: SPSS Inc.; 2009.
- 60.
Hair JF, William CB, Babin BJ, Anderson RE, Tatham RL. Multivariate data analysis. New Jersey, NJ: Pearson University Press; 2006.
- 61.
Gravetter F, Wallnau L. Essential of statistics for the behavioral sciences (8th ed.). Belmont, CA: Wadsworth; 2014.
- 62. Kenny D A, Kaniskan B, McCoach DB. The performance of RMSEA in models with small degrees of freedom. Socio Meth Res. 2015; 44: 486–507.
- 63.
Kline RB. Principles and Practices of Structural Equation Modeling (3rd ed.). New York, NY: Guilford Press; 2010.
- 64. Raykov T. Coefficient alpha and composite reliability with interelated nonhomogeneous. Appl Psychol Meas. 1998; 22: 375–385.
- 65.
Satorra A Bentler PM. Corrections to test statistics and standard errors in covariance structure analysis. In von Eye A., & Clogg C. C., Latent variables analysis: Applications for developmental research Thousand Oaks, CA: Sage; 1994. pp. 399–419.
- 66. Kenny D A, Kaniskan B, McCoach DB. The performance of RMSEA in models with small degrees of freedom. Socio Meth Res. 2015; 44: 486–507.
- 67. Ziegler M. Stop and state your intentions! Let's not forget the ABC of test construction. Eur J Psychol Assess. 2014; 30: 239–242.
- 68.
Arbuckle JL. AMOS (Version 23.0) [Computer Program]. Chicago, IL: IBM SPSS; 2014.
- 69. Anderson JC, Gerbing DW. Structural equation modeling in practice: a review and recommended two-step approach. Psychol Bull. 1988; 103: 411–423.
- 70. Kim TH, White H. On more robust estimation of skewness and kurtosis: simulation and application to the S&P 500 index. Financ Res Lett. 2003; 1: 56–70.
- 71. Fornell CG, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981; 18: 39–50.
- 72. Ziegler M, Hagemann D. Testing the unimensionality of items: pitfalls and loopholes. Eur J Psychol Assess. 2015; 31: 231–237.
- 73. Kim H, Ku B, Kim JY, Park YJ, Park YB. Confirmatory and exploratory factor analysis for validating the Phlegm Pattern Questionnaire for healthy subjects. Evid Based Complement. Alternat Med. 2016; 2696019. pmid:27051447
- 74. Cho E. Making reliability reliable: a systematic approach to reliability coefficients. Organ Res Meth. 2016; 19: 651–682.
- 75. Russell DW. In search of underlying dimensions: The use (and abuse) of factor analysis in Personality and Social Psychology Bulletin. Pers Soc Psychol Bull. 2002; 28(2): 1629–1646.
- 76.
Gravetter F, Wallnau L. Essential of statistics for the behavioral sciences (8th ed.). Belmont, CA: Wadsworth; 2014.
- 77. Chen F, Curran PJ, Bollen KA, Kirby J, Paxton P. An empirical evaluation of the use of fixed cutoff points in RMSEA test statistic in structural equation models. Socio Meth Res. 2009; 36: 462–494.
- 78.
Kline RB. Principles and Practices of Structural Equation Modeling (3rd ed.). New York, NY: Guilford Press; 2010.
- 79. Altman DG, Bland JM. Measurement in medicine: the analysis of method comparison studies. The Statistician. 1983; 32: 307–317.
- 80. Wu JF. Reproducing an alternative model of intercultural sensitivity for undergraduates from Taiwan and Hainan. Inter J Mod Educ Comput Sci. 2016; 1: 1–7.
- 81. Pannucci CJ, Wilkins EG. Identifying and avoiding bias in research. Plastic Reconstructive Surgery. 2010; 126(2): 619–625. pmid:20679844
- 82.
Root D, Blismas NG. Increasing questionnaire responses from industry: practices surrounding the use of postal questionnaires. 19th Annual ARCOM Conference, Vol. 2 (pp. 623–631). Brighton: Association of Researchers in Construction Management; 2003.
- 83. Chankrajang T, Muttarak R. Green returns to education: Does schooling contribute to pro-environmental behaviours? Evidence from Thailand. Ecol Econ. 2017; 131: 434–448.
- 84. Zsoka A, Marjaine SZ, Szechy A, Kacsis T. Greening due to environmental education? Environmental knowledge, attitudes, consumer behaviour, and everyday pro-environmental activities of Hungarian high school and university students. J Cleaner Prod. 2013; 48: 126–138.
- 85. Chen X, Peterson MN, Hull V, Lu C, Lee GD, Hong D, Liu J. Effects of attitudinal and sociodemographic factors on pro-environmental behaviour in urban China. Environ Conserv. 2011; 38: 45–52.
- 86. Evans-Lacko S, Knapp M. Importance of social and cultural factors for attitudes, disclosure and time off work for depression: findings from a seven country European study on depression in the workplace. PLoS One, 2014; 9. pmid:24622046
- 87. Newman TP, Fernandes R. A re-assessment of factors associated with environmental concern and behavior using the 2010 General Social Survey. Environ Educ Res. 2016; 22: 153–175.
- 88. Schröder T, Wolf I. Modeling multi-level mechanisms of environmental attitudes and behaviours: the example of carsharing in Berlin. J Environ Psychol. 2017; 52: 136–148.
- 89. Le Hebel F, Montpied P, Fontanieu V. What can influence students’ environmental attitudes? Results from a Study of 15-year-old students in France. Int J Environ Sci Educ. 2014; 9: 329–345.