Ethics position theory

Both philosophical and psychological analyses of morality have consistently identified two critical determinants of moral decision-making: the degree of harm or benefit produced by the action and the consistency of the action with standards that define what is considered moral. Philosophers have traditionally contrasted moral theories based on principles (deontological models) and models that stress the consequences of actions (teleological models). Similarly, psychological researchers (e.g., [1,8,9]) have explored the interplay between rule-based judgment and sensitivity to welfare of others.

Ethics position theory agrees that actions that cause harm or are inconsistent with commonly accepted standards of morality trigger moral scrutiny, but the theory maintains that people differ in their orientation toward both influences. Some people are more idealistic than others, for they believe that “If an action could harm an innocent other, then it should not be done.” Some people, too, believe that “what is ethical varies from one situation and society to another,” but others disagree with such moral relativism. Schlenker and Forsyth [10] initially identified these two themes in their study of people’s moral judgments of Milgram’s [11] studies of obedience, and Forsyth [3] extended these findings by developing the EPQ to measure both idealism and relativism. The idealism items stressed humanitarian, visionary ideals, such as avoiding others’ harm, whereas the relativism scale includes items that question the value of relying on moral rules or codes of ethics when making moral judgments.

Ethics positions. Ethics position theory assumes people range along two continua defined by their degree of idealism and relativism, but it also suggests individuals’ positions along each continuum defines their ethical type. The two-dimensional model of ethical positions shown in Table 1 considers both principles (relativism) and outcomes (idealism), only principles, only outcomes, or neither. Exceptionists are relatively conventional in their moral orientation. They disagree that morality is purely personal but recognize that moral principles do not always minimize harm. Subjectivists are skeptical about basing moral judgments on trans-situational and trans-personal rules (high relativism), but they do not strongly endorse the “do no harm” mandate (low idealism). Absolutists, unlike both the exceptionists and the subjectivists stress harm minimization, but as the name suggests, endorse moral standards. Situationists, in contrast, do not believe that moral standards provide a bright line between what is morally good and bad (relativism), but they are humanitarian in their orientation—they are committed to promoting human well-being.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The four moral types identified in EPT.

https://doi.org/10.1371/journal.pone.0251989.t001

Factor structure of the EPQ. The EPQ items were selected to measure idealism and relativism—and only idealism and relativism. A number of investigators have tested this assumption using exploratory or confirmatory factor analysis (e.g., [12–26]). These investigations generally confirmed the EPQ’s two-dimensional structure, but some studies suggested the original EPQ could be improved through additional scaling and psychometric refinement. Several studies, too, suggested that certain items on the EPQ are not linked to either idealism or relativism, but instead formed a coherent subcluster within these two domains. For example, Davis and colleagues [17], using confirmatory factor analysis, identified a third factor which included the two items that pertained specifically to lying (veracity). Moreover, several investigators, after administering the EPQ, found that certain items were not highly correlated with the other items, and so removed those items to achieve acceptable internal reliability. When Forsyth et al. [4] meta-analytically reviewed a set of 83 studies that used the EPQ between 1980 and 2006 with 140 different samples, they identified 25 (18.1%) that used a shortened version of the idealism scale, and 30 (22.9%) that used a shortened version of the relativism scale. In consequence, observed differences in results across studies may be due to true differences in participant features or study contexts (e.g., occupation, nationality, generational cohort), but these differences may simply be statistical artifacts that result from the use of altered versions of the EPQ.

Overview of studies

Prior research leaves unanswered several significant questions about the inventory that measures these variations—the EPQ—and the taxonomy of moral types it describes. The EPQ was designed to assess two latent constructs—idealism and relativism—but previous researchers have not consistently confirmed the measure’s psychometric and conceptual adequacy.

We sought to reduce this uncertainty in the current research by conducting an empirically driven revision and validation of the EPQ. In Study 1, we evaluated the EPQ’s factor structure, item adequacy, scale length, and psychometric properties to determine the most accurate and efficient means by which to measure individual’s ethics positions. In Study 2, we replicated the factor structure, tested the EPQ for measurement invariance across men and women, and used the revised measure to classify individuals into the four proposed categories. In Study 3, we gathered information about the convergent and divergent validity of the EPQ. The EPQ has been used by several researchers in their studies morality and moral judgment, and the findings are generally support the validity of both the idealism and relativism scale. However, the relationship between the EPQ and other measures of individual differences in moral values requires further investigation. If the EPQ is a valid measure of people’s level of idealism and relativism, and these two sets of beliefs influence how they think about moral issues, then the EPQ should successfully predict individual differences in related moral beliefs and values, including Schwartz’s [27] social values scale and the moral foundations questionnaire developed by Haidt and his colleagues (e.g., [28,29]).

Methods and materials

Results and discussion

The CFA results are presented in Table 2 and more detailed discussion of the various model specifications are provided in the Electronic Supplemental Materials. The full, 20-item scale results of the EPQ were middling whether modelled as the theorized two-factor solution (Two-factor-I & R) or with the inclusion of veracity modelled as an additional factor (Three-factor-I, R, & V). The RMSEA and SRMR were generally adequate, but CFI values all fell under the .95 threshold [37]. Average variance extracted (AVE) for these models was also marginal and a number of factor loadings in both the two and three factor full scale specifications were less than .40. Modification indices noted sizable, correlated residuals for several of the items.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Confirmatory factor analyses for Study 1.

https://doi.org/10.1371/journal.pone.0251989.t002

Excluding the two items in the relativism scale that pertained to lying/veracity (Two-factor-I & R, no V) did little to improve overall model fit. In particular, the number and magnitude of the correlated residuals among several items indicated that some of the items not only share variance that is attributable to latent levels of idealism and relativism, they also share variance attributable to external artifacts. These external artifacts can be substantive (e.g., two items prone to socially desirable responding) or a function of the measurement (e.g., two overly complicated items that might elicit uncertain responses at the midpoint of the Likert scale).

Item analysis.

As the CFA findings would suggest, item analysis indicated the two scales of the EPQ lacked sufficient internal consistency: despite their content being aligned with EPT, the signal to noise ratio for the 18 items that remained after eliminating the veracity items was unacceptable, and merited further EPQ refinement. We therefore examined each item’s content to ensure its alignment with EPT as well as consulting past research on the EPQ’s dimensionality. We investigated the five items on the idealism scale and three items on the abridged relativism scale with the lowest factor loadings. Upon examining these items, they included content that was narrow in scope or not clearly related to the conceptual meaning of the scale. For example, the primary items on the idealism scale emphasize harm avoidance (Items 1, 3, 4, and 6). Item 2, however, raises the issue of risk—“risks to another should never be tolerated”—but the word risk can mean very different things to different people, including uncertainty, hazard, probability and even danger [38]. Similarly, Item 10 suggests one strive to reach the ideal of “perfect action,” and makes no mention of harm at all. Several other items on the relativism scale were similarly narrow in scope or include extraneous content, such as Item 18: “Rigidly codifying an ethical position that prevents certain types of actions could stand in the way of better human relations and adjustment.” This item, in addition to using loaded wording (“rigidly codifying”) does not pertain to the core focus of the relativism inventory: skepticism regarding moral rules. Other low-loading items were wordy and cognitively demanding (e.g., “Deciding whether or not to perform an act by balancing the positive consequences of the act against the negative consequences of the act is immoral”). We speculate that these features were the reasons for their poor loadings, correlated residuals, and were likely the primary contributing factor to the middling overall model fit.

Both idealism and relativism are unidimensional constructs, thus these items do not represent meaningful facets of idealism or relativism. Rather, they are imperfect indicators of overarching constructs. Based on the collective evidence, we concluded that their exclusion would not result in a loss of construct validity, create a deficiency bias, or detriment the EPQ’s other psychometric properties. The result of this scale revision was that idealism and relativism would both be measured with five items each (henceforth referred to as the EPQ-5, to indicate the number of items on each subscale and each subscale is a distinct measurement index).

The EPQ-5. A CFA of the shortened EPQ (Table 2) indicated the EPQ-5 performed exceptionally well and exceeded all traditional thresholds (CFI = .951, RMSEA = .057 (.052; .063), SRMR = .041). Table 3 provides more detailed information at the item and factor level. All factor loadings exceeded .40 and all but one item was .50 or greater. The deletion of any item resulted in a decrease in Cronbach’s alpha, and despite composite reliability being strongly influenced by the number of items, composite reliability decreases were minimal for the EPQ-5 relative to the full version (Δ idealism = .01, Δ relativism = .03). Further, compared to the original scale, the average variance extracted (AVE) improved greatly for both idealism and relativism (AVE-Idealism = .45; AVE-Relativism = .33).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Psychometrics for EPQ for Study 1 (left) and Study 2 (right).

https://doi.org/10.1371/journal.pone.0251989.t003

The instructions and items for the EPQ-5 are included in the Appendix. The mean and median scores for this sample were 3.83 (sd = 1.06) and 4.0 for idealism, and 3.22 (sd = 1.15) and 3.0 for relativism, and the scores show a slight negative skew; -0.77 and -0.17, respectively. Both scales are internally consistent (αs = 0.80 & 0.70, respectively) and both correlate with the longer, original measures of idealism and relativism; rs = 0.93 and 0.89, respectively. As with the original scale and consistent with EPT, the correlation between idealism and relativism scores was only -0.04.

Methods and materials

Results and discussion

Measurement invariance across the sexes.

When we compared the responses of the 4,087 women and 6,153 men who completed the EPQ-5, we discovered the women’s average scores for both idealism and relativism were higher than the men’s scores; Fs(1, 10,235) = 524.34 and 23.57, ps < .001. The idealism and relativism means (and medians) were 3.73 (3.70) and 3.15 (3.10) for women and 3.32 (3.30) and 3.06 (3.00) for men. However, it is unclear whether these mean differences reflect true differences between the sexes on idealism and relativism or differences in interpretation of the EPQ-5. Determining this requires tests of measurement and structural invariance (MI/SI).

Using procedures outlined elsewhere [41,42], we tested for configural, metric, scalar, and residual invariance tests for MI and factor variance and factor mean invariance tests for SI were conducted in Mplus 8.0 [43]. The summary fit indices are presented in Table 4, the standardized factor loadings for men and women are presented in Fig 1, and the full details and model results are reported in the Electronic Supplemental Materials.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Standardized factor loadings across sexes.

Note: I = Idealism, R = Relativism. Item numbers correspond to the ten items of the EPQ-5 supplied in the Appendix.

https://doi.org/10.1371/journal.pone.0251989.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Tests of measurement invariance and structural invariance (Study 2).

https://doi.org/10.1371/journal.pone.0251989.t004

The EPQ-5 demonstrated strict measurement invariance meaning that men and women had equivalent factor loadings, intercepts, and error variances. That is, observed differences in item means between men and women were attributable to factor (true score) mean differences only. Regarding structural invariance, men had slightly larger variances in both relativism and idealism, and women had significantly higher latent mean levels of idealism (women higher). Latent mean differences for relativism were negligible. Using decision rules outlined in Chen [44], the retained model was the fully constrained measurement model that only allowed latent mean differences in idealism to vary (χ2(97) = 1329, CFI = .964, RMSEA = .050 [.047; .052], SRMR = .062). We conclude that the EPQ-5 is equally interpreted across the sexes, women are higher in idealism than men, and regarding MI and SI, the EPQ-5 is psychometrically valid.

Classification analysis.

Ethics position theory assumes differences in moral thought are determined, in part, by differences in idealism and relativism, but it also suggests that these two dimensions combine to define a person’s ethics position. As Table 1 summarized, situationists are relativistic and idealistic, subjectivists are relativistic but not idealistic, absolutists are not relativistic but idealistic, and exceptionists are neither relativistic nor idealistic. These moral types are assumed to be qualitatively different from one another and, to some extent, mutually exclusive.

Although the theory predicts that individuals’ level of idealism and relativism determines their distinct ethics positions, the four groups identified in Table 1 may not correspond to the natural groups that exist within the population. Rather than four categories, people may naturally cluster into fewer categories, more categories, or no categories at all. Fortunately, researchers have developed several statistical procedures to identify clusters of individuals within a more general population, including Bayesian expectation maximization algorithms, cluster analysis, latent class modeling, and taxonometric procedures [45]. We used one of these methods, cluster analysis, to compare the theory’s recommended taxonomy of ethics positions to the naturally occurring distribution of individuals into cluster or subgroups. A high degree of correspondence between these two classification methods would provide support for the theory’s type-based approach to ethics positions.

Classifications into types. We first classified each individual, a priori, into one of the four ethics positions using a normative procedure. Individuals whose idealism and relativism scores were above average, with the average defined by the median for the sample (3.6 and 3.1), were considered to be high idealists and high relativists. Those with scores below the median were classified as low idealists and relativists. The number of respondents in each ethics position were nearly equal: 27% were classified as exceptionists, 25% as absolutists or situationists, and 23% as subjectivists (see Table 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Ethics position as determined by relative idealism and relativism scores (EPQ classification) and by cluster analysis.

https://doi.org/10.1371/journal.pone.0251989.t005

We used hierarchical cluster analysis (HCA, Ward’s technique) to identify naturally existing subgroups within this sample. This method’s agglomerative algorithm calculates similarities, based on Euclidian distances, between possible clusters. It initially classifies the most similar persons into clusters, and then at each subsequent step identifies those individuals who are most siilar to those clusters. The final step in the algorithm combines all clusters.

Inspection of the magnitude of the squared Euclidean distances between the clusters combined in the final 12 steps in this process suggested that a 4-cluster solution best fit the data. The magnitude of the distances remains relatively constant as clusters are combined into fewer and fewer clusters, until 4 clusters are combined into three. This step generated the largest acceleration of distance growth—the “elbow” indicating where to stop combining clusters (see the Electronic Supplemental Materials).

Degree of Congruence. The 4-group classification generated through cluster analysis was congruent with the a priori classification of individuals based on their ethics positions. The two classifications matched for 89.0% of the cases, although accuracy varied across the four ethics positions. The two methods of classification agreed for 85.8% of the exceptionists, 83.8% for subjectivists, 86.3% for absolutists, and 100% for the situationists. Discrepancies, when they did occur, usually resulted from differences on one rather than two of the EPQ factors. For example, 302 individuals were classified as absolutists by one method and situationists by the other. Both of these ethics positions are idealistic ones, although they differ in relativism. Similarly, the two classification methods were generally congruent in their classification of the respondents who were relativists (situationists and absolutists), but they diverged because one identified 322 as subjectivists but the other classified them as situationists.

Methods and materials

Results and discussion

The correlations among idealism and relativism—as measured by the original EPQ and the shortened EPQ-5—and dimensions of moral values assessed by the SVS and the moral foundations assessed by the MFQ are presented in Table 6 (controlling for gender). These correlations provide support for the adequacy of the EPQ-5, as well as evidence of convergent and divergent validities.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Correlational comparison of the original EPQ and the revised EPQ (EPQ-5) in predicting associations with moral values (SVS: Schwartz Value Scale) and moral foundations (MFQ: Moral Foundations Questionnaire).

https://doi.org/10.1371/journal.pone.0251989.t006

Concurrent validity. The correlations between the scores on the SVS and MFQ and the EPQ’s idealism and relativism scores were consistently larger for the original EPQ in comparison to the EPQ-5, but as Table 6 indicates, the differences were relatively unsubstantial, ranging from .01 to .05. These correlations suggest that the revised EPQ, despite including fewer items, yields results that are essentially equivalent to the original inventory. The correlations between the original and revised idealism and relativism scales were .94 and .93, respectively.

Convergent and divergent validity. The correlations in Table 6 indicate that idealism was more closely associated with the self-transcendence values (universalism & benevolence) of the SVS than conservation or self-enhancement and was not associated with either openness or hedonism. Relativism, in contrast, was negatively correlated with conservation, and more substantially correlated with both openness and hedonism. Table 6 also indicates that idealism was more closely associated with the harm and fairness scales of the MFQ than with the three factors of loyalty, authority, and sanctity (all ps < .001 given the large sample sizes). Relativism, in contrast, was more substantially correlated with loyalty, authority, and sanctity than with harm and fairness.

Ethics positions. We also examined the relationship between the two scales of the EPQ and individuals’ values in two separate 2 (idealism: low and high) X 2 (relativism: low and high) X 2 (sex: men and women) MANOVAs—one for the SVS variables and one for the MFQ variables—with age as a covariate. These results, which are presented in the Electronic Supplemental Materials, are consistent with the correlational analyses presented in Table 6. Individuals who adopted ethics positions that are idealistic—absolutists and situationists—differed from more pragmatic individuals—exceptionists and subjectivists—for all the SVS moral values (except hedonism) and all the MFQ foundations (except purity). Individuals who adopted relativistic ethics positions—the situationists and subjectivists—differed from the nonrelativists—absolutists and exceptionists) in the conservation, openness, and hedonism moral values, and the MFQ dimensions of loyalty, authority, and purity. However, the interaction of sex, idealism, and relativism was significant at the multivariate level for the SVS moral value measures, in some cases qualifying the more general effects of relativism and idealism. For example, both idealistic men and women had higher self-transcendence values than low idealists, but exceptionist men were particularly low in self-transcendence, whereas situationist women scores were substantially elevated for this value. In contrast, relativists were more open than nonrelativists, but situationist women had particularly elevated scores on openness, particularly in comparison to exceptionist men. In contrast, the interaction of idealism and relativism was not qualified by sex in the analysis of the MFQ items. As the correlational analyses suggested, the main effect of idealism was more substantial for harm and fairness, whereas the main effect of relativism was more robust for loyalty, authority, and sanctity. In addition, the interaction of idealism and relativism was significant only for loyalty, authority, and sanctity. Exceptionists were not harm or fairness oriented, but more likely than all other ethics positions to value loyalty, authority, and sanctity. Subjectivists matched exceptionists’ relative disinterest in harm and fairness, but they also did not value loyalty, authority, and sanctity; they valued sanctity lower than all other groups. Absolutists and situationists concurred in their emphasis on harm and fairness but diverged in their weighing of loyalty and authority; absolutists were average, overall, in their emphasis on loyalty and authority, whereas situationists were below average. The two sets of idealists, however, diverged more substantially on sanctity; absolutists endorsed sanctity as a moral foundation, whereas situationists did not.

Study implications

Just as reliability is the ceiling for validity, the ceiling for theoretical advances is the accuracy of measurement. The methodological, empirical, and theoretical implications of the research reported here are considerable. Focusing on methodology, this study both affirms the psychometric adequacy of the EPQ, but also refines the measure to index individual differences more accurately in moral thought. The revised scale, given its length and the removal of items that were not reliably associated with the scales’ underlying constructs, will offer researchers the means to assess ethics positions more easily and also reduce the variation in the instrument’s content across studies. As stated above, a common practice with the full length EPQ is the dropping and altering of items [4]. This is often engaged in to reduce survey lengths and/or improve the reliability scale. However, this distorts a measure’s psychometrics as well as impedes the robustness of systematic reviews that cannot tease apart variance attributable true score differences from variance attributable to differences in the measure [58]. The EPQ-5 provides an already concise measure and one that does not and should not require further reduction.

Empirically, the findings support the central tenets of EPT. We found that idealism and relativism are coherent psychological constructs, and the EPQ-5 accurately measures these constructs. Moreover, the findings also support the theory’s prediction of both a direct and a configural effect of one’s level of idealism and relativism on moral thought, including moral values and foundations. Variations in idealism and relativism are independently associated with values and foundations, but in some cases their relationship to this outcome is a conjoint one. These interactive effects of idealism and relativism provide indirect justification for the taxonometic assumptions of the theory.

Theoretically, the findings are congruent with recent advances in moral psychology and provide the means to more fully integrate those efforts. Despite the critical importance of morality for maintaining stable interpersonal relationships in human societies, disagreement over what is moral and what is immoral is common. Many factors contribute to this diversity, but the current work identifies one significant factor: differences in ethics positions. This analysis is consistent with recent advances in moral psychology, particularly the focus on individual differences that have been examined extensively in philosophical analyses of ethics. Ethics position theory explicitly identifies the link between moral deontological and consequentialist philosophies and psychological judgments of what is and is not construed as moral.

Limitations and future directions

The three studies contain several limitations that future research can address. First, we demonstrated and replicated the factor structure of the EPQ-5, but our samples were composed primarily of individuals hailing from English speaking countries. Although past research has repeatedly shown that the original EPQ largely retains its reliability and validity when translated, the EPQ-5 has yet to go through this thorough vetting. We recommend to scholars using the EPQ-5 that along with standard cross-cultural procedures (e.g., back translate; [59]), that the factor structure be examined and, if necessary, demonstrate measurement invariance in new contexts.

Second, we demonstrated convergent and divergent validity with two prominent scales that measure moral foundations and values, but other individual differences such as the five-factor model of personality, dark triad traits, as well as other inventories and scales related to ethics such as the Multidimensional Ethics Scale [60,61] should be examined to determine the EPQ-5’s nomological net. In addition, all participants in these studies were administered the EPQ online. Although there is generally little difference between the factor structures, loadings, and the understanding of item content between online participants and those responding in person, differences in mean latent levels have been documented [62].

Third, we cannot speak to the predictive validity of the EPQ-5 regarding ethical behavior. The EPQ has received mixed support for predicting (un)ethical acts and intentions [63–66] and perhaps with the streamlined and robust EPQ-5, greater headway can be made in understanding how ethical position influences ethical intentions and behaviors. Furthermore, we encourage researchers to consider intervening variables (i.e., mediators) to understand this process.

Fourth, although we conducted measurement and structural invariance tests for men and women, there are other key groups of individuals that could be examined. Significant research using the EPQ has tested and found differences in mean levels of idealism and relativism across professions [67], cultures [4], and racial identities [68], and determining whether these differences are attributable to true differences across these groups or are an artifact of the measurement would further establish the robustness or boundary conditions of the EPQ-5.