HOT and Attractive? The Hazardous Organization Tool as an Instrument to Avoid Attracting and Retaining People with Low Ethical Standards

People low versus high in honesty-humility can be characterized by traits related to deceitfulness, slyness, greediness, and pretentiousness versus sincerity, fairness, greed avoidance, and modesty (Ashton et al., 2004). Previous research has shown that people low rather than high in honesty-humility have slightly worse in- and out-role job performance (Lee et al., 2019; Pletzer et al., 2021) and are more likely to engage in counterproductive workplace behaviors, including misconduct (e.g., knowledge hiding; Ogunfowora et al., 2022), deviance (e.g., absenteeism; De Vries & Van Gelder, 2015; Lee et al., 2019; Pletzer et al., 2019), and unethical business decisions (De Vries et al., 2017). This suggests that people low in honesty-humility are more likely to pursue personal interests at the expense of others when the situation permits (De Vries et al., 2016a, 2016b), presumably due to their indifference to ethics and their ability to morally disengage from their wrongdoing (De Vries & Van Kampen, 2010; Ogunfowora et al., 2022). Importantly, honesty-humility is strongly associated with morality (0.66; De Vries & Van Kampen, 2010). Taken together, this literature suggests that honesty-humility is a sound measure of people’s endorsement of high ethical standards and a strong predictor of their ethical behavior in the workplace.

To ensure that we could identify the characteristics of hazardous organizations as comprehensively as possible, we reviewed the literature listed on hexaco.org, a HEXACO personality research repository that lists most articles on the HEXACO personality model that have been published by the end of 2019.³ To supplement the list, we manually searched for articles published after 2019. Based on this review, we identified six characteristics that correspond to our definition of hazardous organizations. The six characteristics were differentiated based on content overlap agreed upon by the authors. Three of these characteristics reflect the ease with which people in organizations can obtain Individual Ends that are more attractive to people low rather than high in honesty-humility (i.e., sex, power, and money; De Laat, 2020; Lee et al., 2013; Zettler et al., 2020). The other three characteristics reflect configurations of descriptive organizational norms (i.e., commonly enacted behaviors that create an expectation to conform to the same behavior; e.g., Dannals & Miller, 2017) that provide Organizational Means for people low in honesty-humility to pursue their personal ends. We elaborate on each of these characteristics below (see Table 1 for a summary).

The first characteristic of hazardous organizations is Sex Tolerance. People low in honesty-humility are more likely to engage in uncommitted sexual “quid pro quo” relationships (e.g., Ashton & Lee, 2008; Lee et al., 2013). We therefore speculate that applicants low rather than high in honesty-humility may be more attracted to organizations where “loose” sexual behaviors and misconduct are more often tolerated (e.g., exchanging sexual favors to obtain a promotion is more tolerated).

The second characteristic of hazardous organizations is Power Inequality. People low in honesty-humility may prefer a social system that perpetuates inequality and prioritizes enhancing self-esteem through the pursuit of power and success over protecting the welfare of others (Lee et al., 2010; Leone et al., 2012). Moreover, in Lukacik and Bourdage’s (2020) study, applicants low in honesty-humility were more attracted to organizations with Powerful and Achievement-Oriented organizational images. Therefore, vertical organizations that support power inequality and instigate internal competition are more likely to be attractive to applicants low rather than high in honesty-humility.

The third characteristic of hazardous organizations is Money Orientation. People low in honesty-humility feel entitled to higher levels of wealth and are attracted to making more money, even if it means taking financial or physical risks (Ashton et al., 2010; Julian et al., 2021; Lee et al., 2013). The preference for money may be due to the desire to demonstrate financial success and superiority (e.g., through conspicuous consumption; Ashton & Lee, 2005; Lee et al., 2013). Therefore, it is reasonable to speculate that organizations where organizational members are strongly motivated by financial incentives may be more attractive to applicants low rather than high in honesty-humility.

The fourth characteristic of hazardous organizations is Loose Ethical Standards. De Vries and colleagues (2017) found that people low in honesty-humility were more likely to be lured by unethical business proposals when they perceived high corruption in their environment (e.g., their organization fails to stipulate that corruption will be punished). Ogunfowora (2014) found that applicants high in honesty-humility were more likely to apply for jobs in organizations led by a CEO who adhered to high ethical standards while their counterparts were indifferent to the CEO’s ethics. Similarly, Lukacik and Bourdage (2020) found that applicants high in honesty-humility were more attracted to organizations with Benevolent and Universal organizational images (e.g., high relevance of equality, honesty, responsibility; Schwartz, 1992). Taken together, organizations with loose ethical standards provide more affordances for selfish behaviour, and people low in honesty-humility may feel more comfortable to taking advantage of these affordances. Therefore, we speculate that, compared to applicants high in honesty-humility, applicants low in honesty-humility are more likely to be attracted to organizations that demonstrate loose ethical standards.

The fifth characteristic of hazardous organizations is Knowledge Hiding. De Laat (2020) found that people low in honesty-humility were reluctant to share knowledge with colleagues. This is likely because that critical and unique knowledge can endow the owner with power, status, and a large network (Wang & Noe, 2010). Therefore, we speculate that people low rather than high in honesty-humility will be more attracted to organizations where valuable information and knowledge can be exclusively occupied (e.g., knowledge is more accessible to high-level members).

The sixth and final characteristic of hazardous organizations is Entrepreneurial Orientation. Lumpkin and Dess (1996) distinguished five components of entrepreneurial orientation: autonomy, innovativeness, risk-taking, proactiveness, and competitive aggressiveness. Risk-taking refers to an organization’s willingness to invest in risky projects and take bold steps to achieve organizational goals. Competitive aggressiveness refers to an organization’s tendency to take aggressive actions to confront rivals and gain market share (for the definitions of the other components, see Lumpkin & Dess, 1996). Compared to their counterparts, people low in honesty-humility have been found to have a relatively high risk tolerance when pursuing financial interests (Ashton et al., 2010) and are more likely to engage in aggressive behaviors (Zettler et al., 2020). Therefore, we speculate that people low rather than high in honesty-humility may feel more comfortable joining organizations that are willing to engage in aggressive expansion by taking financial risks than organizations that stick to steady—and less financially risky—development (Table 2).

Descriptions of hazardous organizations (Hazardous Organization Tool, HOT) were constructed in accordance with the above six characteristics. As we mentioned above, the HOT can be used to measure the extent to which people are attracted to hazardous organizations (HOT-A) and the extent to which people perceive an organization to be hazardous (HOT-P). The hazardous organizations described by the HOT should negatively influence the pre- and post-entry work attitudes of people with high ethical standards. Therefore, it is preferable to have two separate pieces of validity evidence for the two HOT measures. First, given that hazardous organizations should negatively influence the pre-entry work attitudes of people with high ethical standards, the attractiveness ratings of the HOT (HOT-A; representing the most primary pre-entry work attitude) should be more negatively related to personality traits that indicate high ethical standards (convergent validity) and less strongly related to personality traits that are weakly related or unrelated to ethical standards (divergent validity). Moreover, previous research has found significant relations between personality traits that indicate low ethical standards (though less strongly compared to honesty-humility) and preferences for organizational cultures other than ethical culture. Therefore, the HOT-A should be positively related to preferences for an ethical culture and other somewhat similar organizational cultures (convergent validity) and less strongly related to preferences for dissimilar organizational cultures (divergent validity). Second, given that hazardous organizations should negatively influence the post-entry work attitudes of people with high ethical standards, perceptions of their current employers as hazardous (HOT-P) should be related to employees’ negative (post-entry) work attitudes, especially for those with high ethical standards (low in HOT-A; predictive validity). We followed this logic to assess the (convergent, divergent, and predictive) validity of the HOT measures, which are detailed below (see Table 2 for an overview).

We argue that organizational ethics is not the only organizational characteristic that influences the pre- and post-entry work attitudes of people with high ethical standards, and that the organizational characteristics described in the HOT can encompass broader relevant content than existing measures of ethical culture, e.g., integrity culture as measured by the OCP. This can be examined by comparing semi-partial correlations of both the HOT-A and OCP integrity culture preference (after controlling for their shared variance) with personality traits indicative of ethical standards. Significant and larger semi-partial correlations of the HOT-A than those of the OCP integrity culture preference will demonstrate measurement breadth of the HOT (Hypothesis 10, Table 2).

We aimed for sample sizes of N ~ 300, at which factor analytic (N > 300) and correlational (N > 250) results become stable (Guadagnoli & Velicer, 1988; Schönbrodt & Perugini, 2013). Four infrequency items (e.g., “I never had hair on my head”) were inserted in random places in the HEXACO inventory to detect failure to understand instructions or careless responses (Fekken et al., 1987). We also used a statistical approach to detect noncompliant respondents (Barends & De Vries, 2019). Specifically, participants who failed 50% above of the infrequency questions, whose responses on the six domain scales indicated an average standard deviation above 1.60 (after recoding negatively worded items), or whose responses on the HEXACO-100 items (without the items addressing the Proactivity facet, see also Measures section) had a standard deviation below 0.70 (before recoding negatively worded items), were excluded from analysis (Lee & Ashton, 2018). Note that the sample size of Sample #3 was slightly lower than the sample size we aimed for, but still met the requirements to establish stable factor analytic estimates, when average factor loadings are above 0.40 (Guadagnoli & Velicer, 1988).

We collected four samples. Samples #1, #2, and #4 consisted of working adults recruited via Amazon’s Mechanical Turk (Buhrmester et al., 2011). Sample #3 consisted of university students invited through the university research participation system. University students are a relevant population because they will typically enter the job market within a short time and need to decide whether to join organizations that are more or less hazardous. Other sample characteristics can be found in Table 2.

Participants from sample #1 sequentially provided responses to the HEXACO-104, the Dirty Dozen, the initial 85 HOT-A items (details see the sections below), and demographic questions.

When collecting samples #2 and #3, we adopted a time lagged design. Sample #2 was implemented at two time points with 1 week. At time 1, participants were randomly presented with either the selected 36 HOT-A or HOT-P items besides the HEXACO-104 and demographics. At time 2, they completed all other scales (e.g., job satisfaction; see Table 3 and the “Measures” section below for more details). Sample #3 consisted of university students who completed the HEXACO items in their first bachelor year of the psychology program between the beginning of April and May during three consecutive years (2019–2021) along with other scales (time 1). The students were invited to this study through the university research participation system in 2021 (time 2) and provided responses to the 36 HOT-A items along with other scales (see Table 3 and the “Measures” section below for more details).

Data collection of sample #4 adopted the similar design as sample #2. However, due to the omission of creating a code to match participants’ time 1 and time 2 responses, we decided to limit the data collection of sample #4 to time 1. Among the 308 valid responses, 155 completed 36 HOT-A items (sample #4 HOT-A group) while 153 completed 36 HOT-P items (Sample #4 HOT-P group).

Informed consent was obtained online from all participants.

A summary of which measures were administrated in which samples can be found in Table 3.⁴

In the first step, an exploratory factor analysis revealed the presence of (positive versus negative) wording effects (see Supplemental Material Sect. 2). To help remove the wording effects and determine to what extent the content domains are distinct from each other, we fitted a bi-factor exploratory structural equation modelling (ESEM; Asparouhov & Muthén, 2009) model with target rotation method (Reise, 2012) using Mplus version 8 (Muthén & Muthén, 1998‐2017). Specifically, we fitted a model containing a general factor, six group factors to represent the six a priori content domains, and a method factor to capture variance associated with method effects (e.g., wording effects; see also Biderman et al., 2019; Maydeu-Olivares & Steenkamp, 2018). The general factor was included to capture the shared variance among the six content domains that represent the common organizational characteristic that can satisfy the exploitative needs of people low in Honesty-Humility. The six group factors were specified to capture the unique variance after accounting for other shared variance among the items (e.g., shared variance among all the items captured by the general factor), which is also informative for deciding to what extent the content domains represent statistically discriminant factors (using several indices; see the next paragraph). Each item was set up to simultaneously load onto all factors. Loadings of the items on the general factor and their a priori group factor were freely estimated, whereas loadings on all other non-primary factors were targeted at zero, which conforms to their a priori content domain. Loadings on the method factor were targeted at 0.2 for the positively worded items and at − 0.2 for the negatively worded items. The values of 0.2 and − 0.2 were arbitrary for contrast of the items with different wordings. The differences resulting from setting other values are negligible (see also Biderman et al., 2019; Maydeu-Olivares & Steenkamp, 2018).

The bi-factor ESEM model indicated better model fit indices than the models without the general factor, the method factor, or both (reduced ESEM models; see Supplemental Material Table S5). For example, a chi-square difference test indicated adding the method factor significantly improved model fit,\({x}^{2}\)(78) = 344.23, p < 0.001.⁹ Two indices were calculated: explained common variance (ECV) and percent uncontaminated correlations (PUC; Rodriguez et al., 2016). ECV estimates the proportion of variance accounted for by the general factor to the total variance accounted for by the general factor and group factors. A higher ECV value indicates a stronger general factor. PUC represents the proportion of unique correlations in a correlation matrix indicating only the influence of the general factor to the total number of unique correlations (Rodriguez et al., 2016). As PUC increases, ECV becomes less influential in affecting the potential parameter bias resulting from fitting multidimensional data to a unidimensional model (Bonifay et al., 2015; Reise et al., 2013).

The ECV was 0.60 and the PUC was 0.84. That is, the general factor explained 60% of the total variance and the PUC was sufficiently high so that fitting a unidimensional model to the data might not lead to serious parameter bias. Although the items were developed from six content domains, they failed to emerge as separate factors after removing method variance. Before going into the next step, we selected 36 items (six for each content domain of which two are negatively worded) mainly based on (a) correlation with Honesty-Humility, (b) content diversity, and (c) score distribution (keeping normal distribution as much as possible; see Supplemental Material Table S4).

In the second step, we contrasted the bi-factor ESEM models and the reduced ESEM models of the 36 HOT-A items on samples #1–3. The bi-factor ESEM solutions indicated best model fit indices except for sample #3 (Table 4; see Supplemental Material Sect. 2 for more details). The ECV values of the bi-factor ESEM model were 0.65, 0.54, and 0.50 for samples #1, #2, and #3, respectively. The PUC value of 0.53 was identical for all samples. Model-based reliability estimates \({\omega }_{H}\) and \({\omega }_{HS}\) were calculated (Reise et al., 2013; Zinbarg et al., 2005). The \({\omega }_{H}\), which assesses the proportion of variance in the total score attributable to the general factor, was 0.92, 0.89, and 0.86 for samples #1–3, respectively. \({\omega }_{HS}\) assesses the proportion of variance in the subscale score uniquely attributable to the group factor after removing the variance accounted for by the general factor. The \({\omega }_{HS}\) values were all below 0.03 for the six group factors in all samples, indicating that none of the group factors could account for substantially unique variance beyond the general factor. Overall, the above model-based indices support that the 36 items can be viewed as essentially measuring a unidimensional construct.

In the third step, we used a genetic algorithm to construct the final unidimensional HOT due to the nearly infinite number of alternative solutions. Genetic algorithms can solve any optimization problem if the multiple criteria that need to be evaluated can simultaneously be defined by a custom fitness function (Scrucca, 2013). In our case, the genetic algorithm was set to simultaneously consider four criteria when searching for optimal scale solutions: (a) mean correlation with Honesty-Humility domain scores across samples #1–3 (maximized in the search process), (b) mean factor loading of one-factor confirmatory factor analysis (CFA) solutions across samples #1–3 (maximized), (c) mean SRMR and RMSEA of the one-factor CFA solutions across samples #1–3 (minimized), and (d) a minimum of one-third negatively worded items (an item-ratio cost function that was maximized; see the shared R script at https://​osf.​io/​js7vf/​). Extra constraints were set to ensure desirable scale properties: (a) a minimum of one item and a maximum of three items from each content domain and a minimum of seven items were selected to keep enough bandwidth, (b) no factor loadings were below 0.30 in any samples,¹⁰ and (c) two relatively overlapping items from the Money Orientation domain would not be selected simultaneously.¹¹ The item selection was implemented in the R environment (version 4.2.2; R Core Team, 2022) with the R package GA (version 3.2.2; Scrucca, 2013, 2021; for similar applications see Moshagen et al., 2020, Olaru et al., 2019).

As shown in Table 5, model fit indices and the substantial factor loadings of the final solution (all above 0.40 with one exception in sample #2 and two exceptions in sample #3) support the one-factor structure. A conventional one-factor CFA model was also fitted to the data on the HOT-P. The model indicated perfect fit indices with all factor loadings above 0.40 and the Cronbach’s alpha reliability was 0.87. Overall, both HOT measures (the HOT-A and the HOT-P) indicated a nearly identical factor structure. Moreover, the commensurability of the measures ensures its usage in the subsequent cubic response surface analysis based on polynomial regression analysis (Humberg et al., 2020; see Predictive Validity of the HOT section below). The mean scores of the final 9-item HOT-A and HOT-P were used in the next steps.

In the fourth step, we examined the convergent and divergent validity of the HOT-A. We conducted a mini meta-analysis of the correlations involving multiple samples (Goh et al., 2016; Table 6). Table 7 presents the correlations that are only based on a single sample. Convergent validity was demonstrated by moderate to strong correlations of the HOT-A with Honesty-Humility (negative), dark personality traits (positive), and organizational culture preferences for OCP Integrity (negative), OCP Collaborative (negative), OCP Adaptability (positive), and OCP Results-Oriented (positive), which supports H1, H3, and H5. The correlations of the HOT-A with all work motivation, albeit weak for external regulation and introjected regulation, also demonstrated convergent validity (supporting H4). Divergent validity was demonstrated by much weaker correlations (most \(\left|r\right|\mathrm{s}<.20\)) with theoretically unrelated variables, such as HEXACO personality traits other than Honesty-Humility (supporting H2 and H6).¹²

As shown in Table 8, HOT-P indicated moderate to strong significant correlations with all work attitudes (supporting H7), weak to moderate correlations with four of the five work motivation as hypothesized (partially supporting H8 with H8c not supported). Next, we used cubic response surface analysis, based on polynomial regression analysis, to test whether people—whose attraction to hazardous organizations (HOT-A) converges with perception of their employer as hazardous (HOT-P)—report more positive work attitudes and motivation.

In sample #3, we used both the HOT-A and the OCP to measure organizational culture preferences. The amount of overlap between the HOT-A and the OCP Integrity culture preference was moderate (the correlation is − 0.39). As shown in Table 10, the semi-partial correlations of the HOT-A, controlling for the shared variance between the HOT-A and the OCP Integrity culture preference, with self-report and peer-report Honesty-Humility and Dark Factor of Personality were significant in their hypothesized directions (supporting H10a, H10b, and H10c). The semi-partial correlations of the HOT-A were always larger than those of the OCP Integrity culture preference, although the differences were smaller and insignificant for peer-report Honesty-Humility (supporting H10d and H10f with H10e not supported).¹⁶ Moreover, compared to the simple correlations, there was a small decrease in the semi-partial correlations of the HOT-A with self-report and peer-report Honesty-Humility and Dark Factor of Personality. The decreases in the semi-partial correlations of Honesty-Humility and Dark Factor of Personality with the OCP Integrity culture preference were larger. In general, this is evidence that the HOT can measure meaningful organizational characteristics, not just organizational ethics, that are important to people with high ethical standards.

The current literature mainly emphasizes ethical aspects of organizations when discussing attracting and retaining people with high ethical standards (e.g., Ogunfowora, 2014; cf. Lukacik & Bourdage, 2020). However, organizational ethics is clearly not the only relevant characteristic of organizations. By using semi-partial correlation analysis, we showed that the HOT can measure important organizational characteristics other than organizational ethics that affect the pre- and post-entry work attitudes of people with high ethical standards. Knowing these characteristics is especially valuable for organizations that want to cultivate an ethical culture but find it difficult to do so. Moreover, the current study reveals one potential reason for this—they might have ignored aspects that may seem less related to organizational ethics (e.g., promoting competition for salary raises or promotions that lead to a money orientation). People with high ethical standards are more willing to accept and help accelerate culture change to be more ethical and help sustain an ethical culture through their daily behaviors (Neville & Schneider, 2021a, 2021b; c.f., Chatman, 2021). To attract and retain enough people with high ethical standards to bring about such change, it may be necessary to change not only organizational ethics but also other organizational characteristics. The HOT captures these crucial characteristics.

Given the value-added nature of an ethical culture (Graham et al., 2022; Guiso et al., 2015), we assume that most of the organizations that advertise their commitment to ethical culture, but do not have it, aspire to foster an ethical culture. However, we cannot rule out the possibility that some of them simply want to use it as a marketing strategy. It may seem unlikely that applicants have accurate prior knowledge about an organization’s culture before joining it. However, previous research has shown that applicants are motivated and able to gather information about their potential employers from a variety of sources (Judge & Cable, 1997), including but not limited to recruitment websites, newspapers, social media, and word-of-mouth (e.g., Stockman et al., 2020; Van Hoye & Lievens, 2005). Organizations have limited or no control over these external sources, so applicants are likely to gain insight into aspects of the culture, including negative ones (e.g., sexual harassment), and develop hazardous organization perceptions at a very early stage. Moreover, small amounts of negative information may be sufficient to significantly reduce the attractiveness of an organization and limit people’s job pursuit in such an organization (Rozin & Royzman, 2001).

Finally, our research suggests that most people perceive their organization as more hazardous than they want it to be. In sample #2, only a small subset of participants (5.10%) rated their current employer’s hazardousness as less than their attraction to hazardous organizations. In contrast, more than half of the participants (60.54%) reported working for organizations that were more hazardous than they wanted them to be. Given the virtues of an ethical culture, which requires attracting and retaining people with high ethical standards to help foster and maintain it, organizations should consider using the HOT to examine their characteristics.

Consistent with the propositions of the situation activation and outcome activation mechanisms from the STOA model (De Vries et al., 2016a, 2016b), we found that people with low ethical standards are more attracted to hazardous organizations (i.e., situation activation) and are more likely to report positive work attitudes when working in hazardous organizations, albeit to a lesser extent for work motivation (i.e., outcome activation). Given that the HOT was based on HEXACO Honesty-Humility, these findings collectively provide partial support for the HEXACO domain specific effects of Honesty-Humility (De Vries et al., 2016a, 2016b), and thus engage the literature on situational affordances. A full support calls for research on the HOT-AP congruence effects on behavioral outcomes.

The relations between HOT-AP fit and behavioral outcomes (e.g., unethical workplace behavior) may be more complex than with work attitudes and motivation, and different theories may suggest different hypotheses. Barrick and Parks-Leduc (2019) have postulated a substitution model of fit on behavioral outcomes.¹⁷ They proposed that as long as the beneficial trait or characteristic of either the person or the organization is high, employees are more likely to perform better. As one of the two becomes increasingly high, the relations between the other one and performance measures would be attenuated to near zero. In the case of the HOT, if the substitution effect hypothesis is supported, employees who report either high HOT-A or high HOT-P, regardless of the level of congruence between the two, will be more likely to engage in, for example, unethical workplace behaviors. This substitution effect hypothesis is consistent with the trait activation theory (Tett & Burnett, 2003; Tett et al., 2021) and the trait activation mechanism from the STOA model (De Vries et al., 2016a, 2016b).

Another line of research supports synergistic effects of PO fit on behavioral outcomes. For example, Chatman (1989) proposed that PO fit may lead to better extra-role behaviors. Indeed, in a meta-analytic study, Kristof‐Brown and colleagues (2005) found that PO fit was positively related to job performance. To our knowledge, although both the synergistic effect and substitution effect hypotheses have theoretical underpinnings and received empirical support, no prior research has contrasted both hypotheses simultaneously. The commensurable HOT-A and -P provide an opportunity to compare the empirical support for both hypotheses through precisely formulated mathematical models, which can facilitate theoretical precision (Guest & Martin, 2021; see Humberg et al., 2019). Regardless of which hypothesis ultimately receives more empirical support, this future research can help distinguish “fitting in” from “doing well” processes, as suggested by Barrick and Parks-Leduc (2019), to examine whether and how PO fit is related to job performance beyond pre- and post-entry work attitudes.

Overall, the current study illustrates the value of this personality-centered approach. Future research may want to develop commensurable measures to assess organizational attractiveness and perceptions based on other personality traits (e.g., extraversion). For example, organizations whose front-line employees interact intensively with customers may want to attract and hire more extraverted employees and may benefit most from a culture assessment using an extraversion-focused instrument. These instruments can advance the theoretical development of the PO fit literature and Tett and colleagues’ (2021) so-called personality-oriented work analysis.

The current study has both strengths and limitations. The skewed HOT-A scores suggest that the descriptions of hazardous organization may be construed as socially undesirable. In contrast, the HOT-P scores were more normally distributed and had higher means than the HOT-A scores. This divergence suggests that, on average, people work for organizations that are perceived to be more hazardous than they want them to be. However, we cannot rule out the possibility that participants may be somewhat reluctant to admit attraction to hazardous organizations or that perceptions of their current employer as hazardous may be distorted by self-serving bias.¹⁸ Future observational studies can help to address this issue. For example, by using a vignette-based experimental design or a field study, one might be able to observe how the ethical standards would change at different recruitment stages (e.g., job application, job offer acceptance) when applicants are exposed to low and high hazardous organizations. In addition, as noted above, Neville and Schneider (2021a) have suggested that recruitment and selection practices may lag behind culture changes. Future research might also want to examine whether hiring managers’ commitment to their current (hazardous or not) organizational culture may influence their selection criteria (e.g., the weights assigned to the criteria that indicate ethical standards).

Next, we used the psychometric soundness of the 9-item Hazardous Organization Tool-Attractiveness as the basis for selecting the final items for both the HOT-A and the HOT-P scales. The psychometric properties of the HOT-P may have capitalized on chance. However, if different items had been selected, the measurement of the HOT-A and HOT-P would no longer be commensurable, which would affect the interpretability of the congruence effects of the HOT-A and HOT-P on the outcome variables. To address this concern, we repeated the primary analyses for the top five solutions selected by the genetic algorithm with and without Sex Tolerance items, which had relatively low factor loadings (see Table 5). The results indicated that the measurement properties of the HOT(-A and -P), the correlations, and the HOT-AP congruence effects on the outcome variables (i.e., job satisfaction, intrinsic motivation, job stress, and turnover intention) were highly similar regardless of which HOT items were selected, with only minor variations when Sex Tolerance items were excluded, supporting the robustness of the current findings (see Supplemental Material Sect. 5).

Lastly, we should note that the advantage of the HOT over existing measures of ethical culture (e.g., the updated OCP; Chatman et al., 2014) may be context dependent and that we only assessed the convergent and divergent validity of the HOT-A. Our primary goal was to get a comprehensive understanding of which organizational characteristics influence the pre- and post-entry work attitudes and behaviors of people with high ethical standards. Therefore, we focused on people rather than organizations (this is why we identified organizational characteristics by starting from reviewing the evidence on personality rather than organization). The semi-partial analyses showed that this approach has the advantage of maximizing convergent validity (i.e., the correlations between the HOT-A and Honesty-Humility and dark personality traits). However, the HOT is not intended to replace existing measures of ethical culture in all contexts and in our research we did not yet validate the HOT-P with existing (ethical) culture measures. If researchers or practitioners want to conduct in-depth research on ethical culture or its relations with organizational-level outcomes, other validated measures (e.g., the updated OCP or the Corporate Ethical Virtues Questionnaire; Chatman et al., 2014; Kaptein, 2008) may be useful. However, if researchers want to examine PO fit from the perspective of personal ethical standards, the HOT may be a better choice. Organizations can use it to understand on what aspects, not just organizational ethics, they need to improve if they want to be become a more attractive and desirable place to work for people with high ethical standards.