A review of applicant faking in selection interviews

Since faking is a common phenomenon in personnel selection (Donovan, Dwight, & Hurtz, 2003) there is a long history of research on faking in personality tests (e.g., Birkeland, Manson, Kisamore, Brannick, & Smith, 2006; Viswesvaran & Ones, 1999). Furthermore, the examination of impression management tactics that are used by job applicants in employment interviews is not new (e.g., Gilmore & Ferris, 1989; Stevens & Kristof, 1995). However, it is only in the last decade that we have observed a growing interest in the examination of applicant faking in employment interviews (Bourdage, Roulin, & Levashina, 2017). The long reluctance of prior research to investigate interview faking might seem surprising given its practical implications. Indeed, the vast majority of applicants use faking tactics in interviews but applicants also vary in the extent to which they engage in faking (e.g., Levashina & Campion, 2007). As such, faking can potentially change their rank order, impact who gets hired, and thus represent a possible threat to the criterion-related validity of interviews. The turnaround point concerning interview faking research is probably the publication of Levashina and Campion's (2006) theoretical model of applicant faking in interviews and of their subsequent scale to measure faking in interviews (Levashina & Campion, 2007). Over a decade later, we believe that enough theoretical and empirical work has been accumulated over recent years to warrant a first review. This review makes a number of important contributions to the literature. First, we identify issues for which sufficient data are available and consensus has been reached. Second, we identify relevant limitations of previous research as well as important but still unanswered questions for which more research is needed. Third, we discuss possible reasons for diverging findings in certain areas and describe the necessary research to gain insights into the reasons for these findings. Fourth, we provide suggestions for future theoretical Received: 21 August 2019 | Revised: 18 December 2019 | Accepted: 12 January 2020 DOI: 10.1111/ijsa.12280


| INTRODUC TI ON
Since faking is a common phenomenon in personnel selection (Donovan, Dwight, & Hurtz, 2003) there is a long history of research on faking in personality tests (e.g., Birkeland, Manson, Kisamore, Brannick, & Smith, 2006;Viswesvaran & Ones, 1999). Furthermore, the examination of impression management tactics that are used by job applicants in employment interviews is not new (e.g., Gilmore & Ferris, 1989;Stevens & Kristof, 1995). However, it is only in the last decade that we have observed a growing interest in the examination of applicant faking in employment interviews (Bourdage, Roulin, & Levashina, 2017). The long reluctance of prior research to investigate interview faking might seem surprising given its practical implications. Indeed, the vast majority of applicants use faking tactics in interviews but applicants also vary in the extent to which they engage in faking (e.g., Levashina & Campion, 2007). As such, faking can potentially change their rank order, impact who gets hired, and thus represent a possible threat to the criterion-related validity of interviews. The turnaround point concerning interview faking research is probably the publication of Levashina and Campion's (2006) theoretical model of applicant faking in interviews and of their subsequent scale to measure faking in interviews (Levashina & Campion, 2007).
Over a decade later, we believe that enough theoretical and empirical work has been accumulated over recent years to warrant a first review. This review makes a number of important contributions to the literature. First, we identify issues for which sufficient data are available and consensus has been reached. Second, we identify relevant limitations of previous research as well as important but still unanswered questions for which more research is needed. Third, we discuss possible reasons for diverging findings in certain areas and describe the necessary research to gain insights into the reasons for these findings. Fourth, we provide suggestions for future theoretical work and put forth specific propositions and research questions that may guide future empirical research. Finally, we inform practitioners on factors that promote or prevent faking, and on the usefulness and suitability of existing measures to prevent or to detect faking in interviews.

| S TRUC TURE OF THIS RE VIE W AND IDENTIFI C ATI ON OF RELE VANT RE S E ARCH
Our aim is to review the existing literature on applicant faking in employment interviews to answer the following questions: (1) What is faking in interviews? (2) How common is it? (3) What are its key antecedents? (4) Does it matter? (5) Can anything be done to detect it?
To answer these questions, we conducted an extensive literature search. We began by searching electronic databases such as PsycINFO/PsycARTICLES, Web of Science, and Google Scholar for studies that included the keywords "applicant faking," "interview faking," or "deceptive impression management." Furthermore, we checked reference lists of identified papers as well as relevant recent reviews on personnel selection in general (Ryan & Ployhart, 2014;Sackett & Lievens, 2008) and regarding interviews (Levashina, Hartwell, Morgeson, & Campion, 2014;Macan, 2009).
After eliminating noninterview results (e.g., faking in personality tests or in the workplace) and nonjournal publications (i.e., dissertations, conference presentations), we identified 36 peer-reviewed empirical articles with a total of 51 studies that dealt explicitly with faking in interviews. We also identified 12 nonempirical articles (i.e., reviews about interviews in general, theoretical models, editorials) that discussed interview faking. Furthermore, this review also covers additional relevant research concerning impression management or self-presentation in interviews and regarding faking in general to the extent to which it is relevant for interview faking.

| Faking, impression management, and socially desirable responding
Researchers in both the personality testing and the interview domain have proposed or used many different definitions and conceptualizations of faking. For the present review, we mainly draw on Levashina and Campion (2007), who defined faking as "conscious distortions of answers to the interview questions in order to obtain a better score on the interview and/or create favorable perceptions" (p. 1639). This definition (like the vast majority of faking research) is focused on faking "good." Faking "bad," that is applicants' attempts to decrease their chances of being hired by presenting themselves as worse than they really are, exists but is a rare phenomenon in personnel selection. Faking "bad" can happen when applicants are encouraged or required to apply for undesirable jobs (e.g., to continue receiving unemployment benefits) or in the case of assessments for compulsory military service (e.g., Boss, König, & Melchers, 2015).
In addition, faking has sometimes been assimilated with concepts like impression management (IM) or social desirability. However, we argue that they should be clearly distinguished in the interview context. IM has been defined as trying to create a particular impression in someone's mind (Leary & Kowalski, 1990). In the personality literature, the terms IM and faking have been used interchangeably. This is because a response to a personality inventory item can either be completely honest (e.g., responding "3" when applicants truly believe that their actual level is 3 out of 5) or deceptive (e.g., responding "4" or "5"). However, in the interview context, IM and faking are related but distinct constructs. IM can be represented as a continuum of influence tactics going from complete honesty and modesty in the way applicants present themselves to outright lying about their qualities (cf. Bourdage, Roulin, & Tarraf, 2018;Levashina & Campion, 2007): The lower part of the continuum can be defined as honest IM (i.e., applicants using varying amounts of IM tactics to create a positive impression but relying on facts and truthful descriptions of their traits or qualities). The upper part of the continuum can be considered as deceptive IM (i.e., applicants using more or less deceptive forms of IM tactics to create a positive impression) with slightly deceptive tactics being closer to honest IM and extremely deceptive tactics being closer to outright lies. In interviews, faking thus only incorporates deceptive forms of IM, while in personality inventories, faking is equivalent to IM.
Another related construct is socially desirable responding (SDR).
Social desirability is defined as "the tendency to give answers that make the respondent look good" (Paulhus, 1991, p. 17). In the personality literature, SDR is usually described as comprising two facets: self-deception, which involves an involuntary enhancement of positive qualities and denial of negative qualities about oneself that is not necessarily aimed for an audience, and impression management, which involves giving inflated statements that are explicitly aimed at an audience (Paulhus, 1991). Although faking and SDR are conceptually related, SDR is defined as a stable trait that should (theoretically) not vary across situations. In contrast, and although there is evidence that applicants engage in faking in a somewhat consistent way across interviews , interview faking (or interview IM in general as described in the previous paragraph) is situation specific, that is, targeted to a specific job or a unique interview situation (Levashina & Campion, 2006). Moreover, SDR has been described as an inadequate way to capture applicant faking (Burns & Christiansen, 2011;Griffith & Peterson, 2008). The distinction between faking and SDR becomes clearer when looking at the measures: An example item for SDR is "I never swear" (Paulhus, 1991), whereas an example item for faking is "during the interview, I distorted my answers based on the comments or reactions of the interviewer" (Levashina & Campion, 2007). Empirically, relationships between interview faking and SDR tend to be only small, with nonsignificant correlations for Paulhus' measure of self-deception and only modest but significant for Paulhus' IM measure (Levashina & Campion, 2007).

| Taxonomy and measurement
There is a wide range of faking behaviors that interviewees can use and that are included in Levashina and Campion's (2007) taxonomy and 54-item measure of interview faking behaviors (i.e., the IFB scale). Levashina and Campion considered various sources of information and used qualitative and quantitative approaches. In their taxonomy, they distinguished assertive techniques, defensive tactics, and ingratiation. Assertive techniques can be seen as applicants' deceptive attempts to create a positive impression by highlighting experiences, skills, or opinions. They can be further divided into slight image creation (i.e., applicants' attempts to stretch the truth by exaggerating skills, abilities, and work experiences) and extensive image creation (i.e., inventing-or borrowing from others-work experiences, skills, or accomplishments). Defensive tactics are used by applicants to protect their image of being the ideal person for the job and include hiding weaknesses and insufficient skills or masking negative events and experiences in their work history. Finally, deceptive ingratiation is used to appear more attractive or likeable, by pretending to conform to the interviewer's and/or organization's values, beliefs, opinions, or attitudes or insincerely praising the interviewer or organization.
Given that Levashina and Campion (2007) derived the IFB items partly from the general IM literature, their faking tactics are related to more general forms of IM tactics. However, this literature did not distinguish honest from deceptive IM (e.g., Stevens & Kristof, 1995).
Specifically, slight and extensive image creation are conceptually similar to assertive IM tactics such as self-promotion. Image protection is similar to defensive IM tactics such as image repair. Deceptive ingratiation is similar to other-focused IM tactics. As mentioned above, all those tactics are on a continuum that goes from complete honesty and modesty in the way applicants present themselves to outright lying about their qualities. Furthermore, for observers it is hardly possible to decide whether a specific IM tactic is honest or dishonest. For instance, applicants who tell interviewers "I am fluent in four languages: English, Spanish, French, and Chinese Mandarin" could be using honest self-promotion (if they actually speak those four languages fluently), slight image creation (if they speak four languages, but only two of them fluently), or extensive image creation (if they do not speak Spanish and French at all). This is why, in response to several authors noting the importance of differentiating honest from deceptive IM (e.g., Levashina et al., 2014), Bourdage et al. (2018) recently developed and validated a self-report measure that specifically captures the honest side of IM. They also showed that honest IM and deceptive IM or faking are correlated, but empirically distinct.
Other studies have captured faking using different items and the randomized response technique (RRT, see König, Hafsteinsson, Jansen, & Stadelmann, 2011;König, Wong, & Cen, 2012). The RRT involves respondents using a randomization device (e.g., rolling dices or flipping a coin) before answering each dichotomous faking item (i.e., reporting engaging or not in the faking behavior). Depending on the randomization result, respondents are instructed to simply endorse the item (e.g., if the die shows 1 or 2) or to report their actual behavior (e.g., if the die shows 3, 4, 5, or 6). Researchers can then estimate faking frequencies after a statistical correction for the base rate of instructed endorsements. This technique has both advantages (e.g., the incorporation of a random component adds a layer of protection for respondents and facilitates their honest reporting of faking) and disadvantages (e.g., it is more complex to score and interpret, and one cannot analyze the data on the individual level but only on the group level). Because of these differences, it is difficult to compare findings from studies using the IFB versus an RRT operationalization of faking.
Willingness is influenced by applicants' motivation and personality but can also be fostered by situational aspects like perceptions of unfair treatment during the interview. Opportunity to fake comprises characteristics of the interview type and format that can enable or hinder applicants' faking like, for example, the type of interview (structured vs. unstructured) or question (past behavioral vs. situational). Levashina and Campion's model is a multiplicative model, so that all factors must be present at least to some extent for faking to occur. It predicts the likelihood to fake. However, it does not differentiate between faking behavior (e.g., the use of a certain kind of faking behavior) and faking success (e.g., a good interview rating or a job offer). In addition, Levashina and Campion's model differs from several models derived from the Theory of Planned Behavior (e.g., McFarland & Ryan, 2000Mueller-Hanson et al., 2006) that focus on intentions to fake as the direct antecedent of faking behaviors.
All faking models consider characteristics of the individual (e.g., ability, personality, motivation) and the situation (e.g., selection method, use of a faking warning, applicants' personal situation) as antecedents of faking. Interestingly, some models (e.g., Ellingson & McFarland, 2011;Marcus, 2009;McFarland & Ryan, 2000; also take contextual factors into account, like the attractiveness of the organization, while other models do not consider such factors (e.g., Levashina & Campion, 2006;Salgado, 2016). Snell et al.'s (1999) model, for example, includes the importance of the outcome for applicants' or their perception of others' faking behavior. Furthermore, while most models consider the role of applicants' abilities (e.g., to analyze the interview situation and adapt their behavior to job requirements), some consider these abilities as antecedents of faking (e.g., Snell et al., 1999), whereas others depict them as moderating the relationship between faking motivation or intentions and actual faking behavior (e.g., McFarland & Ryan, 2000) or between faking behavior and interview performance. Levashina and Campion's (2006) model remains relatively silent on the impact of faking on construct-or criterion-related validity.
In contrast, McFarland and Ryan (2000) assumed that faking may influence applicants' scores, but also impact the reliability, factor structure, and validity of selection instruments. However, they failed to clarify the specific nature of the effects. In contrast, making predictions about the influence of faking on test scores and validity is a strength of the theory of self-presentation by Marcus (2009). His model suggests that the effects of faking on criterion-related validity can range from negative to neutral to positive depending on the specific situation (see below). Furthermore, Salgado (2016) recently predicted a negative effect of faking on both the reliability and criterion-related validity of selection instruments. Specifically, he assumes that faking should increase the mean but decrease the standard deviation of applicants' scores on a selection procedure.
Thereby, faking leads to a homogenization of scores and a reduction of systematic variance.
Finally,  proposed a dynamic model of applicant faking. They suggested that applicants' motivation to fake will be stronger when they perceive that they face an intense competition to obtain the job, which derives from perceived competition on the job market, the competitive culture of the organization, and individual-level competitive worldviews. They also describe faking as an adaptive mechanism, so that the success or failure of faking and the interpretation of this success or failure by applicants will ultimately impact faking attempts in subsequent interviews. Levashina and Campion's (2007) taxonomy and measure offer a precise answer to our first question ("What is faking in interviews?"): Faking is neither SDR nor general IM but corresponds to the deceptive side of IM. Therefore, it is important to separate honest IM, which represents behavior that is accepted and expected by recruiters (Jansen, König, Stadelmann, & Kleinmann, 2012) and that can contribute to criterion-related validity (Kleinmann & Klehe, 2011), from deceptive IM (Bourdage et al., 2018;Levashina & Campion, 2006Levashina et al., 2014). However, the measurement and operationalization of faking still requires additional attention. For instance, although Levashina and Campion (2007) demonstrated the factor structure of their measure (i.e., the IFB), subsequent studies often only used a subsample of the IFB items instead of the entire scale. Existing studies have also focused on either a general faking score or the four faking factors (e.g., extensive image creation), but have largely ignored the subfacets (e.g., borrowing, inventing). The detailed IFB structure thus needs to be confirmed. Furthermore, Bourdage et al. (2018) recently proposed a short version (i.e., the IFB-S), with 16 items instead of the full 54 items, but more research is needed to confirm the strengths and weaknesses of the short scale as compared to the complete scale.

| Discussion and future research directions
Interestingly, Levashina and Campion's (2007) taxonomy and the IFB only focus on verbal forms of faking. However, in the general IM literature, studies have also incorporated nonverbal behaviors such as smiling or making eye contact as possible forms of IM (Barrick, Shaffer, & DeGrassi, 2009). Furthermore, a recent qualitative study on interviewer IM revealed a considerable range of nonverbal IM tactics that interviewers use deliberately to create a positive image on applicants (Wilhelmy, Kleinmann, König, Melchers, & Truxillo, 2016). Accordingly, it is possible that applicants also use nonverbal behaviors deceptively (e.g., fake smile or laugh), and future research could explore this form of faking. This leads to our first research question:

Research Question 1 What are non-verbal forms of interview faking?
In addition, the different faking models predict when applicants are more or less motivated to fake but they do not differentiate specific faking behaviors. A possible reason could be that most faking models were developed for the domain of personality testing where applicants have fewer degrees of freedom on how to fake in comparison to interviews. However, the more deceptive forms of faking might only be used when applicants perceive a higher need to do so, for example, when they perceive more competition with others or if previous attempts using honest IM have not led to the desired outcome .
Most faking models are also heavily focused on individual difference antecedents of faking but pay much less attention to external or situational factors that encourage or discourage faking (e.g., competition on the job market, economic conditions). However, as suggested by Ellingson (2012, p. 19), it is possible that "people fake only when they need to fake." Accordingly, models should consider contextual or situational factors more explicitly that make it more or less likely that applicants try to fake.
Interestingly, most of the faking models have not been empirically tested directly and comprehensively (for an exception, see McFarland & Ryan, 2006). However, many key elements of these models have been examined in separate studies, and the key findings are discussed in the following sections. A final limitation concerning most faking models is how surprisingly silent they are about the effects of faking on validity. As noted above, only two of these models (Marcus, 2009;Salgado, 2016) make predictions about the effects on criterion-related validity. However, their predictions are also inconsistent with each other because they assume different processes. Furthermore, they do not distinguish between different performance criteria, such as task performance, extra-role behavior, or counterproductive work behavior. This leads to the following research question: Research Question 2 What are the processes through which faking affects criterion-related validity for different criteria?

| HOW COMMON IS FAK ING IN INTERVIE WS?
Faking in selection interviews is rather common. Table 1 shows faking prevalence rates reported in various studies. Weiss and Feldman (2006) found that 81% of applicants admitted telling at least one lie in their last interview, with an average of 2.19 lies per interview.
And across three samples, Levashina and Campion (2007) found that 93%-99% of undergraduate job applicants had used at least one faking tactic during their most recent interview. However, more recent studies reported lower prevalence rates with both real job applicants and students (Bourdage et al., 2018;Roulin, Bangerter, & Levashina, 2014).
Although most people fake during interviews, faking tactics close to lying such as extensive image creation are used less frequently than less severe tactics such as slight image creation. This is in line with results from Jansen et al. (2012) who found that applicants primarily used those faking and IM tactics that are expected by interviewers, such as honest self-promotion. More generally, evidence suggests that applicants engage in more honest IM than in faking (Amaral, Powell, & Ho, 2019;Bourdage et al., 2018).

| Discussion and future research directions
Extant research suggests that the answer to our second question ("How common is faking in interviews?") is that a substantial proportion of applicants do fake. However, the majority of studies are based on student samples (Levashina & Campion, 2007;Weiss & Feldman, 2006), and studies including more experienced applicants report somewhat lower use of faking (Bourdage et al., 2018;Roulin et al., 2014). Yet, more research is needed to investigate faking prevalence for applicants with differing levels of qualification, under differing economic conditions, or from different industries.
Indeed, faking models suggest that people are less motived to fake when they do not see a need to do so (Ellingson, 2012;. For instance, applicants with more work experience have had opportunities to accumulate more skills, qualifications, or professional accomplishments, which they can rely on when answering interview questions. In contrast, less experienced applicants might be more likely to rely on faking to compensate for their lack of qualifications. Taken together, this leads to the following propositions: Proposition 1 Although faking is prevalent among applicants, it is still less common than honest IM.
Proposition 2 Faking is less common among more qualified applicants than among less qualified applicants, but more likely when the economy or industry is facing difficult times.
An important limitation of research on the prevalence of faking concerns how faking data is collected. Specifically, the common use of self-reports might over-or underestimate the true prevalence of faking estimates because applicants might not honestly report dishonest behaviors, especially in a high-stakes selection context. Thus, correct prevalence estimates may require the use of alternative approaches to measure faking. One option could be to include a few interview questions about seemingly job-relevant but made-up content to potentially identify applicants who fake, similar to the bogus items approach in biodata research (e.g., Levashina, Morgeson, & Campion, 2009). Another way could be to use measures that help dealing with the problem of social desirable responding such as the randomized response technique (e.g., Jansen et al., 2012).

| WHAT ARE THE ANTECEDENTS OF FAKING IN INTERVIE WS?
Extant research has dealt with five major categories of factors that influence the intention to fake and occurrence of faking in interviews: applicant characteristics, interview format, interviewers' behaviors, culture, and contextual factors.

| Applicant characteristics
As noted above, most theories of faking suggested that applicants' personality, beliefs/attitudes, and skills influence faking. Accordingly, most of the empirical research that we found dealt with applicant characteristics (see Table 2 for an overview).
Several studies examined personality and the Big Five traits. Roulin . This is also true for Integrity as a similar personality construct (Levashina & Campion, 2007, Study 4).
Studies have also explored the relationship between self-monitoring and faking, but the existing results are inconsistent. While Weiss and Feldman (2006) found that the number of lies in an interview did not correlate with self-monitoring, Levashina  As noted above, faking models vary in the role attributed to abilities (e.g., Marcus, 2009;McFarland & Ryan, 2000;Snell et al., 1999): direct antecedent of faking versus moderating the relationship between faking intentions and faking versus moderating the relationship between faking and interview performance. However, there is hardly any research on the impact of abilities and the available evidence is mixed. Specifically, two studies by Buehl and Melchers (2017) found no support for a moderating effect of interviewees' abilities. This was true both for self-reported faking behavior in previous selection interviews and for improvements in performance in an interview simulation (i.e., in a selection vs. an honest condition).
Furthermore, with regard to direct effects, Buehl and Melchers (2017) found a negative relationship between cognitive ability and self-reported faking in previous interviews, but no correlation between cognitive ability and self-reported faking in the simulated interview. However, cognitive ability was positively related with improvements in interview performance (in a selection condition relative to an honest condition), which is in line with arguments that ability is relevant for successful faking when applicants are highly motivated to present themselves positively (e.g., Levashina & Campion, 2006 usually not set up to investigate these characteristics, but generally measured them as potential control variables (see Table 3). Three studies in two papers reported correlations between the level of education or employment status and faking intentions or behavior (Bourdage et al., 2018, Study 5;, Studies 1 and 2): Individuals' level of education or their GPA seem to have little effect on faking, even though one study found that more educated individuals reported slightly lower faking intentions , Study 1). Two of those studies also reported that individuals who were employed were slightly less inclined to fake. In addition, while more studies report relationships between demographic variables and faking, correlations were rather small and often not significant. The only consistent findings seem to be that age is negatively related to faking and that men tend to fake slightly more than women (see Table 3).  (2017) 59 Behavior −.12 Weiss and Feldman (2006) TA B L E 2 (Continued)

| Interview format
(PBDI) questions . Specifically for faking, Van Iddekinge, Raymark, and Roth (2005) suggested larger faking effects for SI versus PBDI questions because answers to SI questions cannot be verified. In line with this, Levashina and Campion (2007, Study 6) found more faking in SI than PBDI questions. In contrast, however, Bourdage et al. (2018, Study 5) found that applicants faked less when SI questions were used (vs. not used), but found no difference for PBDI questions. They also generally did not find more faking on less job-related questions (e.g., about strengths and weaknesses or personal preferences). Furthermore, Levashina and Campion (2007, Study 6) found that follow-up questions increased the occurrence of faking, suggesting that probing might not be a suitable way to determine the truthfulness of applicants' answers.
Theoretical models of faking (Levashina & Campion, 2006; argue that more structured interview formats should reduce applicants' opportunity to fake. In addition, a meta-analysis on the relationship between IM and interview performance found larger correlations for unstructured than for structured interviews (Barrick et al., 2009), but this research did not differentiate honest from deceptive IM. Furthermore, Swider, Barrick, Harris, and Stoverink (2011) found no significant correlation between the duration of the rapport-building phase at the beginning of the interview and interviewees' use of faking. And Bourdage et al. (2018, Study 5) found that neither using a panel of interviewers nor conducting longer interviews influenced faking use.
Research on personality inventories suggests that warning applicants that the organization can determine the truthfulness of their responses and punish those who faked can be an effective means to reduce faking (e.g., Fan et al., 2012).
However, only one study investigated warnings in interviews: Law et al. (2016) found that identification warnings (i.e., telling interviewees that the organization can identify how truthful their responses are) moderately but significantly reduced self-reported faking. In contrast, moral warnings (i.e., telling interviewees that being honest is the right thing to do) or the combination of both warning types did not reduce faking. Thus, overall, the effectiveness of warnings seems to be limited to methods requiring to emphasize the potential risks associated with faking. Furthermore, even though Law et al.
found that warnings did not impact procedural justice perceptions or performance anxiety, warnings might have potential negative effects on other applicant reaction variables that were not investigated in their study.

| Interviewer characteristics
There is hardly any research on interviewer characteristics that considered whether interviewers have an impact on interviewees' faking and this research also found hardly any effects. Specifically, Bourdage et al. (2018, Study 5) only found more deceptive ingratiation (but no difference for other faking tactics) when interviewing with the potential supervisor versus an HR professional.

| Culture
In a recent study with participants from 31 different countries, Fell, König, and Kammerhoff (2016) found cross-cultural differences in attitudes towards faking in interviews. They divided intentions to fake into mild versus severe faking (i.e., conceptually similar to slight vs. extensive image creation) and examined relationships with GLOBE cultural dimensions. Intentions to fake had significant negative relationships with uncertainty avoidance (but only for mild faking) and gender egalitarianism, but positive relationships with power distance and in-group collectivism.
In addition, there are a few studies about reported faking in specific countries: U.S. applicants generally tended to fake more than European applicants (who were from Iceland and Switzerland), and slightly more than Chinese applicants . A follow-up study by Husain, Dayan, Pathak, Langer, and König (2019)  and showed that these differences could be partly explained by regional difference in Counscientiouness or competititve worldviews.

| Contextual factors
In line with faking models that consider organizational attractiveness as an antecedent of applicants' motivation to fake (e.g., Marcus, 2009), Buehl and Melchers (2018) found that students reported higher faking intentions in admission interviews for more attractive study programs, but competition (i.e., the selection ratio) did not influence faking. In contrast, Ho, Powell, Barclay, and Gill (2019)

| Discussion and future research directions
Previous research examining which antecedents influence the extent of faking in interviews (our third question) mainly focused on applicant characteristics, even though there is also some evidence about the influence of culture, interviewer characteristics, inter- However, there is no evidence regarding other aspects of structure, and it is also well known that many organizations still use relatively unstructured interviews (e.g., Highhouse, 2008), Thus, future studies could investigate the impact of question consistency (i.e., asking the same questions in the same order), using descriptively anchored rating scales, or interviewer training, on faking behavior and its effect on ratings of applicants' interview performance. Based on the general IM and interview structure literatures (e.g., Barrick et al., 2009;Levashina et al., 2014), we expect that structure components should help to decrease the effects of faking behavior:

Proposition 3 Increasing the structure of the interview reduces applicant faking and its effects on interview performance ratings.
More research is also needed concerning the role technology plays. Applicants' performance is often lower in technologymediated than in face-to-face interviews (Blacksmith, Willford, & Behrend, 2016) and Blacksmith et al. proposed that applicants' use of IM might be impaired in technology-mediated interviews.
However, other aspects of faking might be encouraged in these interviews. As an example, asynchronous video interviews involve applicants video-recording their response to interview questions via an online platform so that these can be evaluated at a later time. Some features of the online platforms include offering applicants time to prepare their responses before answering or allowing them to re-r ecord their responses multiple times. This might make it easier to engage in self-focused faking tactics, such as slight or extensive image creation. In contrast, applicants talk to their webcam and not to an actual interviewer. This means that there is no target for other-oriented tactics like deceptive ingratiation, which might reduce such behaviors. It also eliminates feedback that applicants receive about the appropriateness of their responses (such as interviewers' probing, nodding, or smiling), which makes it more difficult to estimate whether they should engage in more or less faking on subsequent questions. Given all this, we suggest the following research question for future research:

Research Question 3 What influence do technology-mediated interviews have on applicant faking?
Further, more research is needed to examine the benefits and drawbacks of different forms of warnings against faking. Law et al. (2016) found that only one of two types of warnings reduced faking, but it is unclear why. More research is also needed about potential consequences on applicants' reactions, for instance perceptions of interpersonal treatment (e.g., Gilliland & Steiner, 2012). Warnings could make candidates feel more uncomfortable, which may negatively impact their ability to perform (Feeney, McCarthy, & Goffin, 2015).

Warnings combined with announcements of negative consequences
for those who are caught lying could also backfire and damage the recruitment function of the interview. All in all, it seems rather early to provide organizations with viable recommendations to prevent faking, making this issue a relevant topic for future research:

Research Question 4 How effective are different types of warnings for reducing faking in interviews and how do applicants react to such warnings?
In addition, more insight is also needed on contextual factors beyond interview format and content that can influence the occurrence of faking behavior. For instance, labor market characteristics (e.g., the availability of alternative job openings) or the attractiveness of a given job have been discussed in theoretical models (Roulin et al., 2016), but have only received preliminary empirical testing (see Buehl & Melchers, 2018;Ho et al., 2019). As mentioned above, it is possible that "people fake only when they need to fake" (Ellingson, 2012, p. 19). Therefore, we suggest: Proposition 4 Contextual factors such as attractiveness of a job or labor market characteristics influence applicant faking.

| Faking and interview outcomes
Even though some suggestions have been put forth to prefer alternative selection procedures to interviews because of the latter's potential for faking (Koenig et al., 2013), the extant research examining whether and how faking influences actual interview outcomes like interview performance ratings or interview success (e.g., receiving a job offer or an invitation for the next round of selection) reports mixed results ( Table 4). Studies that investigated relationships between self-reported faking and outcomes in actual interviews found correlations that ranged from small and negative to null to moderate and positive (Amaral et al., 2019;Bourdage et al., 2018, Study 5;Buehl & Melchers, 2017, Study 1;Levashina & Campion, 2007, Study 5;Roulin et al., 2014). Furthermore, four studies examined relationships between self-reported faking and interview performance in mock interviews: two studies found small to moderate positive correlations (Buehl & Melchers, 2017, Study 2; and two others found negative correlations (Bourdage et al., 2018, Study 4;Swider et al., 2011).
In addition, four experimental studies compared IM and interview performance in two conditions: interviewees instructed to answer honestly versus instructed to put their best foot forward.
We note that these studies generally refer to a faking condition/ manipulation, and we include them here because they provide valuable evidence for the potential effects of faking. However, we caution the reader that the "put your best foot forward" manipulation may not only trigger faking, but also honest IM. As such, differences in performance between an honest and a "best foot forward" condition can be due to honest IM, faking, or a combination of both. 1 Concerning the results from these experimental studies, Allen, Facteau, and Facteau (2004) found no difference between interview ratings in an honest versus two different "best foot forward" conditions. In contrast, Van Iddekinge et al. (2005)  performance differences between an honest versus a "best foot forward" condition. However, they looked into differences in the use of IM tactics (albeit without distinguishing between honest vs. dishonest IM) and found that individuals in the "best foot forward" condition used more self-focused IM (e.g., self-promotion) and more other-focused IM (e.g., ingratiation) than individuals in the honest condition. conditions. Mean differences between the honest and the "best foot forward" condition were three times larger for the personality inventory than for the interview. This suggests that although interviews can be successfully faked, they are probably harder to fake than personality tests. Furthermore, the divergent findings from Allen whereas the other studies used between-subjects designs.

| Faking and interview validity
A highly relevant practical question is whether faking influences the validity of interviews. Unfortunately, research examining the effect of faking on the construct-and criterion-related validity of selection interviews is scarce. Van Iddekinge et al. (2005) found that the construct-related validity of their structured interview was worse in the "best foot forward" condition than in the honest condition, which is comparable to findings on faking in personality tests (Schmit & Ryan, 1993 Moreover, the level of construct-related validity in their honest condition was unusually high as compared to other selection interview research (e.g., Melchers et al., 2009).
Only two studies investigated the effects of faking on criterion-related validity directly.  found a nonsignificant correlation (r = .15) between self-reported faking measured after a mock job interview and ratings of in-role performance by supervisors. Buehl et al. (2019) manipulated faking with psychology students completing a mock interview designed to predict academic performance once in an honest condition (i.e., instructed to answer honestly) and once in a "best foot forward" condition (i.e., instructed to present themselves in the best possible way, as if they had applied for an attractive graduate program).
They found that academic performance was better predicted in the "best foot forward" condition, whereas citizenship behavior was better predicted in the honest condition. They argued that applicants' abilities (e.g., cognitive ability or the ability to identify the selection criteria) are necessary for both successful faking in interviews and task performance (cf. Van Iddekinge, 2015, for ATIC, andSchmidt &Hunter, 2004, for cognitive ability) and therefore contribute to criterion-related validity.
In contrast, information obtained in the honest condition allows for a better prediction of citizenship behavior, which is more strongly associated with personality and motivation in comparison to abilities.
In addition, the meta-analysis by Huffcutt, Conway, Roth, and Klehe (2004) offers indirect evidence for the potential effect of faking on criterion-related validity. They compared interview criterion-related validity for predicting job performance with two designs: concurrent (i.e., with job incumbents) versus predictive (i.e., with applicants, and thus interviewees who were more prone to faking).
Interviews from both designs predicted job performance, but criterion-related validity was higher for concurrent than predictive designs.
In contrast to Buehl et al. (2019), this suggests that interviews are better predictors of performance when they are less prone to faking (i.e., from concurrent designs) but that interviews are still predictive when applicants can fake. However, there might also be other reasons for the differences such as the passage of time between the interview and measurement in the criterion in predictive designs or better job knowledge by incumbents (cf. Van Iddekinge & Ployhart, 2008).

| Discussion and future research directions
Although practically very important, our fourth question ("Does faking in interviews matter?") is the one that has received the least attention so far. The experimental studies by Van Iddekinge et al. Evidence from correlational studies examining faking in actual interviews is more mixed. It is also unclear whether the null findings from high-stakes interviews (e.g., Bourdage et al., 2018, Studies 4 and 5;Roulin et al., 2014) represent evidence that faking is unrelated to interview outcomes or whether they are due to applicants misreporting faking. This is an important issue for future research.
More generally, we clearly need more field studies with high-stakes interviews that examine faking/IM effectiveness in interviews. A potential way to conduct such studies would be to compare data from a high-stakes selection situation with a more honest condition, for example as part of a validation study in which interview data from employees are specifically collected for research purposes so that interviewees can be encouraged to answer honestly. Furthermore, following the example of Roulin and Bourdage (2017), it is possible to clearly separate the interview from questioning applicants about their actual faking behavior. Nevertheless, given that research on faking in personality tests revealed more faking in the lab (instructions to fake vs. to respond honestly) than in the field (applicants vs. incumbents, see Birkeland et al., 2006, andViswesvaran &Ones, 1999), we propose a similar effect for interviews: Proposition 5 Faking has stronger effects on interview performance in simulated interviews and/or following instructions in experimental studies than in high-stakes selection interviews.
Furthermore, all the studies that looked at actual improvements in a "best foot forward" condition relative to an honest condition used highly structured interviews that were clearly job-or taskrelated. However, it remains unclear whether faking also leads to higher performance in less structured interviews with less job-related questions (e.g., about leisure activities, general strengths and weaknesses, personal goals):

Research Question 5 Does the level of interview structure moderate the relationship between applicant faking and interview performance?
The studies by  and by Buehl et al. (2019) provide initial evidence concerning the effects of faking on criterion-related validity. However, whether the small positive relationship they found generalizes to high-stakes selection interviews must be determined. Because cognitive ability or ATIC are likely to play a role in successful interview faking and are also predictive of task performance Schmidt & Hunter, 2004), faking could contribute to the prediction of task performance. In contrast, applicants usually fake to compensate for a lack of qualification or fit, and thus faking should negatively impact criterion-related validity (e.g., Levashina et al., 2014). Furthermore, Buehl et al.'s (2019) findings suggest that faking can impair the prediction of citizenship behavior. And, because faking is associated with the Dark Triad or low Honesty-humility, it seems likely that faking also impairs the prediction of other contextual performance criteria such as counterproductive work behavior. This would be in line with recent findings suggesting that applicants who fake on personality tests are more likely to engage in deviant or counterproductive work behaviors (O'Neill et al., 2013;Peterson, Griffith, Isaacson, O'Connell, & Mangos, 2011). Accordingly, we propose: Research Question 6 How does faking affect interview criterion-related validity for predicting task performance?

Proposition 6 Faking impairs interview criterion-related validity for
predicting contextual performance.
In addition, it would also be interesting to explore the effects of honest IM versus faking, but also of various types of faking tactics, on criterion-related validity. For instance, honest IM should theoretically contribute positively to validity because it provides interviewers with more accurate job-relevant information about applicants (Bourdage et al., 2018). In contrast, the more severe forms of faking (e.g., extensive image creation) could potentially harm validity because interviewers collect made-up information about applicants' qualifications (e.g., Levashina & Campion, 2007). The impact of less severe forms of faking (e.g., slight image creation) might sit somewhere in between. As such, the positive findings from studies using the "best foot forward" condition described above could have been due to the manipulation triggering mostly honest IM or slight image creation, rather than extensive image creation. Accordingly, we propose the following research question:

Research Question 7 How do different IM and faking tactics
influence criterion-related validity?

| C AN WE DE TEC T FAK ING IN INTERVIE WS?
Decades of empirical research on lie detection in the deception literature suggest that people are generally not effective lie detectors. For instance, meta-analytic results revealed an average of 54% correct lie-truth judgments, which represent a very marginal improvement over the 50% chance level (e.g., Bond & DePaulo, 2006).
Furthermore, research conducted in both social and forensic psychology has failed to identify relevant individual differences in lie detection (Bond & DePaulo, 2008). Importantly, although there are some valid cues or indicators of deception, people tend to rely on the wrong ones (DePaulo et al., 2003). For instance, both lay people and professionals such as police officers often rely on invalid nonverbal cues such as gaze aversion or fidgeting and ignore valid verbal cues such as vagueness or contradictions (Vrij, Granhag, & Porter, 2010).
Examination of interview faking detection is still in its infancy, but the evidence accumulated so far is generally aligned with deception detection research: Interviewers have difficulties to detect faking and effective strategies to improve faking detection have yet to be validated and implemented. For example, Culbertson, Weyhrauch, and Waples (2016, Study 2) found that ratings of interviewees' performance differed based on raters' perceptions of truthfulness but these perceptions were largely inaccurate. Furthermore, Reinhard, Scharmach, and Müller (2013) conducted an experimental study with participants attempting to detect deception (descriptions of true vs. invented job experiences) in recorded mock interviews.
Participants' accuracy was 52.39%, that is, only slightly better than chance level (i.e., 50%). However, this might still be an overestimation of faking detection, as applicants are not just honest or deceptive, but can use a variety of faking tactics. In line with this, Roulin, Bangerter, and Levashina (2015) conducted a series of five experimental studies investigating interviewers' ability to detect various forms of honest IM and faking tactics and found that most of the time interviewers failed in detecting faking. For instance, interviewers correctly detected only 12%-19% of faking tactics used by interviewees. Moreover, in a field study, Roulin et al. (2014) found no significant correlation between applicants' self-reported use of faking tactics and interviewers' perception of faking.
In addition, interviewers' experience and interviewer characteristics do not help to detect faking. For example, Roulin et al. (2015, Study 1) found no relationship between years of experience and detection abilities even though their participants had an average of 9.6 years of experience as interviewers. Furthermore, they found no difference in detection between complete novices (e.g., students) and experienced interviewers. Similarly, in Reinhard et al.'s (2013)  respond honestly. They found significant differences between honest and deceptive answers for 11 of the 14 cues that were considered (e.g., implausibility, uncooperativeness) but again most of these differences were small. And Roulin and Powell (2018)  A second strategy involves changing interview design aspects such as the type of interview question or the interview medium.
However, this strategy seems to have no consequences for the ability to detect faking. Roulin et al. (2015, Study 2) found that interviewers were better at correctly identifying honest IM with situational questions, but no such effect was found with faking. Similarly, the interview medium does not seem to make a difference either.
Ferrán-Urdaneta and Storck (1997) found no difference in the ability to detect faking between face-to-face and video conference interviews. Finally, a third strategy involves warning interviewers that applicants might fake. However, Toris and DePaulo (1984) found that interviewers who received such a warning were not more accurate in their judgments of deception.

| Discussion and future research directions
Despite some preliminary evidence for the potential value of contentbased techniques, existing research has not found any definitive effective strategy to detect faking, and thus the answer to our fifth question ("Can we detect faking in interviews?") is currently "no." Some valid cues of faking (such as less smiling, less silences, and more speaking errors) have been identified, but effects are usually only small (e.g., Culbertson et al., 2016;Schneider et al., 2015). Analyzing the content of applicants' answers (e.g., with the CBCA approach) appears to be a more promising approach (Roulin & Powell, 2018). Yet, future research is needed to systematically examine the application of content-based techniques in high-stakes selection contexts and to explore whether interviewers could be trained to use such techniques or whether automatic computer-based coding could be applied. Can we detect faking? 7. Training based on content-based lie detection strategies is a viable strategy to help interviewers deal with faking

Dos Don'ts
• Understand that applicants can fake and that most applicants do fake (at least to some extent) • Ask follow-up questions as an attempt to reduce faking • Consider that applicants fake partly because of their personality, values, or attitudes but also take cultural differences or contextual factors (e.g., attractiveness of the job, competition) into account • Use warning that faking can be detected (unless ready to deceive applicants about the actual possibility to detect faking) • Increase the degree of interview structure (e.g., use standardized and job-related questions) • Rely on interviewers' intuition, experience, or abilities (e.g., emotional intelligence) to try to detect when/how applicants fake • When trying to assess the veracity of applicants' responses, train interviewers to focus on content and rely on a combination of indicators (e.g., level of details, plausibility) • Rely on non-verbal behaviors to identify whether applicants are honest or faking

TA B L E 6 Summary of recommendations for practitioners about interview faking
Proposition 7 Training based on content-based lie detection strategies is a viable strategy to help interviewers deal with faking.

| CON CLUS IVE COMMENTS
Since the starting point of systematic research on faking in interviews, there has been a fair amount of theoretical and empirical research. We organized our review of this research vis-à-vis five core questions: (1) What is faking in interviews? (2) How common is it? (3) Which are its antecedents? (4) Does it matter? (5) Can we detect it?
In a nutshell, the existing literature quite extensively covers the first three questions, although there still are some gray areas.
For instance, our knowledge of the prevalence of faking in practice is largely based on students, young applicants, or online samples.
And, the research on faking antecedents focusses largely on individual differences (e.g., personality), while empirical work examining situational factors is scarce despite several theoretical models discussing them. As an example, there is very limited evidence about how interview structure influences faking and about the impact of different structure components. Similarly, while the attractiveness of a given job or labor market characteristics (e.g., unemployment, availability of alternative job openings) have been discussed in theoretical models , empirical examination is still very limited. Applicants' ability to fake and the factors that could influence it, such as cognitive ability, social skills, or the ability to identify the constructs being measured (e.g., Melchers, Bösser, Hartstein, & Kleinmann, 2012;Oostrom, Melchers, Ingold, & Kleinmann, 2016), has also not been extensively examined.
Our knowledge about the fourth question is limited for two reasons. First, only a handful of studies examined how faking impacts interview performance, and results are mixed and confounded by study designs. Second, there is limited work (both theoretical and empirical) discussing the impact of interview faking on criterionrelated validity, and whether faking is differently associated with task versus contextual performance. We argue that faking can potentially contribute to the prediction of task performance, especially when abilities (e.g., cognitive ability) are relevant for both faking effectiveness and successful job performance. However, faking to compensate for a lack of qualifications (especially the use of extensive image creation) might impair the prediction of task performance.
And the fact that faking is related to "darker" personality traits (i.e., low Honesty-humility, high Machiavellianism) suggests that it could negatively impact predictions of contextual performance. Finally, although several studies have examined our fifth question in the last few years, the accumulated evidence depicts a quite bleak image of organizations' (and interviewers') success in their attempts to detect (or deal with) faking.
In addition, our review identified new propositions and research questions that should encourage future research and help interview faking research progress further (see Table 5 for a summary). Our review also highlights several practical implications (see Table 6 for a summary of recommendations for practitioners). But for this field to move forward and have an even bigger practical impact, one must reconsider the way studies are designed and conducted. For instance, examining the prevalence of interview faking in high-stakes selection contexts may require the development and validation of alternative approaches to measure faking (in addition to simple selfreports that may scare respondents in such contexts), similar to the use of over-claiming measures in knowledge tests (e.g., Bing, Kluemper, Davison, Taylor, & Novicevic, 2011) or bogus items in biodata research (e.g., Levashina et al., 2009). Alternatively, a clearer understanding of why and how applicants fake may require researchers to conduct experimental studies and examine (and measure) faking at the level of the interview question (vs. the entire interview). For instance, the use of asynchronous video interviews may facilitate a more precise and systematic reporting of faking by research participants (e.g., by rewatching their responses and reporting when they faked, how much they faked, and what tactics they used).
Longitudinal studies are necessary to examine the impact of faking on workplace outcomes. In addition, as noted above, such research should not only focus on predicting job performance, but also consider other relevant outcomes, such as counterproductive work behavior, absenteeism, turnover, or citizenship behavior. For instance, research on faking in personality tests suggests that fakers tend to engage in more counterproductive work behavior once hired (O'Neill et al., 2013), but we do not know whether this also applies to faking in interviews.
More than a decade after Campion's (2006, 2007) seminal work on the antecedents and types of faking tactics used by applicants in employment interviews, we believe that this review serves both as a comprehensive summary of the accumulated knowledge about interview faking and as a roadmap for future research. Given this active field of research, we are looking forward to the new insights that lie ahead of us.