Discrimination in the Labor Market and Audit Stages

Philosophical Framework

Labor market discrimination, as a phenomenon, is deeply rooted in social and economic structures, requiring an interdisciplinary approach to its study. The analysis is based on a critical materialist approach, which views discrimination not as a random deviation but as a systemic result of the distribution of power and resources shaped by historical and institutional factors [Marx, 1867]. This approach allows moving beyond superficial explanations to explore the deep mechanisms reproducing inequality, as well as the role of audit as a tool for identifying and transforming these structures.

Understanding discrimination as a manifestation of "scientific-political power," where capital acts only as its "most important effect" [Marx, 1867], allows considering audit not merely as a technical procedure but as an instrument capable of revealing hidden forms of this power and its influence on labor relations. Audit becomes part of a broader strategy aimed at "reducing the mutilating structures of existence" [Marx, 1867], directed towards creating a more just and efficient labor market.

Introduction

Labor market discrimination remains one of the most acute problems affecting economic efficiency and social justice. Despite decades of research, legal reforms, and activism, inequality in hiring, wages, and promotion persists, sometimes taking hidden and sophisticated forms. Contemporary studies show that discrimination manifests not only in direct refusals based on race or gender but also through subtler mechanisms, such as client bias or even the difficulty of pronouncing a candidate's name [Ge et al., 2024].

The relevance of the problem is underscored by the fact that discrimination harms not only individuals but also reduces overall economic productivity by hindering the optimal allocation of human capital. For example, studies show that the gender wage gap, although narrowing, remains significant, especially in senior positions [Blau et al., 2017]. In response to these challenges, discrimination audits have become a key tool for organizations seeking to identify and eliminate bias in their practices. However, as Zhuk and Melnik note, audit methodology requires continuous improvement to adequately respond to new forms and manifestations of discrimination [Zhuk et al., 2021].

Literature Review

Methodologies for Measuring Labor Market Discrimination

How can something as elusive as labor market discrimination be measured? It is often not about direct prohibition but hidden prejudices influencing hiring, promotion, or pay decisions. This is not merely a statistical question but a methodological one capable of revealing these implicit patterns.

One of the most direct and convincing methods is so-called audit studies or field experiments. Their essence lies in researchers sending pairs of candidates to real job vacancies who are identical in all objective characteristics (education, experience, skills) but differ in a characteristic that may be subject to discrimination—such as race, gender, or ethnicity. If one type of candidate receives significantly fewer callbacks, interview invitations, or job offers, this serves as strong evidence of discrimination. Holzer and Ludwig [Holzer et al., 2003] note that such methodologies, developed for analyzing housing and labor markets, can be very useful for studying discrimination in other areas, such as education. They emphasize that audit studies allow direct observation of employers' reactions to different candidate characteristics, minimizing the influence of unaccounted variables.

However, despite their apparent simplicity and directness, audit studies are not without criticism. Heckman [Heckman, 1998] points out their fundamental limitations, stating that "the validity of audit methods critically depends on untested assumptions about the equality of distribution of unobserved (by audit developers) productivity components across racial/gender groups and how the labor market functions." In other words, if there are hidden differences in productivity that researchers cannot account for, audits may falsely identify discrimination where none exists or, conversely, conceal its true presence. This raises an important question about how fully we can control all relevant variables in real conditions.

Besides direct field experiments, there are statistical methods analyzing large datasets on employment, wages, and promotion. These methods seek to identify unexplained differences in outcomes for different groups after accounting for all available objective factors such as education, work experience, position, and industry. Michael Piette [Piette, 1991], in his work "Economic Methodology and the Analysis of Employment Discrimination," considers the application of such approaches. He likely focuses on regression analysis, which allows estimating the influence of various characteristics on wages or hiring probability, isolating the effect of presumed discriminatory traits. However, as with audits, statistical methods face the problem of "unobserved variables." There is always a possibility that unexplained differences are related to factors not included in the model rather than discrimination.

Clifford Hawley [Hawley, 1995], responding to Piette's work, likely deepens the discussion on statistical methodology, pointing out potential pitfalls and difficulties in interpreting results. He might emphasize that correlation does not always imply causation and that even the most sophisticated statistical models cannot fully exclude the possibility that observed differences are due to nondiscriminatory factors simply not measured. For example, if women earn less than men, it may be related not only to discrimination but also to differences in career choices, willingness to work overtime, or relocate, which are hard to capture in standard databases.

Contemporary research attempts to overcome these limitations by combining various approaches and using more sophisticated experimental designs. For example, some studies examine the impact of hard-to-pronounce names on labor market outcomes, a subtle indicator of ethnic or racial discrimination [Ge et al., 2024]. The authors found that candidates with less "fluent" names have lower success rates, especially if their resumes are not very strong, indicating cognitive biases. This demonstrates how even seemingly minor details can serve as markers for discrimination.

Other studies focus on so-called "customer discrimination," where bias originates not from the employer but from consumers, which in turn affects worker productivity and opportunities. Kelley, Lane, Pecenzo, and Rubin, in their study of online sales in sub-Saharan African countries, found that assigning a female-sounding name to a seller leads to a significant reduction in purchases, even if the seller's behavior does not change. This shows that discrimination can manifest in unexpected forms and requires comprehensive methods for detection.

Valeriy Zhuk and Katerina Melnik [Zhuk et al., 2021], in their work on the subject of audit, emphasize that "the internal dynamism of the foundations that currently form the basis for understanding this category" poses a serious problem for researchers. They argue that the subject of audit should be interpreted so universally that the emergence of new objects is not limited by the "verbal" boundaries of the subject. This is especially relevant for discrimination audits, where new forms and manifestations of bias constantly arise, requiring methodological adaptation. Institutionalism, in their view, offers a suitable doctrine for explaining the development of any human activity through the study of institutional environment and influences, which can also be applied to understanding the evolution of discrimination.

Measuring labor market discrimination is a complex task requiring a multifaceted approach. No method is perfect, and each has its limitations. Audit studies provide direct evidence but may be subject to criticism due to unaccounted variables. Statistical methods allow analyzing large data volumes but face causality and information completeness problems. Combining these approaches, as well as developing new experimental designs that consider subtle bias manifestations such as name effects or customer discrimination, appears to be the most promising path. However, even with the most advanced methodologies, the question remains how fully we can capture all nuances of human prejudices and their impact on economic decisions.

Ultimately, understanding methodological limitations and continuously improving approaches to measuring discrimination are key to developing effective strategies to combat it. After all, if we cannot accurately measure the problem, how can we hope to solve it? This leads us to the next question: which specific candidate characteristics most often become objects of discrimination, and how do these methodologies help us identify them?

Influence of Various Characteristics on Hiring Discrimination

If the previous section discussed methodological approaches to detecting discrimination, now it is worth moving to specific manifestations of this phenomenon, focusing on how various applicant characteristics affect their chances during hiring. Labor market discrimination is not an abstract concept but a tangible reality manifested in unequal treatment based on traits unrelated to productivity or qualifications. These may include gender, race, ethnicity, sexual orientation, socioeconomic status, and even name features.

Starting with gender discrimination, which remains one of the most studied and persistent forms of bias. Studies show that women often face biased treatment even when their qualifications are identical to men's. For example, [Blau et al., 2017] note that the gender wage and career advancement gap persists across many industries. Moreover, discrimination may manifest not only from employers but also from clients. For instance, a study in sub-Saharan African countries measured the impact of client gender discrimination on online sellers' productivity. They found that assigning a female-sounding name led to a 50 percent reduction in purchases (Kelley et al., 2024). This indicates that even when worker behavior does not change, external gender-related factors can significantly affect economic outcomes.

Similar effects are observed in other contexts. [Chavez et al., 2022] studied gender and racial discrimination in hiring accountants before and during the COVID-19 pandemic. They found that white women experienced positive changes in callback rates during the pandemic, possibly related to a reduction in female labor supply. However, this does not negate the overall picture of bias. [Park et al., 2025] also emphasize that gender stereotypes continue to influence hiring, especially in professions traditionally considered "male" or "female."

Racial and ethnic discrimination is another deeply entrenched problem. Studies using audit methods repeatedly demonstrate that candidates with names associated with certain racial or ethnic groups receive significantly fewer interview invitations. [Quillian et al., 2017] conducted a meta-analysis of audit studies and concluded that there has been no change in discrimination levels against African Americans since 1989, although there are some signs of reduced discrimination against Latinos (Quillian et al., 2017). This indicates the striking persistence of racial discrimination in the U.S. labor market. [Boyd-Swan et al., 2018] confirm these findings by studying discrimination in hiring childcare teachers. They found that Black and Latino candidates receive significantly fewer interview requests than equally qualified white candidates. Moreover, their results align with a customer discrimination model, where the racial and ethnic composition of the client base correlates with the characteristics of candidates receiving interview invitations.

Discrimination can be linked even to seemingly minor details such as name pronunciation or complexity. [Ge et al., 2024] analyzed data on economics PhD candidates and found that name complexity negatively correlates with the likelihood of obtaining an academic position or tenure-track job, as well as early-stage research productivity (Ge, Wu, 2024). This raises the question of how deeply ingrained subconscious biases influence candidate evaluation.

Sexual orientation is also a factor in discrimination. [Tilcsik, 2011] conducted the first large-scale audit study of discrimination against openly gay men in the U.S. He sent pairs of fictitious resumes responding to job ads, varying only the indication of sexual orientation. Results showed that openly gay men faced fewer interview invitations, highlighting bias in hiring.

Beyond obvious categories like gender and race, discrimination can manifest in less apparent traits such as class or appearance. [Thomas, 2018] and [Spencer et al., 2020] studied the influence of cultural class signals on employment chances. [Spencer et al., 2020] conducted research in Jamaica, where racial homogeneity allows better isolation of class effects. They found that employers prefer candidates perceived as coming from high-income backgrounds. Paradoxically, candidates with employer-preferred names had lower chances if they had high-quality resumes, possibly indicating an effort to avoid "overqualified" candidates or other class-related biases (Spencer et al., 2020). This study challenges the assumption that racial discrimination is the sole or main form of bias, suggesting that class discrimination can be as significant as racial discrimination.

Appearance also plays a role. [Roehling, 1999] notes that attractive people often have advantages in hiring and promotion, a form of discrimination unrelated to qualifications. These studies show that employers, consciously or subconsciously, use a wide range of characteristics to form decisions beyond objective assessment of skills and experience.

All these examples demonstrate that hiring discrimination is a multifaceted phenomenon affecting various aspects of an applicant's identity. Audit methods discussed earlier allow revealing these hidden biases, but their interpretation requires deep contextual understanding. As [Berrey et al., 2012] note, perceptions of fairness in such cases strongly depend on "situated justice," where specific and private experiences deeply influence assessments of what is fair and unfair in the legal system (Berrey et al., 2012). This means that even if the legal system strives for neutrality, real experiences and institutional conditions can create significant asymmetries.

The influence of various characteristics on hiring discrimination is not limited to simple binary categories. It permeates many aspects of identity and social status, creating a complex network of biases. Understanding these mechanisms is the first step toward developing effective anti-discrimination strategies, which in turn leads us to the question of how context and policy can influence these processes.

Influence of Context and Policy on Discrimination

If the previous section focused on how individual applicant characteristics influence hiring discrimination, now it is worth considering the broader picture. Discrimination is not only the result of individual prejudices but also a product of complex interactions among macroeconomic conditions, institutional norms, and government policy. The context in which hiring occurs can both amplify and mitigate manifestations of discrimination, making it more or less visible.

Macroeconomic changes, for example, can significantly transform discrimination dynamics. Take the COVID-19 pandemic: it was an unprecedented shock to global labor markets. The study by [Chavez et al., 2022] showed that during this period, white women in the U.S. experienced positive changes in employer callbacks, surpassing white men. This change was especially noticeable in regions with significant reductions in female labor supply. This suggests that in labor shortages, employers may be forced to reconsider traditional preferences, even if unconsciously.

However, not only crises change the discrimination landscape. Anti-discrimination policies and legal norms play a key role in shaping expectations and behavior in the labor market. As [Blau et al., 2017] note, policies aimed at combating discrimination, as well as family support measures such as parental leave, can influence the gender wage gap. These measures aim not only to prevent overt acts of discrimination but also to create more equal conditions for workforce participation, ultimately reducing structural barriers.

Nevertheless, the effectiveness of such policies is not always straightforward. For example, [Darity et al., 1998] point out that although certain periods (e.g., 1965–1975 in the U.S.) saw reductions in racial discrimination, this progress later stalled or even reversed. This raises questions about the sustainability of changes induced by legislation and what factors may neutralize their effect. Possibly, laws can change external behavior but not always eradicate deep-seated prejudices.

Cultural context and social integration also matter. The study by [Štangej et al., 2018] showed that for Polish immigrants in Iceland, having qualifications obtained in the host country reduces employer prejudice. This suggests that social integration signals, such as local education, can serve as a "pass" helping overcome barriers related to ethnicity. Employers likely perceive such education as confirmation not only of professional skills but also of adaptation to local norms and values.

Sometimes discrimination manifests in very subtle, almost imperceptible forms related to cultural preferences. [Thomas, 2018] found that cultural class signals in resumes affect callback likelihood for women applying for customer service positions. Displaying "high taste" (signs of higher social origin) led to significantly higher rates than "low taste." This indicates that employers may unconsciously prefer candidates whose cultural codes align with their own ideas of the desired employee image, especially in client-oriented sectors.

Even company statements about commitment to diversity do not always guarantee the absence of discrimination. [Kang et al., 2016] studied the practice of "whitening résumés," where racial minorities conceal or downplay their racial identity in job applications. They found a paradox: although minorities whiten résumés less often when applying to companies claiming to value diversity, these statements do not correlate with reduced discrimination against "non-whitened" résumés. This means declared values do not always translate into real hiring practices, and employers may still prefer candidates who seem more "neutral" or "familiar."

Besides explicit policies, less obvious institutional factors such as discrimination lawsuits also influence discrimination. [Berrey et al., 2012] analyze how plaintiffs and defendants perceive fairness in such cases. They note that for employer-defendants, settling cases—even with compensation payments—is often seen as a way to avoid larger problems and manage overall costs rather than as an admission of guilt or pursuit of justice. This can create a system where discrimination remains hidden, and its consequences are minimized through financial arrangements without systemic change.

Plaintiffs, on the other hand, often seek not so much monetary compensation as restoration of justice or reinstatement. As one plaintiff in [Berrey et al., 2012] stated, "I didn't want any money. I wanted my job back." This highlights the gap between legal outcomes (monetary awards) and what people consider a fair resolution. If the judicial system cannot ensure reinstatement, it may undermine trust in its ability to combat discrimination.

Context and policy form a complex environment where discrimination can manifest, hide, or transform. From macroeconomic shifts to subtle cultural signals and legal mechanisms, all these elements interact, creating a unique landscape for each discrimination case. Understanding these interrelations is critically important for developing effective anti-discrimination strategies. However, to assess the real impact of these factors and uncover hidden discrimination manifestations, systematic approaches such as audits are necessary, allowing in-depth examination of hiring practices and evaluation for bias.

Subject and Methodology of Audit

After considering the influence of context and policy on discrimination, it is logical to move to how exactly we can detect and measure these phenomena. Without a clear understanding of the subject and methodology of audit, any attempts to combat discrimination risk remaining at the level of good intentions. Discrimination audit is not just document checking; it is a deep immersion into organizational processes and culture, requiring specific tools and philosophical reflection.

The traditional understanding of the audit subject, as Zhuk and Melnik note, "focuses" on service-management functionality, which, in their opinion, cannot serve as a sufficient basis for comprehensive analysis [Zhuk et al., 2021]. The classical approach, focusing on evaluating publicly reported information, is undoubtedly important but does not cover the full complexity of discrimination, which often manifests in implicit forms and requires a broader perspective. If audit is limited only to financial reporting or compliance with formal procedures, it risks missing deep causes and manifestations of unequal treatment.

The expansion of audit methodology proposed by the same authors implies interpreting the subject as the study of the nature of audit functionality aimed at reducing uncertainty in the economic space and ensuring sustainable development, built on general scientific philosophical foundations and responding to social expectations [Zhuk et al., 2021]. This formulation is much broader and allows including in the audit scope not only compliance checks but also analysis of social, ethical, and managerial aspects directly related to discrimination. Essentially, this is a shift from a narrowly specialized check to a comprehensive study capable of revealing systemic problems.

Audit methodology, in this understanding, goes beyond a simple set of research techniques. It includes epistemological methodology, practical integrity methodology, and evaluation methodology (axiomethodology) [Zhuk et al., 2021]. This means the auditor must not only be able to collect and analyze data but also understand the deep mechanisms underlying discriminatory practices and evaluate them from ethical norms and sustainable development goals perspectives. For example, James Heckman critically views audit methods, pointing out that they may fail to detect discrimination where it exists or, conversely, find it where it does not, due to dependence on untested assumptions about equality of distribution of unobserved productivity components [Heckman, 1998]. This underscores the need for deep methodological reflection.

In the context of labor market discrimination, audit must be able to detect not only direct violations but also indirect, statistical forms of discrimination discussed by Gary Becker [Becker] and Kenneth Arrow. For example, "blind" auditions in orchestras, as shown by Claudia Goldin and Cecilia Rouse, significantly increase the likelihood of hiring women, indicating unconscious bias in traditional auditions [Goldin et al., 2000]. This is a vivid example of how changing evaluation methodology can radically affect outcomes and reveal hidden discrimination.

However, despite striving for objectivity, the audit process itself can be influenced by various factors. Berry, Hoffman, and Nielsen, in their study of discrimination lawsuits, note that plaintiffs often face a "maze of litigation," "personal difficulties," and "perceptions of defendant favoritism" [Berrey et al., 2012]. This indicates that even with formal procedures, the real experience of those facing discrimination may be far from ideal justice.

Moreover, the authors emphasize that many plaintiffs consider their lawyers incompetent or working against them, questioning the effectiveness of legal protection as such [Berrey et al., 2012]. If such problems arise even at the litigation stage, at the internal audit stage, where there is no independent judge, risks of bias may be even higher. This requires auditors not only professional competence but also ethical incorruptibility.

Financial costs of litigation are also a serious obstacle for plaintiffs, which, according to Berry, Hoffman, and Nielsen, "shocks" them when they realize how expensive pursuing their rights can be [Berrey et al., 2012]. This raises the question of audit accessibility for small and medium enterprises and the need to develop more economical and effective methodologies.

The subject of discrimination audit must cover not only formal aspects but also consider subjective experience, cultural features, and systemic biases. Audit functionality, as understood by Zhuk and Melnik, should be aimed at understanding the internal essence of audit as not only evaluative but also managerial and institutional phenomenon [Zhuk et al., 2021]. This means audit should not just state facts but also propose mechanisms for changing organizational culture and management practices.

Andras Tilcsik, for example, in his large-scale audit of discrimination against openly gay men in the U.S., used pairs of fictitious resumes to detect employer bias [Tilcsik, 2011]. His study showed that discrimination varies by region and is related to attitudes toward homosexuality and anti-discrimination laws. This demonstrates how audit methodology can be adapted to detect specific discrimination forms and their geographic features.

Ultimately, audit methodology must be flexible enough to adapt to various contexts and detect both explicit and hidden forms of discrimination. It should include quantitative methods for identifying statistical anomalies and qualitative methods for understanding subjective experience and organizational culture. Only such a comprehensive approach will allow not only detecting the problem but also proposing effective solutions, which is a necessary condition for preventing customer discrimination and its impact on workers.

Customer Discrimination and Its Impact on Workers

If the previous section focused on audit methodology as a tool for detecting discrimination within an organization, now it is worth considering a more complex aspect: how discrimination originating from clients affects workers and how this can be accounted for within audit functionality. As Zhuk and Melnik note, audit should strive for systemic cognition of social expectations (requirements) and the potential of professional audit capabilities [Zhuk et al., 2021]. In other words, it cannot ignore external factors shaping the work environment.

Discrimination from clients presents a unique problem because it is not a direct result of employer policy or practice but significantly impacts working conditions and employee well-being. Imagine a situation where an employee faces biased or even hostile treatment from a client due to their gender, race, or other characteristics. This not only creates a stressful work environment but can directly affect productivity and career prospects. For example, the study by Kelley, Lane, Pecenzo, and Rubin showed that in online sales in sub-Saharan Africa, assigning a female-sounding name led to a 50% reduction in purchases. This is direct evidence of how client bias can affect an employee's economic performance regardless of their actual skills.

Such "customer discrimination" puts the employer in a dilemma. On one hand, the company seeks to satisfy client needs and maximize profit. On the other hand, it must ensure equal working conditions and protect employees from discrimination. As Gary Becker notes in his taste-based discrimination theory [Becker], if clients have a "taste for discrimination," they are willing to pay more for service from representatives of certain groups or refuse services from others. This creates economic pressure on companies that may be forced to adapt to these preferences even if they contradict equality principles.

However, if the company yields to such pressure, it risks legal consequences and reputational costs. In the audit context, as Zhuk and Melnik emphasize, a selfish strategy benefits the audit system, but with socio-economic development, audit with such a strategy will remain outside the public need for it [Zhuk et al., 2021]. This means audit must go beyond mere formal compliance and consider a broader social mission, including protecting employees from discrimination even if it originates externally.

How can audit detect and assess client discrimination? This requires deeper analysis than traditional internal policy checks. It is necessary to study data on client interactions, feedback, complaints, and employee performance indicators depending on demographic characteristics. For example, if employees of a certain group consistently receive lower client ratings or more negative comments, this may indicate customer discrimination. It is important to distinguish actual low performance from biased perception.

An interesting aspect of customer discrimination also appears in pricing. Zamora, Mantilla, and Blanco [Zamora et al., 2021] found that prices offered to foreign buyers were higher than those for Colombian buyers. This is an example of price discrimination based on citizenship, which, although not always directly affecting workers, demonstrates companies' willingness to adapt to client preferences or prejudices, potentially creating precedents for other discrimination forms. If a company is willing to discriminate against clients based on citizenship, it may be more prone to ignoring discrimination against its employees by clients.

The problem is exacerbated by the fact that employees facing client discrimination often lack effective protection mechanisms. As Berry, Hoffman, and Nielsen [Berrey et al., 2012] note, workplace discrimination plaintiffs often face enormous difficulties, and very few cases end with court decisions based on legal claims. Instead, about 50% of cases end in settlements, often for a "nuisance value," allowing companies to avoid long-term costs but not addressing discrimination substantively [Berrey et al., 2012]. This creates a sense of impunity for discriminating clients and leaves workers without adequate protection.

Discrimination audit must include not only internal processes but also analysis of external factors such as client behavior. This requires developing new methodologies and tools to identify and measure the impact of customer discrimination on workers. Otherwise, as Zhuk and Melnik warn, audit risks remaining at the "19th-century" level, not meeting the challenges of the "digital economy, the age of artificial intelligence," and "sustainable development" [Zhuk et al., 2021].

Ultimately, the question is how deeply audit is prepared to delve into studying the "nature of audit functionality" to not only detect but also propose solutions for such complex and multifaceted problems as customer discrimination. This requires audit not just to state facts but to actively participate in shaping a more just and sustainable work environment, which in turn raises questions about its role and limits of responsibility.

Criticism and Limitations

Uncertainty in Interpreting Audit Results

One of the most significant limitations of discrimination audits is the inherent uncertainty in interpreting obtained results. Even if an audit reveals statistically significant differences in outcomes for different groups, this does not always unequivocally indicate discrimination. James Heckman [Heckman, 1998] emphasizes that "the validity of audit methods critically depends on untested assumptions about the equality of distribution of unobserved (by audit developers) productivity components across racial/gender groups and how the labor market functions." In other words, if there are hidden productivity differences that cannot be measured or accounted for in audit design, observed differences in hiring or promotion may be mistakenly attributed to discrimination. For example, if candidates from one group on average possess unobserved skills valued by employers but not reflected in resumes, a lower hiring rate for another group is not necessarily due to discrimination.

This problem is exacerbated by the fact that audit studies often focus on the "average" firm rather than the "marginal" one that determines the level of market discrimination [Heckman, 1998].

Even if an audit detects discrimination in a sample of firms, this does not provide a complete picture of the problem's scale across the entire labor market. Moreover, Heckman [Heckman, 1998] demonstrates that depending on the level of standardization used in the audit, completely opposite conclusions can be drawn: an audit can "find discrimination where there is none and hide it where it exists." This means audit conclusions can be sensitive to methodological decisions made by researchers, questioning their universality and objectivity.

Limitations of Audit Methodology and External Factors

Despite its value, audit methodology faces several practical and conceptual limitations that can distort results and complicate their application. Edward Larsen [Larsen, 2020] points out that "matched and unmatched designs differ more substantially than previously recognized: design choice can potentially affect estimates because matched design induces competition affecting callback probability for treated candidates when candidate pools are small." This means that in conditions of limited vacancies or candidates, the very fact of sending multiple resumes within an audit can artificially influence each one's chances, creating an effect that does not reflect the real labor market situation.

Moreover, audits often do not account for labor market dynamics and macroeconomic changes that can influence discrimination levels. For example, the study [Chavez et al., 2022] showed that gender and racial discrimination in hiring accountants changed during the COVID-19 pandemic. White women gained an advantage in callbacks, especially in regions with reduced female labor supply. If an audit is conducted under static conditions, it may miss such dynamic changes, leading to outdated or incomplete conclusions.

Audit results obtained in one period may be irrelevant for another, requiring constant updating and adaptation of methodologies.

The Problem of "Whitening Résumés" and Customer Discrimination

Another audit limitation is that it cannot always fully account for strategies used by applicants themselves to minimize discrimination. Sonya Kang and colleagues [Kang et al., 2016] studied the practice of "whitening résumés," where racial minorities conceal or downplay their racial identity in job applications. They found that although minorities whiten résumés less often when applying to companies claiming to value diversity, these statements do not correlate with reduced discrimination against "non-whitened" résumés. This creates a paradox: audits that do not consider whitening may underestimate the true level of discrimination since some candidates have already adapted to it.

Finally, discrimination audits traditionally focus on employer bias but do not always consider the impact of customer discrimination. The study showed that in online sales, assigning a female name to a seller led to a significant reduction in purchases. If an employer faces such client bias, it may influence hiring or promotion decisions even if the employer has no discriminatory intent. Audits that do not consider this external factor may wrongly attribute all responsibility to the employer without offering solutions for the broader problem. How can discrimination be effectively combated if its source lies outside the company's direct control?

Conclusions

• Labor market discrimination manifests in diverse forms, from overt bias to hidden prejudices related to gender, race, ethnicity, sexual orientation, and even name features, as confirmed by numerous audit studies [Quillian et al., 2017].
• Economic theories of discrimination, such as taste-based models [Becker] and statistical discrimination, explain its causes but do not always cover all subtle manifestations, including customer discrimination, which can significantly affect worker productivity and opportunities.
• Discrimination audit is a key tool for identifying bias in hiring, pay, and promotion practices, but its methodology requires continuous improvement and expansion to adequately respond to new discrimination forms and consider external factors [Zhuk et al., 2021].
• Audit effectiveness depends on overcoming methodological limitations such as the problem of unobserved variables [Heckman, 1998] and employing a comprehensive approach combining quantitative and qualitative methods for deep analysis of organizational culture and subjective experience.
• Macroeconomic changes, such as the COVID-19 pandemic, can influence discrimination dynamics by altering labor market supply and demand, which may temporarily weaken or strengthen bias manifestations [Chavez et al., 2022].
• Legal frameworks and anti-discrimination policies play an important role in combating discrimination, but their effectiveness may be limited if not accompanied by systemic organizational culture changes and consideration of complex litigation processes [Sturm, 2006].
• How can audit effectively detect and counteract discrimination originating not from employers but from clients while not violating principles of economic efficiency?