Final Report Summary - 0DISCRIMINATION (Discrimination Norms Enforced and Disrupted by Social Networks: A Research Program Using Simulation and Experiments)
Project context and objectives
The aim of the project was to understand discrimination in social interaction, especially in hiring decisions in labour markets. In real situations, labour markets suffer from imperfections, inefficient outcomes and a large mismatch between employer demands and worker skills. As intrinsic qualities and skills are difficult to measure, there is hardly any perfectly meritocratic case in which the best workers get the best jobs and no group suffers from being discriminated against. This mostly happens because employers cannot observe the individual qualities of employees in advance for hiring and have to rely on signals and external characteristics. When it is difficult to collect information, group reputation and prejudices are used as proxies to estimate and judge the individual abilities of category members. If there are statistical differences between the categories, there is room for "statistical discrimination". The most difficult problem arises when discrimination is not based on any statistical difference. This type of discrimination is a relevant societal problem, which undermines equal opportunities and threatens social integrity and could lead to severe social conflicts.
Project outcomes
By building a computer simulation of a labour market and running laboratory experiments with students as fictive employers, we showed that discrimination can be significant even in the lack of statistical differences between the qualities of the members of different groups. We found that a certain level of discrimination is a consequence of asymmetric information and limited supply of skilled labour. Discrimination drastically increases when employers have higher aspirations, as the effect of sampling bias is even more prominent. Our results provide reasons why we observe higher employment discrepancies in high status jobs.
Furthermore, we addressed a previously overlooked aspect of discrimination: how social networks can contribute to establish discriminative practices. We found that the use of referral networks in getting a job does not necessarily increase discrimination in employment. Hiring via social network contacts, which could either be worker referrals, friendship ties between employers and workers or business contacts, lowers discrimination rates compared to the market where employers do not have relationships. Therefore, we showed that the social embeddedness of employers can help to reduce this type of market inefficiencies.
Finally, our results showed that assertive workers tend to have high-quality jobs even if assertiveness is not correlated with qualities and employers are not concerned with it. Therefore, a quality-wage mismatch in labour markets can be an unintended consequence of standard hiring practices of profit-seeking employers.
Project implications
Although our research was primarily theoretical, as our innovative computational and experimental methods implied a certain degree of abstraction from reality, our results have interesting policy implications. First, they support the idea that anti-discrimination campaigns have been not fully successful because they have followed the idea that individuals are rational agents in perfect rational markets. On the other hand, policy-makers and market regulation authorities should invest in technologies which can help employers and market decision-makers in general in exchanging information and establishing collaborative relationships more effectively. If labour markets lack these communication channels, as often happens in reality, discrimination is likely to proliferate. Secondly, our results can stimulate employers, business practitioners and human resource managers in recognising that sampling biases can cause hiring bias and that this can have negative consequences for their business. For detail and papers, please visit the project website at: http://www.eco.unibs.it/gecs/0Discrimination.htm and/or contact the project fellow at: karoly.takacs@uni-corvinus.hu.
The aim of the project was to understand discrimination in social interaction, especially in hiring decisions in labour markets. In real situations, labour markets suffer from imperfections, inefficient outcomes and a large mismatch between employer demands and worker skills. As intrinsic qualities and skills are difficult to measure, there is hardly any perfectly meritocratic case in which the best workers get the best jobs and no group suffers from being discriminated against. This mostly happens because employers cannot observe the individual qualities of employees in advance for hiring and have to rely on signals and external characteristics. When it is difficult to collect information, group reputation and prejudices are used as proxies to estimate and judge the individual abilities of category members. If there are statistical differences between the categories, there is room for "statistical discrimination". The most difficult problem arises when discrimination is not based on any statistical difference. This type of discrimination is a relevant societal problem, which undermines equal opportunities and threatens social integrity and could lead to severe social conflicts.
Project outcomes
By building a computer simulation of a labour market and running laboratory experiments with students as fictive employers, we showed that discrimination can be significant even in the lack of statistical differences between the qualities of the members of different groups. We found that a certain level of discrimination is a consequence of asymmetric information and limited supply of skilled labour. Discrimination drastically increases when employers have higher aspirations, as the effect of sampling bias is even more prominent. Our results provide reasons why we observe higher employment discrepancies in high status jobs.
Furthermore, we addressed a previously overlooked aspect of discrimination: how social networks can contribute to establish discriminative practices. We found that the use of referral networks in getting a job does not necessarily increase discrimination in employment. Hiring via social network contacts, which could either be worker referrals, friendship ties between employers and workers or business contacts, lowers discrimination rates compared to the market where employers do not have relationships. Therefore, we showed that the social embeddedness of employers can help to reduce this type of market inefficiencies.
Finally, our results showed that assertive workers tend to have high-quality jobs even if assertiveness is not correlated with qualities and employers are not concerned with it. Therefore, a quality-wage mismatch in labour markets can be an unintended consequence of standard hiring practices of profit-seeking employers.
Project implications
Although our research was primarily theoretical, as our innovative computational and experimental methods implied a certain degree of abstraction from reality, our results have interesting policy implications. First, they support the idea that anti-discrimination campaigns have been not fully successful because they have followed the idea that individuals are rational agents in perfect rational markets. On the other hand, policy-makers and market regulation authorities should invest in technologies which can help employers and market decision-makers in general in exchanging information and establishing collaborative relationships more effectively. If labour markets lack these communication channels, as often happens in reality, discrimination is likely to proliferate. Secondly, our results can stimulate employers, business practitioners and human resource managers in recognising that sampling biases can cause hiring bias and that this can have negative consequences for their business. For detail and papers, please visit the project website at: http://www.eco.unibs.it/gecs/0Discrimination.htm and/or contact the project fellow at: karoly.takacs@uni-corvinus.hu.