Blog
Discrimination – the limitations of implicit association tests and the impact of the job market

The Center for American Progress estimates the costs of discrimination at US$64 billion per year, or roughly 2 million annually displaced American workers (Burns, 2012). Discrimination is clearly costly. It is also, almost universally, a unique and puzzling issue. This blog looks at two recent WIDER Working Papers which take on this issue. The first considers whether the increasingly popular ‘implicit association test’ actually tells us anything about how participants are likely to act in a simple giving game. The second looks at the effect labour market conditions have on the prevalence of discrimination.

The implicit association test

The first paper, by Daniel J. Lee, speaks to a recent trend in the social sciences — the claim that discriminatory attitudes stem from implicit biases and associations. The development of the implicit association test (IAT henceforth) has lent support to these claims by introducing a measure of these implicit biases without having to rely on self-reporting, which is known to be unreliable.

The IAT is essentially a series of timed sorting tasks. Subjects match features, such as faces, to highly and lowly associated attributes, such as ‘positive’ or ‘negative’ words. The underlying idea that it is easier to sort a feature with its more closely associated attributes. For instance, a picture of a chair is more closely associated with the word ‘furniture’ than the word ‘food’, and hence likely to be sorted faster. Thus, implicit biases are revealed through the difference in timing that one takes to sort different pairs of concepts.

Lee suggests that there is some common sense validation to this argument. Frequently cited examples of these biases in decision-making are men being more associated with management, or white faces being more associated with pleasant words and feelings. However, as Lee points out, to act on these biases in an IAT is costless. The question then becomes: do the biases identified by IAT predict participant actions in non-costless activities that are more analogous to real-world scenarios?

Race implicit bias does not predict giving

In order to answer this question Lee uses a dictator game where acts of giving are both non-strategic and non-spontaneous, and therefore easily controlled. Do the acts of giving differ if the participant is asked to give to those they have shown an implicit bias against in an IAT?

Lee finds that, contrary to previous literature, implicit bias fails to predict not only the amounts shared in the dictator game, it also does not predict examples of zero sharing or the choice to exit a giving environment altogether. Thus, despite the IAT’s ever-growing popularity, it fails to predict simple economic behaviours.

Lee’s paper is the first to explore the implications of an IAT in an economics experiment. As such, the analysis in this paper represents a necessary step forward in this line of research that aims to develop methods to detect discriminatory attitudes.

Discrimination and the market environment

The market environment in which discriminatory firms operate may be a relevant determinant of their extent of discrimination. In a recent paper, Eva O. Arceo-Gómez and Raymundo M. Campos-Vázquez aim at analysing the effect of local labour market conditions on a firm’s decision to discriminate in early stages of hiring.

The authors use a direct measure of discrimination, relying on using online job advertisements posted by employers wherein they list characteristics such as gender, age, marital status, even physique, to describe their ideal candidates. Do such discriminatory requirements lessen when employment is high, and increase during a slack labour market?

Discrimination increases when unemployment is high

The authors find evidence that firms explicitly discriminate more on the basis of gender when the unemployment rate is higher: a percentage point increase in the unemployment rate is correlated with a 0.7 percentage point increase in the probability that an ad is targeted. In slack labour markets, firms tend to target their ads to men more often than in tight labor markets.

In such labour markets, the authors argue, the fall in the opportunity cost of waiting to fill the vacancy (opportunity cost of engaging in discrimination) outweighs the effect that the lower absolute number of discriminating positions has on strengthening the bargaining power of the dispreferred candidates during high unemployment spells.

The authors also tested whether other types of discrimination respond to unemployment rates. Their findings indicate that beauty and physique targeting decrease as the unemployment rate goes up. In this case, the job destruction effect dominates the opportunity cost effect.

A limitation to the authors work is the fact that, even though overt discrimination is a direct way to measure the phenomenon, this type of discrimination is only present in a job ad — that is, the beginning of the hiring process. As such, they cannot make any claims of the response of discrimination to unemployment in actual hiring, wages, promotions and so on.

Discrimination clearly remains a problem, and these two papers demonstrate that its causes may not always be a simple as we think. It cannot be explained by implicit biases alone, at least not as measured by IATs, but it is significantly influenced by variations in the labour market. These complexities are what make studying discrimination so important, and why those with an interest in the matter should follow the output of UNU-WIDER’s Discrimination and affirmative action project.

 

The views expressed in this paper are those of the author(s), and do not necessarily reflect the views of the Institute or the United Nations University, nor the programme/project donors.

Follow us on twitter @UNUWIDER. Follow James Stewart on twitter @deveconwatcher.

Previous
How electoral cycles affect school systems – evidence from India
How electoral cycles affect school systems – evidence from India
Next
Domestic revenue mobilization and the role of data
Government revenues are central to funding public expenditures in all countries...
Domestic revenue mobilization and the role of data