Taste-based discrimination

Taste-based discrimination occurs when employers, co-workers or customers act as if dealing with members of a group carries a personal cost, measured by a discrimination coefficient. In Gary Becker’s model, employer discrimination lowers the discriminating employer’s own profit, employee discrimination produces segregation rather than a wage gap, and customer discrimination can persist under competition because catering to it pays.¹

Prejudice as a cost

An employer with discrimination coefficient $d$ acts as if hiring a Black worker paid $w_B$ costs $w_B(1 + d)$. With a wage of $18 and $d = 0.25$, the employer behaves as if the worker costs $22.50. The firm still pays $18; the extra $4.50 is psychic, and it shapes hiring.

The same device describes other sources of prejudice. A prejudiced white worker acts as if a wage $w_W$ earned alongside Black co-workers were only $w_W(1 - d)$. A prejudiced customer acts as if a price $p$ paid to a Black worker were $p(1 + d)$. Following Borjas, the model is written for Black ($B$) and white ($W$) workers, but it applies to any groups: women and men, immigrants and the Canadian-born, Indigenous and non-Indigenous workers.

Employer discrimination

Assume a competitive labour market in which Black and white workers are perfect substitutes with the same value of marginal product, $VMP_E$. Each firm takes the market wages $w_W$ and $w_B$ as given, with $w_B \leq w_W$. The owner maximizes profit net of the psychic cost.

Which workers. The owner compares the utility-adjusted cost of a Black worker with the white wage:

• if $w_B(1 + d) < w_W$, the firm hires only Black workers;
• if $w_B(1 + d) > w_W$, it hires only white workers.

Because the workers are perfect substitutes, no firm hires both: discrimination segregates the workforce across firms. The cut-off is $d^* = w_W / w_B - 1$.

How many. A firm that hires Black workers hires until $VMP_E = w_B(1 + d)$. A non-discriminating firm ($d = 0$) hires until $VMP_E = w_B$. The discriminating firm therefore hires fewer workers than the profit-maximizing level $E^*$, and the shaded triangle in Figure 1 is the profit it gives up: output that would have been worth more than its actual cost.

Discrimination coefficient, d

0.20

Cost as the employer perceives it, w_B(1 + d)

$21.60

Black workers hired

48.0

White workers hired

0.0

Profit given up per hour

$21.60

Open in Graph Lab

Figure 1. Black and white workers are equally productive and earn $18 and $24 an hour. The non-discriminating employer hires $E^* = 60$ Black workers, where $VMP_E = w_B$. An employer with $d = 0.2$ acts as if a Black worker costs $21.60, hires only 48 and gives up $21.60 an hour of profit (shaded). For $d$ above $d^* = 1/3$ the employer hires only white workers, at the higher wage.

Discrimination is costly to the discriminator. Profit falls as $d$ rises, because the firm hires too few workers. Past $d^*$ the firm switches to white workers at the higher wage, and its profit drops discontinuously: at $d^*$ the owner is indifferent in utility, but the white workforce costs more in actual dollars.

Discrimination coefficient, d

0.20

Profit

$518.40

Share of maximum profit

96%

Open in Graph Lab

Figure 2. Profit of the firm in Figure 1 as its discrimination coefficient rises: $540 an hour at $d = 0$, $480 at $d^* = 1/3$, then $240 once it hires only white workers. Every discriminating employer earns less than a non-discriminating one.

Key idea

A discriminating employer pays the market wage like everyone else. The prejudice shows up as fewer jobs for Black workers, segregation, and lower profit for the employer.

The market wage gap

The market sets $w_B / w_W$. Rank employers by $d$. Non-discriminating employers will hire Black workers whenever $w_B \leq w_W$, so the demand for Black workers is flat at a wage ratio of one up to the jobs those employers offer. Employing more Black workers requires hiring from prejudiced employers, and an employer with coefficient $d$ hires Black workers only if $w_B / w_W < 1/(1 + d)$, so beyond the flat segment demand slopes down. The wage ratio is where the supply of Black workers crosses this demand curve.

Supply of Black workers (shift)

50k

Jobs at non-discriminating employers

30k

Wage ratio, w_B / w_W

0.83

Black employment

43.3k

Wage gap

17%

Open in Graph Lab

Figure 3. Demand for Black workers is flat at a wage ratio of one for the jobs offered by non-discriminating employers, then falls as more prejudiced employers must be induced to hire. With 30,000 such jobs, supply crosses the falling segment and the wage ratio is 0.83. If supply crossed the flat segment, there would be no wage gap even though some employers discriminate.

Four predictions follow:

1. If Black workers are few relative to jobs at non-discriminating firms, there is no wage gap: Black workers simply work at those firms.
2. A larger supply of Black workers widens the gap.
3. More prejudice, a downward shift of the sloped segment, widens it.
4. The gap is set by the marginal discriminator, the least prejudiced employer still needed, not by the most prejudiced.

Does competition eliminate it?

Non-discriminating firms earn more. With free entry and constant returns to scale, they can expand or enter, hire Black workers at the lower wage, undercut discriminating firms and eventually buy them out. Becker’s prediction is that competition erodes employer discrimination in the long run. It survives where competition is weak: a firm protected by regulation or market power can spend its rents on prejudice.

Employee and customer discrimination

Employee discrimination. If white workers act as if their wage is $w_W(1 - d)$ alongside Black co-workers, an integrated firm must pay them a premium. A profit-maximizing employer avoids the premium by hiring an all-white or an all-Black workforce. With perfect substitutes and enough firms, segregation leaves no wage gap: employee discrimination produces segregated workplaces rather than unequal pay.

Customer discrimination. If customers act as if the price is $p(1 + d)$ when served by a Black worker, a Black worker in a customer-facing job brings in less revenue. Firms place Black workers in jobs without customer contact; a wage gap appears only if there are not enough such jobs. Because catering to customers’ prejudice raises profit, competition does not remove it.

Worked example

$VMP_E = 36 - 0.3E$, in dollars an hour, with $E$ workers. The market wage is $24 for white workers and $18 for Black workers. The owner’s discrimination coefficient is $d = 0.2$.

1. Utility-adjusted cost. $w_B(1 + d) = 18(1.2) = 21.60$, which is less than $w_W = 24$, so the firm hires only Black workers.

2. Employment. Hire until $VMP_E = 21.60$: $36 - 0.3E = 21.60$, so $E = 14.4/0.3 = 48$.

3. Benchmark. A non-discriminating firm hires until $VMP_E = 18$: $E^* = 18/0.3 = 60$.

4. Profit. Profit is the area under $VMP_E$ minus the wage bill. With 48 workers: $36(48) - 0.15(48)^2 - 18(48) = 1728 - 345.6 - 864 = 518.40$. With 60: $2160 - 540 - 1080 = 540$. The prejudice costs $21.60 an hour, the triangle $\tfrac{1}{2}(60 - 48)(21.60 - 18)$.

5. Cut-off. The firm hires Black workers only while $d < 24/18 - 1 = 0.333$.

Second example: the owner’s coefficient is $d = 0.4$ instead. Which workers does the firm hire, how many, and what does the prejudice cost?

Show the solution

$18(1.4) = 25.20$, more than $24, so the firm hires only white workers, until $36 - 0.3E = 24$: $E = 40$. Profit is $\tfrac{1}{2}(40)(36 - 24) = 240$, against $540 for a non-discriminating firm: the prejudice costs $300 an hour. Employment is lower than at a non-discriminating firm and the wage bill per worker is higher.

Evidence

Prejudice and wages. Using survey measures of racial prejudice across US states, Kerwin Charles and Jonathan Guryan found that Black workers’ wages relative to white workers’ are lower where the prejudice of the marginal white person is higher and where Black workers are a larger share of the population, and are unaffected by the prejudice of the most prejudiced. All three patterns are predictions of Becker’s market model.²

Competition and rents. After US banking deregulation, men’s wages in the industry fell by about 12 percent and women’s by only 3 percent, and women’s share of managerial jobs rose. Sandra Black and Philip Strahan conclude that the rents regulation had protected were shared mainly with men, and competition squeezed them out.³ Using US plant-level data, Judith Hellerstein, David Neumark and Kenneth Troske found that among plants with substantial product market power, those employing more women were more profitable, as they would be if firms with market power indulge a taste for hiring men; there was no such link among plants with little market power. They found no evidence, though, that such plants grew more slowly or were bought out, so competition did not quickly remove the discrimination.⁴

Customers. In a survey of employers in four large US metropolitan areas, Harry Holzer and Keith Ihlanfeldt found that the racial composition of an establishment’s customers has sizable effects on the race of the workers it hires, especially in jobs with direct customer contact, with a net reduction in labour demand and wages for Black workers.⁵

Common mistakes

Traps on exams

• Discriminators pay less. In Becker’s model every firm pays the market wage. Discrimination shows up as fewer jobs and segregation; the wage gap is a market outcome.
• Treating $d$ as a payment. The firm’s actual cost of a Black worker is $w_B$. That is why discrimination lowers profit.
• Employee discrimination and wage gaps. With perfect substitutes it produces segregated workplaces, not unequal pay.
• Competition removes all discrimination. It can erode employer discrimination, given free entry. Customer discrimination pays, so it survives.

On the exam

Draw $VMP_E$ with three horizontal lines, $w_W$, $w_B$ and $w_B(1 + d)$. Mark $E^*$ where $VMP_E = w_B$ and the discriminator’s employment where $VMP_E = w_B(1 + d)$, shade the lost profit, and state the hiring rule with $d^* = w_W/w_B - 1$.

Check yourself

1.

White workers earn $20 an hour and Black workers $16, and the two groups are equally productive. An employer's discrimination coefficient against Black workers is 0.22. Whom does the firm hire?

Discrimination coefficient, d

0.20

Cost as the employer perceives it, w_B(1 + d)

$21.60

Black workers hired

48.0

White workers hired

0.0

Profit given up per hour

$21.60

1Only Black workers, because $16(1.22) = 19.52$ is less than $202Only white workers, because the discrimination coefficient exceeds the 20 percent wage gap3Some of each group, until their marginal products are equal4Only white workers, because a prejudiced employer never hires Black workers

Check answer

2.

True or false: in Becker's model, a discriminating employer who hires Black workers pays them less than the market wage $w_B$.

1True2False

Check answer

3.

White workers dislike working alongside Black workers, but employers and customers are not prejudiced. The two groups are perfect substitutes. What does Becker's model predict?

1Segregated workplaces with no wage gap between the groups2A wage gap in favour of white workers at every firm3A wage gap in favour of Black workers at every firm4Integrated workplaces in which white workers are paid a premium

Check answer

Practise more questions on this topic

Sources

1. Becker, G. S. (1957). The Economics of Discrimination. Chicago: University of Chicago Press. Second edition, 1971. ↩

2. Charles, K. K. and Guryan, J. (2008). Prejudice and wages: An empirical assessment of Becker’s The Economics of Discrimination. Journal of Political Economy, 116(5), 773 to 809. NBER working paper 13661. ↩

3. Black, S. E. and Strahan, P. E. (2001). The division of spoils: Rent-sharing and discrimination in a regulated industry. American Economic Review, 91(4), 814 to 831. ↩

4. Hellerstein, J. K., Neumark, D. and Troske, K. R. (2002). Market forces and sex discrimination. Journal of Human Resources, 37(2), 353 to 380. NBER working paper 6321. ↩

5. Holzer, H. J. and Ihlanfeldt, K. R. (1998). Customer discrimination and employment outcomes for minority workers. Quarterly Journal of Economics, 113(3), 835 to 867. ↩