Discrimination

class: center, middle, inverse, title-slide

# Discrimination
## EC 350: Labor Economics
### <a href="https://kyleraze.com">Kyle Raze</a>
### Winter 2022

---

# Bertrand and Mullainathan (2004)

## **Discussion**

**Q.sub[1]:** How does the study measure discrimination in the labor market?

**Q.sub[2]:** What are the strengths of the research design?

**Q.sub[3]:** What are the weaknesses of the research design?

**Q.sub[4]:** What are the main findings?

**Q.sub[5]:** What does the study tell us about employers?

**Q.sub[6]:** What did *you* find most interesting and/or depressing?

---
# Discrimination

## **Theory**

Economics has a lot to say about discrimination in the labor market and other settings..super[.hi-pink[1]]

While they do not explain all forms of discrimination,.super[.hi-pink[2]] the two most prominent economic models of discrimination are

1. **Taste-based discrimination:** Prejudiced employers willingly sacrifice resources to avoid contact with workers from certain groups.

2. **Statistical discrimination:** Unprejudiced employers use group characteristics to make inferences about an individual worker's productivity.

.footnote[.super[.hi-pink[1]] Kevin Lang and Ariella Kahn-Lang Spitzer (2020), [Race Discrimination: An Economic Perspective](https://pubs.aeaweb.org/doi/pdf/10.1257/jep.34.2.68), *Journal of Economic Perspectives*. .super[.hi-pink[2]] Mario L. Small and Devah Pager (2020), [Sociological Perspectives on Racial Discrimination](https://pubs.aeaweb.org/doi/pdfplus/10.1257/jep.34.2.49), *Journal of Economic Perspectives*]

---
# Taste-based discrimination

Models of **taste-based discrimination**.super[.hi-pink[&#8224;]] posit that **prejudice** (or **animus**) causes discrimination in the labor market. 
- **The premise?** Some economic agents would **willingly sacrifice resources to avoid contact** with certain groups of people.

.footnote[.super[.hi-pink[&#8224;]] Developed by Gary Becker in [The Economics of Discrimination](https://alliance-primo.hosted.exlibrisgroup.com/permalink/f/to8ro2/CP71244526180001451), *University of Chicago Press* (1957).]

### **Setup**

Two groups of equally productive workers: 
1. **In-group** workers (*e.g.,* White workers) who receive the wage `$w_\text{W}$`.
2. **Out-group** workers (*e.g.,* Black workers) who receive the wage `$w_\text{B}$`.

---
# Taste-based discrimination

### **Setup**

A **discrimination coefficient** `$d$` captures the **disutility of out-group contact** for three types of prejudiced economic agents:
1. **Employers** who perceive hiring out-group workers as `$d \times 100$`-percent more costly than `$w_\text{B}$`.
2. **Co-workers** who perceive their wage as `$d \times 100$`-percent lower when working with the out-group.
3. **Customers** who perceive prices as `$d \times 100$`-percent when buying from an out-group seller.

---
# Taste-based discrimination

## **Employer discrimination**

Since both groups of workers are **equally productive**, they are **perfect substitutes**.

- The level of output simply depends on the number of workers: `$q = f(E_\text{W} + E_\text{B})$`.
- A firm with 25 in-group workers and 25 out-groups **produces the same output** as a firm with 50 in-group workers or a firm with 50 out-group workers.
- `$\text{MP}_E$` does not depend on in-group/out-group status!

**Q:** How would a non-discriminatory employer maximize profit?

- **A:** By hiring from the cheaper group of workers until `$w = \text{VMP}_E$`.
    - If `$w_\text{W} > w_\text{B}$`, then hire `$E^*_\text{B}$` out-group workers such that `$w_\text{B} = \text{VMP}_E$`.
    - If `$w_\text{B} > w_\text{W}$`, then hire `$E^*_\text{W}$` in-group workers such that `$w_\text{W} = \text{VMP}_E$`.

Going forward, we will assume that `$w_\text{W} > w_\text{B}$`.

---
# Taste-based discrimination

.pull-left[
<img src="17-Discrimination_files/figure-html/unnamed-chunk-1-1.svg" style="display: block; margin: auto;" />
]
.pull-right[
## **Employer discrimination**

Non-discriminatory employers simply **hire out-group workers**.

- **Why?** Both groups of workers are equally productive (*i.e.,* same `$\text{VMP}_E$`), but out-group labor is cheaper (*i.e.,* `$w_\text{B} < w_\text{W}$`).

]

---
# Taste-based discrimination

## **Employer discrimination**

Discriminatory employers perceive the cost of employing an out-group worker as `$w_\text{B}(1 + d)$`.

- If `$w_\text{B}=10$` and `$d=0.1$`, then the employer will act as though the out-group worker costs `$10(1 + 0.1) = 11$`.
- The "utility-adjusted" cost of hiring an out-group worker exceeds the actual cost!

**The result?** Segregation! A discriminatory employer will

- Hire only in-group workers if `$w_\text{B}(1 + d) > w_\text{W}$`
- Hire only out-group workers if `$w_\text{B}(1 + d) < w_\text{W}$`

---
# Taste-based discrimination

.pull-left[
<img src="17-Discrimination_files/figure-html/unnamed-chunk-2-1.svg" style="display: block; margin: auto;" />
]
.pull-right[
## **Employer discrimination**

**Case 1:** Employer hires only in-group workers.

- The employer **overpays for labor!**
    - `$w_\text{W} > w_\text{B}$`, but `$\text{VMP}_E$` is the same for both groups of workers.
- Because in-group labor is relatively expensive, the employer **hires too few workers!**.
]

---
# Taste-based discrimination

.pull-left[
<img src="17-Discrimination_files/figure-html/unnamed-chunk-3-1.svg" style="display: block; margin: auto;" />
]
.pull-right[
## **Employer discrimination**

**Case 2:** Employer hires only out-group workers.

- The employer **hires too few workers!** 
 - Actual hiring `$E^*_\text{B}$` occurs where `$w_\text{B}(1+d)= \text{VMP}_E$`, even though the actual wage is `$w_\text{B}$`.
- The employer **pays the out-group worker too little!**
 - The marginal worker receives less than her contribution, as `$w_\text{B} < \text{VMP}_E = w_\text{B}(1+d)$`.
]

---
# Taste-based discrimination

## **Employer discrimination**

**Q:** Is discrimination profitable?

**A:** No!

- To indulge their distaste for out-group workers, **prejudiced employers sacrifice profit!**
- By hiring too few workers, discriminatory employers fail to operate efficiently!

**The implication?** In a perfectly competitive market, non-discriminating employers will eventually drive discriminating employers out of business.

- In-group and out-group wages will eventually equalize.

**Q.sub[1]:** Are wage differentials actually decreasing? **Q.sub[2]:** Are markets actually perfectly competitive?

---
# Taste-based discrimination

## **Co-worker discrimination**

Suppose instead that employers are unprejudiced, but in-group workers dislike working with out-group workers.

- In-group workers receive `$w_\text{W}$`, but act as though they're paid `$w_\text{W}(1-d)$`.

To offset the disutility of working with out-group workers, the employer would have to pay in-group workers an additional `$w_\text{W} \times d$` dollars.

- The total wage paid for an in-group worker would rise to `$w_\text{W}(1+d)$`.
- If the marginal productivity of in-group and out-group workers is the same, and there are no discriminatory employers, then `$w_\text{W} = w_\text{B} < w_\text{W}(1+d)$`.

**The result?** Segregation that persists *even with perfect competition*, but no wage differential.

---
# Taste-based discrimination

## **Customer discrimination**

Now suppose that workers and employers are unprejudiced, but customers dislike buying from the out-group.

- A prejudiced customer faces the actual market price `$p$`, but feels as though they are paying `$p(1+d)$`.

An employer with out-group workers would have to reduce `$p$` to compensate prejudiced customers.

- This assumes that the employer is unable to reallocate out-group workers away from customer-facing roles within the firm.

**The result?** The employer decreases wages for out-group workers, creating a wage differential that persists *even with perfect competition*.

---
# Statistical discrimination

Models of **statistical discrimination**.super[.hi-pink[&#8224;]] posit that discrimination arises from employer **uncertainty** about difficult-to-observe productive attributes of workers.

.footnote[.super[.hi-pink[&#8224;]] Developed by Edmund Phelps in [The Statistical Theory of Racism and Sexism](https://www.jstor.org/stable/1806107), *The American Economic Review* (1972), and Kenneth Arrow in "The Theory of Discrimination" in Orley Ashenfelter and Albert Rees, eds., [Discrimination in Labor Markets](https://alliance-primo.hosted.exlibrisgroup.com/permalink/f/to8ro2/CP71344164760001451), *Princeton University Press* (1973).]

**The premise?** Employers use a worker's race or gender to make inferences about the worker's productivity.

- In these models, **employers are unprejudiced**&mdash;they do not have a taste for discrimination.
- **The basis for discrimination?** Group differences in past performance (*e.g.,* achievement gaps).
    - When considering job applicants with the same observable productive traits, but different group characteristics (*e.g.,* race), employers will often rely on the **past performance of groups** to predict **difficult-to-observe productive traits of individuals**. 
    - Discrimination occurs when the employer systematically favors applicants from higher-productivity groups.

---
# Statistical discrimination

Employers can test workers for productive traits, but it is unlikely that a test can perfectly predict productivity.

To set a worker's wage, an employer uses a weighted average of the worker's test score and the average score of the group to which the worker belongs: `$$w = \alpha T + (1-\alpha)\overline{T}$$`

- `$T$` is the individual's test score.
- `$\overline{T}$` is the group average.
- `$0 \leq \alpha \leq 1$` represents how well the test measures productivity.
    - If `$\alpha = 1$`, then the test provides a perfect measure of individual productivity and `$w=T$`.
    - If `$\alpha = 0$`, then the test provides no meaningful measure of individual productivity and `$w=\overline{T}$`.

---
class: inverse, middle

# Ban the box

---
class: clear-slide

**Q:** Why do employers ask about criminal history?

---
# Ban the box

.more-left[
.center[**States with ban-the-box laws (2021)**]
<img src="17-Discrimination_files/figure-html/unnamed-chunk-5-1.svg" style="display: block; margin: auto;" />
.footnote[*Source:* National Employment Law Project]
]
.less-right[
**The policy?** Forbid employers from asking about criminal history on job applications.

**The objective?** Expand employment opportunities for people with criminal records and **reduce racial disparities** in hiring.

]
 
---
# Ban the box

**Q:** How might an employer respond to a ban-the-box law?

- Take a chance on candidates with criminal records?
- Avoid candidates who are more likely to have a prior conviction?

**Intended consequence:** Applicants with prior convictions get their foot in the door, increasing their odds of getting hired.

- Other things being equal, this would increase the probability of employment for young men of color.

**Unintended consequence:** Employers could respond by using race as a proxy for criminal history.

- Other things being equal, this would decrease the probability of employment for young men of color *without criminal records*.
--

- **Statistical discrimination!**

---
# Agan and Starr (2017)

## **Discussion**

**Q.sub[1]:** What is the research question?

**Q.sub[2]:** How does the study address the research question?

**Q.sub[3]:** What are the main findings?

**Q.sub[4]:** How does the study advance our understanding of racial discrimination in the labor market?

**Q.sub[5]:** What are the policy implications?

**Q.sub[6]:** What did *you* find most interesting?

---
# Housekeeping

**Problem Set 4** is due by **Friday, March 11th at 11:59pm**.

**Final Exam** is scheduled for **Tuesday, March 15th at 2:45pm**.

- In-person!