What Works: Gender Equality by Design - Comprehensive Notes

The Power and Promise of Behavioral Design

• The Orchestra Case Study: Blind Auditions

  • Context: In 1970, women made up only $5\,\%$ of musicians in the top five U.S. orchestras.
  • The Intervention: Orchestras (starting with the Boston Symphony Orchestra) introduced blind auditions where musicians performed behind a screen (curtain).
  • Result: Using a screen during preliminary rounds raised the probability that a female musician would advance by $50\,\%$ and significantly increased the actual hiring of women.
  • Timing: Despite high artistic standards, it took until 1997 for the Vienna Philharmonic to admit its first female player.
  • Insight: Behavioral design decisions (the curtain) doubled the talent pool by removing visual bias, as directors previously believed sound was their only criteria, while research shows they were heavily influenced by visual cues.

• Cognitive Illusions and Visual Cues

  • The Checkershadow Illusion: Most people perceive square B as lighter than square A due to its position in a checkerboard pattern and shadow. In reality, both are the same color. By isolating square B, the mind is liberated from the category-based rules of the pattern.
  • Connection to Bias: Just as the curtain liberated orchestra committees from visual patterns, behavioral design helps minds see candidates for their objective skill rather than their category (gender, race).

• The Timing of Decisions: Israeli Parole Study

  • Finding: Israeli judges were found to rule more leniently immediately following meal breaks.
  • The Mechanism: Hunger, fatigue, and the depletion of cognitive resources led judges to revert to the "easy solution" (the status quo of denying parole) right before breaks.
  • Design Implication: The design of schedules (timing and frequency of breaks) has unintentional consequences on fairness and justice.

• Defining Behavioral Design (Choice Architecture)

  • Origin: Referred to as "choice architecture" by Richard Thaler and Cass Sunstein in the book Nudge.
  • Purpose: To create environments that help people achieve their goals without defining those goals or relying solely on laws/incentives.
  • Example from Denmark: Research on $41 \times 10^6$ observations showed tax subsidies had a tiny impact on savings because $85\,\%$ of people failed to act. In contrast, automatic employer contributions to retirement accounts significantly increased savings.

The Economic and Moral Case for Equality

• Aggregate Economic Gains in the United States ($1960-2008$)

  • Productivity: Improved allocation of talent accounted for a $15\,\%$ to $20\,\%$ growth in aggregate output per worker.
  • Medicine and Law: In 1960, $94\,\%$ of U.S. doctors and lawyers were white men. By 2008, that figure dropped to $62\,\%$. Casting a wider net to include women and minorities paid off economically.

• Global Labor Force Implications

  • Japan: Given a current labor force participation of $63\,\%$ for women and $84\,\%$ for men, Japan's labor force will shrink by over $10\,\%$ in twenty years without change. Achieving gender parity would increase Japan's GDP by almost $20\,\%$ over that period.
  • Low-Fertility Countries: Similar benefits are projected for Germany, Italy, Singapore, South Korea, and Spain.
  • Global Loss Simulation: Excluding women from the labor force entirely would result in per capita income losses of nearly $40\,\%$. Regional losses are highest in the Middle East and North Africa at $27\,\%$.

• Education Gap Reversal

  • In the United States, women have held more than half of all bachelor's degrees since the mid-1980s. By the early 21st century, the number reached almost $60\,\%$.

• Micro-level Productivity: Ivory Coast and Collective Intelligence

  • Crop Spending: In Ivory Coast, when "male" crops (coffee, cocoa, pineapple) have high yields, spending on alcohol and tobacco increases. When "female" crops (plantains, bananas, coconuts, vegetables) yield well, households spend more on food.
  • Collective Intelligence: Laboratory studies on group intelligence show that gender-diverse teams score higher than all-male or all-female teams. This group intelligence is only moderately related to individual member intelligence; diversity makes the team more than the sum of its parts.

• The Moral Case and Gendercide

  • Missing Females: The UN estimates up to $200 \times 10^6$ women and girls are "missing" worldwide due to sex-selective abortion, infanticide, and neglect.
  • China Statistics: By 2020, Chinese Academy of Social Sciences calculated a surplus of $30\,\text{--}\,40 \times 10^6$ young men without marriage prospects. Low female-to-male ratios correlate with increased trafficking, domestic violence, and honor killings.

Behavioral Insights in Global Settings

• Rural India and Call Centers (Rob Jensen Study)

  • The Treatment: Provided recruitment services to women in random rural villages for three years.
  • Result: Even a small increase in employment ($2.4$ percentage points) led to five-to-fifteen-year-old girls having better health and higher school attendance.
  • Shift in Beliefs: Seeing women work in counter-stereotypical roles (call centers) allowed parents to imagine new futures for daughters, proving that the game of equality is "positive sum."

• Counter-stereotypical Challenges: Male Teachers

  • Bias Example: Iris Bohnet identifies her own biased snap judgment (sexism) when seeing a male teacher at a Harvard day-care center.
  • OECD Data: At age 15, boys are $50\,\%$ more likely than girls to lack basic proficiency in math, science, and reading. Male role models (only $10\text{--}20\,\%$ of elementary teachers are male) are vital for boys' self-belief.

Identifying the Problem: Unconscious Bias

• The Heidi vs. Howard Roizen Study

  • The Original Case: Written by Kathleen McGinn in 2000 regarding successful entrepreneur Heidi Roizen.
  • The Experiment: Students read a case study of a venture capitalist. Half get the name "Heidi," half get "Howard." All other facts are identical.
  • Result: Both were rated as highly competent. However, students liked Howard but disliked Heidi. Howard was seen as a visionary; Heidi was seen as self-promotional and arrogant.
  • The Trade-off: Women face a "double bind." If they conform to feminine stereotypes of nurture, they are liked but not respected. If they show agency/competence in male domains, they are respected but disliked for violating norms.

• Intersectional Bias and Double Jeopardy

  • Black Women: Robert Livingston found that Black women did not experience the same backlash for dominance as white women, possibly because they aren't seen as "prototypical women."
  • Black Men: Black male CEOs benefit from physical features expressing warmth/deference, whereas these hurt white male CEOs.

• Biased Evaluation Patterns

  • Observable Performance: Successful women are rated as less likable than men.
  • Ambiguous Performance: Successful women are rated as less competent than men.

• Field Experiments in Discrimination

  • Phantom Students: Professors were sent emails from fake students. White men received an $87\,\%$ response rate in business admin, while women and minorities combined received only $62\,\%$.
  • STEM Hiring: Even when candidates provided arithmetic performance data, evaluators favored men $2:1$ unless forced to look at specific past-round scores.
  • Gender Hierarchy Threat: In the U.S. military, male officers gave lower scores to female subordinates whose pay grades were close to their own (violating gender norms).

Cognitive Mechanisms of Bias

• Survivorship Bias (Abraham Wald)

  • Mathematician Abraham Wald realized scientists should not study where returning planes were hit, but rather the empty spaces (where the planes that didn't return were likely hit).
  • Leadership Application: We lack data on successful women because they rarely get the chance to fail or prove beliefs wrong.

• Statistical Discrimination

  • Riddle: The surgeon who says "this boy is my son" is the mother ($1/3$ of surgeons in the U.S. are female).
  • Economics Definition: Basing assessments of individuals on group averages.
  • Car Negotiation Study: Sellers demanded higher initial prices from African Americans and women because of a statistical belief they are less informed. Once the initial price is set, the gap is almost impossible to close.

• Heuristics and Modes of Thought (Daniel Kahneman)

  • System 1: Fast, automatic, intuitive, seeks consistency/confirmation (WYSIATI: "What You See Is All There Is").
  • System 2: Slow, deliberative, effortful, logical.
  • Representativeness Heuristic: Florida residents are stereotyped as elderly. In reality, $82\,\%$ were under 65 in 2013, compared to $86\,\%$ in the U.S. overall. Small differences in relative frequency create large mental archetypes.
  • Continuum Model (Susan Fiske): We categorize people by sex, race, and age first, then maintain that category through "confirmatory categorization."

• Measuring Bias: The Stroop Test and IAT

  • Stroop Test: Naming the color of words (e.g., the word WHITE printed in GREEN ink) slows down reaction time as System 2 overrides the automatic reading of System 1.
  • Implicit Association Test (IAT): Developed by Anthony Greenwald in 1994. Measures the speed of associations (e.g., Career/Male vs. Career/Female).
  • Neurological Insight: Eric Kandel guesses $80\text{--}90\,\%$ of the mind works unconsciously.

Design Solutions and Failed Interventions

• The Failure of Diversity Training

  • Self-Serving Bias Study: Plaintiffs vs. Defendants interpreting the same 15-page legal case. Plaintiffs estimated a jury award of $\$260,000$ while defendants estimated $\$180,000$. Simple awareness of bias does not eliminate it.
  • Statistical Failure: Analysis of $800$ U.S. companies by Frank Dobbin ($1971\text{--}2002$) showed diversity training had no relationship to workforce diversity.
  • Moral Licensing: People who feel they have done something good (like taking a multivitamin or endorsing a diverse candidate) feel authorized to do something bad (smoke or discriminate later).
  • Rebound/Suppression: Instructing people to suppress stereotypes often makes those stereotypes more salient and intrusive.

• Effective Strategies

  • Consider-the-Opposite: Forcing legal teams to write down the weaknesses of their own case reduced impasse rates from $35\,\%$ to $4\,\%$.
  • The Crowd-Within: Averaging three different guesses made by the same person at different deliberative intervals improves accuracy.
  • Norms (Rwanda Study): A radio soap opera in Rwanda did not change personal beliefs but successfully changed behavior by altering social norms around cooperation.
  • Unfreeze-Change-Refreeze (Kurt Lewin):
    1. Unfreeze: Use an IAT to make people realize their own biases.
    2. Change: Offer specific tools (comparative evaluation, blinding).
    3. Refreeze: Institutionalize the change (e.g., hotel cards that automatically turn off lights).

• Checklist for Designing Gender Equality

  • Stop diversity training focused only on awareness.
  • Follow the unfreeze-change-refreeze framework.
  • Train people in reasoned judgment (consider-the-opposite, crowd-within).
  • Use People Analytics (data-driven HR at Google, Credit Suisse, etc.) to replace intuition.