By Jacob Kosoff

1
Testing Theories of DiscriminationEvidence from
The Weakest LinkSouth Africa

• By Jacob Kosoff

2
Chapter Outline

• Chapter I Introduction and Literature Review
• Chapter II Background on the television show
  The Weakest Link (SA)
• Chapter III Strategic Considerations of
  Contestants
• Chapter IV Empirical Results
• Chapter V Conclusions

3
Introduction

• Motivation for Research
• In South Africa, race and discrimination is an
  important subject due to South Africas current
  and past political and economic situation.
  Determining the type of discrimination emerges as
  a fundamental question.
• Motivation to use Evidence from the Weakest Link
• The show is an unusual data source that gives
  the chance to separate two types of
  discrimination. The strategic incentives flip is
  makes this source invaluable.
• Primary goal of the Research
• To determine the significance of age, gender,
  and race in both taste-based and incomplete
  information-based discrimination in South Africa.
  
• Secondary Goal of Research
• To compare the results in South Africa to the
  results from a similar study completed in the
  United States by Professor Steven Levitt

4
Literature Review

• Discrimination
• If discrimination does exist, then understanding
  the source is essential
• Two leading theories of discrimination
• Taste-based Discrimination
• Some individuals prefer not to interact with
  members of a certain group
• This discriminators are willing to pay a
  pecuniary cost to avoid these interactions
• Becker 1957
• 2. Incomplete information-based Discrimination
• Incorrect beliefs about another groups skill
  level
• Certain groups signaling of ability is less
  revealing
• Negative beliefs become self-fulfilling
• Less or no animosity to members of that group in
  statistical discrimination
• Phelps 1972, Arrow 1973, Aigner and Cain 1977,
  Lundberg and Startz 1983, Coate and Loury 1993,
  Fryer and Jackson 2002

5
Caveat

• Limited Data Sources
• Due to the difficulty in separating types of
  discrimination, an unusual data source is used.
• Qualifications for using data from a television
  show
• Not a market
• Contestants not fully representative of society
• Voting in a very public arena
• Highly stylized nature of interactions
• Non-discriminators who believe others are
  discriminators would vote for the discriminated
  group
• Absence of observed discrimination is this
  artificial context
• Due to presence of real-world discrimination
  towards a group, has sensitized South Africans to
  the importance of not appearing outwardly racist,
  regardless of inward disposition
• Contestants then would shy away from
  discriminating on a nationally televised program

6
Chapter II Background on The Weakest Link

• X person level observations, Y person-votes.
• 65 X person-level observation
• 65 Y person-votes
• Prize money is an increasing function of the
  total number of questions correctly answered
• Examining behaviour in banking is interesting.
  This is beyond the scope of this paper.

7
Chapter III Strategic Considerations

• Complexity of the circumstances precludes a
  formal model.
• Given that players have 20 seconds to determine
  their votes, it is highly unlikely most players
  would complete rigourous optimisation. Players
  use rules of thumb and subjective criteria
• Contestants have four consideration
• Building size of prize pool
• Dominant strategy in early rounds is to vote off
  players with low probability of answering
  questions correctly.
• Eliminate strong players to win the penultimate
  head-on-head round
• Dominant strategy in later rounds is to vote off
  players with high probability of answering
  questions correctly.
• Voting similar to others
• Being the sole voter against an individual brings
  retribution to you in later rounds
• Danger of voting for a stronger link who casts
  deciding vote and has tendency to survive and
  therefore vote in later rounds
• Pressure by host to vote of the weakest link
• Host coerces players to vote for weakest link
  with forceful language a sole contestant voting
  for a stronger link is signaled out by aggressive
  host

8
Chapter IV Empirical Results

• Utilizing Regression Analysis
• Hypothesis
• Characteristics such as race, gender, and age are
  not primary determinants of voting behaviour

9
Chapter V Conclusions

• TBA