Chapter 39 Quiz: Race, Representation, and Data Justice
Multiple Choice
1. The "differential undercount" in the Census refers to:
a) The fact that the Census is conducted every ten years rather than annually b) The systematic pattern by which some racial and ethnic communities are counted at lower rates than others c) The statistical adjustment applied to Census data to correct for respondent error d) Differences in how the Census counts people in urban versus rural areas
2. Which of the following communities had the highest estimated net undercount rate in the 2020 Census Post-Enumeration Survey?
a) Non-Hispanic white population b) Asian American and Pacific Islander population c) American Indian and Alaska Native people on reservations d) Black population in suburban counties
3. The "thin cell" problem in polling refers to:
a) When survey interviews are conducted with too many respondents in the same geographic area b) When demographic subgroup sample sizes are small enough that a few unusual respondents can significantly distort weighted estimates c) When mobile phone penetration in a population is insufficient to support a cell-phone-only sample d) When survey questionnaires are too long to fit on a single page
4. Ruha Benjamin's concept of the "New Jim Code" argues that:
a) Racial bias in algorithms is always the result of explicit racist intent by programmers b) Computer code cannot reproduce racial hierarchy because it is race-neutral in syntax c) Algorithmic systems can reproduce racial hierarchy without explicit racist instructions, by training on data generated by a racially structured society d) The solution to algorithmic bias is to remove all race-related variables from predictive models
5. Which Supreme Court decision significantly weakened the Voting Rights Act's preclearance mechanism, affecting its applicability to algorithmic redistricting?
a) Bush v. Gore (2000) b) Crawford v. Marion County Election Board (2008) c) Shelby County v. Holder (2013) d) Department of Commerce v. New York (2019)
6. "Algorithmic redlining" in political targeting refers to:
a) The use of red lines on digital maps to define campaign targeting zones b) Systematic patterns in which modeled persuasion scores are lower for voters in majority-minority areas than for voters with similar individual characteristics in majority-white areas c) The practice of drawing congressional district lines using algorithmic rather than manual methods d) FEC regulations limiting the use of race in campaign targeting
7. Joy Buolamwini's Gender Shades research is relevant to political analytics primarily because it demonstrates:
a) That gender is more important than race in predicting political behavior b) The principle that measurement systems can perform significantly worse for populations underrepresented in their training data c) That facial recognition technology should never be used in political contexts d) That automated systems are more accurate than human interviewers for measuring political opinion
8. "Representational harm," as developed by Safiya Umoja Noble, refers to:
a) Harm caused only by physically preventing someone from voting b) Damage from being counted inaccurately in official statistics c) Harm from being represented in ways that reinforce stereotypes, diminish political standing, or limit political voice d) Legal liability arising from misrepresenting survey results in campaign communications
9. ODA's three-question data justice framework asks: whose data and consent; who benefits; and:
a) What is the sample size b) Which statistical method is most appropriate c) How are accuracy limitations for minority communities being handled and disclosed d) What is the projected return on investment for the analytical project
10. The "surveillance asymmetry" as described in the chapter refers to:
a) The gap between what campaigns know about opponents and what they know about their own voters b) Detailed data about minority communities in formats serving strategic control, combined with limited data in formats that would serve political responsiveness c) The difference in data collection practices between urban and rural campaign operations d) Asymmetric access to satellite data for ground-truth validation of voter file addresses
Short Answer
11. Explain in 2-3 sentences how the differential undercount in the Census can affect federal spending distributions to minority communities — specifically identifying the mechanism through which a population undercount translates to reduced funding.
12. Describe one "affirmative data practice" that a polling firm could implement and explain, in 2-3 sentences, how it specifically addresses a racial equity concern documented in Chapter 39.
13. What is the distinction between using race directly as a targeting variable and using racial "proxy" variables in a campaign targeting model? Why does this distinction matter for both legal and ethical analysis? (3-4 sentences)
True/False with Justification
For each statement, indicate True or False and provide a one-sentence justification.
14. Demographic weighting in survey research completely resolves the differential response rate problem by adjusting sample composition to match population proportions.
15. The Voting Rights Act has been definitively interpreted by courts to cover private campaign targeting algorithms that produce racially disparate effects on turnout.
16. Adaeze Nwosu's testimony in 2019 was about the potential undercount consequences of adding a citizenship question to the 2020 Census.
17. A targeting model that does not explicitly include race as a variable cannot produce racially biased targeting outcomes.
18. Multilingual survey fielding is described in Chapter 39 as an affirmative data practice — a methodological investment that actively advances equity rather than simply avoiding discrimination.
Applied Analysis
19. A campaign analyst argues: "We don't have a racial equity problem in our targeting because our model doesn't use race as a variable. We use purchasing behavior, media consumption, and geographic location — all race-neutral predictors." Using concepts from Chapter 39, identify at least three specific ways this argument fails. Be specific about mechanisms, not just general claims about bias. (200–300 words)
20. Adaeze tells Sam Harding: "Our clean methodology doesn't fix the voter file. It doesn't fix the Census. It doesn't fix the algorithmic bias in commercial targeting models." Sam pushes back: "So why bother?" Write Adaeze's response, in 150–200 words, defending the value of single-organization affirmative data practices even when systemic problems remain unresolved. Then write a 100-word reply from Sam that concedes Adaeze's strongest points while maintaining that organization-level action is insufficient. Your response should reflect the arguments actually developed in Chapter 39.