Chapter 22 Quiz: Down-Ballot and Global Forecasting

Multiple Choice

1. The primary national-level input used in House seat projection models is:

a) Presidential approval rating b) The generic ballot — the national question about congressional party preference c) Senate polling averages in the same cycle d) Incumbent candidate fundraising totals

2. The "efficiency gap" concept refers to:

a) The difference between the margin of error in House vs. Senate polling b) The resource gap between well-funded and poorly funded campaigns c) A measure of partisan advantage in legislative maps based on differential vote "wasting" d) The lag between polling data collection and publication

3. Which of the following best describes why a 4-point Democratic generic ballot lead does not reliably translate to a specific number of House seats?

a) The generic ballot is systematically biased against Democrats b) Seat outcomes depend on the geographic distribution of votes across districts, not just the national total c) House polling is more accurate than generic ballot polling d) The generic ballot only measures two-party preferences and ignores independents

4. In MRP (Multilevel Regression and Poststratification), the "poststratification" step involves:

a) Conducting additional surveys in underrepresented geographic areas b) Applying regression-based predictions to Census demographic data to produce weighted district estimates c) Correcting for partisan nonresponse bias using recalled vote choice d) Weighting the regression toward the most recently observed election outcomes

5. YouGov's MRP model accurately predicted the 2017 UK hung parliament while standard polling averages predicted a Conservative majority. The primary mechanism for MRP's accuracy was:

a) A larger sample size than other polling firms b) Better telephone sampling methodology c) Capture of a demographically concentrated swing toward Labour among young, educated voters d) Historical correction for Conservative underestimation after 2015

6. Which of the following is the most distinctive challenge in forecasting German elections compared to American Senate elections?

a) The absence of polling infrastructure in Germany b) Multi-party vote share estimation combined with coalition formation prediction c) German voters are more likely to engage in spiral-of-silence effects d) German economic voting patterns do not follow international norms

7. Brazil's 2022 polling challenges were structurally similar to American 2016/2020 polling failures primarily because:

a) Brazilian pollsters used the same methodology as American pollsters b) Bolsonaro, like Trump, encouraged his supporters not to participate in polls, producing partisan nonresponse bias c) Brazilian polling is conducted exclusively online, where Bolsonaro supporters are underrepresented d) The Brazilian electoral college produces the same geographic concentration effects as the American system

8. In a parliamentary system, why is a vote-share forecast insufficient for a complete "election forecast"?

a) Parliamentary elections use different polling methodologies than presidential elections b) Parliamentary systems have more candidates and therefore more polling error c) The ultimate object of interest — government formation — depends on post-election coalition negotiation, not just vote shares d) Parliamentary vote totals are not publicly released until weeks after Election Day

9. MRP "borrows strength" across geographic units. This means:

a) Wealthier districts contribute more information to estimates for poorer districts b) Statistical information from respondents in one geographic area improves estimates for other areas with the same demographic profile c) MRP combines multiple surveys from different geographic regions to reduce noise d) Geographic units with more polling data downweight units with less data

10. Parallel vote tabulation (PVT) is used in some international elections as an alternative to pre-election polling because:

a) It is conducted before voting begins and captures voter intent b) It samples actual vote tallies at randomly selected polling stations on Election Day, producing estimates independent of survey methodology c) It is cheaper than polling and available in all countries d) It measures economic conditions rather than stated vote preference

True/False

11. House forecasting relies primarily on individual-race polling rather than national structural signals because House districts are smaller and therefore easier to poll accurately.

12. A party that wins by very large margins in its core districts is using its votes "efficiently" by maximizing seat totals.

13. MRP can produce accurate district-level estimates from a national survey of 3,000 respondents even if only 10–15 of those respondents live in a particular district.

14. The seats-votes relationship in American House elections has a swing ratio greater than 1 near the 50-percent threshold, meaning a 1-point national swing produces more than 1 percent change in seat share.

15. State legislative races are easy to forecast because they receive extensive professional polling coverage and have stable incumbency patterns.

Short Answer

16. A forecaster claims to predict exactly 213 Democratic House seats. Explain in two sentences why this level of precision is inappropriate and what a more honest forecast statement would look like.

17. In one to two sentences, explain why "Who Gets Counted" is both a methodological and a democratic concern in international election forecasting, using a specific example from the chapter.

18. What is the key difference between a first-past-the-post system (like the UK House of Commons) and a proportional representation system (like the Dutch parliament) in terms of the vote-to-seat translation, and what does this imply for forecasting methodology?

19. Describe in one to two sentences why MRP would perform poorly in a district where local economic conditions or candidate-specific factors, independent of demographic composition, are the primary driver of the electoral outcome.

20. The chapter distinguishes between forecasting in "data-rich" and "data-scarce" environments. Name two strategies for producing useful probabilistic forecasts in a data-scarce environment where little or no polling is available.

Answer Key

  1. b
  2. c
  3. b
  4. b
  5. c
  6. b
  7. b
  8. c
  9. b
  10. b
  11. False (House forecasting relies primarily on generic ballot, historical baselines, and fundraising rather than individual-race polling, because most races receive little or no direct polling)
  12. False (very large margins in core districts is inefficient — "wasted" votes that do not produce additional seats)
  13. True (this is the core purpose of MRP: demographic disaggregation allows district-level estimates without requiring large local samples)
  14. True (historically, the seats-votes curve in the U.S. has had swing ratios of 2:1 or higher near 50%, though geographic sorting has reduced this in recent cycles)
  15. False (state legislative forecasting is among the most challenging because individual-race polling is essentially nonexistent and forecasters must rely almost entirely on structural signals)
  16. Projecting exactly 213 seats implies a level of precision that no model can support — House seat totals are subject to substantial uncertainty from generic ballot variation, individual-race polling error, and turnout dynamics. A more honest statement would be: "Democrats are projected to win 200–225 seats, with a median of 213."
  17. Accept answers that explain the connection between survey underrepresentation (sampling bias) and democratic representation — e.g., Brazilian rural voters underrepresented in telephone polling are the voters whose preferences most needed to be measured for an accurate forecast; their systematic exclusion from the measurement was simultaneously a methodological failure and a democratic blind spot.
  18. In FPTP, small changes in vote share can flip dozens of constituency seats simultaneously because plurality rules mean second place gets nothing; in PR, each percentage of vote translates approximately linearly to seats. For forecasting: FPTP requires district-level geographic modeling to capture geographic concentration effects; PR seat forecasting can be done primarily from national vote share estimates.
  19. MRP predicts district preferences from demographic composition; where local economic conditions or candidate factors produce outcomes independent of demographics, the model has no information about these factors and will miss them — the error cannot be corrected by having more survey respondents or better poststratification weights.
  20. Accept any two of: expert elicitation (structured probabilistic assessment from country specialists); historical analog matching (identifying past elections with similar structural conditions); cross-national parameter import (using economic voting coefficients from comparable countries); social media and alternative data signals; scenario analysis with explicit uncertainty bounds.