How Would the US Vote for a Candy?

By Max Candocia

|

October 28, 2020

If you would like to respond to the survey for a future update of this project, click here to open a survey form. This will help tremendously with the US map visualizations.

In the United States, there are several different ways that people vote:

1. Popular (plurality) vote for most local, statewide, and federal elections
2. Ranked-choice voting is used by Maine for senate, house, and presidential elections. It is also used in limited capacities elsewhere in the US, mostly at a local level

There are also some different ways that political representation is given at the federal level:

1. Each state gets exactly 2 senators, regardless of size
2. Each state gets Electoral College votes equal to the total number of senators and representatives for that state (and 3 for Washington DC, which doesn't have any senators and only has one, limited member of the House)
3. Districts within each state choose House representatives and, in a few cases, Electoral College electors

Almost everyone in the US is probably stressed right now about politics, so why not apply this math to something more fun:

What Candy would the United States Choose in an Election?

What candy would the US choose based on on different types of elections? I asked 545 different individuals in a survey across the United States what their favorite candies were in 2 different ways given a list of over 40 candies:

1. For each candy, indicate if you like, dislike, or have no opinion of the candy
2. Rank your top 5 candies

With these two questions combined, in addition to the state and demographic information of each response, wecan look at how candies would be chosen given different voting methods and different types of representatio (popular, electoral college, and senate-based)

Voting Algorithms

• Plurality - only one vote per person. Winner is the one with the most votes
• Ranked-choice instant runoff - Individuals can rank their choices. Winner is determined as follows:
  1. The number of first-place votes for each candy (candy-date) is counted.
  2. The candy/candies with the lowest number of votes are eliminated
  3. Any votes that were lower-ranked than an eliminated candy on a ballot are shifted up
  4. Repeat the above process until the appropriate number of winners (usually just 1) remain
• Borda count - Individuals rank their choices. Each ranking adds a certain number of points. For this example, 1st place=5 points, 2nd 4, 3rd 3, 4th 2, and 5th 1.
• Approval voting - Individuals choose any number of candies they like. The candy with the most likes wins.
• Disapproval voting - Individuals choose any number of candies they dislike. The candy with the last dislikes wins
• Approval-disapproval hybrid - Individuals choose any number of candies they like and/or dislike. The largest difference between the likes and dislikes is the winner. I do not believe this is a very common method of voting.

Tap the map to cycle through results by voting method:

Statewide Summary

As you can see from the above map, Reese's Peanut Butter Cups are the usual favorite among most in the united states when using plurality or instant runoff voting. There are a lot of ties due to the large number of candies listed, large number of states, and limited sample size.

The borda count doesn't suffer from any ties in this case, as it creates a lot more variation, even with a small sample size.

There are also a lot of ties for the approval and disapproval voting rankings. It is interesting, though, to see that Reese's Peanut Butter Cups are not usually the most popular in each state: Kit-Kat Bars and M&Ms are.

Overall Results

While the above is a bit messy because of a lot of ties, the overall results are more robust and tell an obvious picture.

National Results

The national-level results are likely the most accurate, as ties are not issues, and weighting can be done to achieve a proper estimate. Tap the image below to cycle through different methods. The instant runoff vote is animated to show the final 6 steps of the process, since votes are reallocated at each step.

Electoral College Results

These results resemble the national results, since there is still a high correlation between electoral votes and the size of a state. Votes are currently divided evenly in the event of a tie. Tap the image below to cycle through different methods.

Senatorial Results

These results are a lot more varied than the national results, since smaller states influence the results the same as larger states. Votes are currently divided evenly in the event of a tie. Tap the image below to cycle through different methods.

Sampling Methodology

Sample Size: 545 responses were collected, with at least 5 from each state.

Time: Responses were taking during the month of October 2020.

Medium: Responses were taken via a form in Google Forms.

Survey Distribution: Responses were primarily gathered from Reddit, Twitter, Facebook, and sharing via email/other messengers. Some responses were directed from LinkedIn, Pinterest, and other sources. Sampling on Twitter and Facebook targeted individuals more likely to be interested in Halloween, fall, candy, cooking, and baking.

Compensation: Three random participants will receive $13 Amazon gift cards via email.

Requirements: Respondents must be ages 18+ and be a resident of the United States

Demographic Breakdown

The following visualizes the demographic breakdown of the sample:

Candy List

In case you were wondering what the comprehensive candy list is: 100 Grand; 3 Musketeers; Air Heads; Almond Joy; Blow Pops; Bubble Gum; Butterfingers; Candy Corn; Circus Peanuts; Cotton Candy; Cow Tails; Crunch Bar; Gummy Bears; Gummy Worms; Hershey's Chocolate Bar; Hershey's Kisses; Hot Tamales; Jelly Beans; Jolly Ranchers; Junior Mints; Kit-Kat Bar; Kit-Kat Bars, flavored (e.g., grape); Laffy Taffy; Lemonheads; Lifesavers; M&Ms; Milk Duds; Milky Way; Nerds; Peanut M&Ms; Reese's Peanut Butter Cups; Reese's Pieces; Rock Candy; Skittles; Smarties; Snickers; Sour Patch Kids; Starburst; Swedish Fish; Tootsie Pops; Tootsie Rolls; Twix; Twizzlers; Warheads

Caveats

• For the first 8 responses, "Skittles" were not present in the ranked candies section. This was shortly fixed.
• Some individuals did not choose a 5th choice for candy due to the UI of Google Forms sometimes cutting off the column on mobile devices
• Although always allowed as a blank response, a "no opinion" column was added partway through the survey to demonstrate that a like or a dislike was not required for each candy
• Raking was used for weights, which counts towards national-level vote outcomes. I did not weight statewide (senatorial/Electoral College) votes, however, as I would need a much larger sample.
• Arguably, the national winner could be considered to add 1 vote to the senate, as the Vice President of the United States acts as a tiebreaker vote. This is not reflected in the above analysis.
• Some write-in candies were also collected, in case favorites were not represented in the list. A few examples that have multiple mentions are dark chocolate, Baby Ruth, Payday, Take 5, and Heath Bars. There were not any particularly common ones, though.
• "3 Musketeers" was misspelled as "Three Musketeers", which may have confused some individuals with the ordering.
• The "Other" gender group was estimated to be 0.6% of the total US population. If anyone has any input on how this should be done, see my Statistics Stackexchange Question.

Source Code

The current version of source code can be found in https://github.com/mcandocia/CandyRanking. The data will be updated once I have more thoroughly analyzed the data.