— Overview
This reading reads two Korean Gallup readings on Shingori units 5 and 6 as a lesson in sampling, a bandwagon effect and nonresponse, and they turn to a deliberative poll to mend ordinary survey limits.
— Contents and key concepts
Four candidates in 1824, and a missed Truman
Polling, this reading writes, grew from the realisation that the accuracy of sampling matters more than the size of the sample. The first recorded poll, in this telling, was a local newspaper’s reading of four candidates in the 1824 United States election, and the prediction missed. From 1850 everyday opinion was also surveyed, and Debow set down standardised interview procedures, interviewer training and statistical analysis. From 1880 a wish to shrink the gap between survey results and actual vote shares took hold, and Taylor’s exit poll was introduced. In the 1930s Robinson analysed past election surveys and drew out seven kinds of error. George Gallup founded the American Institute of Public Opinion in 1935 and used proportional stratified sampling, dividing the population by demographic traits. In 1948, failing to count the undecided, he missed Truman’s victory, but afterwards, they write, correctly predicted eleven presidential elections. Those identifications stay as given in this reading.
Two Korean Gallup readings, a bandwagon, and a nonresponse rate
On 14 July 2017, they write, Korean Gallup put construction stop for Shingori units 5 and 6 at 41 percent, continue at 37, and reserve at 22. In the same survey, using nuclear power was favoured 59 to 32, and 54 percent said plants were dangerous. On 4 August the gap had narrowed: continue 40, stop 42, reserve 19. In the August survey, Democratic Party supporters favoured a stop at 56 percent; Liberty Korea Party supporters favoured continuing at 76. Men favoured continue at 50 percent, women a stop at 46, a split by gender. Both surveys were of 1,004 adults nationwide, at a 95 percent confidence level, with a sampling error of ±3.1 percentage points. The difference between the two readings, they say, can be found in a bandwagon effect and in the nonresponse rate. Those figures stay in the window this reading gives.
A deliberative poll, and 20.4 billion won in 2010
A deliberative poll, they write, combines a scientific opinion survey with small-group discussion and is taken more than once. It was devised to fill doubts about the quality of opinions that rest on short information and a momentary impression. Its seven merits are representativeness, equality, expertise, rationality, accountability, the prevention of conflict, and a raising of democracy. A scientific sample is drawn for a first survey; information is provided; then, over two or three days, groups of about ten hear expert explanations and debate, and a second survey is taken. Ordinary polls measure opinion at the moment of the survey, but they are limited by indifference, conformity, and the reflection of press coverage. A deliberative poll was introduced to follow how a person’s view moves after enough information and discussion.
A telephone survey takes little time and money, sampling and access are easy, and interviewer error can be reduced, but question length is limited, audiovisual material cannot be used, and some groups are hard to reach. An online survey is fast and cheap, free of limits of time and place, and can offer video and other aids, but securing a representative sample is hard and duplicate answers are possible. A mixed survey applies two or more methods to one research problem, to mend each method’s weakness. The 2010 Population and Housing Census, they write, introduced an internet survey and saved 20.4 billion won in budget; mixed surveys, they add, reduce nonresponse error and coverage error. That saving stays on the dates given in this reading. The same Korean report, at fuller length, is on CampusDown.
Sources
- This reading, a student report on Korean Gallup’s 2017 readings of Shingori units 5 and 6, with the history of polling, a deliberative poll’s seven merits, and a 2010 census saving left on the dates given in this reading.
Last updated: 15 November 2026
Frequently asked questions
What figures do they give for the two Korean Gallup releases in 2017?
On 14 July, they write, construction stop stood at 41 percent, continue at 37, reserve at 22; in the same survey, using nuclear power was favoured 59 to 32, and 54 percent called plants dangerous. On 4 August the gap narrowed: continue 40, stop 42, reserve 19. Among Democratic Party supporters stop was 56 percent; among Liberty Korea Party supporters continue was 76. Men favoured continue at 50 percent, women stop at 46. Both surveys, they write, were of 1,004 adults nationwide, at 95 percent confidence, with a sampling error of ±3.1 percentage points. Those figures stay in the window this reading gives.
Where do they place the difference between the two readings, and how do they tell the history of polling?
The difference, they say, can be found in a bandwagon effect and in the nonresponse rate. Polling, they add, grew from the realisation that accuracy of sampling matters more than sample size. The first recorded poll, in this telling, was a local newspaper’s reading of four candidates in the 1824 United States election, and the prediction missed. From 1850 everyday opinion was also surveyed, and Debow set down standardised interviews, interviewer training and statistical analysis. From 1880 came a wish to shrink the gap with actual vote shares, and Taylor’s exit poll was introduced. In the 1930s Robinson, analysing past election surveys, drew out seven kinds of error. George Gallup founded the American Institute of Public Opinion in 1935 and used proportional stratified sampling. In 1948, failing to count undecided voters, he missed Truman’s win, but then, they write, correctly predicted eleven presidential elections.
What is a deliberative poll in this telling, and what do they say about mixed surveys?
It combines a scientific opinion poll with small-group discussion, and surveys more than once, to mend doubts about opinions built on short information and a momentary impression. Its seven merits, they list, are representativeness, equality, expertise, rationality, accountability, the prevention of conflict, and a raising of democracy. A first survey is taken from a scientific sample; information is given; then, over two or three days, groups of about ten hear experts and debate, and a second survey is taken. A telephone poll is cheap and quick but limited in question length and cannot use pictures or sound; an online poll is fast and can show video, but representation is hard and duplicate answers are possible. A mixed survey applies two or more methods to one problem. The 2010 Population and Housing Census, they write, saved 20.4 billion won by introducing an internet survey. That saving stays on the dates given in this reading.
Refund policy
- Change of mind / refund requested after partial use / more than 7 days after purchase
◆ Exceptions
- File errors / cannot download or open the file
◆ How to request a refund: send a 1:1 inquiry via Customer Service at the top