a loophole in opinion polls
The long-used election polls are now perfect for Brexit in the U.K. 먹튀검증사이트 and the 2016 U.S. presidential election, leading to the opinion polls being useless. CNN also reported live coverage of “U.S. polling companies should close down if they want to produce this kind of result.” Then let’s find out why these findings were forced to come out, and the problems of existing polls.
At least too few samples
Do you happen to know how many American voters do you have? And do they know how many samples do they predict? If you are not an expert, you may well not know. We don’t need to know that because we have unlimited confidence in large polling agencies. But that belief is slowly crumbling these days. Election polls we know can be likened to “the tip of the iceberg” in short. Polls frequently reported on the air show the size of the entire iceberg without seeing the massive iceberg lurking in the water. So predictions may be right, but they are not only more likely to be wrong, but the error rate is also high. Let’s take a closer look at the poll conducted in the 2016 U.S. presidential election. Looking at the method of presidential poll by ABC News, one of the most prominent research institutes, the results were predicted for more than 1,000 voters using home phones and mobile phones.Let’s think about it. It’s the result of more than 1,000 voters. If there are between 220 and 30 million voters in the U.S., and 60 percent, or between 120 and 30 million, cast their ballots, would it make sense to survey public opinion by the figure of more than 1,000? The number is less than 0.00001 percent of the total voters in the population Moreover, since there are 50 states in the U.S., a conservative estimate of 1,000 people is about 20 per U.S. states. Can we say that 20 people a week, and that the results of the poll by phone really reflect the state’s public opinion? There is an old saying that the elephant touches different parts of the elephant and claims that it looks different, and at this level, it is not just touching the elephant’s legs to see what it looks like, but touching its fur.
Low response rate and sample bias The second problem with existing election polls is low response rate and sample bias. At one point or another, there will be readers who have received phone calls from election polls. But what percentage of the respondents who answered the phone said they listened to everything and answered it sincerely? Most people probably knew it was a spam phone as soon as they got the phone, and some of the respondents who decided to answer the phone’s questions were answered roughly once and then hung up. The low response rate of these telephone surveys is nothing new, and election experts are already well aware of these problems. Nevertheless, the reason this method has been used so far is because it is more cost-effective.
In Korea, telephone interview and automatic response survey ARS: Automatic response service are being combined, which is being scaled back due to the cost of personnel. The U.S. telephone poll also showed that the response rate is currently less than 10 percent, compared to 30 to 40 percent, and that the response rate is estimated to be much lower than that for automated response surveys. In particular, telephone surveys in the U.S. showed that respondents were more reluctant to respond because of the lengthy survey. True, the low response rate issue has emerged as a problem not only in Korea but also in the United States, but the bigger problem is the sample bias that arises from it. Getting the right samples, even if they are small, is not a big problem with the election prediction, but if the respondents responding to the telephone survey fail to represent the right population, it can also cause incorrect predictions of the election results.
