by Bushra Shaikh
The New York Times Magazine conducted a poll on Twitter last month: If you could go back and kill Hitler as a baby, would you do it? It represented it as a pie chart in the form of a baby rattle. 42% of the participants voted in favor of killing baby Hitler, 30% said no and 28% said that they weren’t sure.
Polls in general have a margin of error. So in my opinion the kind of sample used for the poll was a result of cluster sampling: the readers of the magazine. The object in the data set here are individuals and the essence of the data here is categorical.
One can clearly say that this is an observational study where the NYT magazine in no way tried to influence the decisions of the subjects.
However, the poll here is a result of a voluntary response. So here we need to understand the audience that was involved had the following properties: The median audience age here is 55, where 60% of the readers are male and 40% are females. A large part of the readership is people who are 35 years or older and have an undergraduate or post graduate degree. These specifics clearly mention that these are educated masses. Another thing to remember is that the magazine has 3,000 subscribers.
However, the margin of error can easily exist here. There are possible issues with just the sample alone. The methodology does not explain the following factors:
- This poll was conducted online, therefore, measurement error, non-response error and coverage error are more likely to take place.
- The poll could represent a biased sample because it is a product of voluntary response. It could be that particular types of respondents are being over presented.
- The poll does not provide you with any specifics of the situation. The respondent could’ve assumed a very different scenario like he/ she would send someone else to kill Hitler and that would still count as being responsible for the killing.
- The poll decision could’ve been made in a happy, sad or angry state of mind by respondents
- The sample size is quite small if you’re trying to mirror a mass opinion. This also counts as convenience sampling.
- The margin of error could be great because the sample has a narrowed audience
- Behavior, roots and knowledge also play an important role. It does not mention how many of the respondents are Jews or a Germans who are likely to be biased. This may skew the opinion a large deal. Socio-demographics matter.
- Values also pay an important role here and it can’t be quantitatively recorded. To come up with a conclusion for a result like this, acknowledging factors like these also play an important role.
- Someone who strongly feels as opposed to someone who feel moderately about it isn’t portrayed here as well.
- It assumes that respondents understood the question, which may not be right because twitter has opinions where people ask if the question was killing Hitler as a baby (the voter being a baby)
- The respondent could’ve assumed that killing baby Hitler means no world war, again making uncertain claims
- The number of men and women are not equal
- The poll is susceptible to extrapolation because of the type and size of sample used.
- No way to derive that every reader took part in the poll
- Possibility of agree-response bias or acquiescent effect appears here.
If this poll were to be carried out again, I think more uniformity would be needed in something like this. It is important to choose a different sample, something like a stratified sample.
Also, in a poll like this age makes a huge difference. There could be war veterans as well families who already have a stand. A poll like this should have the sound benefit of making an objective judgement.
Mentioning no consequence for killing Hitler is also a huge factor that can alter the result. Setting up a conditional clause may alter the outcome here.
Another important thing which needs some focusing is the way the question has been asked. The wording need to be precise.
Taking a larger sample is also important because it decreases the percentage error. Taking precautions to eliminate measurement error, non-response error and coverage error is of utmost importance.
Four steps are important to conduct this poll: opinion, direction, intensity and the salience of the issue to the respondent. What I do appreciate about the poll is that it acknowledges non-attitudes. People who aren’t sure or may not have an opinion on the question. But this does reduce the sample size. Using a number-scale would also portray a more crystal opinion.
Though the poll represents opinion of its readers, there is no way to know if this opinion reflects the stand of a larger population. It also fails to provide any solutions.
Non-statistical fun facts and food for thought
In response to the poll, other publications measured the stakes. The Atlantic realizes the consequences in the story titled The ethics of killing baby Hitler. The article raises an important point: “Beyond the Holocaust, removing Hitler from history would be a gamble with the highest stakes imaginable. Any theoretical attempt to prevent World War II must also reckon with the possible course of history in its place. Without the war’s economic and military toll, would Britain and France have been better positioned to prevent decolonization, or to at least better able to resist nationalist movements in Africa and Asia with force?”
Jake Silverstein, the magazine’s editor-in-chief, was surprised by the results. “I thought more people would say they would,” Silverstein told CNNMoney.
There were articles written according to philosophical as well as ethical perspectives and some articles mentioning that going back in time is not possible. However, BuzzFeed came up with its own poll.