Surprising Items

The Relativity of Poll Results to Time

One of the most shocking findings was that surveys are sometimes overtaken by time (Sheldon and Evans, 2006). Ideally, I have always believed that a concept developed concerning people’s views should reflect the state and belief at the moment. As much as time passes, poll results should not change just because of it. For example, if the majority of citizens interviewed believe that the current government’s leadership is weak, the results as of the date of the interview should remain relevant, albeit confined to the past time, if the government is changed or the leadership improves after, say, two years. That means that becoming better at a future date should not substitute when the government was worse in the past.

The Impact of Sampling Error on Poll Results

Secondly, the fact that a sampling error leading to different results of the same issue can be accepted based on the premise of acceptable error level (Sheldon and Evans, 2006). Ideally, if a researcher samples 1000 respondents to enquire an opinion about a given political candidate for presidency, and the results show that the preference for such a candidate is 70%, hence projecting the results to a country’s population of about 40 million will imply that 28 million people can vote in favor of that candidate. Surprisingly, if another researcher samples equal number of respondents (1000) and after interviewing them finds that only 40% of the citizens prefer that particular candidate, implying 16 million people, that the two poll results should be adopted based on the possibility of sampling error. Ideally, if a sampling error should be tolerated, the the percentage difference should be very minimal.

The Limitations of Sample Selection in Polls

Thirdly, it is surprising that a sample for 1000 for instance can be a true representation of 120 million people (Sheldon and Evans, 2006). Given that people are motivated by different things and can express their personal opinions, sample selection may always not give accurate results unless it is 100% of the population. Finally, it sounds surprising that the order of asking questions can also be a factor that can make a poll yield different results if conducted using different arrangement despite the fact that the questions actually remain the same. In my thinking, honest respondents should be able to express their honest opinions despite the order of questions and in such that should guide a researcher on whether to rely on the poll results or not.

Key Factors in Research Analysis in Political Science

The issue on who actually undertakes a research is an influential factor towards the results. This is due to the ethical concerns and possible influence to both the respondents and a researcher. Secondly, the fact that polls should be evaluated based on who paid for them is a relevant consideration on possible bias and influence. Thirdly, knowledge that the more the respondents in a poll the smaller the error is an elaborating factor on political research analysis. Finally, consideration of past polls or same topic and results is a good practice especially in analyzing the reasons for deviation if any. As such, these points give a clear explanation and build further on conducting research in political science (Geer, 2004).

Challenges and Limitations of Polls in Representing the Population’s Thoughts

In conclusion, polls should actually not be used as a good representation of how the population thinks. Polls are usually marred with external influence, irrespective of the reputation of the conducting firm. Fundamentally, politicians have so much influence especially in less democratic nations. Additionally, results will depend on respondents’ political affiliation, ethnic backgrounds, religion as well as social class (Geer, 2004).

References

Geer, J. (2004). Public opinion and polling around the world: a historical encyclopaedia. Santa Barbara, Calif: ABC-CLIO.

Sheldon, R. and Evans, W. (2006). 20 questions a journalist should ask about poll results. NBC News Poll.