This paper will discuss the power of statistics and how they are used in real life. It will relate the example of ex-president George Bush’s pre-election polls and what happened that reduced his estimate in the election. Several errors in forecasting will be analyzed along with the reasons why those errors exist. Finally some of the sampling techniques will be explained giving the best technique that can be utilized in pre-election polls. Differences in Polls According to the article by Kudlow (2000) “The weak-end of polling,” the author clearly states that polling done by different agencies are subject to a wide range of errors.
And it is because of these errors that Bush’s victory in the polls has been forecasted in all the ranges between 4% to 12% points. Kudlow says that there are two primary reasons why the polls done give varied results. One of the problems is that polling done on the weekend gives a completely wrong picture and it is cannot be trusted. Furthermore, some polls are also conducted overlapping week days and weekends. For example a poll conducted over Friday, Saturday, Sunday and Monday will still yield an error in forecast, however since Monday is included in the polling days, some error margin is reduced.
But in order to get more accurate results the days through Monday to Thursdays will yield the best possible figures. How far this hypothesis is true cannot be completed ascertained but judging the reliability of the people saying this, this statement can be said to be quite valid. The secondary problem with polling by different agencies is the target audience with whom polling is conducted on. According to Kudlow, the polls conducted on ‘registered voters’ will yield an untrustworthy estimate and one should be careful to agree with it.
Yet another problem with polls is the sampling by the polling agency. Since in the USA, every region and state has different requirements for seats to win, it is imperative for the agency to get a representative sample from each region and state to get an accurate estimate. When these three errors are combined as in when polls are conducted on weekends and on registered voters, the resulting estimate is confusing and even more unlikely. Thus when different combinations of these two major errors are taken, then we understand why there is a huge difference in the percentage points favoring Bush.
Apart from these three big factors identified in the Kudlow article, there always exist standard errors in the estimate which can occur due to any reason. For example, every single polling agency will have different expertise, different sampling techniques and different methodologies to analyze data. Also biases mostly creep in such estimates either from the voters or from the data collectors thus the final result is bound to have a certain percentage of error. Combining these generic errors with two big errors identified by Kudlow, it is no wonder that there are so many different estimates flowing around the political campaign.
Sampling Techniques Statistics have given us a lot of sampling techniques that we can utilize in making our estimates for the election victory. However, psychology has also taught us that sampling alone will not reveal an accurate result to various biases that creep in when data is being collected. So in order to negate the biases, a certain error figure is kept aside from the final result to give what we call a realistic estimate. The election system in the USA is divided into regions and then within regions are various states. A certain majority of seats is required to get the electoral vote and the popular vote.
Sampling done to get the most accurate figure will require studying historic patterns of that particular state and region while also seeing any changes in population trends. To make an accurate estimate, a multi-stage sampling technique will be used starting by clustering the USA into all the different regions. Within each region further clustering of the each state will be done within which different townships and segments will be made. These townships and segments will be divided into homogenous groups based on their voting history.
A small random sample will decide which areas to mark as of homogenous nature (Dobson, & Young & Gibberd, n. d. ). When all the different strata are identified, another random sample will be run to identify which portion of these localities to question. After the localities are decided, the random number of people within each locality will be identified to be interviewed about their preferences in the elections. Since this is the foremost issue in estimating, emphasis should be given to this phase. Equal representation for the men and women should be given.
Since within US many different sects, races and religions live, it is vital that all of these segments be equally represented. Since our sampling technique is based on geographical sampling, it must be ensured that these races, and sects are included in the sample. Based on the number of these races living in a particular locality, a quota for these races will be allotted (Probability Sampling, n. d). This stratified random sampling of each segment will yield the number of people within the states choose the options given below. • Party A • Party B • Will not vote • Undecided
When data from each segment of a state is received, it will be compiled and itemized into the localities that were identified to give the final result. While this technique may not be completely accurate, it is a method that will give the most representative sample. This is because this technique uses the multi-stage sampling method whereby stratified samples are taken, quotas are kept, random samples are taken and clusters are made to identify each and every type of person that will likely vote. References Dobson A. , & Young A. & Gibberd B. (n. d. ). SurfStat Australia.
Retrieved November 12, 2008, from http://surfstat. anu. edu. au/surfstat-home/surfstat-main. html Probability Sampling, Retrieved November 12, 2008, from http://www. socialresearchmethods. net/kb/sampprob. htm Statistical Sampling Terms, Retrieved November 12, 2008, from http://www. socialresearchmethods. net/kb/sampstat. htm Statistics Glossary, Retrieved November 12, 2008, from http://www. stats. gla. ac. uk/steps/glossary/sampling. html Kudlow, L. (2000) The Weak-End of Polling, Retrieved November 12, 2008, from http://www. nationalreview. com/kudlow/kudlow071800. html