lFigure 9-2.—Sample size increase vs. accuracy.What are the time and cost constraints? Doyou have the time, manpower and money toaccomplish whatever survey project you areattempting? This is important. Set deadlines andgive specific job assignments or else piles ofsurvey junk may end up stashed in someone’sdesk drawer.SAMPLE SELECTIONHow large should a sample be from any givenpopulation? This question takes us into the mathematicsof probability. Do not worry, you will not have tounderstand the statistics behind national surveysproduced by the likes of George Gallup. But perhaps aquote from Mr. Gallup might help put the question ofsample size in perspective. “Both experience andstatistical theory point to the conclusion that no majorpoll in the history of this country ever went wrongbecause too few persons were reached.”Gallup conducted a number of experiments on theeffects of sample size. In 1936, he used 30,000 ballotsto ask the question: “Would you like to see the NationalRecovery Act (NRA) revived?” The first 500 ballotsshowed a “no” vote of 54.9 percent. The completesample of 30,000 ballots returned a “no” vote of 55.5percent. In other words, the addition of 29,500 ballotsto the first 500 ballots only made a difference of 0.6percent. (See fig. 9-2.)Through the mathematics of probability, we knowthere is a real but unknown distribution of all possibleanswers to a question. If we then know that our sampleis random (meaning that every person in our audienceis just as likely to get a survey as any other person) andthat our techniques are capable of obtaining a reliableresponse (without bias) from each person, we will beable to tell how representative the responses are.For the purposes of this chapter, the SociologyDepartment at the University of West Florida supplieda quick and easy formula often used in social scienceresearch. It is shown in figure 9-3.Figure 9-3.—Sample size formula.9-6