Narrower confidence intervals are better than wider confidence intervals. One way to achieve a narrower confidence interval is to use a larger sample size. So, why do statisticians interested in estimating a parameter value for the population of all adult Americans not always use a sample size of 1,000,000 people?