A researcher conducts a survey to estimate a population's take on various issues. Each question provides a statement presenting a position on the issue, and the respondent must answer on a Likert scale: -3 (completely disagree) -2 (strongly but not fully disagree) -1 (somewhat disagree) 0 (neutral) +1 (somewhat agree) +2 (strongly but not fully agree) +3 (completely agree)

Given that the results will be used to inform the design of an important policy, the researchers worry about ending up with confidence intervals that do not contain the true parameter for each question (i.e., the value of the average answer to each question for the entire population). Assume that they are ready to conduct a new survey that is adapted to this objective. If they want their confidence intervals to keep the same width as with their original results, what should they do? Be specific and state why.