A potential solution to this is the following: when asking whether they support or do not support a given candidate, give the people you are polling the following instructions: flip a fair coin privately, and if it comes up HEADS, answer honestly; if it comes up TAILS, flip another fair coin and if it comes up HEADS, answer 'support', if it comes up TAILS, answer do not support'. In this case, the person being polled can always claim that whatever they answered was the result of the coin - in a sense, the results are anonymized and the people being polled are protected. Let p be the probability that a randomly polled person using this method says 'support'; let q be the true probability a random person actually supports the candidate. We would like to know the value of q, but we can only estimate the value of p: let pn be the fraction of N people who answer support' using this method. We have that E[N] = P₁ as before. 7) What is the relationship between q and p? 8) Construct an estimator qy from px (i.e., a formula for qy in terms of pN) so that the expected value E[N] =q. What is the variance of ĝN? 9) How many people N should you poll to guarantee the actual error between ĝN and q is less than €, with 90% confidence? Note, q is unknown, so you cannot use it to determine N.