uppose that X has a Binomial(n, p) distribution, where 0 < p < 1. Define p = X n When p(1 p) is smaller than its expected value, the quantity p(1 p)/n is typically also smaller than expected. When this occurs, the (random) Wald confidence interval p 1(/2)p(1 p)/n for the unknown proportion p tends to be smaller than it ought to be; and, as a consequence, the coverage probability of this interval is lower than desired value of 1 . Similarly, if p(1 p) is larger than expected, this confidence interval tends to be larger than it ought to be; and, therefore, the coverage probability tends to be higher. So: if p 0 or p 1, does the Wald confidence interval tend to have a probability of coverage that is too high, or too low