E=Z*sqrt (p(1-p)/N), where E= error margin, p=proportion, N=sample size
Katrina's margin error at 85% confidence interval: E=1.96*sqrt (p(1-p)/100) = 0.196 sqrt (1(1-p))
Mathew's margin error at 99% confidence interval: E= 2.58*sqrt (p(1-p)/400) = 0.129 sqrt (p(1-p))
Since both obtained same estimate of proportion (that is, value of p), it can be seen that Mathew's estimate will have a small error (That is, 0.129 is smaller than 0.196). This can be attributed to larger sample size although a wider confidence (99%) interval was considered.