What is the primary effect of decreasing the sample size $n$ on a confidence interval, assuming the confidence level and population standard deviation remain constant?