A health survey samples households and then interviews several people in each selected household. People from one household may have similar health behaviors. What is a likely consequence of ignoring this clustering when calculating standard errors?