In the social sciences, researchers often use surveys to try to find out about the incidence or causes of some social phenomenon. For example, they might use a survey to find out about how common drug use is among various segments (for example, people of different ages or income levels) of the population. Since they cannot survey every member of that segment (for example, all 15 year olds in the country) they survey only a sample and base conclusions about the whole population on the responses elicited from the smaller sample.
When doing this, it is very important to avoid bias in the sample. Household bias is one possible sort of bias that can skew a sample. Household bias occurs when a researcher samples the same number of people from different groups even though those groups are of different sizes. For example, if you wonder about drug use at a high school and you poll an equal number of athletes and non-athletes, you will likely get household bias because there are more non-athletes than athletes in the whole student body. To avoid this, it is important to poll a given type of person in proportion to their presence in the overall population. If the school is 70% non-athlete and 30% athlete, you need to have your sample consist of 70% non-athletes and 30% athletes.