How can convenience samples be biased?
Convenience sampling (a type of non-probability sampling) involves taking a sample from part of a population which is close at hand. This can lead fairly quickly to bias, though the manner in which the bias surfaces may vary depending on the manner of "closeness" used.
For example, suppose that a university student living on campus wishes to gather data on the political affiliations of people in his country. If the student uses convenience sampling, then the closest population (both in terms of physical proximity and social circles) is likely to be the student population at the university. If the university itself has a particularly strong political leaning (for example, either liberal or conservative), then the data would most likely reflect that, even if this leaning is drastically different from the country-wide data.
Often, it is preferred to only use convenience sampling as a preliminary study, and to follow it up with a more reliable method, such as simple random sampling or cluster sampling.