Now let's have a look how exactly we would sample these 1,000 voters at random. A simple way would be to select 1,000 voters from your neighborhood, say your hometown, this is called the sample of convenience and it's not a good way to sample. The reason is that the voters of your hometown will likely be different from the population of US voters. For example, voters on the West Coast and on the East Coast tend to be more Democratic voters, whereas voters in the Midwest tend to vote more Republican. This will introduce what's called a bias, that means the sampling will favor a certain outcome. There are various kinds of biases. Let's look at the three most important ones. We already looked at selection bias. For example, a sample of convenience makes it more likely to sample certain subjects than others. Non-response bias means that the people who choose to respond to the question may be different from the non-responders. For example, parents are less likely to answer a survey request at 6 pm because they are busy with children and dinner. Finally, you've probably seen websites that allow you to rate businesses. The reviews you see there are not very representative because those reviews are more likely to come from customers who had very bad or very good experiences and therefore feel strongly about writing a review. This is called a voluntary response bias. The best way to avoid such biases is to use chance in some planned way in the sampling. The most well-known way of sampling is called a simple random sample. It means that subjects are selected at random without replacement. So if you sample 1,000 voters out of the yes population, that means each sample of size 1,000 is equally likely to be selected. How would one choose a simple random sample? A popular way to do that is using random digit dialing, that means the computer will dial telephone numbers at random. That worked pretty when every household had a landline. It's more complicated now with many people using cell phones instead, and the polling companies are figuring out ways to get around that. A more sophisticated way to sample is stratified random sampling. There, the population is divided into groups of similar objects called strata. For example, the yes voters would be divided in urban voters, suburban voters, and rural voters. Then a simple random sample is chosen in each stratum and the results are combined. This sampling can result in a more precise estimate than with simple random sampling. However, it's more complicated to execute.