We tend to say that "everybody should vote, no matter what". And of course, democracy cannot work if nobody votes. But as long as enough people vote, does it make a difference when the number of voters increases?
In other words, can we say that a votation where only 30% of the people votes is biased (e.g. "older people tend to vote more than younger people, and they are usually more conservative"). Or is it only marginal? Would the results be much different if 100% of the people was voting?
Let's posit a counterfactual world where voters randomly determine whether they will vote or not. Each voter is independent of each other and who they vote for is independent of whether or not they vote.
In this case, it's trivial to say that (so long as an appropriate sample size is met) more voters do not really influence the outcome of the vote.
This isn't true in the real world.
The first reason is voter turnout: the most important factor in determining whether people vote is the social pressure to do so. This can be "manufactured" by using direct marketing (people knocking on doors), telephone calls, mailing, etc. - but it works even better when people in your own social group pressure you to vote. This pressure is significantly increased when voters feel that the vote is going to be close, meaning that their votes are particularly important.
There can be institutional factors which provide incentives to vote. Powell and Jackman provide a decent list in this article. And it's worth noting that Powell and Jackman are well known in the field of voter turnout.
If you have access to academic journals, here are some articles about voter turnout that you might be interested in:
A second reason would be that our voting preferences are not distributed evenly. The most important factors in who we vote for are all demographic (age, race, religious background, your preferred measure of socioeconomic status, etc.). This is the basis of the 'funnel model' popularized by "The American Voter". If you want to read more about voter behavior, this is the single best resource available. It was re-published in the 2000's and can be had on Amazon.
In an extreme case, even an election or referendum with 100% voter turnout would not be unbiased - it would be biased in favor of registered voters.
Many people never register to become voters. I was unable to find any literature on why people do not register to vote, but my suspicion is that it is not random.
There are also examples of large swaths of the public being unable to vote by law. For example, minimum voting ages prevent the vote from being representative of people in those age categories.
The difference between an autocracy and a democracy is that a democracy has the regulation mechanism of regular voting for political leadership. This regulation mechanism encourages political leaders to make decisions which are in the interest of the general public and not just in their own interest.
Democratic systems are built on the assumption that when people are unhappy with the political leadership, they vote for someone else. Not voting at all sends the message "Everything works well, continue like you already do".
However, many people who don't go voting are often those who are the opposite of satisfied. They are often those who feel misrepresented by all major political candidates. But if a democracy is supposed to improve, it is imperative that especially those people are voting who don't feel represented. The option is usually to vote one of the many chanceless minor candidates which usually run in elections. When the dissatisfied people spread their votes on the no-name candidates, they punish the major candidates. The major parties visually lose percent-points which will make them wonder what happened and if they might do better if they change their stance on certain issues.
So far the theory. But in practice, you have to look at who exactly tries to mobilize voters to vote no matter what. You will often see people behind such campaigns who expect that it will lead to more support for their favorite candidates. Others think that preventing people from going to vote can in fact be good for their political success, so they do not actively encourage non-voters to vote and do the exact opposite. For example, in Germany the current leading party CDU followed an election campaign strategy in the past years which was dubbed asymmetric demobilization by outside observers. It means they try to make the election as boring as possible by intentionally dodging all controversial topics. They know they have the largest amount of loyal core voters who will vote them on principle no matter what. So when they can persuade all the switch-voters to stay home at election day, they will benefit.
