actually miniscule, and here’s why: If you think about the expected value of voting the way you think about the expected value of a lottery ticket, you have to multiply the probability that you’ll get something out of your action times the benefit of your action. The benefit of your action is being able to pick the president. Let’s just stipulate that picking the president is something that most people would not spend more than $10,000 for. So what are you doing when you’re voting? When you’re voting, you’re trying to change the outcome of the election. You’re trying to get your candidate to win. But here’s the thing: If the election is won by more than one vote, than nothing you can do will change the outcome. If you vote, the same person wins as when you don’t vote. So the only time that a voter actually makes a difference is when there is an exact tie. So the probability of influencing the outcome of an election is the same as the probability of a tie, and the probability of a tie in the U.S. presidential election is about one in 10 million. That number has actually been estimated. So you have to multiply one over 10 million times whatever you’re willing to pay to pick the president, and that’s the expected value of voting. So $10,000 divided by 10 million is less than 1 cent. So if it costs you even a penny to put gas in your car to go to the polls, you shouldn’t do it. A lot of economists don’t vote for this very reason. But most people don’t think about the world this way.
I started thinking about other literature that was about social effects. People like Bob Huckfeldt [at UC Davis] who have done work on whether or not friends and family who vote tend to influence the people they are directly connected to. There is some evidence that if I vote my spouse is more likely to vote, but these effects are small. In an economic context, if I only affect you by increasing your likelihood to vote by 10 percent, than my decision to vote is my vote plus 10 percent of your vote and that’s not going to change the calculus very much. So there has to be something else going on here. We have to be making this decision not based on this economic calculus, but on something else. It hit me that even though the interpersonal effects, the direct effects, between people can be small, that the network effects can be large. So that one action can actually turn into many, spreading from person to person to person.
So I told my adviser, [Harvard University social science researcher] Gary King, and it clicked with him that he just heard a talk by Nicholas Christakis. Nicholas had given a talk about a very important social science question that he’s working on, which is why do widows die. One of the oldest social science correlations that we know about is that when people pass away, their spouses are more likely to die themselves. Our best guess in the United States, from very large data sets, is that when the man dies the woman loses about two years of her life. And when the woman dies, the man loses about seven years of his life. Nicholas was an end-of-life care doctor. He told a story about getting a call from someone. He worked in hospices where people were dying. And the person on the other end of the phone says, “You’ve got to help me doctor, I’m really having a hard time with this death, I’m feeling depressed, can you prescribe me something?” And Nicholas says, “I can prescribe you something but I have to apologize, I don’t know who you are.” And the person says, “Oh, I’m not [the spouse of one of your patients]. I’m a friend of one of the spouses of a person who is dying.” It was at that moment for him that it clicked that his social science program could be addressed in the context of a broader question about how health effects spread from person to person to person. So we were both really interested in the same question, and my advisor Gary King realized this and he introduced the two of us, and ever since then we’ve been fast and furious friends and collaborators.
