Gery Ryan, Steve Borgatti, and I built a decision model for this question: ‘‘Think about the last time you had an empty can in your hand—juice, iced tea, soda, beer, whatever.

What did you do with it?’’ (Ryan and Bernard 2006). We interviewed a convenience sample of 70 people in California, North Dakota, and Florida and asked them 31 questions about the event, but we were able to predict 90% of reported decisions with just the questions in figure 17.4.

Read figure 17.4 as follows: Start by asking whether the informant at home when they made the most recent decision about what to do with that can. If they were at home, then ask if they recycle other products besides cans. If they were, then guess that they claimed to have recycled the can. This will result in two errors. That is two out of 23 people in this condition claim not to have recycled. Read the rest of figure 17.4 similarly. The whole model gets 63 out of 70, or 90% right—and that’s 77% better than assigning people a result (recycle or not) randomly.

Figure 17.5 shows what happened when Ryan and I tested our ethnographic model in a nationally representative telephone survey of 386 people in the United States. We got 84.5% right, and that was 59% better than chance. It may not work for everything, but on this model, the ethnographic model was a proxy for a national model of the decision on what to do with an empty can.