The article looks at how getting the details wrong in a model can affect the results. They studied the U.S. money demand model and found that when a certain time trend is left out, the results are strong. But when that trend is included, the results are not good. By using a specific method, they found that the preferred models are usually not connected with their causes. This shows that it's important to be careful when setting up a model, or the results might be wrong.