Eric Husman writes to me:
I have a personal theory I call the Alphabet Diet. I always begin with the fact that everyone who is eating anything is "on" a diet, and that when people do what is considered to be "going on" a diet, they are really *changing* their diet. I think the reason that most diets work at first is that they require you to change your eating habits. Since you are unfamiliar with the new rules, you basically cut back on the number of calories because you don’t know what’s "legal" and are confined to collections of suggested recipes based on a best-selling author’s preferences. As you discover foods within the diet that you like, you gradually get back to your previous calorie intake, i.e. you learn to "game" the diet. So I suggest that if you picked five letters at random from the alphabet and confine your diet to foods whose name does not contain those letters, you will see the same initial effects as the Atkins or any other diet. If the diet ceases to be effective, pick 5 new letters. It’s hard to write a best-selling book based on a principle that simple because there is no pseudo-scientific justification for random letters that will dazzle your would-be readership.
This is simple to graph with indifference curves. If you deny a person her ideal point, given previous income and prices, that person will then eat less. Over time, learning effects can counteract this tendency to some degree.
Here is my previous post on why the diet you choose does not seem to matter much.