This is the first of a series of responses by Mario Rizzo and Glen Whitman, authors of Escaping Paternalism, for my Book Club on their treatise.

Present Bias and Time Preference

We should begin by thanking Bryan for his many kind words about our book, and also for hosting this book club.  We appreciate his critiques almost as much as his praise!  Given the many excellent points raised by Bryan and others, we’ve decided to break our reply into a few posts – beginning with the issue of time preference and present bias.

In both of the first two installments, Bryan argues that any time discounting ought to be considered irrational – i.e., that future utility ought be weighted exactly as much as present utility.  The most important thing to realize is that, while Bryan’s position is certainly defensible, it is highly uncommon among both neoclassical and behavioral economists.  So in that sense, his position is orthogonal to our arguments.  Neoclassical and behavioral notions of rationality both allow discounting of the future relative to the present.  For them, the red flag of irrationality is inconsistency of discount rates, not discount rates that differ from zero. That is the argument that we have engaged with (and attempted to refute) in our book.

That said, it is worth considering Bryan’s position on its merits.  The idea of zero time-discounting is based on the idea that “utility is utility is utility,” and that discounting on the basis of anything else – such as uncertainty about the world, potential change of one’s preferences, or declining enjoyment because of age – is a separate matter.  In other words, once we have controlled for these other factors, the remaining level of time discounting should be zero.

Our first reply to this is to ask whether it’s really possible to separate out everything else that matters.  In any real-world measurement of time preference, you can’t rule out uncertainty. You can’t rule out the participants’ awareness that the passage of time means they might die, their preferences might change, their ability to experience pleasure may decline, and so on.  So measured time preference will always incorporate at least some of those factors.

The word utility itself may be misleading in this context.  What is this utility that might be discounted?  Presumably it is something real – a feeling of happiness or satisfaction.  But we don’t feel in the abstract; we feel about specific things, such as consumption of food, socializing with friends, watching movies, and so on.  It is all contextual, including the experience of the passage of time. And time itself is not a homogeneous medium except in idealized models. So outside of a mathematical model created for predictive or heuristic purposes, it’s not clear that utility (and discounting utility) means much at all.  Humans do not make trade-offs between abstract amounts of utility to be received at different times.  They make trade-offs between various types of consumption and well-being to be experienced at different times.  It is these specific sources of satisfaction that may be discounted.  This is why the notion of an abstract rate of time preference, which applies directly to “utility” rather than emerging from the process of comparing different forms of satisfaction, is a chimera.

In addition, the idea of zero-discounting presupposes a specific concept of the self as it endures through time.  We would echo the comments by B K, who mentioned philosophical arguments to the effect that your distant-future self is arguably not even you anymore, or is you to a lesser extent, so aside from altruism it’s not clear why you should care so much about that person.  Discounting the utility of a distant future self is arguably no more irrational than discounting the utility of a stranger.  While this argument may sound like austere oxygen-free philosophy, we think it’s closely connected to the issues of uncertainty and preference change above.  The more likely is future-me to have different preferences or values from my own, the less identity I feel with him, and thus the less inclined I am to be concerned about his welfare.  This would seem to be an ethical-value  question, not a question of rationality.

Finally, Bryan makes a different sort of argument when he says, “The whole idea of time inconsistency is that people predictably change their minds. If this is not a strong sign of irrationality, what is?”  Here, Bryan is saying something closer to what the neoclassical and behavioral economists say:  that the issue is inconsistent time discounting, not discounting that differs from zero.

But is it true?  Does predictably changing one’s mind really imply irrationality?  As we conceive it, the issue is that we humans have conflicting preferences within ourselves – a desire to indulge, and a desire to delay gratification – and we are in a process of negotiating between those desires.  Rationality does not, or should not, mean that we’ve already concluded the negotiation and landed on the final answers at the beginning of time.  When there is not a “correct answer” dictated by pre-existing well-defined preferences, then choice reversal can reflect a person’s process of forming and discovering their preferences over time.

Learning may also be taking place, including learning about both oneself and the outside world.  Individuals can make a plan, and then when it comes time to carry it out, they may realize the plan is higher-cost than they thought, or unexpected factors have intervened, or new options have become available, or their preferences and beliefs have evolved in the meantime.  It might takes someone a few failed plans to realize that specific kind of plan isn’t optimal for them.  By introspection, we’ve seen this in ourselves:  Well, I wasn’t able to carry through with that dieting strategy, so maybe I should try another.  Now, maaaaybe if someone makes the same plan again and again, and breaks that plan again and again, that could be a sign of failure to learn.  We haven’t seen the literature on time preference demonstrate this, although addiction could be one example.  Our personal experiences hew more closely to a model of experimenting with different strategies of self-management and learning more about ourselves over time.