In defence of a positive discount rate
After the random and unprovoked battery I recieved by a group of men, in my (usually quiet suburban) neighborhood, I have reflected on the nature of anxiety. My anxiety level, elevated in the wake of the ambush, has now receded to its norm. Yet the fact that anxiety spikes AFTER the distressing event has already occurred, rather than before, suggests it's an irrational and useless phenomenon. If anxiety was protective throughout our evolution, then why do we feel fine until danger strikes? If anxiety helps us avoid danger, then this pattern is suboptimal.
Perhaps I am confusing anxiety, anticipation of harm or a negative future shock, with adrenaline. Very likely, the surge in levels of this hormone was necessary to hedge against pain from aches, and to promote a swift recovery. Yet it's precisely norepinephrine that's responsible for our endogenous “fight or flight” behaviours that characterise anxiety. Without this neurotransmitter, you cannot feel anxiety, although the reverse isn't true.
If this ex-post transitory surge in acute anxiety was linked to a meaningful change in the posterior probability of an adverse shock, then it can be justified. Yet you're much more likely to be harmed by crossing the road, or a car accident, than a random violent crime. You're much more likely to be a victim of violent crime on a booze-filled night out in a vibrant city-centre bar than on a quiet suburban street. Yet I engage in these activities on a regular basis, without any obvious hesitation. If anxiety was rational, these events rather than walking the streets alone should be more troublesome, yet my emotional concern is for the latter.
Therefore, anxiety is a profoundly irrational phenomenon uncorrelated with risk with little predictive power. Such biases are costly however, so rational agents seek to minimise them [1]. How does this apply to anxiety? Obviously, humanity's favourite intoxicant is consumed for this very purpose. The main mechanism via which alcohol produces euphoria is via reducing anxiety [2].
However, even abstracting from the rebound and alcohol's potential for physical or psychological dependence, I posit that the lifetime costs of any alcohol consumption always outweigh its gains. Fundamentally, alcohol is a depressant, thereby it hampers cognition. Given the importance of human capital both to individual and societal outcomes [3], this is not insignificant. Suppose we consider IQ as a constraint in a two-period model. In the first period, you can spend some points on getting drunk, or keep the points and invest in higher lifetime wealth (doing a productive activity). Next period, if you got drunk, your IQ returns to baseline. If you remained sober however, your lifetime wealth increases. Assume that wealth is utility, and you would be better off if you stayed on the wagon [4].
When we drink, we are in effect borrowing from our future utilities to increase our present utility. However, just as we borrow to finance purchases unaffordable to us with our present incomes alone, this borrowing is rational and justified. If we feel irrational anxiety, or any other temporary state of negative affect, and we borrow from a happier future, then this promotes our rationality. Just as with regular borrowing where you're broke if you rely on this, your mental and physical health (therefore your human capital) is depleted if you rely on liquor. Nonetheless, we don't condemn debt on the basis that some irresponsible individuals end up bankrupt, so neither should we condemn drinking or any other activity in which a positive discount rate is involved. Consider time discounting as the intertemporal substitution of utilities.
My conception of time discounting as intertemporal substitution makes sense once you consider how you distinguish time preference from intertemporal consumption. Credit constraints render a bias towards present consumption obvious, yet why should we call one intertemporal substitution along a utility function, and the other time preference, just because of one constraint? Empirically, you're generally unaware of one's financial situation, so you cannot easily make this distinction. Sure, HANK modelling is more structurally identified due to this separation, yet this is precisely what I called mathematical sophistry to fit the data. Ontologically, I think both are just different means to state the same thing, that we hold preferences relating to time and when we want to consume.
Perhaps models that do not explicitly microfound this as an intertemporal tradeoff, or require an extra parameter for discounting to fit the data, are bad models. Yet dynamic game theory relies on time discounting. Positively, discard the discount rate, and we lose many insights regarding how and why agents cooperate. Normatively, any weighting of the future less than the present could be considered as undermining growth via reducing physical and human capital accumulation. Yet it's plausible that recessions cause scarring. More present-minded individuals yield higher mpcs, and their discounting of future taxes undermines Ricardian equivalence. This myopia should aid a fiscal stimulus, so aid recovery, hence mitigate scarring. As such, it's not obvious that a positive discount rate is necessarily damaging for future prosperity once one considers carefully the dynamics, although a model is probably required.
Despite my defence of positive discount rates, both positively and normatively, this phenomenon is also associated with many of our worst pathologies and impulses (including my attack!). History demonstrates that we cannot take the existence of a stable civilisation for granted. Excessive time discounting is obviously pernicious.
Is there an optimal discount rate, and if so, what value does it take? I tend to consider discount rates as the opportunity cost of investment, as if I recieve a smaller sum now in lieu of a larger future sum, I can invest it into the capital markets. In our standard models, this return is the marginal product of capital - equivalent to the risk-free rate. In more complicated models of capital and asset heterogeneity, there are numerous returns that differ according to risk. Indeed, this explains how impatience tends to be correlated with risk-seeking, and if discounting was a bias (as most psychologists think) then we'd expect this to be uncorrelated with risk preference. One might also augment future returns via weighting with the probabilities of existential risk events, which may generate nonlinearities and hyperbolic discounting.
In principle, one can select a portfolio of returns, consistent with one's risk preferences, to match their discount rate. Yet this is not an arbitrary allocation, as not all values of return are possible. It appears that most, at least some of the time, discount at rates far in excess of what can be justified via returns in the capital markets. As a result, whilst we shouldn't weight the future as equal to the present, we should place higher value on future periods than we currently do. Most versions of virtue ethics seek precisely to do this, whilst embedding our imperfections (so realising a rate of zero is unfeasible). Being a former critic of virtue ethics, this is why I'm increasingly embracing the idea.
I like this framing of the expectations operator as a loss minimiser of forecasting errors. Of course such exist, and cognitive bias can produce them, yet over time agents learn. There's no reason to expect errors to be correlated, or cognitive bias to produce correlated errors, so the use of this operator is justified and rationality results. I think this is how we can square the standard rational choice framework with behavioural economics: rationality arises as we actively minimise our errors.
Albeit the mechanism focuses more on stimulating GABA so reducing inhibition, as opposed to suppressing norepinephrine per se. There are also notable increases in dopamine, serotonin, and endorphins, although I consider this a by-product of reduced anxiety.
Although I'm increasingly sceptical of models that imply that any investment into human capital always increases growth permanently. If that was the case, then it would be optimal to devote all resources to human capital. Such a conclusion denies the salient role of tradeoffs or opportunity costs, which is a fatal flaw for an econonic theory. Nonetheless, the returns to investing in human capital are positive, which is the main factor in this analysis.
Of course this is a highly simplistic framwork. What good is wealth for in a world devoid of pleasure, fun and convivality right? I'd argue that other intoxicants can promote cognition, via expanding one's set of perceptions or for (especially with stimulants) nootropic value. As for socialisation, vital for information aggregation so cognition and human capital, our culture of drinking indeed serves as a focal point. Yet this focal point is inefficient: as I've substantially moderated my consumption, I've noticed that much of what we consider socialisation is merely ritualistic bonding. How many conversations are good enough to learn from at a party, or even your median pub? I think the Bay Area rationalist culture of sober socialisation is far superior in this regard.