Internalities, mechanisms, and booze
NZIER has released an “Insight” on internalities – specifically stating that this is one way we could justify some of the costs in the BERL alcohol report as being policy relevant. Eric Crampton discussed the release here.
What are internalities? Internalities is the name for the cost associated with not being able to “commit” to the best set of actions over time (time inconsistency). For example, on a Friday morning you might think about your weekend and decide it would be best if you only drank 2 beers that night. However, once you started drinking you suddenly decided to have more – something you undoubtably regret the next day. In this case, the inability to commit to the “optimal plan” has costs – costs that are not external but ARE policy relevant (contrary to my slogan here).
Now, both the people from BERL and NZIER that raised the internality issue are correct, and both sets of economists have far more experience and knowledge in the area of time inconsistency (I am only a macroeconomist after all). However, there is one important point for me here – internalities matter because of commitment, and so mechanisms to aid commitment can be solutions.
Why is this so important? Well when we look at the impact of a persons action on themselves it is often easier to develop mechanisms to solve this than in the case of “externalities” (where someones action influences an unrelated party).
In the case of alcohol there are a number of “pre-commitment mechanisms” avaliable to improve outcomes. If we can figure out what the private cost of establishing these mechanisms is, and we see people not use these mechanisms, then we have an upper bound on the relevant “internality” associated with commitment issue!!!
So lets ask ourselves – how costly is it to commit to not drinking too much? Some mechanisms are:
- The instash wallet guide suggests leaving your wallet at home and only bringing in a limited amount of cash (when going to town).
- Buying smaller portions of alcohol (when staying home).
- Dressing badly (when going into town).
- Committing to do something the next day which sufficiently increases the cost of drinking during the “temptation” period.
I have seen all these mechanisms in action. They are costly, but not that costly. As a result, my impression is that the policy relevant loss associated with the “internality” will be quite small.
Good post Matt.
The main concern I have with the internality argument is with testability. How do you detect an internality?
The only approach I have seen, and one I think is implicit in the BERL report, is to give the consumer an objective function (e.g. maximising longevity) and then detecting an internality by observing behaviour that is inconsistent with that objective.
But there are two reasons for such a departure in behaviour. One is internality. The other is that the consumer has a different objective! (A third possibility is that behaviour is complex and dynamic and your modeling of that behaviour doesn’t capture what in fact is consistent with that objective. A fourth is that one’s objectives change over time).
So how does one actually test for the difference?
If one sees an appropriate role for government policy to tell its citizens what their objective function should be, then it doesn’t matter: whether by internality or the wrong objective function, that behaviour needs fixing.
But if government’s role is to fix internalities only, then testing really matters. How do you do it?
@ben
That is a reason why it is often preferable to have mechanisms with voluntary opt-in.
All of your mechanisms sound excellent for casual drinkers who don’t want to have a big night. I’m unconvinced that they would be as effective for alcoholics. While I think you may be underestimating how costly addiction is to an alcoholic, I would agree that tax policy is a very crude tool for addressing alcoholism.
One obvious way to detect internalities is to look for evidence of time inconsistent behaviour. Addicts who repeatedly try and fail to quit, for instance. Where intentions differ from future actions there may be internalities at work. Of course, there are other possible reasons for time inconsistency so it’s not certain, by any means.
There is also the external regulation of internalities in the form of licenced alcohol retailers. Namely, retailers can lose their licence for selling to intoxicated people. Licenced managers can and do limit the commitment of the most dedicated lush.
Sky City set the standard for this with their traffic light guide to serving liquor. For all their faults, Sky City have led the way in the voluntary opt-in stakes.
Rauparaha: at my post, I note a very very effective precommitment mechanism for alcoholics who really do want to quit drinking entirely: disulfiram. Take the pill, and you’ll be violently ill the next time you drink. If folks who claim to want to quit nevertheless refuse this very cheap and very effective mechanism, I question their sincerity about wanting to quit.
Ben: You’re right. It’s impossible for the external observer to observe internalities.
Matt: Do note that while internalities can give some justification for counting some portion of internal costs, it gives no justification for what BERL actually did: counting ALL internal costs on an assumption of zero benefits.
@Eric Crampton
The problem with things like disulfiram is that no addict will want to take them because they WANT to drink now. If it prevents them from drinking at the present time then there’s no incentive to take it. If they didn’t presently want to drink then there’d be no inconsistency and no need for a commitment mechanism.
Most theory work using quasi-hyperbolic discounting assumes that they don’t want to drink from the next period onward. When the next period is depends on the precise specification of their utility function. Do you know of any work that empirically estimates what that might be? Knowing when t+1 starts might shed some light on whether these devices are likely to be used for precommitment.
@Rauparaha: First cut answer, there has to be some point at which the alcoholic is sober, wants to be sober tomorrow, and knows he’ll have problems in staying sober. It’s at that point he grabs the disulfiram.
If you want to argue that it’s the St Augustine problem, then how hard would it be to run a modified version of the drug such that it’s implanted today but triggered to start the drip-feed tomorrow? Or, where an appropriate concentration of the drug doesn’t build up in the system ’till a day later? I can’t imagine it would be that tough. Wouldn’t a minor amount of drug development money make a lot more sense than a generalized alcohol tax?
If you want to go even the farther step that the drinker never even wants to bind himself against drinking at t+1, then it’s not an internality problem anymore.
@Eric Crampton
“It’s at that point he grabs the disulfiram.”
Definitely. But will he then continue to take it in future when he desperately wants a drink? I don’t know the exact mechanics of the drug so I’m not certain that this is an issue, but it seems that it could be.
“how hard would it be to run a modified version of the drug such that it’s implanted today but triggered to start the drip-feed tomorrow?”
I don’t know. That’s why I said that I thought research into where t+1 is would be useful. We’d need to know when the addict wants to stop themselves from drinking so that they have a private incentive to continue taking the drug.
“Wouldn’t a minor amount of drug development money make a lot more sense than a generalized alcohol tax?”
Yes. That’s why I said “…I would agree that tax policy is a very crude tool for addressing alcoholism” above. I think that’s because the majority of drinkers are not alcoholic, which distinguished the case of alcohol. In the case of smoking, where most smokers try to quit regularly, there seems to be more of a case for a general tax.
@Matt
Does dressing badly to avoid going out include dressing up as a transvestite?
i agree there is a time inconsistency, but isn’t it up to the individual to discount those costs back to present value, the same as any investment, and compare that to the current benefits now?
for example, if I decide to drink a few more because i’m having a good time. I make a judgement that the cost of my hangover tomorrow, discounted back to tonight,is less than the extra benefit I get from having a few more now with my friends because this party is better than I expected.
The only way I can see a time inconsistency problem occurring in this case is if you assume I am irrational (which is not too unreasonable depending on my level of drunkenness at the time I make a decision to drink more).
So long as your initial commitment, is at a point where you are still rational, there is no time inconsistency problem.
What if my internality is that I under consume? Is my consumption to be subsidised? After all I under consume Rolls-Royces because I can’t not commit to saving enough to buy one.
@steve
You’re right, it is up to the individual to discount back. The reason for the time inconsistency is that the discount rate isn’t constant, so preferences change over time. Check out ‘hyperbolic discounting’ on Wiki 🙂
@Paul Walker
Under-saving is definitely a big issue for people researching hyperbolic discounting. Obviously the proliferation of savings accounts with penalties for withdrawals are a response to these sort of concerns. The government’s response so far seems to be Kiwisaver but, as NZIER points out, the heterogeneity among savers makes it a difficult target for subsidies just as alcohol is a difficult target for internality taxation.
@Paul Walker
Also I am not sure that the lack of buying a Rolls-Royce is a time-inconsistency issue, it is probably an issue of opportunity cost – namely, I am sure you are spending your money on products that offer you a higher lifetime return.
Now, we could tell which of these it is by offering you a voluntary commitment device which you can opt in to in order to buy your Rolls if you don’t trust yourself to save money sitting in your account. Like, a lay-by, or a hire-purchase agreement, or an employee retirement scheme.
If you take the scheme, then we know we’ve got a time inconsistency issue. If you don’t, then we know that either:
1) It is just an income issue and so doesn’t matter,
2) It is a time inconsistency issue, but the welfare loss is lower than the cost of joining the voluntary opt-in scheme.
@goonix
Possibly – it depends on my preference set.
Then where is the policy issue? Why is there any basis for government intervention here?
@Paul Walker
I never said there was. My post was solely saying that if there was a welfare loss stemming from an internality it could potentially be quantified in the face of the rejection of a voluntary commitment mechanism.
Now if it turns out the government can provide a cheaper commitment mechanism a policy intervention could help. And in the extreme case where there is only ridiculously expensive commitment mechanisms even a grossly misapplied tax could help.
I’ve just read the NZIER piece. I think it says a couple of ludicrous things.
Sometimes private commitment mechanisms are not available and that is where government intervention can help people to keep their promises to themselves.
I think this is generally untrue. Alcoholics have support groups and can buy therapy and read books and (I think) receive medication. Similar for smokers and gamblers. There is a demand for support and, entirely unsurprisingly, people have organised a supply of it.
NZIER is postulating a market failure of some kind but it isn’t clear a) why the array of supports for alcoholics and smokers and gamblers doesn’t count, or b) what is preventing people and organisations privately arranging whatever is missing. Does NZIER have any reason to expect a market failure? What barrier to private commitment do they see that requires government intervention? None that I can see.
The problem is that drinkers who keep trying to drink less but never do are probably not doing the best thing for themselves. That alcoholics often admit they would like to drink less in future just reinforces that notion. However, because they are unable to force themselves to drink less, they end up drinking far more than they want to.
If this behaviour is the product of hyperbolic discounting, why is this discounting ignored for the purpose of policy design? Why is a consumer’s undiscounted preferences considered policy-relevant when their utility and decisionmaking is calculated with discounting? It would be nice if NZIER could put forward a view on this – it is fundamental to the argument.
What’s got the booze has to do. 🙂
I am on your side. Very nice article.
Everyone commits internalities. Sometimes they are hard to control and the situation dictates your actions.