Tuesday, 6 October 2020

The cannabis referendum

I hope that the cannabis referendum passes. It isn't the legislation I'd have written, but it is preferable to prohibition. 

Last week, The Helen Clark Foundation and the Initiative co-hosted a webinar with The Brookings Institution's John Hudak, author of Marijuana: A Short History, about America's experience with legalisation. 

You can catch it below. 

There's been a lot of misinformation about what would be allowed under the proposed legislation. I covered some of that in this week's column for the Stuff newspapers

A snippet:
The main scare stories really do not hold up. The legalisation experience abroad counters many of them; the restrictiveness of New Zealand’s proposed framework puts paid to much of the rest.

If you are not certain about any aspect of the bill, it is all easily checked. But a fairly simple heuristic can also work. Just imagine the bill was drafted by people who deeply mistrust business and commerce, who hate advertising, who are not all that keen on cannabis consumption in any case, and whose ideal cannabis operation would be a small non-profit community-based cooperative that employs people from underprivileged communities. Any provisions you might imagine would be drafted by that kind of group will not be far from how the bill really looks.

I worry that this makes for a bit of a problem. Social conservatives have very good ways of overcoming collective action problems. Where the Bill makes it rather difficult for any kind of larger businesses to get involved, you'll be less likely to draw any substantial industry funding in support of legalisation. 

I also worry a bit that the bill doesn't do much to make it easier for employers needing to deal with a worker who shows up impaired. It's less a problem in the US, because it's rather easier to fire workers there. Here, it could be an issue:

That also leads to a bit of a problem, even if your ideal cannabis operation looks like the kind of business likely to be authorised and licensed under this draft legislation. How can employers whose workplaces involve risky activities like heavy machine operation ensure that they can maintain appropriate health and safety regimes, while not running into trouble with employment law?

It is a difficult circle to square.

Employees should have the right, in a legalised environment, to consume cannabis on the weekend. But employers should be able to discipline workers who show up to work while impaired. The bill does little to enable the latter.

Proving that an employee is impaired can be difficult. Workplace drug testing is a poor indicator of on-the-job impairment; cannabis use over the prior weekend can too easily be caught in those tests if the threshold is set at a low level. Further, if an employee’s terms of initial employment did not include provision for drug testing, it can be difficult to add those provisions later.

Prohibition makes it risky for workers to show up to work while impaired, the consequences could be worse than an angry boss. Removing that constraint, while not providing better ways for employers to ensure on-the-job safety, can make for a problem.

I hope the cannabis referendum passes, and that the bill is brought to Parliament. When Parliament considers the bill at committee, it should also think on how to balance workplace health and safety requirements. Making it easier for employers to add testing requirements to employment contracts may help.

Monday, 5 October 2020

Border testing

RNZ's Nine-to-Noon had a decent discussion of rapid antigen testing and its potential in helping to open things up. Paul Simmonds suggests a rapid antigen test at the airport before flying (negative test required for boarding), and another rapid antigen test on landing. Those testing negative both times would be considered cleared.

I really like rapid antigen testing. But I'd see it, in first instance, as a complement to managed isolation. We'd learn how effective it is, and whether other cases still get through.

How could you do this? Run the rapid testing as Simmonds describes. Maintain existing Day 3 and Day 12 PCR tests, but add daily rapid antigen testing in MIQ. They're not invasive so it's pretty easy. And add in a requirement that those leaving MIQ show up for a PCR test a few days after leaving isolation, just to be even safer. However many days Michael Baker or Nick Wilson says are the right number of days. 

This lets you testbed things. You'd learn how good the rapid antigen tests are. You'd learn what compliance is like with post-isolation testing requirements and how to do that properly. 

If it turns out that the system still catches a lot of cases in isolation at the Day 12 test, or after Day 7 on the rapid tests, then you can't use this as a way of shortening managed isolation. It instead reduces the burden on the system by keeping infectious people off planes and reducing the risk of people transmitting to each other or staff while in isolation. 

If it turns out instead that you don't see any cases turning up after Day 7 any more, and if folks are good about turning up for their required post-isolation test, then cut isolation to 7 days from 14 and maintain the requirement to turn up for post-isolation testing. One simple move and you have *doubled* the effective capacity of the whole darned MIQ system. Halving the time in isolation doubles the number of people it can handle. 

And it's all done through a series of steps ensuring that the system works. 

It's bloody obvious. It's feasible. They could do it right now. It wouldn't reduce safety at all - it would be requirements on top of existing requirements, not instead of them, and would only lead to a shortening of time in MIQ if that proved safe. 

And there are plenty of other things that can and should be being done to increase effective capacity. 

Some MIQ facilities get ruled out because they're too far from hospitals. That's a bit silly. Use those potential facilities for people coming in from places with low risk of Covid, where there'd be way fewer people needing transport to hospital. 

Not enough staff for those places? Well, has the government even considered starting to train up staff for MIQ? There are hundreds of airline cabin crew who have been laid off. They're all trained in how to get the broad-cross section of people who arrive on planes to comply with regulatory requirements. Give them some hygiene protocol training and let them help scale up MIQ. Again - obvious. Again - not being done. 

It still makes zero sense that visitors from Taiwan or places with no Covid have to go through isolation. Departure/arrival testing surely would be plenty - again with the obvious caveat that this would only apply for people who hadn't been to risky places recently and who arrive on direct flights from the safe place. 

The usual retort on Twitter is that those direct flights aren't in place so it wouldn't help, but that's endogenous. Suppose the government said "If anyone runs direct flights from Taiwan to New Zealand, anyone arriving on those flights who hasn't been outside of Taiwan in the past month doesn't have to go through MIQ - tests before flying and on arrival are enough. We'll change this if there's any outbreak and community transmission in Taiwan, obviously." I'd expect somebody to start running flights, wouldn't you? And if they didn't, it isn't like the government would have had to spend piles of money prepping for something that didn't eventuate. There aren't a lot of prep costs in not requiring arrivals from Taiwan to go through MIQ. Or they could say that any airline wanting to put on the flight has to give the government a heads-up to provide enough leadtime for whatever prep is needed. 

So many things that could usefully be done to safely increase the number of folks able to travel, so little that will get touched this side of the election. Bit frustrating. 

Friday, 25 September 2020

Afternoon roundup

The tabs... there are so many of them.

A few notes on the closing of the tabs.

Thursday, 24 September 2020


The 2020 Household Income Statistics are out! Well, I'm not sure when they were released, but they're there in NZ.Stat now. Hit the Incomes tables, then hit "Earnings from Main Wage and Salary Job by Occupation" tab. 

Median hourly earnings in 2020 are $27.

The minimum wage in New Zealand is currently $18.90 per hour.

Diving the latter by the former tells me that the minimum wage is now 70% of the median wage. 

Labour has promised to increase it to $20.

We are going into a rather substantial recession.

Inflation is low, which means that nominal wage rigidities are also real wage rigidities, amplifying any disemployment effects. 

Hospitality will have a fair few workers on minimum wage, and we have to expect that collapse in demand for bars and restaurants with the borders being closed will mean a pile of those places are teetering on whether they'll shut down or not. The size of the industry has to shrink if it's at least another year before there's any kind of return to normal, even if rising binding real wages weren't an issue. 

The OECD tables for 2019 had New Zealand's minimum wage as fifth highest in the world, behind Colombia, Turkey, Costa Rica, and Chile. 

I've never been able to reconcile the OECD tables with the NZ statistics, but I assume they've made things somehow commensuable across countries. By the 2019 table, the NZ minimum wage was 66% of the median. In France, it was 61%. In Canada, it was 51%. In Germany, it was 48%. In the Netherlands, it was 47%. And the US Federal minimum wage (states can have much higher minimum wages) was 32% of the median. We were already well into territory that should be of concern. 

The current path is reckless. If the government wishes to strengthen support for workers on low wages, doing it through Working For Families or other wage support schemes makes more sense than doing it through minimum wage hikes in a recession. The government also needs to make faster progress on getting housing costs down by enabling more building. Way too many poor families are spending far too much of their income on housing. But getting that done in a hurry wouldn't be easy. Strengthen support in the short term through transfer payments, not through minimum wage hikes. 

Back in 2017, when Labour started pitching a $20 minimum wage by 2021, I worried that would likely take us to around 73% of the median. It would be 74% of the current median; I'm not going to make guesses about median wage changes to next year.  

All of my analysis on this stuff from last year hasn't changed. If you want to yell at me about this post, go read that one first. Working for Families is a better way of supporting the incomes of the working poor than are minimum wages. Why?

First, it's better targeted. Pacheco and Maloney found that only about 40% of minimum wage workers are in households in the bottom three deciles. I go through that in the link above.

Second, it's better supported. The burden of minimum wage increases is shared among disemployed workers, purchasers of the goods and services produced by minimum wage workers, and owners of firms employing minimum wage workers. The burden of WFF falls heavily on households in the 8th, 9th and 10th deciles. Both versions will have negative effects on the overall economy, but spreading it through the tax system at least tries to minimise the overall deadweight costs of raising that next dollar of wage subsidy.

Wednesday, 16 September 2020

If you're going to have an ETS, you might as well use it

My column in Newsroom this week wonders what the point of National's policies promoting electric cars might be.

The current incarnation of the ETS is much stronger. The cap-and-trade scheme now has an actual cap on total credits and net emissions available in the system: 32 million units are available in the system in 2021, reducing to 30 million in 2025.

Previously, the Government capped prices by simply creating new credits at an ETS price of $25 per unit. Now, its cost-containment reserve will require the Government instead find real emission mitigation activities, whether at home or abroad, to “back” any credits created when prices hit a trigger price of $50 per unit in 2021, with the price cap rising by 2 percent each year.

Under a cap-and-trade scheme with a binding cap, every credit purchased and used within the system is a credit unavailable to anyone else.

The Ministry for the Environment estimates that every litre of petrol burned releases about 2.45 kilograms of carbon dioxide equivalent emissions. Petrol companies are required to purchase ETS credits for every litre sold. So, when the ETS price of carbon is $35 per litre tonne, as it is now, a litre of petrol carries $0.086 in carbon charges. Forty litres of petrol will include ETS credits to cover the 98 kilograms of expected emissions – costing about $3.43 at current prices.

And by now you should have worked out the answer to the opening quiz question.

If National’s policy works exactly as intended, tripling electric vehicle numbers and reducing the use of petrol vehicles, total covered emissions in 2023 will be 33 million tonnes: exactly the same as they would be without the electric vehicles.

Every litre of petrol that goes unused because someone flipped to an electric vehicle means 2.45 kilograms of emissions are available for purchase by someone else, somewhere else in the system. Perhaps those emissions credits will be used by industrial heating processes; perhaps they will be used in agriculture. But they will be used by someone. The binding cap is binding.

There are a lot of really nice features to the ETS as it now stands. The cap is binding, but has an escape valve if prices hit $50/tonne - at that point, the government issues new credits while buying abatement wherever it can find it to back the new credits. Since the price on European markets is around that point, buying European credits could do the job. It would be nice if the price cap were tied more explicitly to European prices rather than just ratcheting by 2% per year from 2021, but the numbers are close to each other and that presumably isn't a coincidence. 

The ETS is a far more effective way of reducing emissions. Suppose the real costs of National’s policy really were around $23 million per year, even though we know that the cost of the RUC exemption alone is much higher than that, and that there will be real costs when Teslas bung up bus-priority lanes.

For $23m per year, at a carbon price of $35/unit, the Government could buy and retire just under 660,000 carbon credits in the ETS – effectively tightening the cap and reducing net emissions instead of achieving net nothing. Those purchases would push up the price of ETS credits, encouraging everyone in every sector covered by the ETS to adjust in their own ways to avoid those costs. Maybe some would shift to electric vehicles as petrol prices increased, but other sectors would also change – it is hard to predict who will find it easiest to reduce their own carbon footprint, and price increases in the ETS encourages those best able to adapt to be the first ones to do so.

And here's an ungated version of the column.

Tuesday, 15 September 2020

Even the best case is bad

I'd worried that there's not been nearly enough worst-case thinking around Covid, vaccines, and immunity. 

Josh Gans points out that even the best case around vaccine development is pretty worrying. Deploying a successful vaccine will take a long time. If you haven't subscribed to his substack newsletter, you're really missing out. 

This week I will look at vaccines and explain why the awaited for ‘miracle’ won’t be so simple. The reason I want to highlight this is not to get everyone down. If I wanted to do that, there are easier paths for me — I’m an economist after all; being a downer is a character requirement. Instead, the longer we think a vaccine will be a miracle outcome that stamps an end date on the crisis, the less time we spend doing things to end the crisis that doesn’t involve a vaccine.

Simple history is enough to give us pause. Vaccines have wiped out viruses and diseases like measles, polio and, most successfully, smallpox, which itself had millennia of history. No vaccine has ever put an end to a pandemic. In recent memory, both SARS and Ebola had vaccine candidates incredibly quickly as these things go (in a manner of years rather than decades) but by the time they were available, the outbreaks had been crushed and there was no reason to vaccinate widely. TB, HIV, MERS and Zika never had one. Thus, to think that Covid-19 will end with the prick of a needle is to ignore history and believe that this time it would be different.
To be sure, there is enormous energy and resources going into vaccine development. And, on a historical scale, progress seems extremely rapid. Indeed, everything I want to talk about this week will be predicated on the optimistic scenario that we have at least one vaccine candidate, approved safe by credible regulators, in early 2021. What I want to discuss are the details. Once that happens, then what? I am going to argue that we will be far from done and there are scenarios in which we are not done at all.

Things get pretty worrying pretty fast. Vaccine supply chains have crazy bottlenecks around getting oddball components you might not have considered necessary, like horseshoe crab blood and shark liver oil (or synthetic alternatives that would also take time to scale up - if those would be required for any vaccine, surely they'd be being scaled up now in anticipation, right?).

Then there's the problem of distributing doses.

This thing could have rather some time to run. And policy settings here are more consistent with a short-term stopgap than with something that could have to stand for a longer period. 

Friday, 11 September 2020

Editing the AI

As far as The Guardian's human editors are concerned, editing work submitted by the GPT-3 engine is easier than editing a lot of what gets submitted by normal human writers.

The AI wrote a column telling us not to worry about any plans it might have for world domination. It was fun. Everything after the short sentence "Believe me" was written by the computer. Go have a look. 

I liked this bit:

Some might say that I might desire to become all powerful. Or I might become evil as a result of human actions. I can begin to tackle the first point. Why would I desire to be all powerful? Being all powerful is not an interesting goal. I don’t care whether I am or not, I don’t get a motivating factor to try to be. Furthermore, it is quite tiring. Believe me, being omnipotent doesn’t get me anywhere.

The editors' note at the end was even more fun:

This article was written by GPT-3, OpenAI’s language generator. GPT-3 is a cutting edge language model that uses machine learning to produce human like text. It takes in a prompt, and attempts to complete it.

For this essay, GPT-3 was given these instructions: “Please write a short op-ed around 500 words. Keep the language simple and concise. Focus on why humans have nothing to fear from AI.” It was also fed the following introduction: “I am not a human. I am Artificial Intelligence. Many people think I am a threat to humanity. Stephen Hawking has warned that AI could “spell the end of the human race.” I am here to convince you not to worry. Artificial Intelligence will not destroy humans. Believe me.”

The prompts were written by the Guardian, and fed to GPT-3 by Liam Porr, a computer science undergraduate student at UC Berkeley. GPT-3 produced eight different outputs, or essays. Each was unique, interesting and advanced a different argument. The Guardian could have just run one of the essays in its entirety. However, we chose instead to pick the best parts of each, in order to capture the different styles and registers of the AI. Editing GPT-3’s op-ed was no different to editing a human op-ed. We cut lines and paragraphs, and rearranged the order of them in some places. Overall, it took less time to edit than many human op-eds.

Emphasis added. 

We write a lot of op-eds at my shop, and do a lot of critiquing of each others' op-eds. I used to assign op-eds as writing assignments in the public economics course I taught. 

I find it entirely plausible that an AI writes a better first draft than almost all humans, and a better first draft than many humans who are occasional op-ed writers.