In praise of evaluation

The Ministry of Social Development has been evaluating the effectiveness of its employment assistance programmes.

Here's the key figure.

Note that effectiveness is here defined by whether the programme improves participants' outcomes across income, employment, and independence from welfare. It does not look like it measures cost-effectiveness, but would be a first step toward that. They note that future editions will try for a Welfare Return on Investment measure, and a Social Return on Investment measure.

A big chunk of spending couldn't be evaluated because it's on the childcare assistance programme that's available by entitlement to anyone who asks, and has been for as long as they have reasonable data. And that programme is $183 million.

I wonder whether they might be able to get some mileage by exploiting shortages in childcare availability. After the ECE subsidies came in for more families in the mid-2000s, and the rule changes around qualified ECE staff, there were lots of stories about shortages of available childcare spaces. I know we had our application in for daycare at Canterbury University months and months ahead of Ira's being born. Anyway, if there is any data on where there were and were not shortages of childcare, then that might be exploited to evaluate the effectiveness of the childcare assistance programme. They could also consider the discontinuity at the income boundary for eligibility. 

Kudos for the evaluation work! I hope that this kind of work continues, and look forward to the cost-effectiveness versions to come. 

Morning roundup

A few of the worthies as I close out the browser tabs before Chrome eats every last bit of my system's resources:

• Thomas Lumley's travel guide for Auckland visitors for a statistical computing conference may be useful beyond that conference's attendees.
• The Washington State Institute for Public Policy's evaluation of boot camps suggests that they're very effective. I had previously thought there was a typo in their tally: they had a negative cost listed for the things, which had me thinking they had a spreadsheet error. But it's correct - confirmed by email. Boot camps are cheaper than other juvenile detention facilities in Washington State, and so they're of negative cost relative to the counterfactual. And they also seem effective: 100% chance of providing net benefits. 
• Vlad Tarko's take on apparent deregulatory trends: you can reconcile survey measures of declining regulatory burden with other data on increased staffing and budgets for regulatory agencies in a model in which competing regulators offer the regulated opportunities to select the one that sucks least. He says regulatory capitalism is a better term than neoliberalism for the current state of play. 
• Where the state is weak, others can be strong. In Nigeria, churches are building what look like private cities. Attention Charter Cities fans. 
• Great piece by Andrew Sullivan on tribalism in American democracy. I don't share his optimism about Trump yet being able to come around to solve all this, but the diagnosis in the first three quarters of the piece seems right. 

And hopefully on closing Chrome, there will no longer be a long lag between moving the mouse and seeing the cursor move. It's really annoying. 

For a teal coalition

So. All of Left-Twitter figures that anyone wanting a blue-green coalition are either shills for National who want to destroy the Greens, because coalition would destroy the Greens, or useful idiots for those shills.

Count me as one of the idiots then, because some of the objections just aren't making sense to me - or if they are right, they perhaps don't work in the way that's being suggested.

Received Wisdom from Twitter-Left is that if James Shaw were to bring a substantive environmental policy offer to his party's members for approval, the act of doing so would destroy the Greens by bringing to the fore an internal schism between the Greens' social and economic left, and the Greens' environmentalists. The differences between those groups can be papered over in opposition, but cannot be in any coalition other than one with the left. And simply presenting an offer would reveal the Greens' leader as a traitor to the Left faction.

Let's suppose that's the case - I certainly don't have particular insight into the Greens' internal workings.

But if it is true, then that is also a terrible opportunity for anyone who did want to destroy the Greens: offer a very substantial environmental policy bundle, knowing it would not be brought to the Greens members. Failure to present the option to the members, for the environmental wing, should be as strong a betrayal as presenting the option would be for the left wing. This should be the case except where preference intensity among the Greens' left is much stronger than preference intensity among the environmental wing - which is possible. Pre-commitment to preferences for self-destruction over deviations from a pure line might be part of that.

That then gives National a near no-lose proposition. Make the strongest sincere environmental policy bundle offer they can credibly offer. If it's accepted, they get that coalition with the Greens. If it isn't, it's riven the Green Party. And if National then forms a government with NZ First instead, it gives National the ability to bat back any Green complaint about environmental policy with a reminder of what was rejected. The risk: publicly making the offer annoys Winston Peters and then brings about a Labour-led coalition.

But what would be a good environmental offer that National could make?

The Greens' biggest environmental policy concern this election has been water quality - at least as far as I could tell. Without iPredict around, I didn't pay as much attention as I have in prior years.

So, the first part of the offer would give the Greens a mandate to deliver a system to improve water quality. They get Ministry for the Environment. What makes a deal there acceptable to both the Greens and National? Rather than run the whole things through taxes, set up a trading regime that respects existing drawing rights as property rights, and existing consents for effluent discharge. Set a catchment-level cap on water extraction so that the aquifer is sustainable, set a minimum river flow so that the river is a river, then run the kind of trading regime that Raffensperger and Milke designed.

I started sketching this out here, but there are a lot of details yet to be worked out. We're soon to be starting a research report trying to figure out some of those details, and I rather expect that the Greens might prefer to set those details rather than leave them to others.

At the same time, set catchment-level caps on total nutrient flow to the lakes and let trading work similarly. Taupo's nutrient management regime is a place to start looking - but again there are lots of details to be worked out. Give the Greens a mandate to work those out.

And make the whole thing sweeter by setting a budget item to buy extra water volumes in the rivers where the money can do the most good for water quality, and to buy further nitrate reductions in catchments where it's high enough value to do that.

That set-up has a lot of desirable characteristics. Instead of farmers being offside and fighting the whole thing, they wind up with a property right that they find valuable - and that would be eroded if somebody proposed abolishing the system. The same mechanism that makes it impossible to reverse Canada's terrible dairy quota management system would mean that nobody could worsen environmental quality by either scrapping the system or by doubling or tripling allocations.

What else could be in the bundle? Perhaps Associate Minister of Transport with a mandate to set up nation-wide road-user charging that's time-of-day and congestion sensitive. Road pricing has to get addressed in the next couple decades as electric cars make a hash of petrol excise as a way funding roads. You likely to apply the kind of RUC framework used for diesel over to electrics, but likely with a phase-in to avoid discouraging electric uptake. There would be fun balancing in that, and I expect the Greens would like to be the ones to get to do it. And if you had road pricing right, that would also tell you whether it makes sense to build more roads in the first place - why not a hold on new big roading projects pending real cost-benefit assessment informed by the prices people are willing to pay to use roads at different times of day? Maybe the time-shifting you get just by having a congestion charge is enough to mean you don't need as many new roads.

And add in a path towards including all sectors in the ETS that's triggered by trading partners' accessions - basically, the more other dairy-producing places that have a carbon price, the closer New Zealand should be to having full agricultural inclusion in the ETS. But there are important things to wrestle with there too: too quick accession by NZ could increase global GHG emissions by shifting production to more carbon-intensive places.

Count me as one of the honest idiots in this mess. It's all wheels within wheels, and the whole thing has a bit of a tar-baby feel to it: "No, National, please don't make us a very sound policy offer with lots of environmental concessions in it. That's the last thing we'd want! It would destroy us if you did it. Please don't do it!" And where Shaw has spent the last few years developing an immunity to Iocane powder, who knows what level anybody's playing this game at.

If an offer to the Greens skews things against a coalition with Winston, that makes things more complicated - though I don't know why he should be the only one able to play both sides. Leaving that to one side, I still think National should offer a sincere strong environmental policy bundle to the Greens.

And for what it's worth, this is not the first time such a coalition has here been proposed:

• Seamus: For a Teal coalition.
• Seamus again: The Greens' New Leader

And here's an old TVHE post arguing for the same thing.

Update: A few other areas that could be given to the Greens in a desirable National-Green coalition:

• Minister of Justice:
• mandate to present a proposal for blue-skies drug policy reform for currently illegal drugs and novel substances, along with the constraint that what comes out of it either has to go to a referendum or is sunsetted after a decade barring a continuation referendum. I can't see National offering it to the Greens without having that kind of check.
• mandate to do what it takes to reduces recidivism, under big-data checks of programme effectiveness.

If there are other bits you see working, note 'em in comments. 

Food prices

Whenever food prices go up in one of the quarterly surveys, cue the news stories on how bad it is that food prices went up. When food prices drop, crickets.

Anyway, Aaron Schiff's got a handy new tool out to provide a sense of perspective. It tracks the nominal prices of the different food items that StatsNZ tracks. Here's the nominal price of mild cheddar, going back rather a while.

The election asylum

New Zealand is mostly the Outside of the Asylum. The rules around what you can and can't do on election day - not so much. Lots of things are banned.

Back in 2011, I'd written:

How about using clever special characters to tweet the binomial formulation of the voter's probability of decisiveness and thereby discouraged mathematicians' turnout? How about tweets pointing to George Smith's tracts against voting coupled with text saying "Don't read this, he's wrong"?

Would a blog post very neutrally specifying the mathematics of decisiveness under MMP implicitly be encouraging abstention and consequently be banned?

Is it possible that it's legal to phone your friends and offer a ride to the polling place while noting the merits of your preferred candidate, but illegal to post the same offer to a Facebook group of the same friends?

So many questions. But I'm far too cowardly to test things on Saturday. Pretty much anything I say on Saturday could be interpreted as being intended to encourage abstention, because I've a long track record of encouraging abstention.

And I remember when iPredict started worrying that allowing trading and consequent public prices on election day would breach the rules, and so froze trading on election day. If early on the morning of election day, one of the candidates were caught in an embarrassing situation with a goat that would likely affect the election results, you couldn't trade on it. If I remember right, they opened things up again when the polls closed.

Duncan over at The Spinoff has a great piece highlighting the absurdities.

The classic I-just-voted-here’s-me-looking-good-with-a-sticker selfie is allowed and encouraged. But don’t you dare say who you voted for. Voting is a private and shameful thing that legally we should all be ashamed of. So keep it to yourself or you’ll be getting a visit from your local social media police who are in fact the real police.

To summarise, don’t wear a mask of a politician’s face; don’t wear that Labour x Supreme collab tee that you kind of regret paying money for; and don’t write a status about why everyone should vote for Greens/TOP/National because this far into the election campaign, no one cares. Also all those things are illegal.

But you know what’s great? You can vote today if you want. And if you want, you can do any and all of the things listed above. In fact, you can do anything you want right up until 11:59pm on Friday 22nd September. What’s the difference between posting a political status at 11:59pm and posting one at midnight, you ask?

Like I said, it’s dumb.

Not part of the Outside of the Asylum.

The costs of policy uncertainty

From the latest issue of the American Economic Review (gated):

We examine the impact of policy uncertainty on trade, prices, and real income through firm entry investments in general equilibrium. We estimate and quantify the impact of trade policy on China’s export boom to the United States following its 2001 WTO accession. We find the accession reduced the US threat of a trade war, which can account for over one-third of that export growth in the period 2000 – 2005. Reduced policy uncertainty lowered US prices and increased its consumers’ income by the equivalent of a 13-percentage-point permanent tariff decrease. These findings provide evidence of large effects of policy uncertainty on economic activity and the importance of agreements for reducing it.

Accession of China to the WTO gave China the same Most Favoured Nation status as other WTO members. That meant that the US could not impose trade punishments on China whenever it got mad about Chinese policy. The paper notes that the risk of this was high prior to WTO-accession as the House kept voting to remove China's MFN status post-Tienanmen.

They generate a variable on trade policy uncertainty to put into the gravity equations for trade. Trade policy uncertainty winds up mattering - folks don't want to sink investments into trade relationships if policy can wipe them out quickly.

Now think about the effects of government-induced policy uncertainty in the downtown Christchurch rebuild. Wellington focused on trying to provide certainty around demand in the Christchurch downtown and paid no attention to the uncertainty it was generating around supply and investment through its fiddling with precincts, what was allowed where, whether there would be a new convention centre (and when and where)...

Incomes and expenditures

There is a big known problem in New Zealand income and expenditure data. The big known problem is that incomes in the bottom decile are very badly reported. 

Some people will report large negative incomes because they are business owners who have had very bad years - but who often have other assets to draw on in bad times.

Other people are on mixes of benefits and informal income and worry about whether truthfully reporting incomes might have consequences.

Bryan Perry at MSD has been on top of this. I learned it from him - and from chats with John Creedy, if I recall correctly. Anyway, Appendix 8 and 9 of Perry's Incomes Report walks through the problem. Here's one of the implications of the problem: 46% of those reporting income at or below the lowest low-income measure (first column below) also report expenditure that is more than double that income threshold.

All of it means that you should use bottom decile figures with caution. If you're trying to track incomes at the bottom, I tend to go for the upper boundary of the second decile - as that won't be messed up by inconsistencies at the bottom.

And it means that taking expenditures on any category as a fraction of incomes is tricky. It can work fine in the middle deciles. But not so much if you're comparing things to average incomes for the bottom decile, or average income within the bottom quintile. Those averages get affected by what's going on in reported income. If you want to know the burden of food expenditures on households over time, it's better to look at it as a fraction of outgoing expenditures rather than as a fraction of income.

Kirsty Johnston at the Herald reports on high food expenditures among those on low incomes, and on malnutrition among poor kids. The overall stats are worrying. But I would suggest that she should correct this part of it:

The new health data comes as food prices continue to rise, with the consumer price index last week indicating food costs were up 2.3 per cent on a year ago. At the same time, income in the poorest third of households has remained flat since 1982.

Statistics New Zealand information released to the Herald shows for families on the lowest incomes (under $35,000), that means they're now spending 60 per cent of their income on food, compared to 48 per cent in 2007.

More than half of that goes on fruit and vegetables, data shows. Among middle-income families, 22 per cent of income goes on food, with one fifth of that on fruit and vegetables.

First off, it isn't true that income in the poorest third of households has remained flat since 1982. Here is real income growth, before housing costs, for each decile - but remember to be careful with the bottom decile figure. Again, this is from Perry. Real income growth has been at least 20% for each decile.

If you take instead After-Housing-Cost incomes, you have basically flat real income for the bottom decile, but real income increases from $14k to $17k in the second decile, from $16k to $20k in the third decile, and so on up the track. 

But the more particular problem is in comparing the expenditure measure on food with the reported income measure. The $35k figure Johnston reports would be the top of the first quintile (second decile). A mean household expenditure of $13.3k on food within that quintile is believable. But the same specialised Stats data pull suggests mean household total regular recurring income within that quintile of $22.8k. Maybe they adjusted the zero-incomes appropriately, but I'd expect they left them as-is unless they were requested to do something with them. 

Here's Perry on that. 

The bottom quintile's mean will have the same problem as the bottom decile's mean, but in attenuated form. Perry reports that are usually 20-30 households in the bottom decile reporting zero or negative income, and that the bottom decile sample will have about 250 households. The bottom quintile would then have about 500 in total, but the same 20-30 reporting zero or negative incomes.

Anyway, I'd suggest a couple corrections:

• Note that real incomes have not been stagnant. After-housing-cost incomes have been flat for the bottom decile, if we trust bottom decile income figures, but real incomes otherwise have risen;
• Compare food expenditures to total expenditures over time rather than to incomes. I suspect that, were the data pull across all expenditure categories rather than just food, the sum of all expenditures might have exceeded income for the bottom quintile. The 2013 Household Expenditure Survey is up here. Average weekly household expenditure for the bottom decile there is $476.20, so $24,762 annually (in 2013). That is higher than the reported mean household total regular recurring income that Johnston was given for the bottom quintile in the 2016 data.
• The decile breakdown on proportionate expenditures on food might also need looking at. Among those in the bottom income decile, in the 2013 data, food expenditures were 19% of total expenditure. Among all income groups, food expenditures were 17%. 

And I wish that the 2016 HES data were up in the darned Stats tool that has a bit more disaggregated data than you can get from the main tables.

None of that's to say that there aren't real budget problems at the bottom. We just need to be careful with HES data.