Diversity in the Boardroom

I just loved Isabelle Solal and Kaisa Snellman's piece over in Organisation Science. 

They look at the effect of the gender of a company board appointment on subsequent sharemarket performance (Tobin's q). They summarise the existing literature, concluding (in line with the rest of the academic literature) that there is no particular effect of boardroom gender on company performance.

But then they do something rather neat. They look at how things vary based on company rankings on the KLD corporate social performance index. That index ranks companies by their commitment to corporate social responsibility objectives, and includes an index ranking a company's commitment to gender diversity.

They find that the gender of a board appointee has no effect on sharemarket performance, among firms with a low ranking on the KLD index. But firms with a high ranking on the KLD index, the appointment of an additional female director reduces Tobin's q. They reason that investors infer different things in the two cases:

We examine investor responses to board diversity and highlight a previously unexplored mechanism to explain negative market reactions to senior female appointments. Drawing on signaling theory, we propose that an increase in board diversity leads investors to update their beliefs about firm preferences. Specifically, we argue that a gender-diverse board is interpreted as revealing a preference for diversity and a weaker commitment to shareholder value. Consequently, firms with more female directors will be penalized. We test our argument using 14 years of panel data on U.S. public firms. We find that firms that increase board diversity suffer a decrease in market value and that this effect is amplified for firms that have received higher ratings for their diversity practices across the organization. These results suggest that observers respond to the presence of female leaders not simply on their own merit but as broader cues of firm preferences and that firms may counteract any potential signaling effect through careful framing.

I liked the piece enough that I made it my column over in the Fairfax papers today.

It is too easy to convince ourselves of things that are not true.

We all do it and it is hard to avoid. Some beliefs, from religion to sport, are just comforting. And when some comforting beliefs are very popular, being the one to say otherwise can be a bit risky.

But, at least in business, believing things that are not so can eventually get you into trouble. Businesses face the market test. And, for publicly listed companies, share prices can provide a quick signal that you may have made a bad decision.


The latest issue of Organisation Science, a top academic management journal, provides a wonderful case study.

...

And that helps us to understand why investors might respond in the way that Solal and Snellman discovered.

It may now be a bit passé to say it but improving shareholder value is a company's ultimate responsibility. There are plenty of wonderful things that companies do to help the communities they serve, but shareholder value is a hard bottom line. It is part of the market test.

Companies that keep that sharp focus on the bottom-line will make board appointments that they think will do the most to improve the company's performance.

When investors have little worry that the latest board appointment was made for any reason other than improving the company's performance, the gender of the latest board appointment has no effect on share prices. Investors simply expect that the best candidate was chosen.

But among firms highly rated for their commitment to gender diversity, an additional female appointment to the board reduced the firm's market value relative to the value of the company's physical assets (Tobin's q) by almost 6 per cent.

Solal and Snellman suggest that investors infer, in those cases, that the company is less worried about shareholder value than about other objectives. And that can be a worry if you care about your portfolio's returns.

While the literature showing no particular effect of boardroom gender on corporate performance is rather substantial, Solal and Snellman's results are still just one study. Others could yet overturn it.

But it does provide a bit of a warning for companies that put substantial effort into advertising their corporate social responsibility credentials.

If Solal and Snellman are right, then investors can be quick to infer that companies demonstrating their commitment to popular but mistaken beliefs have taken their eye off the ball.

The market test matters. And mistaken beliefs can be costly. 

The piece has not met with universal acclaim.

Over at LinkedIn, Sky director Rob Campbell writes:

I’m not sure whether Dr Crampton has ever worked in a listed company or been a director of one. As an academic he would not get much traction by grabbing one study in a quite active literature, even one based on meta analysis, and building an argument on it. The academic world has its own rigour.

But back in the listed corporate world the thinking runs a good deal deeper than “let’s appoint a woman or two and see if it lifts the share price”. We are all looking to improve the performance levels of our boards and management and it would be foolhardy to ignore the negative impacts of mono gender, mono ethnicity, mono experience on that. Not least because our shareholders (rightly) demand that we make the shift. So sell your shares in any business I’m involved with Dr Crampton, we will keep searching for better solutions.

It's a bit funny really. I note how desperately people want to believe something to be true that isn't true, and pointed to the metastudies of dozens of prior studies that show that there is no effect of boardroom gender on company performance, and that's the reply.

I hadn't known about the KLD index before. Someone's likely already done this study, but if it hasn't been done, it would be a lot of fun.

I'd be keen to know whether firms subject to greater regulatory risk show up differently in the KLD index. Maximising shareholder value, for some firms, also involves making sure that the regulators have warm feelings towards a firm because the regulatory risk is substantial. Anticipating the social preferences of those regulators and working to demonstrate shared values could be a way of buying friendlier relationships with the regulators.

Different firms and industries will face different regulatory risks from different administrations. There'd then be potential for identifying effects by looking at the composition of congressional oversight boards, or changes in the Presidency as the regulatory agencies will be affected by the tone set in the Executive, or firms that face regulatory risks in different states as well as federally.

You might think that, in general, firms that depend more on friendly relations with regulators will invest more in CSR efforts where doing so buys friendlier relations. It would be neat to see whether that's the case. One does hear incredibly interesting anecdotes consistent with that kind of story, but does it show up in the data?

Household wealth and housing wealth: first quintile oddities edition

In my column over at Newsroom this week, I noted one strange feature of New Zealand's household wealth statistics:

Unfortunately, data on wealth is far worse than data on income – the Government gathers a lot less data on wealth. For example, Statistics New Zealand reports that people in the least wealthy 20 percent have $1.75 in property debt for every $1 in property assets – but no bank in the country would extend a loan on that basis. It is more likely that the survey data misses some houses owned through family trusts where, for example, a 25-year-old takes over the mortgage and effective ownership of their parents’ second home held in the family trust. The mortgage payments and mortgage debt are noted in Household Economic Survey data, but the ownership may be missed. This means their parents’ net wealth will be overstated, their own net wealth will be understated, and measured wealth inequality among younger cohorts would be somewhat understated but overall measured wealth inequality would be somewhat overstated. 

There's an ungated version of the column here.

The effect wouldn't be large because there aren't many households in that situation. But it's odd. So I asked Statistics New Zealand what's up with that, wondering whether the trusts explanation might be what's going on. I'd also asked whether Stats were able to ask follow-up questions of their HES respondents to see what's going on.

Stats' Statistical Analyst Michelle Griffin helps me out:

We’ve had a look at the distribution of owner-occupied property assets and owner-occupied mortgages for those in net worth quintile one. There are about 30 more households (unweighted) with mortgages than homes, and because of this difference, you need to go further along the mortgage distribution to get to the median, resulting in a larger mortgage median and hence a larger debt to asset ratio. So it’s looking like the high ratio is due to a mix of under-reporting and distribution differences.

We’ve had a look at only those with both a home and a mortgage to see what they look like (this is excluding those with a home in a business or trust), and this brings the ratio down to $1.44. This drops even further when you take the median of each household’s individual debt to asset ratio (to $1.26).

We’ve had a quick look at the households with only a mortgage to try and see what might be happening. They appear to be a mix of households with the home in a trust or business which is assigned to someone else (which could be a situation like you’ve described below), and households which say the home is in a trust or business but then don’t mention this trust/business later on (could be partly due to misreporting or respondent burden leading to households not answering fully). I think the fact that there are so few households in net worth quintile one with owner-occupied property is making this quintile more sensitive to these issues. Unfortunately, we can’t follow up with the respondents later on to confirm these situations – although it would certainly be interesting!

So restricting things to those households in the first quintile fixes some of the problem, but we're still at a debt-to-equity ratio on housing debt well out of step with sane banking practice, let alone existing LVR rules. And while there aren't many households with mortgages without homes, it's hard to tell whether debt to asset ratios for a broader set of respondents elsewhere mightn't be wrong too. 

Michelle suggests that keeners might apply for microdata access to really drill down into what's going on. It would be a great Honours thesis for somebody wanting to get microdata experience who has a qualified supervisor.

Many thanks to Michelle and to Statistics NZ for the helpful and detailed response.

Update: An informed reader writes:

A couple of thoughts about the data on wealth:

• There is a wider problem with data for people at the top and bottom of the distribution. Surverys include lots of trade offs between clarity, comprehensiveness and accuracy. For instance survey data on the benefit population substntially under-reports income from second and third tier payments (that is all loans, one off payments, disability allowance and accommodation supplement). Since this is a relatively small part of the population, SNZ makes the reasonable decision not to overcomplicate their survey to deal with the problem, but the consequence is continuous understatement of income for many people. The same applies at teh top where it is hard for surveys to capture the many different assets people with many assets have.


• More technically there is problem using a small number of  discrete categories like quintiles to describe a continuous variable. The general point is that if someone makes a mistake they have have to go somewhere in the discrete categories, so effectively there is a cut off at the top and bottom that will skew the results. It's easy to think of this through an intuition. If a person's income is roughly in the middle, say $50 000, and they miss a zero then they drop to the bottom of the distribution ($5000) while if they add a zero by accident ($500 000) they go to the top of the distribution. However a person whose income really was $5000 who makes a mistake either stays in the same place or goes up (ie income is $500 or $50 000); while a person with income at $500000 either goes down, or stay at the top. Even if we assume the liklihood of a person making an error and its direction are independent of income, the measured income will be pushed to the extreme quintiles. "On average" it may be right, but it will overstate the dispersion. My guess is this will be even more pronounced with wealth because it is so much harder to measure and thus more likely to generate error-prone answers.


• Finally, they are using snapshots for variables that change over time. I know this is a problem with flows like income and employment, but I sometimes wonder what it means for wealth. In particular, there is an issue with measuring transitions. The following example is made up, but gives a clue what might be happening. Say some % of people selling a house to buy another spent a couple of weeks where they had two mortgages while the paper work was settled. For the individuals the cost might be a few hundred dollars - which is not a lot when you get real estate agents involved! - but two weeks is approx 4% of the year so you would expect your snapshot survey to include some small % of people with double the mortgage debt relative to their asset. It is not many people, but it does not need to be because they will be a small percentage of the population but they will concentrated in decile/quintile. 

To me the deeper problem is that surveys tend to be designed to be "on average right", whereas the data is often used to make distributional statements. And Bryan Perry’s incomes report notes the problems with the extreme tails of the distribution.

Fun studies for someone else to do: contraceptive access

The Dom points to a natural experiment waiting to be studied:

"While current dispensing restrictions limit how often a person can collect their Levlen ED prescription, this should not stop patients from accessing their medicine or taking it as directed.

"We thank pharmacies and wholesalers for their assistance in helping to manage this stock issue and apologise for any inconvenience," Fitt explained.

Concerns about the availability of Levlen in New Zealand were raised earlier this month.

Recently, pharmacies had been asked to dispense the pill in one-month courses, rather than the three-month courses usually given out.

But on Friday that request will become mandatory as officials step up efforts to maintain sufficient supply until the medication was restocked in late March.

So a fun Masters thesis for somebody: use regional variation in the timing here to check the extent to which the fixed costs of going to the dispensary affects both the likelihood of scripts being filled and unplanned pregnancies. Prescriptions like this should be in Pharmac data accessible in IDI.

I'm not sure what you could do with this second natural experiment though, noted in the same story:

In April, medical professionals were asked to ration prescriptions of Durex condoms for men with large penises, as increased demand contributed to a global shortage.

Durex Confidence 56mm condoms and Gold Knight Larger were the Pharmac-funded brands affected by dwindling supply.

At the time, Pharmac said the Durex condoms would be temporarily unavailable in New Zealand, but the brand's five other varieties in differing sizes were available.

In July, a disagreement between the brand's owner and its Indian partner cut up to 60 per cent of supply.

Of course, condoms remain accessible (at a somewhat higher cost) at the supermarket; it's just the subsidised ones that are short. No non-prescription options for birth control pills.

The Stuff comments thread is more amusing than most.

The quake vote

Were I still in the business of proposing future honours research projects at Canterbury, I'd be pitching this one. Anybody please feel free to run with it: I'd love to know how it goes.

The National Party decisively won the 2014 election, taking the Party Vote in all but the poorest of Christchurch suburbs. While National polled well even in Christchurch's more earthquake-affected Eastern suburbs, National polled well everywhere. 

Take 2008, 2011, and 2014 polling-place election data. Combine it with Census data for the meshblocks forming the natural catchment for each polling place. Predict National's 2008 party vote on the basis of underlying demographic characteristics. See whether changes in those characteristics predict changes in local-level National Party support in 2011 and 2014 for areas outside of Christchurch. Then, add in CERA data on land earthquake status in Christchurch as additional explanatory variable for Christchurch in 2011 and 2014. Christchurch's poorer neighbourhoods were also its more quake-affected ones. Was voting in quake-affected poorer Christchurch neighbourhoods substantially different from voting in poorer Auckland or Wellington neighbourhoods?

Can we reasonably interpret Christchurch's pro-National vote as endorsement of the handling of the rebuild? If more quake-affected polling places returned a higher-than-otherwise-expected National vote, then perhaps we can. If National fared more poorly than expected in the more quake-affected polling places, then perhaps not. You could also exploit split-ticket voting here. Informal Twitter discussion after the election suggested Labour voters would support their local Labour MP while giving their Party Vote to National as expression of disapproval with the Labour leadership or of support for National's overall policy agenda. Was split-ticket voting in Christchurch different from split-ticket voting elsewhere, correcting for the pre-quake Christchurch split-ticket vote in 2008? Was Gerry Brownlee's electorate vote stronger or weaker than we might expect for a government Minister?

If it is possible to get EQC and CERA data on the proportion of completed earthquake claims in each meshblock, that would be a fine additional variable. Finally, if you have time, give some thought to how migration from 2008 to 2014 might have affected results: would the out-migration of the worst-affected residents introduce attenuation bias?

I would love to know how EQC completion rates affected National's Party and Electorate votes.

An Uber experiment

Reason asks an excellent question: is Uber helping to cut drink driving rates in the US? When it's cheaper and easier to get a cab, maybe more people will do it instead of chancing a drive home when they shouldn't.

Reason points to some preliminary work on the topic done by Uber, looking at Uber's entry into Seattle with San Francisco as control. The work's suggestive, but hardly conclusive - especially when there are dozens of cities that could have been chosen as treatment or control.

So, here's the Masters thesis for somebody. Get a city-level panel of DUI rates and of taxi fares. As a first step, just run fixed effects with Uber entry and exit dates. Then run a few more complicated versions, like matching cities by probability of Uber entry based on city characteristics and taxi fares (comparing those of similar ex ante probabilities with different ex post resolution). Or exploit the city-by-city variation in pre- and post-Uber prices. There's loads of potential here, if city panel data on DUI arrests is available. There's loads of wonderful not-related-to-DUI variation in whether cities allow Uber or not making for something close to a natural experiment, though that will be less true if Uber's started using DUI-effects in its lobbying.

In other news, I had my first ride in an Uber cab in Auckland a couple of weeks ago. The cabbie was very enthusiastic about telling me all about it: he's on an hourly rate, but flips to commission when it's busy enough (says he gets 80% of the take). It makes sense that Uber would put cabbies in new markets on hourly to make sure that there's enough supply there when new customers come in.

If you sign up with Uber on code ericc294, you get $10 off your first ride and I get $10 in credit too.

Occupational licensing: NZ Edition

David Smith points out that New Zealand isn't as pure as I'd like. In comments over on last week's post on New Zealand's generally "less stupid" policy stance, he wrote:

Eric,I'm afraid your comment on NZ occupational licensing is not correct. You can see the official list here.This just refers to regulations that have a specialist body. (Even so, it includes Real estate agents!) You will find many interesting anomalies in the way these organisations work. For instance, of particular interest to a Cantabrian would be the building regulatory bodies that have set up a system that means the average age of an apprentice is 28 (twenty eight). There is no overview whatsoever of how these bodies go about their regulatory task. Such overview is not about specialist knowledge, but simply asking those organisations how they implement legal obligations to protect consumers. e.g. I do not know how to be a dentist, but they could be obliged to show how they assess patient safety and the criteria they use to decide whether or not a procedure could safely be done by a non-professional. No regulatory body in New Zealand is asked to do this.However, this is scratching the surface. The two tricks in New Zealand are highly restrictive "health and safety" laws that in practice exclude many low skilled people from jobs because they have not done low value courses costing a few hundred dollars. Low cost to you and me, high cost to a kid on welfare if they have no guarantee of a job. The other restrictions are demanding academic qualifications. Three decades ago it was possible to be an an academic with a masters degree. Now a PhD from an overseas university is needed. That's what I call a restrictive practice!

David Smith

Here's the list provided by Immigration NZ:

SM19.5 Occupations requiring registration

In New Zealand registration is required by law in order to undertake employment as one of the following:

Architect Barrister Barrister and solicitor Cable jointer Chiropractor Clinical dental technician Clinical dental therapist Dental hygienist Dental technician Dental therapist Dentist Dietitian Dispensing optician Electrician Electrical appliance serviceperson Electrical engineer Electrical inspector Electrical installer Electrical service technician Financial adviser

Immigration adviser Line mechanic Medical laboratory scientist/technologist Medical laboratory technician Medical practitioner Medical radiation technologist Nurses and midwives Occupational therapist Optometrist Osteopath Pharmacist Physiotherapist Plumber, gasfitter and drainlayer Podiatrist Psychologist Real estate agent Cadastral (land title) surveyor Teacher Veterinarian

I last week also discussed how doctor licencing in New Zealand works to support what's effectively a cartel.

I don't know how binding a barrier any of these licensing requirements prove in practice. It would be a pretty interesting research project for someone like the New Zealand Initiative to find out. But in all of these with which I've had any experience as consumer, the barriers here seem lower than those in the States, with evidence mostly coming from prices.

Here's one example. Dentists are pretty cheap here relative to the US. I typically pay about $150, including GST, for me and the two kids when we get a check up and cleaning - we have no dental insurance, and there's no kid subsidy.* This survey of NZ dentist fees says an exam and x-ray is $95 and a filling is $160. Susan had a root canal, under sedation, for about $600. Dentistry is an undergrad degree here with a registration exam after your degree. In the US, it's a graduate degree after a biology-heavy undergraduate programme. Here's the US recommended undergrad prep for entry into dental school. Raise the entry bar, you lower the number of people passing through and so hike the fees.

Dentistry still does have restricted entry: the only school in the country allowed to teach it is Otago, and they only take 54 domestic students per year. But foreign dentists are allowed to practice here; I have no sense of how onerous the examination and registration requirements are for those wishing to do so.

While I agree with David that we'd do well to have a much better sense of the scale and severity of occupational licensing here, it's also worth noting that the problem's really rather worse in the States. Colorado's list** includes acupuncturists, addiction counselors, athletic trainers, barbers, funeral home operators, massage therapists, private investigators and social workers, for example. Don't try braiding people's hair for money in Utah, or many other states. Here's the Reason Foundation's work on occupational licensing. Kleiner and Krueger estimated that 29% of employed Americans were fully licensed by the government for work in their profession; licensing brings a 14% wage premium. I wouldn't be surprised if the wage premium in New Zealand were on that order, but I would be pretty surprised if the proportion of Kiwis working under occupational licensing were over 20%.

File under future honours projects.




* There is a no-cost-to-patients dental service for kids, but we don't use it. We don't want to clog up the public system when we can afford to pay and I don't want the hassle of having different appointment times for me and for the kids. For a while, we kept getting voicemail from the government-provided dental service. They sounded pretty annoyed that we're not using their service. I'd left a message on their voicemail saying the kids are with our family dentist; that didn't stop the calls. I don't know whether our message didn't get through to them or whether their KPIs involve having all kids seeing a government-provided dentist rather than just seeing a dentist. There's also cheaper medical visits in general for those qualifying for Community Services Cards, eligibility for which is based on household income. The price I'm quoting is the full-price, no-subsidy version.

** Not trying to pick on Colorado: they just came up first on a Google search.

Youths, booze and crime

Sir Geoffrey Palmer recommends a set of anti-alcohol measures to reduce crime rates.

''We've got something like 8000 prisoners and rising, and that needs to be broken, but it takes enormous political courage to do that if you're feeling the heat of public opinion at election time.''

Sir Geoffrey, as a retired politician, was comfortable labelling prisons as ''universities for crime'' and advocating unpopular law reform.

''I would increase the price of alcohol by increasing excise tax by 50%; put up the purchase age to 20; ban advertising of alcohol on television and control advertising very carefully; regulate promotions and sponsorship; have tighter trading hours; and make it compulsory for all local authorities to have an alcohol policy,'' he said.

''Those measures would reduce the risks. Otherwise, we're going to produce more binge drinkers and addicts. I'm not a wowser but having read all the literature I am extremely worried about where we're going as a society about alcohol.''

If Palmer is right about the alcohol purchase age part, then we would expect a surge in offending by 18 and 19 year olds relative to 20-21 year olds in the period subsequent to the drop in the alcohol purchase age. Doing it up properly would require crime rates by age group over the period.

The quickly available StatsNZ figures aggregate by age group 17-19, 20-24, 25-29. If the purchase age had effects, we'd expect a jump in total sentences (which includes imprisonment, home detention, community work, and everything else) among 17-19 year olds from 2000 onwards relative to their older peers. Note that this is just a first cut: we would need to standardise by population in each age cohort to do it properly. But here's the non-standardised first cut: The left axis has the total number of sentences by age group, with the blue line showing sentences among 17-19 year olds. That series dropped from 1980 through to 1992/1993, then showed a slow rise through 2008/2009, then a sharp drop. The number of sentences provided to those aged 17-19 is lower now than at any time since 1980.

The green line shows the ratio of sentences among 17-19 year olds to those among 20-24 year olds. That series dropped from 0.94 in 1981/1982 to a minimum of 0.6 in 1992/1993, rising to 0.91 in 2006/2007, then falling to 0.55 in 2012/2013. 17-19 year olds are now receiving a smaller fraction of sentences, relative to those given to 20-24 year olds, than any time since 1980. The purple line gives a similar ratio, but with 25-29 year olds as comparison group. Again, we have a sharp drop from 1980 through the early 1990s, a slow rise from 1994 through 2002, a sharper rise from 2002 through 2006, then a rapid fall through 2012/2013. The ratio this year is, again, lower than in any prior year going back through 1980.

I can't rule out demographic changes driving changes in the aggregates. I have a harder time imagining demographic changes driving changes in the ratios. You could maybe build a case for a break in sentencing relative to 25-29 year olds starting around the time of the change in the alcohol purchase age, but the drop since 2008 has been pretty remarkable.

InfoShare provides estimates of population by age going back through 1991. Assigning 1991 data to the 1991/1992 fiscal year sentencing figures, I get the following sentencing rates by age group for 1991 to present. Again, I have a hard time seeing any kind of break at the year 2000 for 17-19 year olds. There's a big spike upwards from 2005 through 2009, followed by a sharp decline. Sentencing rates among 17-19 year olds are lower now than they've been since 1991.

I'll have to see whether I can get better age-disaggregated data in this, the season of thinking about future honours projects. It's certainly fodder for a quick diff-in-diff study. And it looks like something else was going on in crime policy in the mid 2000s that will throw in a nice confound for a student to puzzle out.

Are the All Black Selectors biased against Canterbury?

This is a guest post by Scott Brooker.

Does the Canterbury rugby team's recent run of success suggest that the All Black selectors are biased against Canterbury players? I don't know. But it's certainly an interesting question. As a result of their away victory in the final against Wellington last Saturday night, Canterbury has now won six national rugby premierships in a row. The NZRU keeps changing the competition format on a regular basis, but at a minimum there have been seven teams who are in the top division and therefore eligible to win the premiership each year. These teams are generally strong, and one weak team at least has the potential to get relegated to the lower division in any given year. To win six championships in a row is an exceptional achievement. It's easy to conclude that Canterbury is simply a very strong team, but there are at least a couple of factors that should naturally assist with spreading the talent pool more evenly and therefore contribute to making it difficult for the same team to win year after year:

• The ITM cup is run at the same time as the best players in the country should be in the All Black environment. This can take as many as 30 players out of the ITM cup at any one time. The number of players who would be considered All Blacks who are playing in the ITM cup is few and decreasing every year. Very good players should be selected for the All Blacks, which should disproportionately hurt the chances of winning of the provinces with the better talent base.

• It could be argued that the above only affects the very top players, but there is another factor that affects the rest of the player base. It is reasonably common for players to shift to different teams from season to season. A common reason is to get more opportunity. If there are too many good players that play your position in your current team, you think about moving to a team that is weaker in that position, thereby weakening your former team and strengthening your new team. This generally occurs between seasons.

Neither of these factors would perfectly balance out the competition. But they would assist with increasing the probability of a different winner each year, and making six wins in a row very unlikely.

So what are the candidate explanations for Canterbury's consistent success? I can think of a few:

1. Pure luck. Although winning the title by luck alone six years in a row must be extremely unlikely.
2. As inferred in the title, perhaps All Black selectors are for some reason more reluctant to pick Canterbury players than players of similar ability from other provinces, which would  diminish the effect of the first bullet point above. They might require a longer period of good performance to be picked, or they may be ignored entirely. I'm not really suggesting that the All Black selectors sit there with their clipboards and put big red lines through player names simply because they play for Canterbury, but it is possible that Canterbury selects players with particular attributes (eg. Goal kicking ability, percentage of missed tackles etc) that contribute to success which are undervalued by the All Black selectors.
3. Canterbury is a fantastic place to live, which encourages players to stay here regardless of their selection opportunities. The counterpoint to this one would be that Canterbury has probably never been a more difficult place to live than it is currently.
4. Canterbury is no better than other teams in terms of individual player ability, but has a team culture that leads to a more cohesive style of play and helps the team to win more often. A counterpoint to this would be that with player movements between provinces, other teams have an opportunity to learn and improve their own team cultures.
5. Canterbury is no better than other teams in terms of individual player ability OR team culture, but has uncovered a secret formula for how to play in play-off matches. If we assume that the team that finishes top in round-robin play is likely to be the best team, then Canterbury has been the best team in only three of their six successes. In the other three successes, they played the top qualifier in the final and had to play the final away from home.
6. Referee bias in favour of Canterbury (unlikely as their performance is assessed).

What other possibilities are there to explain this run of success? With a well-designed framework, some of these are probably testable using variables including player movement rates, All Black selection rates and how long a player lasts in the All Blacks after initial selection. Any other ideas on how to distinguish these hypotheses in the data are welcome.

What best predicts success in test cricket?

A commenter by the names of Peter (initially Tyrion) at an English cricket blog, Declaration Game, asks an interesting question: are averages in first-class cricket better correlated with how a batsman does in test cricket than are his averages in ODI cricket? I started to reply there, but the comment got too long, so I thought I'd do a separate post.

It is an interesting question. First-class cricket has the same general format as test cricket (no limitation on overs, so the emphasis on bowlers is to get batsmen out rather than simply restrict runs, and the emphasis on batters is to bat for long periods). On the other hand, we can all think of players who have been successful playing for their province, state or county, who were unable to make the step up to international level. So performance in first-class cricket is perhaps the better measure of a player's skill set while performance in ODI's the better measure of his temperament, and also of whether there are technical flaws in a player's game that will be found out at the highest level. 

I don't know what the correct answer to this is. My guess would be that the stronger is the competition at sub-national level, the better would be first-class cricket as a predictor. So, for instance, I would expect that first-class cricket is a better predictor of Australian players' success at test level than it is for New Zealand players. And I suspect that, overall, even in country's with weak domestic competitions, first-class cricket would be the better correlate, but it is just a guess. 

But this rasies a different question: What measure of performance in ODI cricket would be the best measure of success to use in a correlation with test averages? This rises again because of the different format. Batting and bowling averages are a very good measure of performance in test or first-class cricket. Yes, we need to adjust for the conditions in which different players have played, and the quality of the opposition, but in general maximising your batting average and minimising your bowling average is the way to maximise the chance of your team winning. In ODI cricket, this is not the case. The relative importance of runs and wickets changes depending on the game context, so that averages are not a good measure of a player's contribution to his team. Naturally, I would prefer a measure like the player's contribution to the WASP, as maximising that is what a player should be doing to help his team win.* But let's say you had two players, one with a better average and one with a better WASP-based performance. Does the difference arise because the latter is a better player in general, or does it indicate that the former has a skill set more suited to test cricket. I really don't know. Maybe that is a future Honours project. 

* Actually, WASP-based measures would only be useful for comparing players at a similar point in the batting order. My student, Marcus Downs, has been writing an Honours dissertation this year developing an adjusted WASP-based measure that enables better comparisons across players with different positions in the order, but this is secondary to the main point. 

Are wickets more likely on hat-trick balls?

A student of mine has emailed me asking if I know anything about whether in cricket a wicket is more or less likely on a hat-trick ball then on any other ball. (Note for Eric and others similarly challenged in the finer nuances of cricket, a hat-trick occurs when a bowler takes a wicket with each of three consecutive deliveries. Note for followers of other sports: this is the original use of the term "hat-trick" in sport.)  The student and his flatmate have surmised that taking a wicket on a hat-trick ball is more likely than on any other randomly chosen ball. I don’t know what the data say on this, but I think the students are almost certainly right, mostly for statistical reasons. It is fun to think about how to formalise the hypothesis, and then how to test the effect of different forces. Maybe it could be a future Honours project to take this theory to the data.

Take a set of games in a particular format (say test cricket), and find the total number of deliveries and the fraction of those that resulted in the bowler being credited with a wicket. Then find the total number of deliveries in all those matches where, if the bowler had taken a wicket he would have achieved a hat-trick, and find the fraction of those deliveries where a wicket was in fact taken. Our guess is that this latter fraction will be higher than the general fraction of deliveries with wickets, and that that difference would be statistically significant. I am fairly confident about this purely because of sample selection:

• Pitches vary considerably across matches; if a bowler has already taken two wickets in two balls, it is likely that the pitch for that game (and that point in the game) is an easier one for taking wickets than the average.
• Bowlers (and their supporting fielders) vary in ability; if a bowler has already taken two wickets in two balls it is likely that he is a better bowler (with better supporting fielders) than the average.
• Batsmen vary in ability and batter ability is both correlated within the batting order and correlated within teams; if a bowler has taken two wickets in two balls it is likely that the batting team has below average quality batsmen and that it is one of the weaker batsmen in the team who is facing the hat-trick ball.
• Statistically (I can confirm this from test-cricket data), batsmen are more at risk at being dismissed early in their innings than later on; there is a high likelihood that the batsmen facing the hat-trick ball is facing his first ball of the innings.

So let’s control for these sample selection issues and consider instead a conditional probability question: Given the ability of the bowler and fielders, the batsman, how early it is in the batsman’s innings and the state of the pitch, does being on a hat-trick change the probability of a wicket? The question here becomes whether the unusual situation leads players to change their behaviour in some way. On the bowling side, the captain might set more aggressive wicket-taking fields on a hat-trick ball, but the bowler might try too hard and lose his rhythm. Similarly, the batsmen might be more conscious about not giving his wicket away, but at the same time the pressure of the situation might lead to his having leaden feet.

I would expect that the psychological effect would be greater on a batsman new to the crease than a bowler who has had a chance to find his rhythm. And in test cricket, I think that batsmen are always concentrating only on wicket preservation on the first ball they face. So If I had to guess, I would say that in test cricket the net result would still be that wickets are more likely on balls where the bowler is on a hat-trick, but the effect would be very small (and probably not discernible with statistical significance in the data). In limited overs cricket, I would expect the effect to be much smaller or even zero.

Now, if only I had ball-by-ball data for the entire history of test cricket! 

The scream test

Imagine two policies.

Policy A would have government nationalise all liquor companies and distribute their product, for free, for anybody who wants some, with costs borne out of general taxation. There would be no compensation of the liquor companies.

Policy B would strengthen alcohol intervention programmes for prisoners with alcohol abuse problems. They'd spend a fair bit on it and work hard with prisoners, both while in prison and during their re-integration back into the community, to help them to avoid falling back into substance abuse.

Now Alcohol Action NZ proposes a "scream test" to tell whether some proposed policy would reduce alcohol harm. In their model, industry profits are increasing in the harm imposed by alcohol, and industry screams when profits are threatened. Anything that reduces harm reduces profits, so the scream test tells us which policies are likely to reduce harm.

I suggest instead that industry would scream a lot about Policy A, and would likely support Policy B. I also suggest that Policy A would increase harm and Policy B would reduce harm. Note further that National is already implementing something like Policy B - kudos to them. I've yet to hear screams from industry.

The scream test is a bad one. Profits are perhaps increasing in total consumption, but they're likely decreasing in alcohol's harms not only because of the policy reaction function but also because moderate consumers' consumption is likely decreasing in perceptions of harm.

The one spot where Sellman's scream test could be right would be policies potentially hitting the very heaviest consumers. There are discrepancies between median reported consumption figures and total alcohol available for consumption suggestive of that there is a small group consuming booze by the barrel. Policies disproportionately hitting that group could potentially reduce both harm and profits. But I hardly expect that mandatory 1 am bar closing times fit the bill.

Knowable, but not known to me

In the futile hope that maybe, just maybe, folks' views about welfare policy might just stand to be informed by data, here are a few testable hypotheses I've seen floating around. They posit things that are knowable, and I'm sure data exists to resolve things. Let's walk through a few of them.

First, how do poor people use money? I tend to say we ought to just give money to poor people if we want to make poor people better off. Other folks think that they'll just waste it on booze and cigarettes rather than helping their kids. I don't discount that that's also possible; it's an empirical question.

@ericcrampton Is there any research on how people on low incomes spend their money, and what changes when get more cash per week?
— Jackson James Wood (@_jjw_) May 28, 2013

Now why does this matter? If you think that parents will waste money given them, you might prefer in-kind benefits provided directly to the children of poor parents rather than cash transfers. School breakfast programmes can fall into that category, despite that they're rather ineffective and largely go towards feeding kids who would have been fed anyway. I think that some of the support for wrecking the GST by exempting merit goods also comes from this kind of view, though I think this rather misguided: vouchers for merit goods could be a rather less ruinous way of achieving the desired end.

So, the test. Get household consumption survey data, look for some shock to benefit payments, and check the effects on different consumption categories. If extra money going to poor households disproportionately increases consumption of lotto tickets and booze, then the paternalists who want to make sure that money given to the poor is used for particular things are right in wishing for more in-kind benefits; if not, then the paternalists should back down on such assertions.

I can't imagine that this empirical test has not been done by somebody somewhere; I just don't know the results. I also don't expect that it will change many minds. Paternalists will want paternalism for its own sake, and anti-paternalists won't mind that poor people enjoy some consumption goods. I'm one of the anti-paternalists, but if the data showed little benefits to kids of cash transfers to families intended for kids, I'd shift towards preferring rather more in-kind benefits to kids. Any readers able to point to relevant NZ studies are welcome to do so in the comments.

Second, "can't feed 'em, don't breed 'em". Twitter and the NZ blogs have a bunch of folks yelling at each other about whether the main problem in child poverty stems from poor people's unwillingness to engage the prudential constraint or whether it's bad luck. Those on the right note that if poor people stopped having kids they couldn't afford, then child poverty would be less of an issue. People on the left instead remind those on the right that birth control can fail and that people in good financial circumstances can fall on hard times for reasons outside of their control and after they've set their family size.

So, a test. Start with DPB numbers. What is the current fertility rate of women receiving the Domestic Purposes Benefit, and how does it compare to the fertility rate of women of similar age and marital status who are not receiving government support for the raising of children? If the fertility rate among women on the Domestic Purposes Benefit is roughly what we would expect given known rates of contraception failure, then score a point for the left. If women on government support are instead choosing to have more children while in poverty, then score a point for the right. I would bet that the data shows rather more childbearing than would be expected from contraception failure alone, but less than the fertility rates among similar-aged women not on the DPB, but I've not seen the data.

Again, I'd be surprised if this kind of data didn't exist somewhere. I expect that the data could actually potentially make some difference here, though it depends which way it goes. If current rates of childbearing by women in poverty are consistent with failure rates of reliable and available birth control methods, I don't think many on the right would shift to demanding abstinence and abortion. But if current rates of childbearing by women in poverty are consistent with deliberate choice to bring more children into poor households, I expect that most on the left would shift to a fairness argument about that those in poverty shouldn't be constrained against choosing to have more children. And then there'd be the obvious counterargument about how it's a bit perverse that richer households deciding to have fewer children because of the costs are compelled to subsidise the fertility decisions of those happy to raise a kid in very bad circumstances; those on the right then might wish to advocate for that reliable birth control be a precondition of welfare receipt. Those are values-based arguments I can't adjudicate, though I expect that if, for many, the point of social insurance is to insure against bad outcomes, and if there were reasonable evidence of that many on the DPB were choosing to have many more children, there could be reasonable support for that birth control be among the conditions of welfare receipt.

A second test: what is the elasticity of childbearing among the poor to changes in benefit rates? Those on the right worry about paying women to have children they can't afford and think that paying more to benefit existing poor kids does a lot to bring more poor children into the world; those on the left think that the elasticity is pretty low and that we need to focus on the potential first-order benefits of higher transfers to existing poor children. I don't know if this elasticity is known, but it's definitely knowable. Find some shock to the generosity of payments to poor children and see whether it has any effect on subsequent fertility decisions. If little to no effect, score a point for the left; if things are reasonably elastic, score one for the right.*

Again, those with data or studies that might help resolve the second question are welcome to provide pointers in the comments.

If this stuff turns out to be in the "knowable, but not known to anybody" category rather than just "not known to me", file this under "future honours projects".


* I'm on some orthogonal dimension where I reckon it's good that more kids be brought into existence conditional on their enjoying their existence, even if they are poor. I worry instead about net effects when higher income people forbear from having their third child because of the income effects of the taxes taken from them to subsidise the bringing-into-being of a lower income person's third child. And then we get into the empirical question of relative elasticities and some rather thorny questions about trade-offs.

Measuring the influence of golf caddies

Over at the Dismal Science, a correspondent, Ross, commented on my post about the wage contracts for golf caddies, syndicated from here on Offsetting.

Ross suggests a neat natural experiment that could potentially identify an exogenous effect of caddy on a golfer's performance. The imposed exogeneity comes from the fact that for years all players in the Masters were required to use an Augusta National Club caddy, whereas in the other three major tournaments, golfers could use their own caddy. So one way of looking at the effect of caddying is to see how much predictive power player rankings at the other tournaments have for performance at the Matser's relative to how much predictive power they have for each other. If there had been no change of rule, such an exercise might be suggestive but hardly conclusive, as  there might be something about the Master's that encourages a different kind of player. (My little bit of golf knowledge suggests that this might not be a huge deal--the major tournament that is the most different from the other three is the British Open.) 

But the rule did change in 1982, after which golfers could use their own caddy at the Master's. So here is my suggested test. Take the 5 years prior to the rule change and the 5 years after. For each major tournament in those 5 years, find the subset of players who played in that tournament and each of the previous three (i.e. all 4 majors over a 12-month period), and note the ranking out of that subset of players in each tournament. Then for every pair of players the player with the highest average ranking in the three previous tournaments also had a higher ranking in the 4th. If caddying is important, we would expect to see that the fraction of pairs where the pairwise ranking stayed the same was lower for the pre-1982 Master's than for the other three tournaments (due to the effect of randomly  assigned caddies), but that that difference disappeared after 1982. 

I don't have time to do this, so I put it out there if someone wants to jump in as a co-author and do the legwork. My guess is that golf is such a high-variance game that there simply won't be enough data to tease out any statistically significant result, but it might be interesting. Or maybe it could be an Honours project. 

Efficient excusing

I have twice been excused from jury service when the jury duty conflicted with exam preparation and supervision; fortunately, I haven't been called during times of the year when I'm not teaching.

The Herald wonders whether lawyers Googling jurors helps them craft more convincing arguments for that particular jury. But that's gotta be a massively second order effect relative to Googling's effect on jury selection. Any Crown attorney Googling me would find I support jury nullification* in drug cases or for any other victimless crime and would knock me out before I made it onto the jury. Any Defence attorney would find I support that those actually imposing harm be made to make the victim whole, and likely would knock me off the jury on that basis. So it's efficient that I not be in the jury pool at all: the odds that I'd fail to be knocked out by one side or the other are exceedingly low, my time is worth something, and the process of being excused from a series of trials because of challenges is lengthy.

But think about the pool of people who would not be challenged by either side in a world where a rather substantial portion of potential jurors blog, tweet, or put stuff on Facebook.

It actually makes for a fun option value / optimal stopping rule problem. Here's the game, as described on the Justice website:

 The sequence of events at the beginning of the trial is as follows: the persons who have been called for jury service ("the jury panel") are brought into the courtroom where they sit at the back.⁷ Once court staff, counsel, the accused and the Judge have entered the court room the counts in the indictment are put to the accused, who pleads to them. Then the jury is chosen by ballot from the jury panel. The Registrar calls out each name as it is chosen, and that person walks from the back of the court to the jury box. If the person sits down before being challenged, they become a member of the jury unless they are discharged.

The page above is out of date; the number of peremptory challenges has dropped from six to four; I trust that the rest of it hasn't fallen out of date. But here's the basic game. Each side can challenge potential jurors as they are drawn from the pool of 30-40 who sit at the back of the room. If a juror is actually not qualified to serve, he can be challenged. Each side can challenge four without cause - peremptory challenge. I am not sure if there is a limit to the number of challenges for cause that can be made, but the judge has to be happy with the cause. Presumably my "anybody caught smoking a joint, and who admits having smoked a joint, is nevertheless innocent regardless of the law" views, as well as my three-day-a-week-anarchist status, would have me challenged for cause.

Suppose you're a lawyer who has to decide when to use your challenges to best serve your client's interests. So, when you got the list of potential jurors a couple days before the trial, you Googled them and gave each a score on the [-1, 1] interval indicating how hostile they were likely to be to your client's interests; you might also have noted your uncertainty about each score.** If you're the defence attorney on a drug trial where it's only a possession or trafficking charge, you give Crampton a +1 with zero variance (but you also know the Crown will veto); if you only find that somebody has 420 on his Facebook page, maybe you give him a +0.5 with a wider confidence interval. When you walk into the room, you check to see who is in the jury panel for the day; you then array the jurors from most to least hostile along with the scores you'd already given them. As each juror is called, the composition of the panel and of the jury changes, and so too then should the value of a held peremptory challenge: you have to weigh the hostility of the juror you reject with certainty against the odds that a more hostile juror might come up when you've run out of challenges.

I'd be pretty surprised if this hasn't been modeled before. But if it hasn't been, it would be a fun honours project. It would also be fun to see whether you can make an efficiency case for the reduction in peremptory challenges from 6 to 4 as lawyer certainty about juror views tightens up when they can Google potential jurors. File it under "Potential Honours Projects", if it hasn't already been done.

* UPDATE: It occurs to me that many Kiwis will have no clue about jury nullification. In short, it's the idea that a juryman must evaluate both the facts and the law. If the law is unjust, the jury must acquit no matter what the judge says. Juries are only a bulwark against tyranny if they are willing to overturn unjust laws by refusing to convict peaceful people for doing peaceful things. That's the point of juries. Expert judges will beat juries in assessing the facts, and especially so in complicated cases. Read Lewis Carroll on the competence of juries. But juries can assess whether it would violate the community's norms if the law were upheld. Consequently, there are no circumstances under which I could render a guilty verdict in a case involving victimless crimes. And so I know that I would never be chosen to be part of a jury. I do not expect or purport that most people here or anywhere else agree with me about the desirability of nullification; if they did, there wouldn't be prohibitions in various places on telling juries about nullification.

** Doubt lawyers do this explicitly; would expect decent ones do it implicitly such that a Friedman "as if" move works. Note that most of my knowledge of actual jury selection comes from an old Al Pacino movie. Keanu Reeves couldn't really Google the jurors in 1997. But things have changed....

Kiwi Freedom

New Zealand is the best country in the world if you weigh up a bundle of economic and personal liberties. I've argued this more than a few times, and I've teased American libertarians about their commitment to liberty if they're unwilling to consider emigrating here because of the drop in income or difficulty in convincing relatives to come along. This is no problem for a pluralist who weighs liberty up among other values, but it is a problem for folks who claim to put a very strong weight on freedom and who wear Live Free Or Die t-shirts.

For rather a while, we've had world economic freedom rankings that have placed New Zealand at or near the top. The Fraser Institute, the Cato Institute, and the Liberales Institut have today released an aggregate Human Freedom Index.

Using indicators consistent with the concept of negative liberty—the absence of coercive constraint—we have tried to capture the degree to which people are free to enjoy classic liberties in each country: freedom of speech, religion, individual economic choice, and association and assembly. The freedom index is composed of 76 distinct variables including measures of safety and security, freedom of movement, and relationship freedoms such as assembly or legal discrimination against gays. In this preliminary index New Zealand ranks as the most free country in the world, followed by the Netherlands and then Hong Kong. Australia, Canada, and Ireland follow, with the United States ranking in 7th place.

I have a few quibbles with the index (here is the full index), but I expect that they're all things that would be tough to incorporate with data for any large number of countries.

The index of government threats to individuals rightly includes extrajudicial killings, torture, political imprisonment and disappearances. But it doesn't include the number of individuals imprisoned for victimless crimes like prostitution and drug use. The US would fare poorly here, but New Zealand wouldn't do well on the drugs side either. Potential variables here could include an indicator for whether sex work is legal or illegal; an indicator for whether drug possession is legal or illegal; annual expenditures on drug enforcement; proportion of the prison muster whose offending relates to drug use or drug trafficking.

Threats to private property rightly include theft, burglary and inheritance takings, but miss civil asset forfeiture. I can't easily see how this could be quantified in a big cross-section.

One of the biggest regularly experienced differences in personal freedom between New Zealand and America is airport security. There is zero risk that this kind of thing happens in New Zealand. Again, I don't know how you could quantify this for any large number of countries.

Further, our police are unarmed and remain so despite the police rather frequently asking the government that they be allowed to carry weapons. One option that might capture the overall level of "police state" impositions would be to include the total police budget as a bad while also counting experienced crime rates as a bad. I suppose that America's ranking of 5 on "Extrajudicial killing" captures some of this (NZ scores a 10).

Finally, there's no accounting for the growing scourge of paternalistic regulation around alcohol, tobacco, and the like. Sin taxes as proportion of aggregate government revenues could be a start, and should be feasible, but it would be harder to get comprehensive data on whether you're allowed to brew your own beer, ease of starting a brewery, and restrictions on smoking on private property. I love that, in New Zealand, people can move really easily from goofing around in their garage with a completely legal home still or beer-making kit to selling their product to willing customers. In America...

The index a great start though, and I'm especially glad that this data has come out in this, the "setting honours projects" time of year. I'm consequently putting this up as a potential honours project; incoming Canterbury honours students, take note, but also note that this is my working draft of the project and that I might improve it based on comments.

The weight of freedom: economic and personal liberties in a gravity model of international migration.

We typically assume that people move from country i to country j because doing so makes them better off. But what precisely proves attractive? Karemera et al (2000) show that a modified gravity model can explain a decent share of international migration: origin-country population, destination-country income, and origin-country restrictions on emigration explain much. Lewar and Van den Berg (2008) show that institutions and distance also matter. Ashby (2007) shows that, within the United States, those states with greater economic freedom draw more migrants, but only because of increased expected income rather than because of the direct effects of freedom per se.

In this project, you will start by figuring out gravity models. Your supervisor has no experience with them either, but we can likely work things out.

Next, you’ll take the Fraser Institute’s newly created Freedom Index which compiles both economic and personal liberties to develop an aggregate freedom score: note that New Zealand is the best country in the world by this measure. The index compiles 76 separate variables relating to economic and personal liberties.

You will add these personal and economic liberties as explanatory variables into a gravity model of international migration to find out:

1. Does liberty enter positively into migration decisions? 
2. What’s the elasticity of migration with respect to economic and personal liberties?
3. What is the value of freedom? A standard deviation increase in personal or economic liberty counts as much as what in a gravity model? I want you to be able to say something like “All else equal, a standard deviation increase in personal freedom [economic freedom] is the equivalent of reducing (or increasing, who knows) the distance between two countries by XXX kilometres or increasing the expected income jump from migration by $YYY.” How much less attractive would New Zealand be to international migrants were we to fall in this ranking? People deciding to emigrate often have choice among country destinations. Singapore is richer than New Zealand but ranks 39^th in overall freedom and scores only a 6.6 in personal freedoms. How much extra income does a move have to provide, in expectation, to make it worth dropping a point in personal freedom?

You’re going to have to sort out how to work with the OECD migration database, figure out what scope is feasible given that data within an honours project, and then run things. Do not select this project unless you have done reasonably well in undergraduate econometrics. But it’s going to be hellafun and I think it could be publishable if you do a decent job of it.

Initial Sources:

Ashby, N. 2007. “Economic freedom and migration flows between U.S. states”. Southern Economic Journal 73:3 (January), 677-97.
Karemera, D., V.I. Oguledo, and B. Davis. 2000. “A gravity model analysis of international migration to North America”. Applied Economics 32:13, 1745-55.
Lewer, J. and H. Van den Berg. 2008. “A gravity model of immigration”. Economic Letters 99:1 (April), 164-7.
OECD Migration Database available at http://stats.oecd.org/Index.aspx?DatasetCode=MIG [or, if you can find a better database, go for it]
Watkins, T. and B. Yandle. 2010. “Can freedom and knowledge economy indexes explain go-getter migration patterns?” Journal of Regional Analysis & Policy 40:2, 104-15.

I love setting honours projects.

Flynn effects

Professor Flynn is trying to help you improve your mind. My review of his latest book should be in today's Christchurch Press. They gave me 450 words; I took 480. Here they are.

Professor Jim Flynn sets a clever trap in his latest book, “How To Improve Your Mind”. Promising that all readers will “be far more able to defend their position after reading it than before”, Flynn instead provides the critical tools necessary instead for reassessing the irrational beliefs each of us hold and, in so-confronting, end them. For your own good, you should let yourself be trapped. Jim Flynn is arguably New Zealand’s best social scientist. I disagree with him on some aspects of economic policy, but Professor Flynn is a truth-seeker. And, even better, he’s very often a truth finder. Here, he helps the rest of us be better truth-seekers. This is no mean task. Thinking rationally is hard and, often, unpleasant. Worse, there is no payoff to rational thinking in many areas of life; you will hardly be less successful in most occupations if it gives you pleasure to believe that the earth is only a few thousand years old. But I’ll return to this.

Flynn ably demonstrates the methods of rationality through a series of case studies of beliefs that contravene the methods. He also warns against the tricks used by policy advocates against those not appropriately armored against them.

I disagree with a few of the finer points in Flynn’s chapters on economics. For example, even if the reservation wages of second earners are lower than those of primary family earners, this will not drive down female wage rates unless we add assumptions around segmented labour markets. Further, decent labour standards seem to owe more to higher incomes and increased productivity than they do to Twentieth Century regulatory interventions. In developing countries, child labour is very sensitive to family income; when parents can afford to do so, they send their kids to school. And he is perhaps a bit too optimistic about the feasibility of ameliorative regulatory reform in financial markets. But Flynn has a sharp nose for the failures of rational thinking about economics among those whose values he shares – he correctly sees negative income taxes as better than a minimum wage for helping those he would wish to assist, though he misses that cleverer forms of the negative income tax solve some of the problems he highlights.

The bigger question, and the one harder to answer, is why to choose rationality in the first place. Flynn writes, “If you learn how to use logic and evidence to examine your own principles, and the principles other people urge upon you, you can enlist in the ranks of mature moral agents rather than in the army of stones.” But it’s an expensive endeavour – not unlike taking the Red Pill offered Neo in The Matrix. I’d be surprised if more than a quarter of the population even tries to choose rationality over comforting illusions. But if you’d like to try to try, Flynn’s book is a good start.

I should be reviewing Eliezer Yudkowski's "Harry Potter and the Methods of Rationality" for next weekend's Press.

One bit that wasn't relevant for the Christchurch Press but I found awfully interesting was this observation:

"Academic competition for grades in a particular course selects out an elite that gets an A+. But within some areas, the correlation between courses is far more perfect than within others. Within mathematics, the best student in differential equations is likely to be the best student in algebraic geometry and in number theory, and so forth. Within political studies, the best student in political behaviour is less likely also to be the best student in areas as diverse as international relations, political philosophy, and quantitative methods. So, talented math students will come to scholarship committees with a string of A+s, and talented political studies students with  a mix of A+s and As. Rather than putting all the math students at the top, the obvious thing would be to alternate science/math students with arts/social science students. Observe how quickly math professors forget what they know about regression to the mean when such a proposal is put."

I'm on the Scholarships Advisory Committee as the Commerce rep; we help advise on policy around scholarships. I wonder how much this affects our University-wide awards. Economics probably doesn't do horribly out of this kind of overall effect: more than any of the other arts or social sciences, there's likely an underlying e-factor - thinking like an economist - that explains variance across students. At least in microeconomics. But arts and social sciences in general will do worse than the bench sciences and maths.

I wonder if anybody has run the test. If Flynn is right, then we expect lower variance in course grades for maths and science majors than for arts and commerce majors. We can use incoming high school grades as a measure of baseline ability. We could observe higher variance among the best students in arts courses than among the best students in maths courses simply because of grade truncation issues if the very best of all students pick maths over arts. I think I'll add the "future honours project proposals" tag; this seems testable.

Update: Oops...didn't see it in today's paper. Likely in next week's then.