Variability in returns to education

The headline in the LA Times was ‘College is a bad financial bet for some, study says’. The story focused on the cases in which students had a negative return on investing in higher education:

A surprising 14% of high-school graduates earn at least as much as people with bachelor’s degrees, and 17% of those with bachelor’s degrees outearn compatriots with professional degrees, the authors found.

The study in question is here, a Brookings Institutions report about the variability of returns to education.

The main thing I wanted to point out was the framing of these numbers. Research has shown that the way that percentages are presented changes how people react to them. Is it a 20% chance of failure or 80% chance of success? Is it a 1% probability of damage or a 1-in-a-hundred chance? It matters.

So let’s flip it around. Are you surprised that 86% of high school graduates earn less than people with bachelor’s degrees? How about that 83% of people with bachelor’s degrees earn less than graduates with professional degrees? If you were playing the percentages, would those results encourage you to get a degree?

What the authors are telling us is that earnings by degree have a distribution around some mean. There is some distance between the means, and the overlap of the distributions isn’t all that large (15%-ish). I haven’t gone through the report, but the results would be affected by whether they are doing a sort of t-test of the two distributions, or doing something like analysing joint distributions of two random variables.

Does this mean we are sending too many people to university? I’d suggest we don’t have enough information. If we think of it as a comparison of two distributions, what would we be trying to do? Are we trying to:

1. create enough distance between the means so that the overlap is small? But why should we encourage a larger premium for education when on average the benefit-cost ratio of education is already around 5?
2. shrink the left-hand tail of the distribution for the more-highly educated? But how do we reliably identify these students, and should we give up on majors or degrees that don’t have a high enough return on investment?
3. do something with the right-hand tail of the high school graduate distribution? But what do we do with them? They have done well as high school graduates — it doesn’t then logically follow that they should have more education.

I don’t see that there’s necessarily a problem. The fact that a small-ish percentage of people don’t get much from a university education means that we are casting the net wide enough to bring in most of the people who potentially would. The fact that some high-school graduates can still make a good living shows that there are still opportunities for all kinds of people, not just top STEM graduates from top schools.

Bets don’t always pay off; investments sometimes fail. But if I were playing blackjack and winning 86% of the time, I’d be at the table all night.

Give the drummer some

James Brown’s funk is tight. On a track like ‘Licking Stick’, the music threatens to break loose at any time, barely contained by Brown and the beat. The little I’ve read about Brown suggests that this is no accident. He was apparently a difficult and demanding band leader, but listen to the result.

On several recordings, James Brown calls to Maceo Parker to take his solo. Oh, man, can Maceo play — the pacing, the expressiveness, the musicality — no wonder he’s gigged with everyone.

On ‘Cold Sweat’, as Maceo is finishing up, Brown asks, should we give the drummer some? Wikipedia says this is the first recording in which Brown does this. This call for a solo highlights the importance of the drummer for the whole enterprise. Maceo can play with the rhythm and Brown can give us all his famous ‘uhs’ and ‘good Gods’ because that drum is keeping things together, keeping it tight.

Now, let’s shift to some economics (sorry, but you knew it was coming). I’m involved in a few projects right now that are mainly modelling projects. We aren’t doing primary research in the sense of going out and collecting data and producing new empirical findings. Instead, we are organising existing information. We are using not only economic data, like price elasticity of demand, but also information from other disciplines, like dose-response functions for medicines or nitrogen leaching rates for different land uses.

It occurred to me that we are the drummers in these projects. We have a particular set of skills — keeping information organised and finding ways of making different types of data fit together. But the value of the drummer isn’t the particular beat they’re laying down. Their value is to provide a groove that the rest of the music can revolve around.

The drums provide a solid structure, and that’s what a good model does. As a result, the rest of the information makes more sense, in the same way that a horn solo makes more sense once the beat is established. A good model also demonstrates which parameters are important or which relationships determine the outcomes, just like a solid beat lets the singer shine.

Sometimes, we modellers even get the spotlight; sometimes, even the drummer gets him some.

Info on the regions

The Ministry of Business, Innovation & Employment released a new report a couple of days ago — the Regional Economic Activity Report. Let me recommend it — it’s a useful, easy-to-read summary of, well, of economic activity in the regions. A lot of the information can be found elsewhere if you know where to look, but this puts it all into one tidy, 82-page package. People are interested in how their cities and regions are doing — I often get questions about local economies — so I expect this will be a good resource.

Canterbury gets a couple of extra pages for the earthquake impacts. Very sobering to see tourism and education series just plummet.

Here’s one graph I felt like sharing. It appealed to me, thinking about the New Zealand economy like a Fibonacci sequence:

NZ as sum of regions, MBIE Regional Economic Activity Report

 

Metaphor of student as customer

A colleague sent me a link to Professor Scott Galloway’s advice to a student who dared take a metaphor seriously:

When the student arrived an hour late to the professor’s brand management class, Galloway told him to leave. Later the student emailed Galloway, explaining that he was shopping around for classes, which is why he was late: “It was more probable that my tardiness was due to my desire to sample different classes rather than sheer complacency.”

The student, it seems, really did believe he was customer of the university. He was shopping around, trying to find the product that best met his needs. And why shouldn’t he think that? The metaphor has been around for a long time, and the more students are asked to contribute to the cost of their education, the greater currency it has. I came across this article in the Times Higher Education, from 1999:

Universities face a wave of student litigation because of a failure to grasp their changing contractual relationship with fee-paying undergraduates, an academic lawyer has said.

…

Mr Birtwistle, a principal lecturer, found that only a minority of universities surveyed understood the potential impact of the introduction of fees. “There can be no doubt now that students hold a consumer contract with their university,” he said. Mr Birtwistle said that now students pay their fees directly means they are, in legal terms, buying a service. They are therefore entitled to private law redress for breaches of contract.

‘Students are our customers’ goes from being a business-school metaphor to being a statement of civil law. Students who are dissatisfied have therefore appealed to the courts for redress — for not providing promised support, for a poor grade (the case failed), for not getting a job, for failing to provide courses in a timely fashion [can't find a link].

Looking through the various cases, it is clear that courts do not want to meddle with academic issues. They tend to side with universities, who defend themselves by saying they are upholding academic rigour. Interesting, Prof Galloway did not appeal to academic rigour, or even professorial authority. He appealed essentially to accepted standards of behaviour.

Now, we could view the NYU episode as an example of the teaching/learning that can happen at a university beyond the subject matter. The professor is trying to teach the student how to behave in, as it were, polite society. But that doesn’t break the metaphor. It just means that the lessons the student is buying are more than the lectures and slides. They are behaviour lessons — the so-called ‘soft’ employment skills — that make a difference to how people get along in the workplace.

Once again, we get back to the purpose of a university education. Is it to produce graduates who know how to behave? Is it to teach them specific areas of knowledge? To get them jobs? It’s probably a bit of all those things. But that also gives universities a lot of ways they can fail their students, and a lot of potential grievances to be redressed.

Preferences for avoiding death

Slate blogger Matthew Yglesias has been getting flak for his post that appeared quickly after news of the factory collapse in Bangladesh. In it, he explained that economics was all about diff’rent strokes for diff’rent folks:

Bangladesh may or may not need tougher workplace safety rules, but it’s entirely appropriate for Bangladesh to have different—and, indeed, lower—workplace safety standards than the United States.

Reactions in Western corners of the internet have been fierce and occasionally funny:

Corey Robinson questions whether Yglesias is right about the collective preferences of Bangladeshis:

‘Hundreds of thousands of garment workers walked out of their factories in Bangladesh Thursday, police said, to protest the deaths of 200 people in a building collapse, in the latest tragedy to hit the sector.’

Would it not be easier for Matt Yglesias to dissolve the Bangladeshi people and elect another?

Justin Zachary at Daily Kos points out that the factory was in fact in violation of local safety laws:

What happened in Bangladesh was the result of the safety standards that are currently in place not being enforced. As Kalpona Akter, executive director of the Bangladesh Center for Worker Solidarity, told Democracy Now!, Bangladesh “already has some rules and regulations for safety,” with which some politically powerful owners are not complying.

Maha Rafi Atal at the (UK) Guardian tries to walk a middle ground of increased safety but continued employment for Bangladeshi workers:

that should be about making a distinction between wages, which do not have to be the same everywhere, and workers’ rights, which should.

It may look on the surface like Yglesias is being all ‘realist’ and ‘sensible’, but in fact he gets the economics wrong. He forgets three things:

• preferences are only half the story. The other half is the choice space in which preference can be expressed. It is the combination of preferences and available options that lead to the choices made. Ascribing the choices to preferences alone gets the theory wrong; one can just as legitimately point to the limited options
• the market theory that Yglesias uses to underpin his ideas — that there are market transactions deciding the prices of garments and safety — assumes freely available and perfect information. A large economic literature then explores the impact of relaxing that assumption. But that’s the post-grad course, and Yglesias is stuck in 101. Here’s the thing: we could make it perfectly obvious to Western consumers how their garments were made, what the working conditions were. Then we could talk about a market solution. Let me put it another way: is Burger King going to launch a horse-burger because people were buying them before they found out what was in them?
• supply and demand do not exist outside the institutions that help shape the economy. An analysis that doesn’t account for politicians who can override police edicts and flout safety regulations is incomplete. We should recognise that, for example, agreements and regulations help set the conditions in which the market is operating. So, there are trade agreements around clothing that promote its production in poor countries, but much less international recognition of professional qualifications for doctors, lawyers, accountants, etc. (On that front, economics is like the Wild West — anyone can hang out a shingle.) It is at best disingenuous to throw your hands up and say

in a free society it’s good that different people are able to make different choices on the risk–reward spectrum.

In a free society, it’s also good that people can express different opinions. Even when they haven’t got a clue what they’re talking about.

The dreaded spreadsheet error

I posted flippantly last week about the Reinhart and Rogoff (R&R) re-assessment by Herndon, Ash, and Pollin. There’s been more bytes spilled since then. The Economist says it’s not such a big deal, because  ’Ms Reinhart and Mr Rogoff acknowledge in their academic work that this conundrum “has not been fully resolved”, but have sometimes been less careful in media articles.’ Paul Krugman counters that, yes, it is a big deal and provides some links. Matt Nolan at TVHE provides more links and more perspective:

it has been used as an inconsistent marketing tool by people for selling their own unrelated ideological policies….

I’m going to be careful here. Media interviews are not the same as academic writing. Keeping my thoughts straight while listening to someone else’s questions, and then controlling the random thoughts that spring to mind whenever (ask my poor students — I don’t censor digressions quite the same way in lectures) while not babbling — hey, it’s fun and energising but only approximately accurate. So, I’m not going to pile on.

I’m fascinated that it was a spreadsheet error, at least in part. Most economists I know proudly and loudly avoid Excel for anything analytical. Grunty programmes like Stata, sure, and nerdy open-source stuff like R (thanks, Auckland!), absolutely. I mean, these are guys (yes, guys) who sneer at SPSS. To find out that R&R were relying on Excel is like, I don’t know, seeing a celebrity chef eating at Burger King.

There’s a lesson for consultants here. Excel is the sort of programme that gave rise to this:

To err is human, but to really foul things up requires a computer.

Nevertheless, I like Excel a lot. Despite all the stupid and paranoid security controls that Microsoft has added, it is still a portable way to give clients the analytical details of what I’ve done. It also allows me to build dynamic tools to help clients tweak the analysis for their own questions. And, I can show them exactly which number is multiplied by which other number, and then transform it all into pretty pictures clearly and transparently. Throw in some macros and buttons, and it’s really powerful.

The best advice I’ve heard about building those sorts of files is to treat them like programming tasks. You are essentially programming a new bit of software. There are established protocols for tracking versions and checking code — that’s a place to get some tips on good design processes.

It’s the best advice, but I’ve generally ignored it (like a lot of good advice). It’s just too hard. So, let me offer my own advice:

• do it differently — there are always multiple ways to make calculations. I like to make calculations two different ways, and then check whether they have the same values (‘=A3=B3′ will give a TRUE or FALSE; or, use an IF statement)
• back-of-the-envelope — just the other day, we were looking at a spreadsheet model (again, portability is important), and we did some back-of-the-envelope calculations to check whether they were sensible. It’s similar to the idea of an elevator pitch — can I explain in simple language and logic why we get these results?
• have someone check — give it to someone else. Let them see everything, get them to check everything. Make sure they have the chops, too, to do it right. Now, that can be expensive, several hours of work. So ask yourself, do I feel lucky today is it worth it for the job or the client? I mean, if I’m going to recommend unemployment for a few million people, I want to make sure my cell references are right. But not all clients warrant that level of scrutiny.

After all that, though, mistakes will happen. The best thing to do is be a mensch — I’m not sure what the New Zild translation is. Own up, walk the client through the impacts, and do as much work as you need to do with the client to restore some credibility.

And then, add it to your bag of tricks. You’ve just learned an expensive lesson.

What’s the point of academic research?

I’m still thinking about MOOCs. A university is supposed to be involved in research and teaching, and MOOCs potentially cut into the teaching side of the business. Even if they aren’t as good, they may still take a big chunk of market share. One can buy hand-sewn shirts, but mass-produced shirts are much more common.

So that leaves the research side of the university. What’s the point? Is it to be ‘critic and conscience of society’, which is the New Zealand job description for an academic? Is it to advance knowledge and understanding?

What got me thinking about the topic was this profile of Noam Chomsky by Glen Greenwald. Greenwald, a journalist, has been a relentless critic of the security state that the US has put in place over the last two presidencies. Chomsky, an academic, has been a critic of American hegemony for decades. It is likely that academic tenure has helped Chomsky speak his mind. That is, the economic security of his job allowed him to have ‘a room of one’s own’ (Virginia Woolf) and be a critic of society.

University research, then, might be about providing an environment in which individuals and teams can pursue research, whether that research is criticising society or supporting it. The university buffers researchers from that same society — providing them time for the research to come to fruition, shielding them from reactions when their opinions or findings are unpopular. The uneasy bargain is that society pledges resources to the university — even when it bites the hand that feeds it — because of a belief that ultimately it will be for the social good.

But is it? Or, more precisely, is it at the margin?

And that question takes me to findings like those discussed here:

Consider this tally from Science two decades ago: Only 45 percent of the articles published in the 4,500 top scientific journals were cited within the first five years after publication. In recent years, the figure seems to have dropped further. In a 2009 article in Online Information Review, Péter Jacsó found that 40.6 percent of the articles published in the top science and social-science journals (the figures do not include the humanities) were cited in the period 2002 to 2006.

So it seems that much university research isn’t even of value to researchers themselves.

There is also discussion of the ‘need’ for academics to contribute more, be more engaged with society, adopt more of a public intellectual stance. Those discussions suggest that society — government, business, the chatterati — might feel that academics aren’t pulling their weight.

Where I’m getting to is this: if MOOCs call into question the near-monopoly of universities for delivering advanced education, then universities will have to lean more heavily on the research function to justify their existence. But, the research side seems anemic, at least at the margin. The additional contribution of the extra dollar of spend seems to deliver little in the way of engagement or criticism. Oddly, the crisis in teaching raises the title question: what’s the point of research?

The source of the world’s ills

A colleague, a fellow economist and Lincoln grad, has this morning sent through an article by Dean Baker (also an economist but not a Lincoln grad). He points to a fascinating dispute:

The basic [Reinhart and Rogoff (R&R)] story was simply the result of them getting their own numbers wrong.

After being unable to reproduce R&R’s results with publicly available data, [Herndon, Ash, and Pollin] were able to get the spreadsheets … that R&R had used for their calculations. It turns out that the initial results were driven by simple computational and transcription errors. The most important of these errors was excluding four years of growth data from New Zealand in which it was above the 90% debt-to-GDP threshold.

So, to summarise: the problems with the R&R analysis were caused by:

• spreadsheets
• New Zealand
• economists.

You have been warned.

One from the archives

Publishing academic articles is complicated. You have an idea, you do the research, you turn it into a paper. Then what?

Well, unpublished papers represent your resources, which you have to turn into products/commodities. It’s a bit like exporting. You could ship off raw logs to whomever will take them — there’s always some obscure journal that will take your paper. Or, you can fashion the logs into high-end furniture and try to find a market. Those markets are more difficult. There are fashions to follow and a certain amount of protectionism from established interests, but the returns are much greater if you are successful.

It comes down to four rules:

• chunking — some people advocate figuring out the least publishable unit so you create the greatest number of products from your raw resources. I must admit that I don’t worry about this, because time is more a limiting factor than ideas. Getting time to turn ideas into polished papers is much more of a problem, so having more than the minimum in each paper is a better strategy for me
• aim high — a good piece of research, once published, has been sold. The higher the price you get for that research, the better your return on the investment. You want to get it into the most exclusive journal you can, given the quality of the research and the time you have for working through the publishing process. That’s why I tend to aim just above where I think a paper could be published, and then settle for whomever finally accepts it
• diversify — research, even your amazing research, is of variable quality. Journals, too, have great diversity in the speed of review and publishing. I’ve always advocated a portfolio approach to publishing: try to have one or two winners, some middling articles for profile, and some pieces in lower-ranked journals to keep the numbers up
• know your limits — my limiting factor is always time. There is never enough time to write up results, submit, resubmit, chase stuff up, etc. I’m better off shoveling stuff out the door and getting it through the process with minimum fuss. But that’s me. You’ll be different — different objective function, different constraints. Figure out what your limiting factor is and work around it or relax it — that’s the key to shifting your production possibility frontier

Going through some files on the weekend, I came across an old paper. I rather like this paper — I think it’s one of my better ones. Unfortunately, my publishing strategies rather failed with this one. I aimed high and started through a painful revise and resubmit process, then ran out of time, and then the topic was stale and I gave up. The lesson? Well, I’m not sure there is one. In a diversified portfolio, some projects/investments/papers will fail. It’s the value of the portfolio that counts, and focusing only on a specific failure may lead to the wrong conclusions.

This paper also failed because it hit a specific confluence of difficulties. It is a paper on choice modelling of genetically modified food. Both the method and subject matter were controversial, and putting the two together was just too much:

• on the choice modelling side, I was dinged because I used the wrong equations. Not, the reviewer noted, that the equations were wrong, just that they belonged to That Tribe Over There*, whereas This Tribe Here had better notation
• the reviewer cried for ‘more maths!’ I have come to realise that this is the economics equivalent of ‘more cowbell!‘
• GM articles in the ag and food literature can be divided into ‘pro’ and ‘other’. My experience was that the burden of proof for ‘other’ submissions was greater than that for many published ‘pro’ articles.

Since this raw log has just been sitting in the yard for years, I’m posting it here. Not much of a publishing strategy, I admit, but it’s the best I can do with limited resources.

Kaye-Blake 2007 Mixed logit as data torture

–

*who eat taboo food and have sex with their relatives.

Mass-produced education

MOOCs — Massive open online courses — are the latest Next Big Thing in education. Technology has made it cheap to reproduce and transmit information. The hope is that it can spread education far and wide.

The discussion of MOOCs reminds me of other technology discussions. Back in the early days of Web commercialisation, there was a lot of jostling and experimentation to try to figure out how to use the Web and make money from it. Some models boomed, some failed, and some limped. MOOCs look like the same sort of process — trying to figure out how to make a profitable mass education business model.

They also remind me of MP3. The analogue proponents say that compressed digital music doesn’t provide the quality that vinyl can. Listening to poor-quality songs from my smartphone, I know they are right. But then, I can’t carry around 1,568 songs on vinyl in my pocket. The criticism that MOOCs are providing poorer quality education — which is likely to be accurate — ignores that there are other considerations. Some people in some situations are willing to trade quality for convenience.

The criticisms of MOOCs seem to revolve around their commercial focus, which is just the latest fight over commercialisation of universities. This Lawyers Guns and Money post on US and UK universities discusses MOOCs  as part of a larger discussion of

the commercialisation of academia and the erosion of academic freedom [which] are tightly interwoven.

In particular, critics are concerned about profiteering by the course providers and the rise of superprofessors —  seeing MOOCs as ways to stroke the egos of people who are already successful while creating profits only for the companies involved.

These courses are revealing an important split in the role of universities — the production of new knowledge, which is expensive and time-consuming, and the dissemination of knowledge, which needn’t be. And that suggests the possibility of greater division of labour, which has historically made things less expensive and more available. These changes don’t tend to be (are never?) unequivocally good (or Pareto improving). This was Rousseau’s critique, as it was Marcuse’s, but people continue to buy the newer, cheaper stuff.

I am in the middle of some lecturing. I have two sections of about 250-300 students each. I pace about at the front of the lecture halls — purpose-built to have that many students — and talk them through basic statistics. I get the occasional comment or chuckle as feedback, but I’m not interacting with the students in any meaningful way. I could be performing in front of 300 or 3,000. They could be watching me in person or on-line.

These large lecture halls show that universities already recognise the efficiencies to be had in transmitting information. Universities are already mass-producing education, and students’ experiences are already inferior to mine of 25 years ago. MOOCs are not just a new technology that breaks with the past; they are also a continuation of it.