← More or Less
More or Less
Tuition fees
00:28:03
Tim Harford and the team examine examine tuition fees, drugs testing and inflation.
Loading summary
Transcript78 lines
- [00:00]
A
Thank you for downloading the More or Less podcast from the BBC. You can find out more about our programme on our website, BBC.co.uk radio4. But before you do that, here's Tim Harford.
- [00:15]
B
Hello and welcome to More or Less Speaking Truth To Power. Armed only with a slide rule this week, the U.S. supreme Court weighs in on the subject of statistical significance. How could we resist? And we speak to a man who thinks progressive taxes are unfair. And of course, there's the response to last week's item on cutbacks in Trumpton.
- [00:39]
C
Last week's dismissal of Dibble from the Trumpton Fire Brigade has provoked widespread concern and anger in Trumpetonshire. There's been a Twitter campaign initiated by Mr. Clamp and Ms. Lovelace to bring back the Trumpton one and to sack Treasurer Harford, whose actions have been described as being motivated by narrow economic dogma. Meanwhile, there's anxiety about the former fireman's state of mind. He was last seen drunk and disorderly near the bandstand.
- [01:06]
B
The Great Dibble controversy refuses to go away. But first, the other side of university tuition fees. During the last series, our reporter Wesley Stevenson immersed himself in this debate, calculating that about a quarter of students will pay back less than they would have done under the old system, and that about half of students won't have to pay back the full value of their loans. Now. Now, let's think about the other side of the equation. Not how much students pay, but how much the taxpayer pays. Hi, Wes.
- [01:35]
C
Hi, Tim. And yes, some think the government might end up paying a lot more than it bargained for. The Guardian reported that the new university system might cost the taxpayer an extra billion pounds each year beyond what the government had anticipated. Which starts to raise the question that it might even cost the taxpayer more than the current system. At least that's what our listener Chris Bissell reckons.
- [01:57]
B
I'm not surprised the new system doesn't save the taxpayer as much as one might think, given you'd already found that many students won't be paying their fees back in full. But it seems bizarre that the new system asks so many students to pay £9,000 a year and might still end up costing the taxpayer more.
- [02:16]
C
Well, it would be, but it may be a little too early to say. As each university announces how much it will be charging next year, we will get a better picture of how much this might cost. At the time the new fee structure was announced, the Business Secretary, Vince Cable, said that the £9,000 fee would be the exception. But as universities have announced what they're going to charge. This looks less likely. Of the 39 that are declared so far, 27 have set the maximum fee. So it's no wonder. The Guardian says it's going to cost a billion pounds a year more than expected. But when this was put to David Willits on BBC One's Politics show, he said this.
- [02:51]
B
When you look at all these behind these specific reports of headlines, I mean, London Met, which is a very large university with tens of thousands of students, they're coming in at about 6. The University of Aston, which is a much smaller university, has talked about a top price of nine, but it's about a tenth of the size. So when you look at the average, you have to take account of these sorts of factors and it's still very early days.
- [03:15]
C
So when we think about these fees, we need to think about the average fees per student, not per university. So if David Willits is right, then some large universities will charge low fees and bring down the average.
- [03:25]
B
Ok, well, fine, but he gave just a couple of examples. So is there enough information out there to find out whether those examples are actually typical?
- [03:32]
C
Well, we had a look at the claim about the average, and if you average out the fees by the number of students, as Mr. Willits would prefer, the average fee actually goes up slightly to around the £8,500 mark.
- [03:44]
B
Well, hang on, that's the opposite of what David Willits was saying, right?
- [03:47]
C
Yeah, kind of. I mean, he's right to say you should average out across students, not universities, but unfortunately that makes things look worse, not better, at least based on the institutions that have set their fees so far.
- [03:58]
B
The thing that feels slightly false about this, though, is that the headline fees, they're not the average fees, are they? They're the maximum fees. And I thought that universities were supposed to be offering students fee waivers as part of an effort to boost access.
- [04:11]
C
Yes. We don't know what the impact of those fee waivers are going to have on average fees yet either. Now, David Willetts raised another interesting point. As long as students actually pay back their loans, then the higher than expected fees isn't a problem for the taxpayer, it's a problem for the students. But the government does have to cover a shortfall in student fees because not everyone will pay back the whole of their loan. When this was first announced last year, the government was working on the idea that universities would charge, on average, £7,500, as we said, at this price, according to the IFS, the Institute for Fiscal Studies, the government would have to cover a 25% shortfall. But with the back of an envelope average being closer to 8,500, a lot of this extra cost will fall on the government making the shortfall around 30%.
- [04:57]
B
So the Guardian reckon this is going to cost the taxpayer a billion pounds a year more than expected. What are your numbers, Wes?
- [05:05]
C
Well, I went to the Institute for Fiscal Studies to help with this one. They've looked at the numbers and they say that if the average stays around this mark, the 8500 mark, then the government is facing a bill that is about half a billion pounds bigger than they originally thought.
- [05:20]
B
I guess we don't really know how many students are going to pay back their full loans and it's also going to depend on the kind of course they do, the kind of university they graduate from, as long as that has an effect on their salary.
- [05:34]
C
Yeah, no, absolutely. I mean, if Cambridge or Imperial charge? 9,000, then due to the success of their high flying students, graduates are more likely to earn enough to pay back their fees. But if they go to a lower ranked university and do a rubbish course, then their career prospects may not be so good. So the chances of them earning enough to pay back the entirety of their loan may be lower and therefore the government will be liable for more of their fees.
- [05:57]
B
Ok, so the taxpayer is still on the hook. Now, didn't Vince Cable say that the 9,000 pound fee was going to be the exception? Is that realistic? Was it ever likely that most universities were going to charge much less than £9,000?
- [06:08]
C
Well, if universities charge the minimum, which would be £6,000, then the institute for Fiscal Studies have calculated that these institutions would actually be £1,500 per student worse off relative to the current system. That's across the whole of a three year degree course. And universities charging £9,000 will be almost £7,000 per student better off over three years. So some of this will go to helping poorer students with the fee waivers and the bursaries. And universities also argue it will help replace some of the money they lose through research grants and things like that. But overall it does seem that the universities are the winners, not the losers in this scheme. The but it does seem very likely the whole system will be cheaper for the taxpayer than what we've currently got.
- [06:54]
B
Well, thanks, Wes. You know, that was really interesting. And to think one disloyal listener, Patrick French, had the nerve to describe our conversations as like a 1950s government propaganda film.
- [07:05]
C
That was very hurtful.
- [07:06]
B
Hurtful and wrong. Also wrong. We don't sound anything like that. I don't think we ever did, not even in the 1950s. Good evening, Mr. Stevenson.
- [07:19]
C
Good evening, Mr. Harford. What is this remarkable piece of mechanical apparatus?
- [07:24]
B
Why, this newfangled contrivance is a powerful supercomputer. The more or less 1. What is it for automated arithmetic, Stevenson? Simply dial in the calculation. 2, 2 equals and. And the answer is quickly printed in the adjacent room.
- [07:47]
C
Remarkable. Harford.
- [07:49]
B
Yes. And so easy even a lady could do it. You're listening to More or Less with me, Tim Harford, in association with the open University. It's 2011 and of course, we no longer stand for such sexist bigotry on the programme, nor indeed do we let any supercomputers into the studio. Not after the last time. Now, the Prime Minister clashed with his former university this week after commenting at a public meeting in Harrogate that it was disgraceful that Oxford University had only admitted one black student last year. Oxford wasn't impressed. This is a persistent misunderstanding, launched, perhaps inadvertently, by Labour MP David Lammy. Last year. Mr. Lammy said correctly, but with a curious specificity, that Oxford had admitted only one British black Caribbean student in 2009. Oxford replied at the time that 27 black students had been admitted and that 22% of its students were non white. Is that enough? And if not, who's to blame? To answer that question, why not listen to our item on Oxbridge admissions, which we broadcast last year and which is available via our website, BBC.co.uk more or less. Now, the US Supreme Court recently issued a judgment about what might seem like an unlikely subject, the uses and abuses of statistical significance testing. Before we come to the court's decision, let's think about statistical significance itself. Imagine that you have a jar full of Smarties. I only like the orange and blue ones. And being the diva I am, I like to get the newest member of the more or less team to pick them out for me, who at the moment, is Hannah Barnes. Go on, Hannah, get on with it. Thank you very much.
- [09:35]
A
I should state at this point that other forms of confectionery are available.
- [09:38]
B
Yes? Yeah, yeah. They're not as nice. Okay, so once sorted out, I want to figure out what the mixture is. Are there more orange ones or more blue ones? And I can get Hannah to help me here again. So, Hannah, we've got this very big jar in front of us containing only orange and blue Smarties. Now, without looking, please take 10 of them at random out of the jar, place them on my plate here beside the glass of champagne, okay? Okay.
- [10:08]
D
Right.
- [10:09]
B
So two are blue, eight are orange. So now the question is, does that prove that there are more orange than blue smartest in the jar?
- [10:23]
A
I think what Tim is trying to say is that statistical significance testing works in the following way. You assume that in fact the proportion of orange and blue Smarties is the same. Then you calculate how likely you would be to draw eight or more orange Smarties out of 10 if the proportion really is the same.
- [10:42]
B
So you might think that seeing eight orange Smarties, they're delicious, and two blue ones was pretty good evidence that there are more orange Smarties. But it turns out the chance of drawing eight or more orange Smarties from a huge jar with an equal number of orange and blue Smarties is. Is a little over 5%. The statistical significance test says that leaves too much open to coincidence, and there is not statistically significant evidence that the proportions are unequal.
- [11:08]
A
Yes, but if we'd drawn only one blue and nine orange, that would have been much less likely if the proportions were equal, and then we'd be able to say that there are more orange Smarties. And our conclusion is said to be statistically significant at the 5% level. And the real world relevance of this kind of test is pretty clear.
- [11:28]
B
These are great.
- [11:29]
A
Let's say that instead of Smarties, will you stop eating them, Tim? Instead of Smarties, you're giving some people a new medical treatment and some people an old treatment. Statistical significance is a test that assumes the treatments are the same and then looks for evidence that, in fact they are not the same.
- [11:49]
B
Okay, but here's the tricky thing. While statistical significance testing is incredibly common in science and social science, a lot of statisticians themselves think it's clumsy and inappropriate. I spoke to Steve Zilliak of Roosevelt University, co author of the Cult of Statistical Significance, who has a number of complaints.
- [12:06]
D
One reason I do not like the test of statistical significance is that it doesn't tell you what you want. What you want to know is the probability of your hypothesis being true. Given the available evidence, that would be very, very useful. For example, when you're taking a pain relief pill, you'd like to know the probability of the benefits and costs of taking that medicine when that effect is actually there. The test of statistical significance does not give us that information. It reverses the problem entirely and says the following. What is the probability of seeing the data given that there's no difference between the two medicines that we're examining?
- [12:48]
B
But in fact, there may well be a difference.
- [12:50]
D
There might well be a difference. And you can see. Secondly, there's a major problem with the test of significance is that it does not adequately deal with reasonable human beliefs about the odds of events and the consequences of following or not following the available and knowable odds.
- [13:10]
B
And here's where the US Supreme Court came in. Matrix Initiatives, a drug company, was being sued by investors for not revealing what looked like bad news about side effects from an important product, Zycam. But Matrix asked the Supreme Court to rule that they weren't at fault for not disclosing these possible side effects because they hadn't reached the level of statistical significance. In other words, they might be a coincidence. So the Supreme Court had to rule on whether lack of statistical significance was a good enough reason not to disclose reports of adverse effects. Steve Zilliak filed an amicus brief with the US Supreme Court itself. In other words, he gave the court his friendly advice on the matter. And I think we can guess what he thought about the case. I asked Steve to tell us what the court concluded.
- [13:53]
D
The Supreme Court decided unanimously that statistical significance is neither necessary nor sufficient for proving a commercial, scientific, medical or adverse effect results.
- [14:07]
B
Now just briefly summarize for us the Matrix Initiatives Zicam case that went before the Supreme Court. What was the argument about?
- [14:16]
D
Well, the argument seemed to be mainly about whether or not the nasal form of Zicam cold remedy medicine was highly correlated with or was it causing a permanent loss of smell among the users. It's called anosmia. Starting in 1999, a doctor initiated reports by telling Matrix Initiatives that a cluster of my patients can no longer smell and they're all taking Xycam. And he warned the company that there were known studies showing that zinc glutamate, the active ingredient, could cause loss of smell.
- [14:59]
B
Just explain to us briefly how Matrix Initiatives were using statistical significance to justify not releasing these reports of potential problems to their investors.
- [15:13]
D
Matrix was doing what 8 or 9 of every 10 scientists using the test of statistical significance does. And what they said is that the adverse effects that were discovered in double blind placebo controlled randomized trials were not statistically significant at the 5% level. And they said therefore, that it was correct under the 1934 securities Exchange Law not to disclose that information.
- [15:46]
B
In short, if you assume that their drug had no adverse effects, then these cases do not conclusively prove otherwise. But that's not necessarily a reason to stop investigating. You might well have cause for concern.
- [16:02]
D
Absolutely. Put it this way, the scientists who created Zycam knew that there is a difference between zinc glutamate at a certain proportion compared to the others, otherwise they wouldn't have used it. But the more oomph you get from the benefit oftentimes means the more oomph you risk getting on the downside.
- [16:25]
B
And Steve, it seems to me that on the one hand you're asking for different statistical tests. Statistical tests that directly address the hypothesis, is there a problem? Rather than indirectly address the hypothesis, assume there's no problem. Is there evidence otherwise? But it seems to me you're also just arguing for common sense that some results are so potentially important that simply because they do not pass this bright line test of statistical significance doesn't mean that we can ignore them.
- [16:56]
D
That's absolutely correct, because the third major problem with the test of significance is that it's possible to get your significance test your P value, if you will, below that 5% threshold, simply by adding observations to your study. If you increase sample size, other things equal, you increase statistical significance. But let's think about what that means. That means that individual scientists, let's say, especially if they have a commercial interest in the project, they really want to find a statistically significant difference between two things. Well, one way they can do that is just get more observations. But that doesn't sound very sensible. That's not what we want. What we want to know is what's the economic or clinical or medical importance, the practical importance of the quantitative difference between two or however many items it is that you're comparing at once. There's a big difference between magnitude and likelihood of magnitude, and strangely enough, 90% of publishing scientists do not make that distinction.
- [18:14]
B
Professor Steve Zilliak of Roosevelt University this week we were told that by either of two common measures, inflation has fallen, with the biggest drop coming in the Consumer Prices Index. It fell from 4.4% to 4%. Here's how Radio 5 Live broke the news.
- [18:32]
E
According to statistics released within the past half hour, the Government's favoured indicator, the Consumer Prices Index. And that is the favoured indicator because it doesn't include property prices, does it?
- [18:45]
B
It doesn't include housing costs.
- [18:47]
E
It doesn't include housing costs. It's a smaller measure.
- [18:50]
B
So that's very clear. The only difference between CPI and RPI, or the Retail Price Index, is housing costs. But is that all there is to it? Paul Lewis sent us this from the moneybox yacht.
- [19:04]
F
The way 5Live explained the difference between the two measures of inflation is the way most journalists have since the government decided that two price indices meant double the fun. But the difference between the Consumer Prices Index and the Retail prices index is not mainly due to the fact that CPI excludes housing costs. At the heart of the difference is mathematics. Each month, 300 agents record 180,000 prices in 20,000 shops and they are turned into an average for the price of, say, bananas. But just what is an average? The RPI uses the simple arithmetic. You add up all the prices for bananas and divide by however many there are. But the CPI uses the geometric mean. You multiply all the prices together and then you take the nth root, where n is the number of items. That method takes account of the fact that if the price of bananas rises, we buy fewer of them or we switch from organic to an own brand bunch. In other words, as inflation bites, we keep the share of what we spend on bananas the same. And it's the different maths which accounts for most of the difference between the two indices. This month, for example, the RPI was 5.3% and the CPI 4%. And that difference of 1.3 percentage points was made up of 1 from the formula and just 0.3 from housing costs. And because the geometric mean is always less than or equal to the arithmetic mean, CPI is almost always lower than rpi. Which is why the government prefers CPI when it raises pensions or benefits, and chooses RPI when it comes to putting up the duty we pay on petrol and alcohol.
- [21:01]
B
Paul Lewis, you're listening to More or less in association with the Open University. I'm Tim Harford and I'm on Twitter as Tim Harford. And you can reach us via More or less@BBC.co.uk or via our website, BBC.co.uk more or less. The website is also the place to listen to our archive and I've no doubt you'll hear many items there in which we've tried to figure out whether some particular policy is progressive or regressive. We just can't seem to leave the subject alone. Progressive taxes, you'll remember, are taxes which make the rich bear a greater burden as a percentage of their income. It seems that not a week goes by without a politician claiming to be progressive or claiming that the other guy is is regressive. So we wondered, although everyone seems to assume that progressiveness is self evidently a good thing, perhaps that isn't true. For instance, if brain surgeons earn more than prison inmates, it's progressive to tax those brain surgeons and send the proceeds to those prison inmates. So we searched high and low for an alternative view. And we found one inside the brain of Jamie White, libertarian philosopher and author of books such as Bad thoughts. Jamie began by arguing that one of the problems with progressiveness is that it's simply not fair.
- [22:19]
E
Basic notion of fairness, until at least socialism got going, was equal treatment. So it's fair to treat people equally. And it's very clear that progressive taxation violates that general principle. So in the case of income tax, the more you earn, the higher rate of tax you pay, which seems clearly unfair. You haven't done anything wrong by earning a lot of money. Why should you be punished?
- [22:42]
B
Can't we just say that a progressive tax is fair, does treat people equally in that it treats all millionaires the same and it treats all poor people the same. What's unfair about that?
- [22:51]
E
It's not unfair in the sense of treating individuals. It doesn't pick on individuals. You're right. It picks on people who have certain characteristics. So, for example, if we taxed all black people a bit more on your account, that would be fair because all blacks get treated the same. We're not picking on you as an individual black person, just as a black person. So the question about whether it's fair is whether the feature that's being picked on is rightly picked on. Now you, I assume, would think it was wrong to tax people more because they're black. You don't think it's wrong to tax people more because they have high incomes? It seems to me arbitrary to think that somebody's done something wrong by earning a lot of money and ought to be punished. Seems almost as odd to me as the idea that they've done something wrong by being black.
- [23:34]
B
I don't think it's obvious that to tax somebody more because they have high income is a punishment or a value judgment that they are bad people. Although of course, in the case of bankers bonuses, that probably is the value judgment that's being made. I think it's simply a case that anybody fortunate enough to find themselves in that category can afford a little bit more. Are you really suggesting that's equivalent to, say, a racist tax or a tax on the bald or a tax on the short or any other offensive.
- [24:03]
E
Whether it's intended as a punishment or not is irrelevant. It is a punishment. That's to say, if I pay double the rate of tax that you do, that's unpleasant for me. That's painful. That's a punishment. Whether it's a value judgment or not, it doesn't make any difference. Now, it was odd that you suggested that people who have high incomes are fortunate. This, I suspect, is the idea lies behind high taxes being fair. It's not really a matter of fortune. People who win the lottery are fortunate and interestingly don't have to pay any tax. Whereas people who earn very high incomes, almost never fortunate. It's usually a combination of skill and application. They may be fortunate to be intelligent, but that's a different issue. And in fact that leads on to my second argument for why progressive taxation is a bad thing, which is that, and this has nothing to do with fairness, it's to do with economics. As people income increases, the marginal value of each new pound or dollar, as I'll say, reduces a declining marginal value of money. To put it simply, £1,000 is worth less to a millionaire than to a pauper. Now there is in that fact, quite an argument in favor of redistribution of wealth. So if I transfer that 1,000 from the wealthy person to the pauper, the loss to the wealthy person is less than the gain to the pauper. However, offsetting that is the fact that my incentive to work reduces. As I earn more, each extra dollar is worth less to me. I'm less inclined to go through whatever pain is involved in getting it. The natural response to that, if you want people to keep being productive, is to reduce the tax burden rather than to increase it. Whereas in fact we do the complete opposite and exacerbate the problem by making the tax rate higher as people earn more. The effect of this is that we disincentivize the most productive people in society to work while trying to encourage the least productive to work by giving them lower rates of tax. That's not strictly how it comes out in the end, because actually marginal tax rates are very high for low income people because of benefits and so on. But take that away. And a progressive tax system basically encourages the least productive to work and discourages the most productive, which seems insane.
- [26:15]
B
Jamie White, send your hate mail or your praise to more or lessbc.co.uk. that's almost all we've got time for. But before we go, there is some unfinished business from last week's programme when we examined the macroeconomic consequences of cutbacks in Trumpetonshire. Many people emailed or tweeted to express their concern for fireman Dibble, who, regrettably, I had to let go. In fact, if you'll excuse me, I must return a call from Mrs. Minton.
- [26:50]
C
Hello, Mrs. Minton here.
- [26:52]
B
Mrs. Minton, it's Tim Harford speaking. I hear Dibble's very upset. People tell me he's been on a week long bender.
- [26:58]
C
I've never seen anything like it.
- [27:00]
B
Is he with you now?
- [27:01]
D
No.
- [27:02]
C
No sign of him, I'm afraid.
- [27:03]
B
I feel responsible, Mrs. Minton. I had no idea he'd take it so badly.
- [27:07]
C
Of course, you didn't know.
- [27:08]
B
I heard he spent most of last night trying to upend the bandstand, but he couldn't lift it.
- [27:12]
D
Yes.
- [27:13]
B
It's too big. It's so tragic to see a former pillar of the community reduced to such a savage, incontinent beast.
- [27:20]
C
Yes, it is. What a pity.
- [27:23]
B
Well, if you missed the story tale, you can, of course, listen again on our website, BBC.co.uk more or less. That's also where you'll find a link to our alternative census. We're closing the survey at midnight on Friday 15th April, so if you're listening to our Sunday repeat, sorry, too late. But we'll be back next week with an analysis of the results. Until then, goodbye.
- [27:47]
A
More or Less is presented by Tim Harford, the Financial Times undercover economist. The More or Less team is Richard Varden, Richard Knight, Wesley Stevenson and me, Hannah Barnes. More or Less is made in association with the Open University.