Monday, November 28, 2011

The confidence trick

This is the heart of the narrative that holds the government’s economic policy together:

Towards the end of Labour’s time in office, government borrowing was threatening to spiral out of control. Questions were being raised about our national credit rating, and the bond markets were pushing up the cost of borrowing. The coalition’s determined programme of larger and faster spending cuts has rescued the UK, creating much-needed confidence in the prospects for deficit reduction and, as a result, economic growth.

So: what’s the evidence?

I’ve looked at the yield on 10-year government bonds – a standard measure of the interest rate the government pays to borrow – and worked out how it changed when a select half-dozen events happened. I’ve taken the average bond yield over five trading days before each event and over five days after, to smooth out the effect of momentary blips, and compared each pair:

  1. On 29 March 2009, credit-rating agency Standard & Poor’s warned that without a stronger deficit-reduction strategy, the government’s AAA credit rating might be at risk. Between the five days before this and the five days after, 10-year bond yields rose by 0.06%, making borrowing a bit more expensive.
  2. On 21 May 2009, Standard & Poor’s formally moved the UK government from a “stable” outlook for its credit rating to “negative”. Bond yields rose by 0.38%.
  3. On 26 January 2010, the fund management firm Pimco (a major player in the global bond markets) said that UK national debt was a “must to avoid” as it was “resting on a bed of nitroglycerine”. Bond yields rose 0.002%.
  4. Coalition negotiations meant that the change of government took longer than usual. But between the five days before the election (6 May 2010) and the five days after David Cameron became PM (11 May), bond yields fell by 0.21%, making it cheaper for the government to borrow.
  5. On 22 June 2010, George Osborne produced his first Budget. Bond yields fell by 0.18%.
  6. On 22 November 2010, the government published its full spending review. Bond yields fell by 0.04%.

Well, this all seems to support the government’s case. There’s only one small problem: all these numbers are for US government bonds, not UK ones.

Here, by contrast, is what happened to UK government bond yields:


So in all six cases, when there was supposedly bad news for UK bonds, US bonds took a bigger hit, and when there was supposedly good news for UK bonds, US bonds got a bigger boost.

[Update: For a longer perspective than five days, I’ve averaged over the six months before the election (November 2009–April 2010) and the following six months (May–October 2010). From the first period to the second, UK government bond yields fell by 0.64%, but this was beaten by US bonds, which fell 0.75%. This simply does not support the claim that the coalition and its policies have been key to keeping UK borrowing costs down.]

In fact, US and UK government borrowing costs are very strongly correlated these days (since 2008, a massive +0.9):


So, unless the US markets are hypersensitive to what happens in Britain, something much bigger has been going on.

Before I get into what that is, let’s see how the actual amount of government borrowing relates to the cost of borrowing. The government’s story says that these are positively correlated – as does common sense. Surely, the more you borrow, the bigger a credit risk you become and people will charge higher rates to lend to you. So, if the amount of money the government is expected to borrow rises or falls, bond yields should move in the same direction.

Let’s see.

Every month, the Treasury collates the latest set of independent economic forecasts. Sometimes these look only two years ahead and sometimes they look farther, so the data I’ve put together is a bit irregular.

First of all, in Labour’s final year, there were five sets of independent forecasts of government borrowing that covered 2010/11 and 2011/12. This chart shows these alongside monthly averages for government bond yields (yes, UK ones):


As Labour’s time in office drew to a close, independent expectations of government borrowing were falling, not rising. The correlation between the level and cost of borrowing (albeit from a tiny sample size) was strongly negative. This means that fears of higher deficits could not have been responsible for the rise in bond yields that happened during this time. (And bear in mind that US bond yields also rose over this period.)

Secondly, over the lifetime of the coalition, here’s a chart showing monthly bond yields alongside independent forecasts of government borrowing across 2011/12 and 2012/13:


Again, the correlation is negative. As borrowing expectations have risen over the last six months, the cost of borrowing has fallen. This means that increased confidence about lower deficits could not have been responsible for the fall in bond yields. (And bear in mind that US bond yields also fell over this period.)

I’ve shown one chart for before the election and one chart for after. Wouldn’t that miss the change between the two governments, when both borrowing forecasts and the cost of borrowing both fell? Yes. But this fall in UK bond yields was matched – in fact, exceeded – by the simultaneous fall in US bond yields, which really can’t be explained by changes to UK fiscal policy. Something bigger is going on. But what?

The answer has two parts. For the first part, we can turn to Nick Clegg.

After the election, the Lib Dems abandoned their more gradual manifesto proposals for reducing the deficit and signed up to the Tories’ plans for faster, bigger cuts. Clegg said that the situation in Greece, which exploded shortly before the election, had changed things. In that, at least, he was right. However, he meant that there was a danger of the sovereign debt crisis spreading to us, so we had to be extra cautious with our finances. But the real lesson is different.

Investors have to put their money somewhere. Bonds issued by developed-country governments are normally seen as ultra-safe, and the yields paid on these bonds are normally quite low, as the safety is their main selling point. For higher returns, you can make a higher-risk investment elsewhere, if you dare. But the crisis in the eurozone meant that the risk-averse large institutional investors lost their appetite for Greek (and later Portguese, Spanish and Italian) government bonds.

So they needed to invest somewhere else.

On 23 April 2010, the Greek government requested an IMF/EU bailout. Between the five trading days before this desperate act and the five after, UK government bond yields fell by 0.07% and US bond yields by 0.04%. Rather than a fear of contagion hitting all sovereign debt, those countries that the markets saw as nowhere near the danger zone – including the UK – found that they could borrow more cheaply. Increased demand from fleeing investors brought down our bond yields.

And this persisted: 23 April 2010 was the day that US and UK bond yields both began a largely uninterrupted four-month fall.

The second part of the answer – what bigger thing is going on to make non-eurozone government borrowing so cheap even while so much of the stuff piles? – is the condition of the economy more broadly. But it relies on the same principle: investors have to put their money somewhere.

Jonathan Portes of the National Institute of Economic and Social Research reasons that lower bond yields would be a sign of confidence and economic strength if they were associated with a better stock-market performance; but they’d be a sign of weakness if they went hand in hand with stock-market falls. He finds that, since the coalition was formed, there’s a “strong, significant and positive” correlation: bond yields fall when the stock market does as well.

He’s right. I’ve checked, and from June 2010, government bond yields and the FTSE have shown a correlation of +0.56. The persistent weakness of the economy as a whole (in Britain and elsewhere) means that private-sector investment prospects look poor. The result of this pessimism – as with the flight from Greek bonds, although at a less frantic rate – is that investors who might have put their money in the stock market are instead going for government bonds, further pulling yields down.

This can also explain the counterintuitive positive correlation I noted earlier between the cost and the predicted amount of government borrowing. As the economy falters, investors move from stocks and shares to the safety of government bonds, driving down the yield on them. But a weaker economy also means lower tax receipts and higher welfare spending, so borrowing forecasts go up.

And that’s exactly what’s happening. The government can borrow so much so cheaply not because its plans (which keep getting rewritten for the worse) inspire such stunning confidence but because the rest of the economy inspires so little. It’s a sign of failure, not success. The story the government’s trying to spin us is entirely the wrong way round.

A few closing remarks, because I’m not in this post advocating any particular fiscal policy over another:

  • While the government’s deficit-reduction plan has had little discernible effect on the cost of its borrowing, that doesn’t mean there isn’t a difference between safe and risky borrowing. There is. Both this government and the last have clearly been seen as being on the right side of the line – by the markets if not by the commentators. Despite the political rows, the difference between the Osborne plan and the Darling plan wasn’t massive.
  • While nobody knows exactly where the line is, it’s clear that a few extra tens of billions of borrowing wouldn’t spark a crisis; indeed, feeble growth and high inflation mean this is exactly what’s happening. But, conversely, it’s far from clear whether a few tens of billions of stimulus would really make much difference to growth.
  • Crisis avoidance isn’t the only reason to reduce borrowing as quickly as we reasonably can. A 12-digit deficit, even on a low interest rate, still costs a lot, and servicing debt interest isn’t the most socially or economically useful thing to do with taxpayers’ money.
  • Economics isn’t the only reason to be sceptical about the virtues of rapid public spending cuts. It’s not just faceless bureaucrats and feckless scroungers that will suffer as a result.


Update: Chris Dillow also ponders what low bond yields mean and what the stock market can tell us. but he looks at the FTSE small caps index rather than the more globalised FTSE 100. I think the lesson is that the economic slowdown that's pushing bond yields down is not confined to the UK - just as my chart of US and UK bond yields would suggest.

Monday, November 21, 2011

Poor school results

One of the biggest things affecting a school’s exam results is not what happens inside the classrooms: it’s the backgrounds and circumstances of the children the school lets through its gates in the first place. The more children from poorer families a school has, the worse its exam results tend to be.

The graph below shows almost every state secondary school in England [1]. It plots how many of each school’s pupils qualify to receive free school meals against how many of each school’s pupils get at least five A*-C grades at GCSE including English and maths. Roughly speaking, it plots poverty against results [2].


There’s a pretty good correlation, of -0.57, between how many pupils are on free school meals and how many get five A*-Cs including English and maths. As poverty rises, exam results fall (although this fall levels out after around the point where a fifth of a school’s pupils qualify for free meals).

The grammar schools skew the picture – but only a little. Among comprehensives, the correlation is -0.55 – only slightly lower.

Most grammar schools are clumped up in the top left corner of the graph. They get very good results, which is hardly surprising given their intake. But selecting for ability also tends to select for wealth. The average selective school has just 2.6% of its pupils on free meals, while comprehensives average 16.7%. In fact, there’s not a single grammar school in all the land with even three-quarters as many kids on free school meals as the average comprehensive.

The next graph tells a similar story to the first, but it just covers comps, and it gets rid of a lot of the noise by grouping together schools by how many children on free meals they have (0-0.9%, 1-1.9%, 2-2.9%, 3-3.9% etc.) and shows the average results score for each such group:


But the correlation figures above understate the strength of the relationship. There are wide regional variations in prosperity across England, and of course parents don’t navigate the education system on a national scale: school choice works locally.

So by breaking the figures down into individual local education authorities (LEAs), we can see how strong the relationship is between poverty and results on a scale more relevant to parents’ choices about which neighbourhood to live in and which schools to apply to.

I’ve checked every LEA with 20 or more secondary schools (the smaller they are, the less reliable the correlation stats become [3]). There are 45 such LEAs, out of a total of 150.

The correlations range from -0.57 (Cornwall) to -0.90 (Leeds) – with an average of -0.76. So if you look into an individual LEA, the local link between poverty and results is usually much stronger than across England as a whole [4].

(LEAs with selective schools appear to have only slightly stronger poverty-results correlations than all-comprehensive LEAs.)

All this just shows how strong the link between poverty and GCSE results is. The exact nature of the causation is more complex to untangle, but it’s going to involve the fact that struggling to make ends meet makes it harder to devote time and resources to one’s children’s development. It’ll also, conversely, involve the fact that schools with good results become oversubscribed, driving local house prices up so that these schools become fuller and fuller of children with well-to-do parents.

One way to stop the education system from segregating children is to prevent popular, higher-scoring schools from cherry-picking the most promising ones. Banding and lotteries are efforts to achieve this, but I’m not aware of any compelling evidence either way on their effects.

The government’s big plan is the pupil premium, Nick Clegg’s favourite boast but something that appeared in all three main party manifestos in 2010. It’s an interesting way to motivate higher-scoring schools to admit more poorer pupils. Whether it succeeds or fails will depend on whether the extra cash can overcome the fear of being dragged down by the chavs.


[1] 2009/10 data, from the ‘Secondary’ tab of the ‘School spending – all data’ spreadsheet produced by the Department for Education. I’ve had to take 23 schools out of the calculations as there aren’t exam results figures for them.

[2] Qualifying for free school meals is an imperfect measure of poverty, related as it is to receipt of certain benefits. But it’s a decent indicator. Also, there are many other ways to measure how good a school’s results are than the number of kids getting five A*-Cs including maths and English: you could raise or lower that bar as you liked. This, however, is the data I have.

[3] Out of curiosity, I also looked at each of the 26 LEAs with under 10 schools in the data set. All have similarly strong negative correlations, for whatever that’s worth – except for the odd (but tiny) outlier of the Isle of Wight, whose five schools form a decent line sloping the other way, for a correlation of +0.51. I don’t know what, if anything beyond random variation, is going on there.

[4] How can pretty much every LEA have a stronger correlation between poverty and results than the national figure? Simple. This chart shows the schools in Bradford and Northamptonshire:


The red dots are pretty clearly grouped around a line (Northants has a correlation of -0.81), and so are the blue dots (Bradford’s is -0.76). But they’re different lines: Bradford is, on the whole, poorer and has more schools getting lower results, but with increasing poverty making less of a difference among its schools than in Northants.

So putting all the dots together blurs the clarity of the lines a bit: the overall correlation across the two LEAs between poverty and results is -0.73 – still high, but lower then either alone. The same thing happens, on larger scale, when looking at all the schools across England together.