Generalising

Following on from the previous post, some more notes on a somewhat odd article on FiveThirtyEight.

The bulk of the article seems to be criticising the system for drawing parliamentary boundaries on four points:

1. …the census data used is quite elderly

2. …the boundaries take account of things other than population distribution, such as council areas
3. …some geographical oddities like islands are dealt with as special cases
4. …Wales, Scotland and NI are required to have fixed seat numbers

The main objection here – barring the first point, which I broadly agree with, but which is hard to improve on in most cases – seems to be with the use of non-demographic boundaries in drawing the lines, and I’m just not sure why that is a problem. When electing a representative, the key point is that they be able to represent a constituency; that it be practical for them to find (and be found by) their constituents, that they can identify a distinct section of the population, etc.

Taking the Isle of Wight as an example, it’s big, but how would we split it? We could split it in two – but these would be unusually small constituencies. We could carve off a third and put it in with a Hampshire constituency – but this would disadvantage that third, who would find it more difficult to be in contact with their MP than those who remained insular. In the north of Scotland, you could easily add Caithness to Orkney and Shetland – total population thus “correct” at around 60,000 – but would this have any effect save to further marginalise voters in Lerwick, already pretty distant from the seat of power, and make it vastly less likely they’d ever see an MP?

Political boundaries matter in the same way as geographical ones; if you have existing subnational boundaries like counties or local authorities, then matching the constituencies to these where possible saves a lot of work. People know, much more easily, who their MP and their constituency are; the MP is much more able to focus on issues relevant to a given city or county rather than two or three. In a more practical sense, constituencies have names not numbers; if you’re going to have constituencies with a fixed name, it cries out for a certain degree of geographical coherence.

This is, of course, saying that the absolute “each vote should count equally” principle doesn’t work, that constituencies shouldn’t be made an absolute size and that some people have a theoretically “more powerful” vote than others. This is true – but it’s endemic in the system anyway. If we were going to carefully sculpt constituencies so as to be identically-sized, and demographically balanced, then we may as well drop the constituency idea altogether and go to proportional representation, as in the Netherlands – an ideal case of a balanced constituency.

All this may seem like a bit of overemphasis on some pretty trivial points, but I think those points are telling. FiveThirtyEight ran excellent, intelligent, detailed and basically interesting studies of the last US election; plenty of people outside the US read it avidly for this, myself included. We had to take a lot of its basic assumptions on trust, though; we didn’t have the underlying understanding of the system to do a quick sanity-check and know whether they sounded right or not.

So, come this election, they write about here, and the problems start becoming apparent. To pick a quote from a subsequent article:

Supporters of the Liberal Democrats, while not ordinarily inclined to vote Tory, may do so to push through proportional representation.

This is the sort of thing that… well, hands up, who’s ever met many committed Lib Dems who say, hey, I might just vote Tory to stop those evil Labour people getting in? Um.

Yes, there will be swing voters from LD to Conservative – though not as many as sometimes thought – but those will be classic swing voters; they’ll swing based on the economy, or other overall political issues, or simply because they no longer see the Tories as… quite as messy as they used to be. They won’t be core LD voters, devoutly wedded to PR, who are trying to produce a hung parliament in the hope that this will somehow extort PR from the resulting trainwreck.

In a comment on the last post, cim points out another delight from the same author:

And then there was this one from the EU elections where they decided that it was meaningful to group Plaid with the BNP as “nationalists”. I conclude that they don’t actually know anything about UK politics, and look forward to more amusement as their election coverage continues.

I think that last comment says all I was tempted to, really.

There’s a slightly odd post on FiveThirtyEight which I noticed: Breaking News: U.S. and U.K. Redistricting Processes Equally Boneheaded. I wonder if this is what we Brits sound like when we pontificate on US politics.

The part I couldn’t help laughing at.

In the UK by contrast, the net effect of these aspects of the mapping system is one-sided against the Conservatives. The Conservatives have little presence in Scotland (where they hold 1 out of 59 seats) or Wales …

…and a nice large illustration of Scotland to show how terribly the constituency boundaries oppress all those hordes of Tories. Um.

This is a problem of first-past-the-post, first and foremost; there are arguably ancillary problems with the system being marginally skewed pro-Labour due to urban voting concentrations etc etc, but they just don’t come into play at this level. Why?

The reason the Conservatives don’t have many seats in Scotland is – let us be sadly honest – that the average Conservative parliamentary candidate is about as popular on a Scottish doorstep as the rent collector. In 2005, they were fourth, with one seat and 15.8% of the votes; in 2001, the same, with 15.6%. In 1997, they took a staggering no seats on 17.5% of the vote – whilst the SNP, on 22%, took a mere six out of 72.

Coming fourth with ~16% of the vote in an overall campaign, without a great deal of concentration of your voters in specific regions, is almost guaranteed to be a damp squib in a first-past-the-post system. It doesn’t matter if you’re the Conservative Party or the Labour Party; any marginal systemic biases the system may have are vastly wiped out by the sheer scale of the problem.

We can see this using a (very non-Tory) example – look at the Liberal Democrats in 1992, across the whole UK. They took 18% of the vote, but only twenty seats – 3%. In 1987, the Alliance took 22.5% of the votes, but still only ~3.5% of the seats; in 1983, 25% of the vote… and still only 3.5% of the seats.

Would it be better under a non-FPTP system? Well, we actually have a perfect laboratory system here – Scotland. Let’s look at the 2007 Scottish elections – the Conservatives took 16.6% of the “constituency” vote and secured 5.5% of the seats, but 14% of the “list” vote – and 23% of the seats. This puts them on an average share of 15.3%, with 13.2% of the seats – not particularly unrepresentative number-wise, but definitely polling around the same as they did for Westminster elections.

In 2003, the same pattern – 16% of the overall vote, 14% of the overall seats – and again in 1999 – 15.5% of the votes and 14% of the overall seats. In both cases, they took the lion’s share from the proportional-representation list votes; indeed, in 1999, they didn’t get a single constituency seat.

So, yes, if there were not a FPTP system the Conservatives would do better; there is one, and they get hammered. But to use the situation of Scotland – where the Conservatives manage to be a third-place party if they do well – as though it is indicative of the rest of the UK is either disingenuous or just confused; the political realities of Scotland just aren’t the same as the nation as a whole. We cannot logically leap from “the FPTP system is wildly biased against a local minor party”, when we know that FPTP is already biased against minor parties, to “the FPTP system is therefore wildly biased against that party when they compete nationally as a major player”.

The government has published its response (PDF) to the review of the 30-year rule.

The rule itself is a bit of a historical oddity – a blanket rule saying that government records should be opened after thirty years is a hangover from before the “on-demand” FOI Act, where you can apply for material regardless of its age. However, it survives in Part VI of the legislation; any record is deemed a “historic record” at thirty years after creation, and as such loses a lot of its ability to be exempt from disclosure. A document which was originally kept secret because of commercial confidentiality, for example, loses this exemption after thirty years.

The review recommended reducing this to fifteen years; the government response has advised twenty. In Scotland, which is subject to a similar but not quite identical piece of devolved legislation, the time limit has recently been dropped directly to fifteen years, in three tranches, with full release being completed by later this year.

The counterproposals are, on the whole, sensible. Other than the fifteen/twenty year change…

• The thirty-year rule will remain for a set of material: anything commercially confidential, anything sensitive touching on the devolved administrations, and the “effective conduct of public affairs” clause. It’s a little vaguely worded, and this worries me that in the eventual legislation we may find a much more flexible (ie, long) term for some material. How long is the average PFI contract for, for example?

• The new system will be phased in a year at a time – so, for example, we would release records from 1981 and 1982 in 2011, then 1983 and 1984 in 2012, letting the threshold creep steadily down to 20 years. This is in contrast to Scotland, which tried to clear the backlog in a couple of years – but then, it’s a bigger problem.
• There is to be a presumption in favour of anonymity for civil servants; names in records released as historical documents will be redacted unless decided otherwise or they’re very prominent.
• The legal position of communications with the monarch is to be clarified; records will be absolutely exempt for twenty years, or five years after their death, whichever is the longer. This also applies to the heir and the second-in-line to the throne, but all other members of the Royal Family have a qualified exemption – they’re presumed closed, but can be released under request.

On the whole, good things. I’m moderately concerned by the alterations to the commercial confidentiality clause – by the time this report turns into a statute, that looks like it may drift into actually making things worse than the current state of affairs – but it is, to mangle the old quote, currently moving in the direction of goodness. The anonymity proposals I am not sure about – there are good reasons for it, but there are also serious impracticalities, especially from a historical perspective. The amount of redaction appropriate at the twenty-year mark may well not be the same as the amount appropriate a hundred years on, but if we only have the one point of release, we’ll be stuck with whatever decisions were made at that point. Hmm.

Some details emerge from the report that reflect things which’re already happening. The government has formally clarified that “special advisors” are indeed legally civil servants, and subject to the provisions of the Act; they’re also grinding slowly forward with making a number of additional bodies, including UCAS and the legally ambivalent academy schools, subject to the Act.

So, twenty versus fifteen. This is an interesting debate.

A key part of the philosophy behind the thirty-year-rule is to avoid governmental records being turned into political ammunition, and relatedly to reassure the authors that they don’t have to keep material off the record for fear of it later being used against them. (It’s assumed that once you’re retired or dead you can live with the opprobrium.) The main example of this is ministerial papers, but from this standpoint, fifteen years is often just a little too short – it would have meant, for example, that papers relating to Callaghan as Chancellor (1964) got released when he was Prime Minister (1979), and in the event Brown wins the coming election, he’d have the same issue.

The current government will have lasted fourteen years, almost at that threshold, and the previous one eighteen – it’s true that no-one held senior office through the Conservative period, but it does demonstrate how prolonged a spell in power can be. If the modern trend is to parties holding power for long periods with occasional grand shifts, which it may well be, twenty years is a little safer; it also covers for the rare case of someone, like William Hague, whose ministerial career seems likely to bracket one of the prolonged spells of opposition.

There’s also a good case to be made that for politically touchy material like, say, Brown’s records from the early chancellorship, it wouldn’t be released regardless of the threshold, due to the section 36 exemption, likely to prejudice the conduct of public affairs. But a fifteen year rule would imply the time and expense of specially auditing an entire tranche worth of departmental and cabinet records to decide if they’re sufficiently sensitive to be restricted, then doing it again a couple of years later once the person in question is out of office in order to clear them for publication. The extra few years would, presumably, sharply cut down the number of such cases; is it a worthwhile payoff from an efficiency standpoint?

I am ambivalent on this one, I admit.

In the Scottish context, of course, this is currently academic. Holyrood is not overly concerned with the prospect of a former Scottish Secretary looking a little silly, and there won’t be any Scottish Executive (as was) records released until well into the coming decade. On current form, it seems likely that with the less monolithic nature of devolved politics we’ll have shorter ministerial tenures, as well, and thus less need for the extra few years.

We shall see. It’s an unfortunately timed announcement, coming as it does just before an election, but hopefully whoever gets in will pick up the ball and run with it. It’s not something either side would particularly benefit from more or less than the other, after all – they all have ticking skeletons in the closet waiting for the cabinet papers to be released.

A couple of interesting blog posts on the BBC – part 1, part 2 – about a recent set of crime statistics publicised by the Conservatives.

The basic gist of the Conservative claim is that violent crime is vastly increased over the past decade; the basic problem is that the method of recording violent crime changed in the middle of the period, to a much more “permissive” approach, where police were obliged to record a complaint rather than dismissing it. Which, unsurprisingly, tends to lead to a lot more reported crime, without actually saying anything about the underlying crime rates.

I suppose in an ideal world Labour would be running a campaign of “Do you really want to be governed by people who can’t read printed warnings on graphs?”, but sadly all we’ll get is a bit of he-said-she-said over the next two weeks and a few more people will be left beliving that the country is a far scarier place now than it ever was.

An interesting new project: Where Does My Money Go? [currently an alpha version; details; announcement]

Basically, interactive visualisations of UK governmental spending, broken down by topic or by region. There’s also a time-series function, which is quite interesting to see – overall government spending, as a proportion of GDP, has just hit the bad old days of 1992.

Things that currently stand out as major issues:

• uncleaned data means hideous governmental terminology – “n.e.c.” everywhere
• expanding on that, the data needs a bit more organising – ensuring you can switch between subdivisions on the national-level, for example, would mean linking the ‘economic > transport’ sections together in the same way that the ‘economic’ sections currently are
• the main charts are a little ambiguous as to which circles are subdivisions of each other
• there’s no way to apply the time-series graph to “second level” data – so you can compare “general public services” spending over time, but you can’t compare the amount spent on debt servicing
• mousing over a column really should display its numeric value

On the whole, though, promising – definitely worth ten minutes playing with. Gets the concept across a lot more clearly than the bare figures might.

In brief: the US House of Representatives passed the health care reform bill. It is called the Affordable Health Care For America Act and expands federal healthcare provision enormously – 36 million more people will be eligible for Medicaid, most employers will be required to provide healthcare coverage for their workers, and there will be a government-funded “public option”. Also notably, health insurers will be prevented from refusing coverage based on medical history (no more gender-based “pre-existing conditions” such as pregnancy, rape and domestic violence) and the exemption for insurance companies from antitrust legislation will be repealed.

So far, so hoopy. The Stupak Amendment, with which this Act has been passsed, is as follows:

“No funds authorised or appropriated by this Act… may be used to pay for any abortion or to cover any part of the costs of any health plan that includes coverage of abortion, except in the case where a woman suffers from a physical disorder, physical injury or physical injury which would… place the woman in danger of death unless an abortion is performed… or unless the pregnancy is the result of an act of rape or incest.”^[1]

In other words, to get this Act passed, someone had to be the sacrificial lamb and 150 million American women were it. (Also, something else I have just spotted – the obvious women are excluded, women who want abortions for what are nauseatingly called “social” reasons, because pregnancy is not the right thing for them, but also, women who have mental illnesses which pregnancy would exacerbate are excluded, too.)

I actually have no further commentary to make on the issue, and I wondered if that were just me, but actually, I think there is nothing very profound to say about it. Institutional politics, particularly in the United States, is boring and it doesn’t yield to analysis. Feminist analysis of the narratives of privilege and oppression, that is interesting; so is sociological thinking about why people think the way they do such that amendments like this are seen as a good idea, but on the institutional level of why, in the specific instance, the House of Representatives has voted like this, I’m coming up with nothing. They voted like this because they’re misogynists, fundamentalists, or spineless; you can lobby them, but to be effective, you either run for the House of Representatives or wait for the current incumbents to die, or both. You can’t argue, you can’t write about women’s rights to their own bodies, you can’t talk about restriction of reproductive options as a form of control of women. Well, you can, but it’s a category error to think you can convince an edifice of misogyny to change their minds because that, I think, fundamentally misunderstands why they hold the opinions they do – it’s not because they arrived at them through logical argument.

(Evidence in point: thirty-nine Democrats voted against the reform bill. Twenty-one of them, besides Stupak, voted for the amendment. Institutional politics defies logical analysis.)^[2]

It will go to the Senate, but I’m not optimistic.

[2] From here. And yes, lawyers are allowed to run a defence in the alternative, but I suspect it’s not the same thing.