Will a hung parliament undermine fiscal discipline? Here’s some data.

May 8th, 2010 by Ben Goldacre in bad science, politics | 36 Comments »

Lots of people have been explaining that a hung parliament would be unable to rein in public spending. Do the figures support this idea? I’m sitting in my kitchen with a political data nerd who’s just pulled this gem out of his back pocket. It’s from this document on this page of the Institute for Government website, and it’d be nice if someone like this could make a less ugly version. The y-axis is how much you borrow, the x-axis is the size of your biggest party, and it seems to show that countries with single powerful parties tend also to have worse government debt. Just a piece of the puzzle, but I give it to you for free.

image


++++++++++++++++++++++++++++++++++++++++++
If you like what I do, and you want me to do more, you can: buy my books Bad Science and Bad Pharma, give them to your friends, put them on your reading list, employ me to do a talk, or tweet this article to your friends. Thanks! ++++++++++++++++++++++++++++++++++++++++++

36 Responses



  1. ellieban said,

    May 8, 2010 at 6:55 pm

    USA, JPN, GBR, AUS and, would that be Greece? Five bastions of electoral fairness and fiscal propriety!

    Let’s get this new collaborative politics underway, roll on a liberal alliance of every party except the Tories.

  2. FelixO said,

    May 8, 2010 at 7:12 pm

    Definition of “General Government Structural Balance”
    stats.oecd.org/glossary/detail.asp?ID=3343

  3. muscleman said,

    May 8, 2010 at 7:33 pm

    Under a series of coalition and minority governments NZ govt has run a budget surplus which was used to retire debt. The argument at the moment is whether or not the current minority Tory govt should borrow to invest in the recession (which has hit NZ very mildly in comparison, the NZ$ is high).

  4. SteveGJ said,

    May 8, 2010 at 7:36 pm

    I’d like to see a scatter chart with more than the borrowings in just one year and one election to be convinced (and a correlation cooefficient would be good too – not just a visual). It would be better to see the long term relationship of public finance to single party domination.

    However, we do, of course, know that some economies manage their public finances whilst running systems with perpetual coalitions. For good or ill, our political system isn’t much used to working that way (at least in the UJ parliament – Scotland and Wales have more experience). As somebody pointed out in an interview, generally the politicians at the top can usually cope, the general public can also understand it. However, it’s the party activists where the trouble lies – with the Conservatives and Labour at least, they are from a winner takes all culture.

  5. aluchko said,

    May 8, 2010 at 7:43 pm

    I’m not sure if I buy this correlation.

    Looking at Canada, one of our bigger majorities came in 93 when the Liberals came in after Progressive Conservative fiscal mismanagement caused the party to implode. The Liberals had several years with on effective opposition, and in that atmosphere they managed to take unpopular actions which turned a huge deficit into a huge surplus. The Canadian province of Alberta, while it does have the advantage of oil reserves, hasn’t had an effective opposition in decades and has the best economic situation in the country.

    I wonder if what you’re seeing is parties being voted in with big majorities as a result of previous fiscal mismanagement. Part of the reason the Democrats won big in the US is because the economy was in such bad shape. And part of the reason Canada has a minority government is none of the ruling parties screwed it up bad enough to cause a major voter reaction.

    My own suspicion is that large majorities allow ruling parties to take unpopular action to keep the government books in order, kind of a benevolent dictatorship. This is my one real reservation about proportional representation, that parties will never have the power to do the unpopular but necessary actions to keep the books balanced.

    It would be interesting to see the correlation between change in budget deficit over a governments term vs. size of largest party. I wouldn’t be surprised if the bigger parties did better.

  6. Thomas said,

    May 8, 2010 at 8:22 pm

    Here are the numbers, if anyone wants to play with them:

    spreadsheets.google.com/ccc?key=0AqIElKUDQl8tdFFBRHBLaHdKemdaWmtjY1NhWmJwamc&hl=en_GB

    (Extracted from the graph using g3data)

  7. shockdoc said,

    May 8, 2010 at 10:38 pm

    lm(formula = Largest.Party ~ GGSBGDP, data = brw)

    Residuals:
    Min 1Q Median 3Q Max
    -19.8290 -7.7681 0.4342 6.6420 15.4465

    Coefficients:
    Estimate Std. Error t value Pr(>|t|)
    (Intercept) 28.1234 5.2224 5.385 2.85e-05 ***
    GGSBGDP -2.7417 0.9293 -2.950 0.00791 **

    Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

    Residual standard error: 10.4 on 20 degrees of freedom
    Multiple R-squared: 0.3033, Adjusted R-squared: 0.2684
    F-statistic: 8.705 on 1 and 20 DF, p-value: 0.00791

    Fairly modest predictor, y’d have to say.

    Ben have a look at the original numbers.. are they anything like what the man (who’s probably failry drunk by now) in your kitchen said they’d be??? (Don’t trust lovely Linux g3data, used it before)

  8. CoralBloom said,

    May 9, 2010 at 1:19 am

    Well what do you know!

    So bribes to the electorate?

    I’m grateful for the data – been wondering where to find decent data to play with.

  9. l0ki said,

    May 9, 2010 at 8:12 am

    I’m not sure if the question is so much whether or not a coalition government would be able to rein in spending so much as whether they would be united enough to make big decisions as and when necessary.

    While getting into lots of debt before a credit crisis isn’t ideal, the national debt as a whole isn’t necessarily a bad thing. The fact that countries with a less assertive governments don’t tend to spend as much money on big projects probably isn’t really that surprising.

  10. Moth Eyes said,

    May 9, 2010 at 10:45 am

    While not particularly strongly correlated, the interesting thing is that they are correlated in a direction opposite to the one the pundits are suggesting.

  11. mo-seph said,

    May 9, 2010 at 1:52 pm

    Slightly less ugly version (with a smoothed mean) here:
    www.mo-seph.com/files/StructuralBalance.png

  12. msjhaffey said,

    May 9, 2010 at 4:17 pm

    I think there are two distinct points here.

    The first is about governments which go on a spending spree. The other is about reining in excessive spending. What we need is the second, but I am not sure that your chart addresses this directly.

  13. SteveGJ said,

    May 9, 2010 at 4:25 pm

    Quite apart from the issue of the longer term issue of the voting system, the immediate problem is one of putting together a deal such that a Government can be formed which will produce a credible plan for public finances. If this isn’t done, then the markets are going to put up the price of gilts which means that the freedom of the next Chancellor to make decisions is going to be severely constrained. That’s despite much of our dept being relatively long dated compared to that of the Greeks and Portuguese.

    This is going to be a major test of our politicians. We’ve had the nonsense of an election where all the parties (to one extent or another) have been afraid to tackle this head on for fear of being slaughtered in the polls (Gordon Brown uses fear of public expenditure cuts in much the same way that past Conservative governments have played on fears of increased taxation – of course the truth is we are going to get both, whoever is in power, it’s just the timing that matters).

    So all this talk of process is fine and dandy – but it does not tackle the immediate issues.

    ps. just looking at my gas and electricity bills and it remainds me that, in his very first budget, GB cut what was to be the rate of VAT on fuel to 5%. That was a short-sighted, pouplist measure which ran against any sensible green agenda. It would make sense to up the VAT on fuel combine with some policies to reduce the impact on the poor. If people are after evidence-based policies, then higher fuel charges are for sure a way of improving energy efficiency.

  14. fontwell said,

    May 9, 2010 at 5:31 pm

    While this is very interesting SteveGJ makes a very point. For instance the USA is presumably showing the majority gained by Obama but I hardly think it is reasonable to correlate that large majority with their current debt which was pretty much run up under Bush, whose majority was much smaller.

  15. SteveGJ said,

    May 9, 2010 at 6:07 pm

    @fontwell

    Whilst it’s nice to think somebody believes I’ve made a good point, I think the position with the US is rather complex. I’m not sure the graph’s presentation of how many seats is directly comparable with the UK position as the systems are very different.

    In the UK we effectively have a single elected parliament with only a very weak separation of powers between the executive and legislature. In effect, if you get in with a big majority then the executive in the UK have a lot of leeway (OK – sometime the back benchers get a bit uppity as they did with Gordon Brown abolishing the 10% tax band). However, in general, the executive get their own way. The second house provides only a very weak level of balance – by constitutional convention it never vetoes budgets for instance and it has no electoral mandate.

    In the US it’s a whole lot different. Quite apart from the whole state system, the federal government setup is a lot different. The executive, in the form of the President, has a lot of powers, but there is a system of checks and balances by Congress. It’s often the case that the president does not command a majority in Congress and there is a lot of bargaining that goes on in setting budgets. In some cases it is almost an uneasy temporary alliance of a president with at least some politicans from the opposing party.

    Also, the various party members in congress tend to be rather more independent than the UK equivalents (ofen reflecting local interests). We do, of course, have a rather patchy and inconsistant set of devolved bodies in the UK (Scots parliament, Welsh and NI assemblies, some city mayors and assemblies), but the powers are widely inconsistant between different parts of the UK. In fact it’s a complete mess.

  16. Nile said,

    May 9, 2010 at 6:24 pm

    Hmmmm…. That doesn’t look like a random scatter: I can see two lines, one above and parallel to your line of best fit (BEL to USA) and one suggesting an opposite correlation in the group from SVK to NZD.

    Admittedly, the human eye is a poor tool for statistical analysis – the mind seeks patterns where none exist – but this data set looks like a case for further analysis, rather than a flat dismissal that no strong correlation was found, or that a weak correlation exists between ‘A’ and ‘B’.

    Consider the case of a clinical trial in which there is a beneficial effect (with a dose-response curve!) on one group of patients, and a negative effect on another group in the trial: overall it’s inconclusive if the study fails to separate out the two groups. Worse, it might be published (or rather, interpreted, as inconclusive trials rarely see the light of day) as ‘The treatment has no effect’ when both the positive and negative effects are significant.

    Returning to our fiscal data set, I would speculate that the upper line – if real – says something about majorities in two-party systems, and the lower line (again, it there’s a real relationship in the data) says something about multi-party polities and coalition governments.

    I look forward to reading a more detailed analysis.

  17. Sqk said,

    May 9, 2010 at 10:47 pm

    Please define: “political data nerd”. There’s a lot that could fit that description.

  18. shockdoc said,

    May 10, 2010 at 1:28 am

    RE: Please define: “political data nerd”. There’s a lot that could fit that description.

    Am I a political data nerd?: Scatterplot with information added on GDP Per Capita (pointsize)and Ratio of Health to Defense spending as a % of GDP (colour, red:blue):

    bit.ly/cuLyvZ

    On the basis that one might hypothesise and increase (or decrease) in wealth, “democratisation” or even “socialisation” with smaller parties … but.. looks like not.

    might as well vote by throwing yer granny down the stairs and seeing which way up she lands

  19. Amsterdamtomsk said,

    May 10, 2010 at 7:52 am

    I’m not a political nerd but I do think about economic data as part of my job.

    I was sent this chart as an “aha” moment by a facebook friend so I thought I’d stop by.

    Some thoughts:

    Firstly, don’t worry about the differing electoral systems – let us not forget that Belgium was without a functional political leadership for several months.

    Secondly don’t worry about applying “seats won at last election” as the measure of stability in these radically differing systems – Australia vs France for example.

    Thirdly, don’t worry about not correcting for currency fluctuations (is this in local currency or USD) or taking any sort of longer term look at the surplus data. Fourthly don’t worry about the relative size of the economies in question.

    Lastly, don’t worry that the graph doesn’t begin to get close to proving that a coalition is better at controlling government spending than a single party over time (no correction for stable near single party coalitions and unstable coalitions, and no attempt to suggest a cause just a correlation.)

    Also, just to throw my cat firmly among the pigeons, anyone want to guess at the countries globally with the largest surplus. Do we think China, Saudi Arabia, and some other size of largest party = 100% of seats won upset the correlation?

    So – what we should all be worried about is that not one of these democracies can run a surplus, not one. Arguing that coalitions are the least worst way of breaking the economy isn’t particularly useful.

  20. anon3455 said,

    May 10, 2010 at 10:25 am

    This is such a stupid graph. As if essentially two-party system could be usefully compared to multiparty system like this.

    Eg. Finnish system is essentially a three-party system. There are three major parties out of which two always form the government with some smaller parties. I would think it is more significant which parties form the government than the vote share of the largest party.
    However, the spending versus _the_government_’s share of votes could be interesting. Of course, using only one year’s data contains so much noise (parties, economic situation) that any trend found can not be taken seriously.

  21. daveneedstoknow said,

    May 10, 2010 at 2:36 pm

    Interesting – I used your spreadsheet, thanks Thomas and plugged in the income inequality figures from en.wikipedia.org/wiki/List_of_countries_by_income_equality I ended up with this graph; sites.google.com/site/wellandhung/incomegap which suggests an equally strong correlation that countries with large majority governments also have the biggest gaps between the richest and poorest 10% of the population. Japan sticks out as an exception here.

  22. elvisionary said,

    May 10, 2010 at 2:38 pm

    @fontwell. The point isn’t about the size of the majority, it’s really about the electoral system. The x-axis is percentage of seats won, not percentage of votes. In virtually all mature democracies, how far along the x-axis a country appears will depend on its electoral system: first-past-the-post and similar systems create a parliamentary majority out of a minority of the vote.

    In this context, the US is probably an outlier anyway – it may have a first-past-the-post electoral system but its separation of powers (and the “swing” power that is often in the hands of a handful of congressmen and particularly senators) means that the government’s ability to control expenditure is highly constrained anyway. The US has a weird mix – traditionally low overall public expenditure allied to a system that is almost constitutionally incapable of reducing it without being held hostage by vested interests.

    I agree with msjhaffey above that there are two different questions. It may be that perpetual coalition is the best way to ensure sustainable public expenditure – in stark contrast to the ridiculous Tory-underspend Labour-splurge cycle we have to put up with in the UK. But if you find yourself in a position where you have to confront vested interests in order to reduce expenditure, more coalition partners means more chance for the vested interests to get in the way. With the possible – and somewhat undemocratic – exception that if you’re in a coalition you can always blame your coalition partner!

  23. Jon d said,

    May 10, 2010 at 5:27 pm

    Which page of the IoG’s pdf document does this scattergraph come from? cos I can’t see it there.

    Just wondered if there was any sort of method governing which countries were included really.

  24. Jon d said,

    May 10, 2010 at 5:47 pm

    Oh right, it’s not from that pdf, it’s from this powerpoint www.instituteforgovernment.org.uk/images/files/understanding_hung_parliaments.ppt#8 – there’s an even dodgier looking scattergraph on the page before correlating large majorities to bad government if anyone fancies having a pop at that.

  25. Diversity said,

    May 11, 2010 at 12:44 pm

    These scattegraphs are important. They show that the there is no prima facie evidence that fiscal discipline and/or quality of governemnt are worse when and where there is no absolute majority in the legislature.

    To put the same point another way; as a remedy for fiscal indiscipline one-party majorities are snake oil; and nothing more.

  26. fontwell said,

    May 11, 2010 at 2:23 pm

    I wasn’t really trying to make a point about the USA other than use it as an example of where the current government has a much bigger/smaller lead (in terms of the units on the graph) than the government responsible the debit.

    So the point was more about the financial situation in a single year vs the incumbent government’s majority being a rather crude measure, which I guess we all new anyway. I mean, the UK is about to leap from right to left on the graph overnight with no change in what what kind of government caused our current condition.

  27. Jon d said,

    May 12, 2010 at 3:17 am

    Well the source of the structural balance data seems to be the 2010 column of table B7 of the 2010 IMF report. First up it’s forecast not measured.
    Also there’s a forecast for 2011, it occurred to me that an alternate measure of financial discipline could be the proportion by which the structural balance is forecast to come down between 2010 and 2011 (or go up… NZ I’m looking at you). If you do that the correlation comes out very weak.
    And another thing, the odd looking selection of countries, the reason it includes countries like Slovenia is, afaict, cos it runs the Euro as a headline economy with all the Euro members as sub economies. Seems to me the Euro countries might not be entirely economically sovereign, i.e statistically independent, least not without looking deeper. Maybe we should include Poland, Brazil, South Korea and a few other places.
    Oh yeah and it’s also not clear to me that less structural balance is better, I’m not an economist but maybe there’s a good level of structural balance that’s not zero?

  28. Filias Cupio said,

    May 12, 2010 at 7:14 am

    One could make a data-nerd quiz show out of stuff like this. Take pairs of variables (like this example), and ask your contestants which way the correlation will be and how strong.

  29. rickogorman said,

    May 12, 2010 at 10:58 am

    @shockdoc: Firstly, whether the graph shows a ‘modest’ predictor depends on your point of reference. Seems to me accounting for about one quarter of the variance i the data, given all the points above, is pretty good. Secondly, as others note, it certainly speaks to the ridiculous coverage I’ve seen on the Beeb so far of a hung parliament/coalition. Not to mention the dumb ‘but the Tories “won”, so shouldn’t they be in government’. As if it were the boat race.

    Finally, perhaps the most prescient part of the Hung Parliament report that Ben posted:

    “But the greater risk is that when it happens, the media will denounce minority government as something fearful, unstable and deeply undesirable.”

    Sadly spot on.

  30. Daniel said,

    May 12, 2010 at 1:02 pm

    Dude I would love to see your geek’s Pearson’s correlation coefficient (I would be surprised if it reached 0.2). His outliers are insane! Oh and Spain & Greece in there, that’s a Joke right??
    Oh and you’re saying Mr. Cameron and his newly-found set of (i would love to put ‘like-minded’ here) political allies might not spend as much as their predecessors. Doesn’t contractionary fiscal policy promote economic recession? A little bit counter-intuitive if you consider UK’s current economic conundrum.

  31. shockdoc said,

    May 13, 2010 at 12:27 am

    well..

    > (hp)
    Country Party Debt
    1 Belgium 19.96310 -4.2559524
    2 Finland 25.49815 -1.8750000
    3 Denmark 25.49815 -1.6964286
    4 Netherlands 27.26937 -5.2083333
    5 Austria 31.03321 -4.3154762
    6 Slovenia 32.14022 -4.4047619
    7 Slovakia 33.24723 -4.7023810
    8 Sweden 37.23247 -0.8333333
    9 Norway 37.78598 -7.3511905
    10 Germany 39.88930 -3.7797619
    11 Portugal 42.10332 -7.0535714
    12 Italy 43.76384 -3.4821429
    13 Ireland 46.30996 -7.8869048
    14 Canada 46.42066 -3.0059524
    15 Spain 48.30258 -7.3214286
    16 New Zealand 48.30258 -2.0535714
    17 Greece 53.28413 -8.8988095
    18 France 54.16974 -4.6130952
    19 United Kingdom 54.94465 -7.5595238
    20 Australia 55.27675 -4.9107143
    21 USA 59.04059 -9.2261905
    22 Japan 64.13284 -7.5000000
    > cor(Party, Debt, method=”pearson”)
    [1] -0.5506873
    > shapiro.test(Party)

    Shapiro-Wilk normality test

    data: Party
    W = 0.9744, p-value = 0.81

    > shapiro.test(Debt)

    Shapiro-Wilk normality test

    data: Debt
    W = 0.9514, p-value = 0.3372

    So, greater than 0.2 then but…. all of the above caveats make this fairly redundant. Just in case anyone was still interested. Yes the r^2 is ok but let’s face it the data isn’t really able to cause inter-ocular trauma (it doesn’t hit you between the eyes).

    Also, I love “R”. If you’re not using “R”. If you’re using something wet like SAS or Stata or worse… ew …. SPSS, then download “R” and play with it…. real men (and real women) (and possibly some real men that think they’re real women) (and vice versa) Use R!

  32. kingshiner said,

    May 13, 2010 at 10:48 am

    The p-value of 0.01 just says it’s unlikely there is zero association, but it’s extremely weak, never mind whether it’s causal, due to outliers etc.

    P-values for correlation coefficients give very small p value from totally uninteresting data, usually they’re a mistake. The ‘limits of agreement method’ (Bland/Altman) is more informative, it’d tell you how far off you could be in predicting debt from size of largest party (hopelessly far off).

  33. Leppo said,

    May 14, 2010 at 9:23 am

    Usually when I see an interesting phenomenon, I look up for the previous knowledge about the issue, instead of a lot of guesswork. In this case, there are numerous academic studies on the subject. Why not consult the previous literature instead of considering simple cross-section correlation?

    As some observers have pointed out, the association illustrated by the scatter plot may well be a dynamic in nature.

    Just a few references, but you get them free:

    The Weak Government Thesis: Some New Evidence
    Public Choice, Volume 101, Numbers 3-4 / December, 1999

    Jakob de Haan, Jan-Egbert Sturm and Geert Beekhuis

    Abstract This paper presents new evidence on the hypothesis that coalition governments will find it more difficult to keep their budgets in line after an adverse economic shock than do one-party, majoritarian governments. The estimates are based on a broad sample of OECD countries, for the period 1979–1995. Using various specifications as suggested in the literature, we do not find evidence that the type of government affects cross country variation in fiscal policy. However, the number of political parties in government affects central government debt growth.

    Government Strength, Power Dispersion in Governments and Budget Deficits in OECD-Countries. A Voting Power Approach

    Public Choice, Volume 116, Numbers 3-4 / September, 2003

    Gerald Huber, Martin Kocher and Matthias Sutter

    Abstract We test for the influence of government strength and dispersion of power among the parties of coalition governments on the size of annual debt accumulation through budget deficits in OECD-countries from 1970 to 1999. Government strength and power dispersion in coalition governments are measured by the Banzhaf index of voting power, respectively the standard deviation of Banzhaf indices of coalition parties. We believe that these are better-suited proxies than most of what has been applied so far. Government strength turns out to be insignificant. However, coalitions with equally strong partners run significantly higher deficits than coalitions with one dominating party.

  34. SteveGJ said,

    May 14, 2010 at 8:03 pm

    @Leppo

    An extremely good idea to look at the research on this and, on a quick scan on the items listed, it would appear the evidence is contradictory and no doubt has kept many social scientists happily working away with their toolbox of statistical tools for many a long year.

    In any event, the thought that a simple little scatter chart, like the one used to illustrate this item, is actually at all likely to shed some blinding shaft of light onto such a complex problem is, to say the least, dubious when so many have struggled on this before.

    Just because it is possible to phrase a simple question does not mean that there is an equally simple answer, although politicians often seek to make it appear that way.

    Personally, I suspect any effect of whether majority or coalition governments are more likely to maintain good public finances is probably going to get swamped by a multitude of other factors.

  35. susu.exp said,

    May 14, 2010 at 8:35 pm

    I crunched a few numbers using data from the German Länderfinanzausgleich from 1952 to 2008 (1995-2008 for the 5 formerly eastern german Länder). In the federal system the Bundesländer running a surplus cover some of the debt of the less disciplined (or merely troubled) Lander. So I´ve looked at the various coalition governments running them and these are mean values of what Länder goverened by coalitions have recieved (negative numbers indicate that they´ve given away funds, number in brackets indicate the number of years these coalitions were in power):
    SPD-greens: -446 (39)
    CDU/CSU-FDP: -333 (88)
    CDU/CSU: -105 (198)
    SPD-FDP: -19 (78)
    SPD: 76 (133)
    SPD-FDP-greens: 291 (1)
    CDU-SPD:505 (74)
    SPD-PDS/Die Linke: 1323 (16)

    Non-coalition governments are in the middle, while both the top recipients and top disciplinarians are coalition governments. This does broadly support the idea that coalitions between roughly equal strenght partners are less fiscally responsible (both the grand coalition of CDU and SPD and the SPD-PDS/Die Linke line up fill that slot), but coalitions with one strong and one weak partner (SPD-greens, CDU-FDP, SPD-FDP) do rather well.

  36. Jon d said,

    May 15, 2010 at 1:31 pm

    yeah a straight line correlation seems like a naieve expectation, I did read something about voting power analysis (of the sort done in the papers leppo mentioned)in the new scientist around 1986-1988 and iirc it works like a series of tipping points where small changes in the proportions of seats can result in large changes in the relative voting power of the players. 50%-100% seats for one party in parliament where a simple majority is sufficient means you don’t need to talk to anybody else to push your programme, if you’ve got ferinstance 2 parties on 40% seats each and one party on 20% seats the party on 20% is really quite powerful.

    if anyone wants to play around with Banzhaf voting power analysis there’s a page on wiki, a download for wolfram (requires mathematica playa) demonstrations.wolfram.com/BanzhafPowerIndex/
    and an excel macro – home.comcast.net/~raagnew/site/?/page/Downloads

    Both are pretty horrible to use tbh – but I managed to get the excel working for the current UK parliament after rolling up my sleeves & increasing the size of a couple of arrays… anyone fancy writing a decent banzhaf programme for R? it’d be fun.

You must be logged in to post a comment.