Facebook Google Menu Linkedin lock Pinterest Search Twitter

Advertisement

General Economics

May 12, 2010

The most important chart in the budget

Tucked away in Budget Paper Number 1 is a fascinating little chart that is arguably the most politically important piece of data in the entire budget, as it justifies not just the very

User login status :

Share

Tucked away in Budget Paper Number 1 is a fascinating little chart that is arguably the most politically important piece of data in the entire budget, as it justifies not just the very existence of the stimulus program and the political baggage that is coming with it, but also shows the likely consequences of the alternative “what if the stimulus was smaller or didn’t exist” scenario.

You can see the Treasury version of the chart over in Statement 2 of Budget Paper No. 1 – “Box 4: Economic growth and fiscal stimulus”. However, that treasury graphic is a bit of a chartcrime, so we can reproduce it ourselves from the raw data cited in the budget papers.

Back in early 2009, the IMF made GDP forecasts for a number of countries based on what the prevailing economic outlook was for those countries at the time that the forecasts were made. Those early 2009 IMF forecasts came in like this:
forecastgrowth

During the period between when these forecasts were made and today, these nations deployed fiscal stimulus packages of varying sizes. You can see the size of the stimulus packages as a proportion of GDP in the table on page 36 of this IMF document – under the 2009 column “Crisis related discretionary measures”. This was the size of the stimulus packages for each of these nations as a proportion of GDP:

stimulussize

The third and final piece of data we need is the actual growth achieved by these nations in 2009, which the IMF also provides over here.

This is what they come in as (using last night’s Treasury data to substitute for the Australian entry):

actualgrowth

So what was the impact of the stimulus on GDP growth across nations? To find out, if we subtract the 2009 IMF forecast growth from the actual growth achieved in 2009, it tells us the forecast error – or the amount by which the original GDP forecast for each nation was out.

That then allows us to compare the size of the various stimulus packages to the size of those forecasts errors via a scatter plot and regression – if the stimulus packages explain the forecast errors (in that, the bigger the stimulus package, the better the economy performed over its original forecast), then we should see a tight linear relationship with nations starting in the bottom left hand corner (representing small stimulus and small errors) going roughly linear up until the top right corner (representing large stimulus creating large forecast errors).

So what does the chart look like?

stimuluseffect

This is a very statistically significant relationship (albeit with only 11 observations), with the budget papers stating that the T-stat on the stimulus package coefficient was 3.3. Using the raw data, my results confirm that – here’s the regression output:

stimpakeq

What this chart demonstrates is that the size of the stimulus packages mattered – seriously mattered. The larger the stimulus package, the better a country performed compared to what its original IMF growth forecasts suggested.

The stimulus packages made a clear difference to the size of GDP growth. The size of the packages made a clear difference to the size of the effect on GDP growth.

As the effect was consistent across nations – we can use this data to get an approximate estimate of what actual Australian GDP would have looked like in 2009 as a function of the size of a spectrum of stimulus package sizes.

If we treat the regression line in the above chart as an approximate estimate of the stimulus effect on GDP, we can slide Australia up and down that line according to hypothetical stimulus packages of different sizes, and recalculate the left hand vertical axis to represent Australian GDP rather than the forecast error.

ozgrowthestimates

According to IMF data and the consistency and significance of the stimulus/forecast error relationship, a stimulus package in Australia around the 2.2% of GDP would have given us zero growth – anything less than that would have delivered us negative growth.

When the Coalition says the stimulus package should have been smaller – they need to be asked how much smaller. Every drop in GDP causes an increase in size of unemployment. How many more people would they have been willing to throw on the scrap heap of unemployment in their pursuit of a smaller package?

Every time the Coalition complains about debt, they need to be asked how much smaller they believe the stimulus package should have been and how many more people would they have been willing to throw on the scrap heap of unemployment in their pursuit of a smaller debt load.

Debt was the cost of growth – growth was and always is the provider of jobs. Pretending that less debt could have provided the same jobs is fairy floss economics. A dollar is a dollar is a dollar.

The size of stimulus packages mattered – the international evidence is in.

The Coalition needs to be questioned about its economic viewpoints – viewpoints which are far from mainstream economics, existing on the very fringes of economic debate and which are completely at odds, completely and utterly at odds, with the international empirical evidence.

You know what an argument without evidence is called?

Making shit up.

UPDATE:

Sinclair Davidson (which some of you may remember from one of the funniest exchanges in Senate Estimates history with Doug Cameron!) over at Catallaxy has redone the same chart with more than just this G7 plus selected countries, but expanded it to the whole G20, finding much more variation in the results, leading to a crashed significance and much reduced slope on the regression line over the full G20 spectrum.

Possum Comitatus — Editor of Pollytics

Possum Comitatus

Editor of Pollytics

Political Commentator and Blogger

Get a free trial to post comments
More from Possum Comitatus

Advertisement

We recommend

From around the web

Powered by Taboola

89 comments

Leave a comment

89 thoughts on “The most important chart in the budget

  1. Lincoln Fung

    You state: “You can see the Treasury version of the chart over in Statement 2 of Budget Paper No. 1 – “Box 4: Economic growth and fiscal stimulus”. However, that treasury graphic is a bit of a chartcrime, so we can reproduce it ourselves from the raw data cited in the budget papers.”
    Then you went on to recreate your own charts and regression results to derive your own nice chart or graph – “Australian GDP growth under alternative stimulus sizes”.
    Then you went on to say:
    “According to IMF data and the consistency and significance of the stimulus/forecast error relationship, a stimulus package in Australia around the 2.2% of GDP would have given us zero growth – anything less than that would have delivered us negative growth.

    When the Coalition says the stimulus package should have been smaller – they need to be asked how much smaller. Every drop in GDP causes an increase in size of unemployment. How many more people would they have been willing to throw on the scrap heap of unemployment in their pursuit of a smaller package?

    Every time the Coalition complains about debt, they need to be asked how much smaller they believe the stimulus package should have been and how many more people would they have been willing to throw on the scrap heap of unemployment in their pursuit of a smaller debt load.

    Debt was the cost of growth – growth was and always is the provider of jobs. Pretending that less debt could have provided the same jobs is fairy floss economics. A dollar is a dollar is a dollar.

    The size of stimulus packages mattered – the international evidence is in.

    The Coalition needs to be questioned about its economic viewpoints – viewpoints which are far from mainstream economics, existing on the very fringes of economic debate and which are completely at odds, completely and utterly at odds, with the international empirical evidence.

    You know what an argument without evidence is called?

    Making shit up.”

    But have you realised or ever considered whether the IMF data is correct based on its own forecasts that were revised on a monthly basis and changes all the time?
    Man, we need some sanity in place!

  2. Tim Watson

    Sinclair Davidsons regression analysis shows a positive stimulus effect- albeit without a high degreee of statical significance at the 5% level I gather.

    What about at the 10% level?

    In any case he should publish the full results of his regression to let people make up their own minds.

    Regardless of statisical significance, what about what his regression doesn’t show:

    – it doesn’t definitively show no multiplier effect from stimulus funding
    – it doesn’t show a crowding out effect from government stimulus

    It doesn’t look at the composition of stimulus money between various countries (i.e. how much was spent on enhancing consumption and investment versus bailing out the financial sector).

    It doesn’t greatly enhance the sample of countries considered from the Treasury analsysis.

    JC- you should also read Barry Eichengreen’s historical paper on fiscal multiplier’s. A little quote from that paper:

    “They suggest that fiscal policy made little difference during the 1930s because it was not deployed on the requisite scale, not because it was ineffective. They suggest a positive impact of government expenditure on GDP during the interwar period, with substantial fiscal multipliers: for example, the first set of VAR exercises suggested that these were 2.5 on impact and 1.2 after one year. Where significant fiscal stimulus was provided, output and employment responded accordingly.”

    Just because I am not enthused by your shallow analysis of fiscal multipliers doesn’t make me a “concern troll”.

  3. Possum Comitatus

    LO went:

    [Sorry to ruin the love-in, but the following statement is just plain wrong]

    Hang on – the chart you’re complaining about is the best, most easily understood visualisation that incorporates all the complexities going on underneath the GDP results that I’ve seen. (incorporates them via the mechanics of the models that produced the baseline projections from which forecast errors actually derive in the first place)

    Do you really think I’d run this graphic if the more complex data underneath it didnt justify it? Well, you might – I think you’ve tried to slap me a few times before for doing the same 😛

    The international evidence *is* “in” – and it’s hardly surprising… it was expected.

    Back before the actual GDP results of 2009 released, but after the IMF’s initial growth forecasts, the IMF ran another set of growth projections incorporating early planned stimulus packages of various countries (early, in that the size of the fiscal injections used were much smaller than what was eventually deployed through 2009).

    Those projections clearly showed the correlation between the size of projected growth and the size of stimulus packages across their forecast ranges:

    Page 18:
    http://www.imf.org/external/np/g20/pdf/020509.pdf

    The size of the stimulus – all things being equal – was expected to correlate with the change in growth.

    And it’s not just some quaint little IMF issue, all of the big structural models suggest the same under a wide spectrum of uncertainty and conditions – at least as far as the G20 nations (or combinations thereof) are concerned.Last week I was reading through this:

    http://www.imf.org/external/pubs/ft/wp/2010/wp1073.pdf

    …which made the chart in the budget (that you’re complaining about) quite fortuitous.If I hadnt just read that paper, I probably never would have done this post. The thing that strikes me about what is happening as the actual GDP numbers are rolling in is how much larger the multipliers appear to actually be, compared to what just about everyone thought they would be going into the crisis.

    To the point, where the size of the multipliers were seeing doesnt appear to be constant across the size of the fiscal injection – but where larger injections seem to provide even larger multipliers.

    The reason I published the regression equation in the post was two fold. Firstly, to demonstrate that the chart I produced was identical to the chart in the budget (identical T-stats on a coefficient don’t come about by accident). The second was so I didnt have to answer questions about the results of the regression line in the chart since all the info is right there.

    I thought the words “albeit with only 11 observations” spoke for itself – though apparently not.

    This isnt, nor was it meant to ever be, the definitive econometric explanation of the behaviour of the global economy – it’s really pretty silly to suggest it was trying to be. What it is, however, is the best visualisation around on what is actually going on in terms of the results we are seeing.

    On your other part:

    [I’m going to use this post as an opportunity to register a broader complaint. The blog seems to have become mainly a vehicle for pointing out the errors and inconsistencies in MSM commentary on political and policy issues.]

    I’m glad you did, because a few people have said this lately.

    Over the last 30 posts I’ve made (excluding the post saying that I’m going on holidays), if we really stretch the definition of pointing out errors in media commentary, there has been 3 posts dealing with that.

    The BER post that dealt with the poor understanding of the audit report, the post dealing with Rio Tinto using the news cycle as a lobbying vehicle and how it blew up in their faces (forcing them to backdown with an ASX statement) and the ETS post that was mostly about the importance of the commencement date in the DD trigger legislation – but which had a go at the commentary that seemed to overlook this rather important bit of info.

    3 out of the last 30 (or 31)

    I think what is happening is that the relatively few articles that point out these errors get much wider net linkage than the bread of butter if this place – which is polling data.During those last 30 posts we’ve had a series on the Roy Morgan Polliegraph data from the health debate, a large piece about electoral demography, an election simulation, migration stats, betting market roundups and large chunks of polling.

    I don’t know what else I can say.

    [While it isn’t unreasonable to spend some of your time doing this, I can barely recall a single post that has subjected the government’s own statements, analysis and policies to the same critical evaluation.Why is this? The evidence base underpinning a host of government policies has been thin at best. Yet you never turn your statistical analysis to scrutinising the government’s claims. Don’t you think such analysis would be important?]

    The last time I attempted to do that was with the insulation program – but the data came in the other way.

    Got a suggestion for any issues to have a squiz at?

  4. EngineeringReality

    Great article Possum.

    Yeah the injection of money into the economy either had an effect or it didn’t. The likelihood of billions pouring into the economy having no effect is about as Likely As Wombat-shit.

    So it almost certainly had an effect.

    An injection of cash never reduces economic activity so any effect it would have had would have been a positive effect on our GDP.

    The only question remains “How much was that positive effect?”. I think your article answers that Possum.

    It leaves the coalition exposed as the financial dunces and bullsh!t artists that we know they are.

    They have unfortunately realised the ease that they can quickly and effectively perpetuate the simplistic myth through the MSM and crap political ads that:

    Government debt = bad

    and therefore

    Government surplus = good

    Without the mainstream public being educated that government fiscal policy is really one of the levers of control of an economy and not like the finances of ordinary people or the balance sheets of corporations then unscrupulous political parties can keep influencing public opinion with this falsehood.

    A government surplus is where the government takes more money off us than it gives back to us. It keeps some of our money – to reduce the amount of money in the economy – to reduce inflation and reduce the need to raise interest rates. Its not the same as saving money for a rainy day that individuals do or raising levels of cash that a company may do to meet upcoming commitments or use in future takeovers or investment.

    The real pity is that no one in the MSM understands this – so they just regurgitate the press releases and comments of the coalition.

  5. Labor Outsider

    Sorry to ruin the love-in, but the following statement is just plain wrong:

    “The size of stimulus packages mattered – the international evidence is in.”

    This is a regression with 11 observations and besides the constant, only 1 regressor!! You can’t make any reliable inference on that basis given the potential size of the omitted variable bias. Take out 3 of the eleven countries (Korea, Australia and China) and there is no statiscally significant relationship at all.

    There are no controls for the size of the intial shock (the regression assumes that the original forecasts made the right assessment of how large the shock was) nor non-fiscal stimulus and crisis resolution measures.

    Now, I think it is likely that the Australian stimulus had a positive short-term impact on GDP. Theory and most of the rigorous empirical literature suggests that it should. But to say that the evidence is in on the basis of this empirical test is a joke. You certainly cannot do something as simplistic as read off the impact on GDP from a smaller stimulus given the potential endogeneity problems. Think about monetary policy for example. If the fiscal stimulus had been smaller in Australia, the reaction of the central bank would have been larger. Take a look at the literature trying to determine the impact of fiscal policy on GDP and the sophisitication of the econometric methods used to try and identify the causal impact. This regression wouldn’t see the light of day in even the world’s least sophiticated economic journal.

    Also, many of the arguments against the stimulus are based on concern that some of the government’s expenditure programmes wouldn’t have passed a cost-benefit test, and hence are likely to reduce the level of potential output. In that sense, there is a trade-off between current and future growth.

    I’m going to use this post as an opportunity to register a broader complaint. The blog seems to have become mainly a vehicle for pointing out the errors and inconsistencies in MSM commentary on political and policy issues. While it isn’t unreasonable to spend some of your time doing this, I can barely recall a single post that has subjected the government’s own statements, analysis and policies to the same critical evaluation. Why is this? The evidence base underpinning a host of government policies has been thin at best. Yet you never turn your statistical analysis to scrutinising the government’s claims. Don’t you think such analysis would be important?

  6. ihatett

    Wow… a statistician here. I never expected to see an article in the media that actually paid some respect to statistical methodology – a fantastic forward. It’s… refreshing to say the least, considering the **** that comes out of some outlets! I would just like to comment on Mr Squiggles remarks.

    1) Statiticians sometimes find reason to remove the top and bottom outlying numbers from a group under analysis as abberations. In the case of your chart, Korea looks very much like an outlyer, and possibly even the UK or Italy. I wonder what the following charts would look if you removed these countries. My guess is the line would be flatter, and as a result the line representing Australian GDP under alternative stimuls packages would move to the left somewhat.

    Mr Squiggles… the decision to remove observations from a regression analysis is never as arbitrary as you suggest. The decision to remove outliers, when made by a statistician, is related to the concept of that point influencing the magnitude and variance of the coefficient estimate. Information that can assist in whether or not to remove observations can be found by plotting the residual errors, determining points that sit above the cook’s distance, dfbetas… blah blah blah too much information.

    The ‘statisticians’ that remove observations in that fashion are usually the ones that did an intro stats course in their masters degree or a ‘research methods’ psych degree and hence think they know everything. It’s quite scary what you can pick up from journal articles where there wasn’t a statistician involved in the experimental design. I don’t remember the title, but I remember there was a journal article floating around the traps about statistical errors (i.e. in terms of methodology and pure mis-statement of results).

    Most statisticians wouldn’t bother removing observations from such a small data set, but then again most statisticians would only use an 11 observation data set to demonstrate the concept of linear regression.

    Like all statistical analysis however, there are a lot more variables at play in the gdp = stimulus size equation. I’d be careful making any major inferences on this point, given all the nested and latent factors involved in a model like (e.g. distribution of stimulus by sector, consumer confidence etc etc). Though major props to you crikey – I think I’ll have to read your news site more often!

Leave a comment