More austerity, more protests?

A recent study by Jacopo Ponticelli and Hans-Joachim Voth (nominated for the Vanguardia Science Award) found a clear relationship between austerity and social unrest. European countries with expanding budgets saw on average 1.5 incidents (ranging from demonstrations to attempted coups) per year; however, incidents more than doubled in countries that cut their expenditure by more than 5 percentage points of GDP.
The dataset used by Ponticelli and Voth spans the period 1919-2009, which means it doesn’t include the anti-austerity protests that spread over Europe since 2009. I created the map below to illustrate how these recent anti-austerity protests are related to cuts in government expenditure (the map may not work in internet explorer). The size of the circles reflects the number of protests, while darker colours for countries reflect government expenditure cuts. Click on the map to navigate, hover the mouse over the map to see the actual data.

Problem is, there doesn’t seem to be a very clear relationship between cuts and protest. One explanation is that there’s something wrong with the data I use. That’s certainly a possibility - I discuss limitations of the data below. But it’s also conceivable that the dynamics of protest have changed during the current crisis.

Select change 2009-2012 in:

The scatterplot above shows how the number of protests is related to changes in government finance (2009-2012). There is a very weak correlation between expenditure cuts and protest. However, the correlation gets stronger if you look at changes in government revenue (e.g. tax rises) or at changes in the budget balance (in other words, the combined effect of changes in expenditure and revenue). Two conclusions can be drawn:

  1. Taxes appear to have become more important in explaining protest, compared to the period studied by Ponticelli and Voth;
  2. The structural government balance (which corrects for one-off revenues and expenditures) has a stronger correlation to protest than the figure you get by simply subtracting expenditure from revenue.

Ad 1. In 2010, Greeks were not only protesting against severe pay and pension cuts, but also against an increase of the value added tax. These measures were introduced by the Greek government to meet the criteria for a bailout. Other countries would also receive bailouts with strings attached. And even before the bailouts, France, Spain, Ireland and Greece had been ordered by the EU to reduce their budget deficits.
All in all, European governments have been under considerable pressure over the past years to reduce their deficits, which can be achieved by cutting expenditure and by raising taxes. Apparently, it’s the combined effect of these two measures (rather than just the effect of cuts) that drove people onto the streets.
(My guess is that raising taxes like VAT, which disproportionately affect people with low incomes, is more likely to incite protests than raising more progressive taxes. But I don’t have the data to prove this.)

Ad 2. I’m just guessing here, but suppose a government makes considerable one-off expenses related to the crisis, while simultaneously cutting it’s social programmes. In such a case the net effect on the ‘normal’ balance would be limited, but people would still feel the pain. So it would seem to make sense that a measure that corrects for one-off budget items does a better job at explaining why people protest.

Of course, these technical nuances don’t change much about the general conclusion as formulated by Voth:

Even if (and it's a big if, given the IMF's latest research) [proponents of austerity] are right, and growth can follow cuts, the pain may be concentrated amongst some groups. If these become massively unhappy... it can start to look pretty ugly out there in the streets, and I doubt that that'll be good for growth.

Obviously, this is just an exploratory analysis and I don’t pretend to provide any final answers on the matter. It would be interesting to see whether this analysis would still hold true if a more complete dataset on recent European protests could be used. (Further I have to admit I’m not entirely sure about the statistical analysis but I reckoned a rank correlation is probably a safe bet...)

The graphs probably don’t work in old versions of Internet Explorer.


Protest events: I used the New York Times API to search for articles containing the terms ‘austerity’ and ‘protest’ and manually removed duplicates and articles that weren’t really reports of incidents (e.g. analyses and op-ed articles). Obviously, there will always be a bias in media reports of protests (this problem probably affects all protest databases). At least the NYT is probably less biased than a news source from a specific European country when it regards the occurence of protests in European countries. While my measure of protest has limitations (e.g., it didn’t pick up on bossnappings in France, the Kitchenware Revolution in Iceland, the Grape Revolution in Moldova and some of the Occupy protests), I think it’s a usable indicator of anti-austerity protest intensity.
Government finance: I used data from the IMF World Economic Outlook (October 2012), which has the advantage of containing data for recent years (obviously, these are estimates). All changes are measured in percentage points of GDP.
The visualisations were created with D3.js as explained in this tutorial and this book.

Blind followers on Twitter

Select group:

On 30 september, I posted the last article on Nieuws uit Amsterdam (News from Amsterdam). The website has been inactive since, apart from a message on 28 October formally announcing that the site is no longer active. As expected, the number of new followers of @nieuwsamsterdam on twitter dropped in October. Intriguingly, it started to rise again after that.

The list of new followers has been compiled from ‘You have new followers’ emails and may be incomplete. Graph may not work in older versions of Internet Explorer.

‘Trade unions should take a much tougher stance’

Dutch trade unions have a reputation for constructive dialogue, but that’s not necessarily what people expect of them. In the LISS Political Values study, some 6,000 panel members have been asked a number of times whether they agree with the statement ‘Trade unions should take a much tougher political stance, if they wish to promote the workers’ interests’. In the latest edition of the study, those who agree with this statement outnumber those who disagree by 2.6 to 1. This support for tougher unions holds for most subgroups (but not the self-employed and people earning more than 4,500 euros per month).

Support for tougher unions over time

Percentage of respondents who agree or disagree with the statement ‘Trade unions should take a much tougher political stance, if they wish to promote the workers’ interests’. Graph may not work with older versions of Internet Explorer. Source LISS, graph dirkmjk.

Support for tougher unions, by group


Values higher than 1 mean that within that group, those in favour of tougher unions outnumber those who disagree. For example, among people with paid employment, the number of respondents in favour of tougher unions is 3.5 times as high as the number who disagree. Hover mouse over bar to see percentages. Graph may not work with older versions of Internet Explorer. Source LISS, results for December 2011, graph dirkmjk.

My first D3 graph

I’m trying to master D3, a javascript library for creating (interactive) web graphics. As an excercise, I redid this graph, which uses Eurostat data on the percentage of the population who have ever written a computer programme.

I can’t say it’s a very good graph: some of the most intriguing aspects of the data have to do with changes over time (decline in some countries, rather large growth in Finland, implausible fluctuations in the Netherlands), which don’t show very well in my graph. Nevertheless, it feels good to have coded my first interactive D3 graph.

P.s. the graph may not be visible in older versions of internet explorer.

ATMs and cycle paths

The habits of cyclists are shaping cities like Amsterdam. “There are many ATMs along the main bicycle path network”, urban planner Marco te Brömmelstroet told Vogelvrije Fietser, the magazine of cyclists’ organisation Fietsersbond.

The map above shows Amsterdam’s main cycle path network (provided by the city as open data) and the location of ATMs. It appears that many ATMs are indeed located near cycle paths. Exceptions include shopping areas such as the Kalverstraat, Gelderlandplein and Bijlmerplein. (I tried to calculate the distance between ATMs and cycle paths but I couldn’t get this to work in QGIS.)