«Jacopo Ponticelli Hans-Joachim Voth UPF ICREA/UPF and CREI Abstract: Does fiscal consolidation lead to social unrest? From the end of the Weimar ...»
Austerity and Anarchy:
Budget Cuts and Social Unrest in Europe, 1919-2009*
Jacopo Ponticelli Hans-Joachim Voth
UPF ICREA/UPF and CREI
Abstract: Does fiscal consolidation lead to social unrest? From the end of the
Weimar Republic in Germany in the 1930s to anti-government demonstrations in
Greece in 2010-11, austerity has tended to go hand in hand with politically motivated
violence and social instability. In this paper, we assemble cross-country evidence for the period 1919 to the present, and examine the extent to which societies become unstable after budget cuts. The results show a clear positive correlation between fiscal retrenchment and instability. We test if the relationship simply reflects economic downturns, and conclude that this is not the key factor. We also analyse interactions with various economic and political variables. While autocracies and democracies show a broadly similar responses to budget cuts, countries with more constraints on the executive are less likely to see unrest as a result of austerity measures. Growing media penetration does not lead to a stronger effect of cut-backs on the level of unrest.
This version: August 11th, 2011 * We thank Jordi Galí for inspiring this work. Financial support by the European Research Council and the Barcelona GSE is gratefully acknowledged.
Electronic copy available at: http://ssrn.com/abstract=1899287
1. Introduction Social unrest has led to key turning points in modern history since, at least, the French Revolution. Marx saw it as the driving force of the transition of societies from feudalism to capitalism and, eventually, communism. The power of unrest as a catalyst for change manifests itself explicitly in regime changes, such as during the “Arab Spring” of 2010-2011, or it operates through expectations: The extension of the franchise in Western societies has been interpreted as an attempt to heed off the threat of revolution (Acemoglu and Robinson 2000).1 What leads to social unrest is less clear. Economic shocks are one important contributing factor: The demise of the Weimar Republic during the Great Depression is a prominent example of how economic hardship can translate into unrest (Bracher 1978).2 In this paper, we examine what leads to social instability and violent protests.
In particular, we ask whether fiscal policy affect the level of social unrest. The extent to which societies fracture and become unstable in response to drastic changes in the government budget is a primary concern for policymakers attempting to reduce budget deficits: From Argentina in 2001 to Greece in 2010-11, austerity measures have often created a wave of violent protests and massive civil unrest. Economic conditions can deteriorate further and faster if political and social chaos follows attempts to reign in spending. Consequently, sustainable debt levels for countries that are prone to unrest may be lower than they otherwise would be.
We use a long panel dataset covering almost a century, focusing on Europe, 1919 to 2009. The continent went from high levels of instability in the first half of the 20th century to relatively low ones in the second, and from frequently troubled economic conditions to prosperity. It thus provides a rich laboratory of changing economic, social and political conditions. In terms of outcome variables, we focus on riots, demonstrations, political assassinations, government crises, and attempted revolutions. These span the full range of forms of unrest, from relatively minor disturbances to armed attempts to overthrow the established political order. We compile a new index that summarizes these variables, and then ask -- for every percentage cut in government spending, how much more instability should we expect?
In a related exercise, Boix (2003) models the incentives of the populace to resort to violence as a function of the wealth distribution and economic development.
The French Revolution has also been interpreted in these terms (Soboul 1974; Doyle 2001). The view is controversial (Hunt 2004; Cobban 1964).
Electronic copy available at: http://ssrn.com/abstract=1899287
The data shows a clear link between the magnitude of expenditure cut-backs and increases in social unrest. With every additional percentage point of GDP in spending cuts, the risk of unrest increases. As a first pass at the data, Figure 1 examines the relationship between fiscal adjustment episodes and the number of incidents indicating instability (CHAOS). CHAOS is the sum of demonstrations, riots, strikes, assassinations, and attempted revolutions in a single year in each country. The first set of five bars show the frequencies conditional on the size of budget cuts. When expenditure is increasing, the average country-year unit of observation in our data registers less than 1.5 events. When expenditure cuts reach 1% or more of GDP, this grows to nearly 2 events, a relative increase by almost a third compared to the periods of budget expansion. As cuts intensify, the frequency of disturbances rises. Once austerity measures involve expenditure reductions by 5% or more, there are more than 3 events per year and country -- twice as many as in times of expenditure increases.
Figure 1: Frequency of incidents and the scale of expenditure cuts Exactly the same relationship can be observed in each of the four main subcategories of CHAOS. The frequency of demonstrations, assassinations, and general strikes rises monotonically with the scale of cuts. Only in the case of riots is there a small decline for the biggest cut-backs. In the case of demonstrations, the frequency of incidents appears to rise particularly fast as expenditure cuts pass the 3% threshold.
The strength of the link between austerity measures and unrest is our first important finding. Is the link causal? Other factors, such as generally depressed economic conditions, could drive up unrest and the need for cut-backs simultaneously. Controlling for economic growth does not change our results. This suggests that we capture more than the general association between economic downturns and unrest. To demonstrate that causality runs from cut-backs to unrest, we refine the data in two ways: First, we analyse a more detailed dataset that gives information about the causes of each incident. Second, we use recently-compiled data on changes in the government budget that follow directly from policy changes (Devries et al. 2011). For both types of additional evidence, we find clear indications that the link runs from budget cuts to unrest. We also conduct placebo tests with other types of unrest – inspired by ecological issues and world peace, for example – and find no effect of budget measures.
Our findings are robust to a wide range of alternative specifications and further tests. Different measures of unrest do not affect our conclusions. We examine if the link between austerity and unrest changes as countries institutions improve. For most value of the Polity2 score of institutional quality, results are broadly unchanged.
However, countries with very high levels of constraints on the executive show a weaker degree of association. Further, we examine if the spread of mass media changes the probability of unrest. This is not the case. If anything, higher levels of media availability and a more developed telecommunications infrastructure reduce the strength of the mapping from budget cuts to instability. We also test which part of the distribution of unrest is responsible for our results, using quantile regressions: The higher the level of unrest, the bigger the relative impact of additional budget cuts.
Finally, we test for asymmetries in the relationship between unrest and austerity.
Reductions increase instability, but spending increases do not cut the number of incidents to the same extent.
Earlier papers on the same topic have typically focussed on case studies, or on subsets of the developing world.3 DiPasquale and Glaeser (1998) analysed race riots in the US in the 1960s and early 1990s.4 They find that ethnic heterogeneity and unemployment rates are a strong predictors of riots, and that poverty is relatively Theoretical work on unrest includes Kuran (1989), Tullock (1971), and Grossmann (1991).
The authors also analyse international data for the period 1960-85. They find that higher GDP reduces the incidence of riots, while urbanization rates are positively associated with them.
unimportant. Bohlken and Serengeti (2010) find that riot probabilities in India over the period 1982-95 dropped sharply when growth rates accelerated. Field et al. (2008) examine religious violence in Ahmedabad, India, in 2002, and argue that tenancy rights created neighbourhoods that were more integrated, and hence more prone to violence.
Comparative work by Alesina and Perotti (1996) suggests that inequality leads to more unrest, and this adversely affects investment. Work on 23 African countries during the 1980s found that budget cuts had typically no effect on political and social stability. IMF interventions, on the other hand, often led to more frequent disturbances (Morrison, Lafay, and Dessus 1994). Paldam (1993) examines current account crises in seven South American countries during the period 1981-90, using high-frequency (weekly) data. He finds that the run-up to new austerity measures is associated with higher levels of unrest, but that actual implementation is followed by fewer disturbances. Similarly, Haggard, Lafay and Morrison (1995) find that IMF interventions and monetary contractions in developing countries led to greater instability. Analysing the period 1937-1995, Voth (2011) explores related issues for the case of Latin America. He finds that austerity and unrest are tightly linked in a majority of cases.
Relative to these papers, we make a number of contributions. To the best of our knowledge, ours is the first systematic analysis of how budget cuts affect the level of social instability and unrest in a broad cross-section of developed countries, over a long period. We also examine the exact causes of instability in a subset of the data where more detailed information is available, and find a strong link with austerity.
This also allows us to perform placebo tests. Using detailed evidence on the motivations for government expenditure changes, we strengthen the causal link still further. Finally, we examine the link between instability and media penetration.
Contrary to expectations, increased availability of telecommunications is not associated with higher levels of unrest.
Other related literature includes work on the political economy of fiscal consolidation, and on its economic effects. The composition of fiscal adjustment has been examined; cutting entitlement programs tends to produce persistent improvements in the budget balance, while revenue measures and capital expenditure cuts have only temporary effects (Alberto Alesina and Roberto Perotti 1995). The timing of stabilization measures has been explored in war-of-attrition models, which view relative bargaining strength of different groups as crucial (Alberto Alesina and Drazen 1991). A rich literature has examined the macroeconomic effects of budget cuts. Giavazzi and Pagano (1990) and Alesina et al. (2002) find that cuts can be expansionary. Amongst the reasons suggested for this finding are a reduction in uncertainty about the course future spending (Blanchard 1990a), and a positive wealth shock as a result of lower taxes in the future (Bertola and Drazen 1993).5 Recently, work by the IMF has suggested that austerity measures may be less expansionary than previously thought; they may well have the standard negative Keynesian effects as a result of lower demand (IMF 2010; Pescatori, Leigh, and Guajardo 2011).
We proceed as follows: Section 2 presents our data, and section 3 summarizes our main results. Robustness checks and extensions are discussed in section 4; section 5 concludes.
2. Data In this section, we briefly describe our data and summarize its main features. We use two datasets – a long-term one which allows tracing out the broad patterns of unrest and austerity since 1919, as well as a short-term one that contains richer information on the causes of unrest. For both, we use information on unrest as well as on economic performance and budget measures.
Five main indicators of domestic conflict in the long-term data will form the main focus of this study – general strikes, riots, anti-government demonstrations, political assassinations, and attempted revolutions. These data are part of the Cross National Time Series Dataset, compiled by Arthur Banks (2010) and his collaborators. The main source of data on unrest episodes are the reports of the The New York Times, while the variables' definition is adopted from Rummel (1974). In addition, we use data on GDP, government revenue, expenditure, and the budget balance from a variety of sources.6 The long-term data has information on 26 European countries and covers the years from 1919 to 2008.7 Once the response of labor supply and capital formation is fully taken into account, these effects may not go through (Baxter and King 1993).
Data on fiscal variables (Total Central Government Expenditure and Revenue) and GDP are from OECD Stat (2010) for years from 1970 onwards, and from Mitchell (2005) for the period 1919-1970.
Data on GDP growth in real terms for the all sample are from Maddison (2010).
The 26 European countries included in the long-term data are: Austria, Belgium, Bosnia and Herzegovina, Bulgaria, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Netherlands, Norway, Poland, Portugal, Romania, Russian Federation, Serbia, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, United Kingdom, Yugoslavia.
Table 1 gives an overview of the main variables and their descriptive statistic for the long-term data. The average number of assassinations and general strikes was quite low in our sample, with less than 2 events in each decade. There were more riots and more demonstrations – 5-6 per decade. Attempted revolutions are quite rare, but some countries registered high levels of instability. The record in our sample is Germany in 1923, with 5 recorded attempts at overthrow (with communist insurgencies in Saxony and Thuringia, the Hitler Beer Hall Putsch, and a separatist movement in the Rhineland). Assassinations and riots similarly show a broad range of observed values.