«Jacopo Ponticelli Hans-Joachim Voth UPF ICREA/UPF and CREI Abstract: Does fiscal consolidation lead to social unrest? From the end of the Weimar ...»
Romer and D.H. Romer 1989). Table 7 gives the results. If we use the same specification as in Table 1 (where we analysed the dataset spanning the period 1919we find similar results. Increasing expenditure lowers levels of unrest (column 1). The key variable driving the relationship between budget balance and instability is expenditure, not taxes (columns 2 and 3). The results are robust to including country and year fixed effects. In column 6, we investigate what happens when we use all forms of demonstrations, not just those associated with austerity. The coefficient is small, positive, and insignificant.
Table 7: EPCD data on unrest and austerity – 1980 to 1995 We can strengthen this result further by conducting a placebo test. In Table 8, we use a set of alternative types of unrest, and test if they can be predicted by the same explanatory variables as in Table 7. Labour disputes and unrest inspired by the state of the economy are more frequent when budgets are being cut, but the link is not strong or statistically significant. Peace rallies, and unrest as a result of education issues, show the opposite sign of the coefficient on austerity – times of rising expenditure also seem to bring these issues to the fore. Overall, the placebo test shows that only in the case of anti-austerity demonstrations is there a strong and significant link with changes in government expenditure.
Table 8: Placebo tests
Another way to strengthen the argument for a causal link is to examine budget measures in more detail. Some of the variation in the budget balance that we have used so far will simply reflect revenue and expenditure changes that are driven by the economic cycle. A simple way to deal with the problem is to use Alesina and Ardagna’s (2010) cyclically-adjusted primary budget balance. In table 9, col. (2), we report the results. The coefficient on budget changes is almost identical to the baseline specification. In col. (3), we use the IMF measure of policy-action based changes in the budget balance.13 This also produces a large, significant coefficient. The closer we get to measuring the impact of policy measures, the larger coefficient becomes. This strengthens the case for a causal link between unrest and austerity.
Since Devries et al. (2011) only report positive changes in the budget balance, data from IMF International Financial Statistics has been used to proxy for negative changes in the budget position in the IMF (2011) series, sign and size of the coefficient are not affected by this assumption.
Table 9: Unrest and alternative measures of budget balance
4. Robustness and Extensions In this section, we examine the sensitivity of our results. We first examine interaction effects with institutional factors. Do countries with more accountable governments weather the storms of austerity better?. We also examine if the effect may be driven by outliers, whether positive or negative changes in expenditure matter more for the effect on unrest, and whether the effect is constant in all parts of the distribution of the dependent variable.
Greater constraints on the executive and more democracy should on the hand reduce social conflict; on the other, there will be less repression by the authorities as Polity scores improve. Which effect dominates is not clear ex ante. Table 10 demonstrates that in countries with better institutions, the responsiveness of unrest to budget cuts is generally lower. Where constraints on the executive are minimal, the coefficient on expenditure changes is strongly negative – more spending buys a lot of social peace. In countries with Polity-2 scores above zero, the coefficient is about half in size, and less significant. As we limit the sample to ever more democratic countries, the size of the coefficient declines. For full democracies with a complete range of civil rights, the coefficient is still negative, but no longer significant.
The link with growth is less clear-cut. Higher output hardly dents the tendency to riot, demonstrate, assassinate, or strike in countries with low institutional quality.
The opposite is true on average in countries with scores above zero, and throughout the range of scores. The only exception is for full democracies, where the connection is weaker.
Table 10: Unrest and Institutional Quality (dependent variable: CHAOS) When does the link between budget cuts and unrest become particularly strong? We examine which part of the distribution of CHAOS shows a particularly large impact of austerity measures. To do so, we estimate quantile regressions, where we estimate the conditional median, and then the effect from the 5th to the 95th percentile of the distribution of CHAOS. Figure 5 shows the size of effects. The estimated coefficient is zero for much of the range. Only from the 80th percentile upwards – for countryyear observations with two or more incidents – is the effect visible. It then grows rapidly as estimated coefficient on expenditure changes (and on output growth) increases at higher and higher percentiles of the distribution of CHAOS. This suggests that unrest reacts particularly strongly to budget cuts and growth when unrest levels are already high.
Figure 5: Quantile Regression Plot, Expenditure and Growth (95% confidence intervals) How much does our main finding depend on the way in which we aggregate unrest?
CHAOS is the simple sum of incidents. Instead, we can use the weighted conflict index, as compiled by Banks (1994) and collaborators. It encompasses a larger set of domestic conflicts including, in addition to the components of CHAOS, purges, major government crisis and guerrilla warfare. It also assigns different, fixed weights to each individual component. The correlation coefficient of the variable with CHAOS is 0.75, significant at the 1% level. Another alternative is to use the first principal component of the five indicators that go into CHAOS. They all enter with a positive weighting. The first principal component explains 0.42 of the overall variance. The correlation coefficient with CHAOS is 0.98.
In Table 11, we use both wci and the first principal as dependent variables. Since the dependent variable is no longer a count variable, we use panel OLS, and obtain large and significant coefficients for expenditure changes and the budget position. As before, the same is not true for tax changes. The results are largely identical in terms of magnitude and significance with the baseline results in Table 3. We conclude that the way in which we measure unrest does not matter for our main finding.
Table 11: Unrest and Budget Cuts – Alternative Indicators of Unrest
An additional factor that can be questioned involves the use of the sum of unrest in the baseline results. The variable CHAOS is designed to capture the intensity of unrest, but it may be that it is influenced by a number of outliers with a high count of incidents. This would then make it easier to find significant effects. To examine this potential issue, we transform CHAOS into a simple dichotomous variable, with unrest coded as equal to unity if there are one or more incidents in a country in a single year.
In table 12, we re-estimate the baseline regression with panel logit using country- and year-fixed effects. We find the same results as before – expenditure cuts wreak havoc, tax increases do so only to a small extent and insignificantly. Overall, the budget balance matters for predicting unrest. We conclude that the role of outliers is not decisive in underpinning the relationship we established in baseline results.
Table 12: CHAOS as a dichotomous variable
Which part of the variation in the explanatory variables is responsible for the link between austerity and unrest? Do increases in expenditure do as much to reduce unrest as cuts increase them? In Table 13, we look at the issue. Column (1) shows the results for expenditure changes that are positive. The coefficient is negative, but not large, and not significant. In contrast, if expenditure changes are negative, they matter a great deal for unrest, driving up CHAOS by 0.19 incidents for each standard deviation of expenditure cuts. Next, we repeat the exercise for output changes.
Increases in output do much to cut unrest (col. 3), with a one standard deviation increase in output (3.77%) reducing CHAOS by 0.2 incidents on average. In contrast, declines do not set off major disruptions to the same degree. Overall, the results in table 12 confirm that the relevant identifying variation for expenditure changes comes from cuts; for output changes, it comes from positive growth, not recessions.
Table 13: Instability, Expenditure Cuts and Growth
Does greater media penetration increase or reduce unrest? Events in the Arab world in 2010 and early 2011 have led many to believe that greater media availability tightens the link between discontent and unrest. Data on media penetration is available in the Banks dataset. Four indicators are suitable – phone penetrations per capita, radio and television take-up, and the number of telegrams sent per capita. Radio and television are unidirectional forms of media, allowing typically government-controlled messages to be broadcast to the population. If anything, they should make it easier for authorities to reduce unrest. Phones and telegrams, on the other hand, allow peer-topeer communication. All else equal, the expected effect is that they facilitate organized protest.
To analyse the data, and to avoid confusing results with the growing availability of broadcasting and telecommunications over time, we rank penetration rate in our sample in each year. We do separately for each category, and then sum the ranks for each country-year. This gives a rank ordering of media penetration in year y.
We then divide the sample at the median. Table 14, col. (1) and (2) presents the results. We find that below-average media penetration is associated with a strong effect of expenditure cuts on unrest. Above the median, the effect disappears. There is also some evidence that the opposite pattern obtains with respect to economic conditions – the responsiveness to output changes increases as media penetration grows. In col. (3)-(6), we differentiate between uni-directional information media (infomedia) and peer-to-peer telecommunications (peermedia). While there is some attenuation of the effect of expenditure changes, it is milder than for all media. For both types, the effect of economic conditions changes from insignificant (in the part of the sample with below-median penetration) to highly significant (above the median). These results do not suggest that countries which, at any one point of time, have greater availability of mass media (relative to their neighbors) experience a higher level of unrest.14
Table 14: Media Penetration and Unrest
5. Conclusions The political economy literature on austerity suggests a paradox. There is no significant punishment at the polls for governments pursuing cut-backs (Alberto Alesina, Roberto Perotti, and Tavares 1998; Alberto Alesina, Carloni, and Lecce 2010), and no evidence of gains in response to budget expansion (Brender and A.
Drazen 2008). Also, the empirical evidence on the economic effects of budget cuts is mixed, with some studies finding an expansionary effect, and others, a contractionary one.15 Why, then, is fiscal consolidation often delayed, or only implemented halfheartedly?
The obvious alternative is to condition on the absolute level of, say, phone penetration. Most of the variation in phone penetration, however, simply reflects GDP growth and the declining cost of telephones relative to all other goods; no clear pattern emerges.
Alesina and Silvio Ardagna 2010; Alesina, Silvio Ardagna, et al. 2002; Pescatori, Leigh, and Guajardo 2011. An early example in the literature is Giavazzi and Pagano (1990).
This paper suggests one possible reason why austerity measures are often avoided – fear of instability and unrest.16 Expenditure cuts carry a significant risk of increasing the frequency of riots, anti-government demonstrations, general strikes, political assassinations, and attempts at revolutionary overthrow of the established order. While these are low-probability events in normal years, they become much more common as austerity measures are implemented. This may act as a potent brake on governments. In line with our results on expenditure, Woo (2003) showed that countries with higher levels of unrest are more indebted. High levels of instability show a particularly clear connection with fiscal consolidation.
We demonstrate that the general pattern of association between unrest and budget cuts holds in Europe for the period 1919-2009. It can be found in almost all sub-periods, and for all types of unrest. Strikingly, where we can trace the cause of each incident (during the period 1980-95), we can show that only austerity-inspired demonstrations respond to budget cuts in the time-series. Also, when we use recentlydeveloped data that allows clean identification of policy-driven changes in the budget balance, our results hold. Finally, the results are not affected by using alternative measures of unrest. Contrary to what might be expected, we also find no evidence that the spread of mass media facilitates the rise of mass protests.
Alesina, Carloni and Lecce (2010) also suggest that implementation of budget measures may be harder if the burden falls disproportionately on some groups. War-of-attrition models of consolidation are one alternative (Alberto Alesina and Drazen 1991).
Acemoglu, Daron, and James Robinson. 2000. “Why did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective.” Quarterly Journal of Economics 115 (4): 1167-1199.
Alesina, A., and R. Perotti. 1996. “Income distribution, political instability, and investment.” European Economic Review 40 (6): 1203-1228.
Alesina, Alberto, Silvio Ardagna, Roberto Perotti, and F. Schiantarelli. 2002. “Fiscal policy, profits, and investment.” American Economic Review 92 (3): 571–589.
Alesina, Alberto, Dorian Carloni, and Giampaolo Lecce. 2010. “The electoral consequences of large fiscal adjustments.” Harvard University, mimeo.