Economic growth and political extremism

Our theoretical model concentrates on purely economic motives and does not consider other, in particular, social motives of political choices that have been discussed in the literature (see, e.g., Lipset 1967; Corneo and Grüner 2000). Our analysis applies to democratic countries, i.e., countries in which exists political competition exists. In that context, we call a political platform extreme if it proposes major redistributions of resources when compared to standard policies. See, for related theoretical analysis, Artale and Grüner (2000).

Glaeser (2005) provides political economy insights on hatred.

Although totalitarian regimes often are founded on strict systems of rules and beliefs - “the rule of law is limited by the rights of leaders to interpret the true meaning of constitutional or other legal statements” (Bernholz 1991, p. 436).

According to Porter (1998): Different groups wore different triangles, and different triangles denoted different crimes. Jews wore yellow stars but also red triangles [and] political triangles. One of the biggest groups consisted of Germans who were made to wear black triangles, meaning saboteurs. Green triangles were worn by murderers. There were other triangles or strips for Jehovah’s Witnesses, vagrants, emigrants, Gypsies, race defiler (male), race defiler (female), escape suspects, special inmates, repeaters (those who were incarcerated more than once), and members of armed forces. A bewildering array of stigmatization.

One such example is the stigmatization of homosexuals during the Nazi era. Porter (1998) writes: “The Nazis’ murder of some homosexuals started earlier than that of the Jews with the murders of Ernst Roehm and other brown shirts in his paramilitary group known as the SA.... Roehm was a major Nazi leader, second only to Hitler as they rose to power in the 20’s and early 30’s. He and his cadre of “brownshirts” were homosexuals, which was not a problem at the beginning for Hitler, but later did prove an embarrassment and a threat. Roehm and other SA leaders were murdered without warning in a famous blood purge which was led by Himmler and other SS officers at the instigation of Hitler and began on June 30, 1934, which has been called “The Night of Long Knives”.”

Our formal model does not capture forms of stigmatization that go beyond group-specific income redistribution, such as harassment, imprisonment or genocide.

Using time series analysis and US data, Durr (1993) finds that changes in consumers’ economic expectations are significantly positively correlated with a survey-based measure that takes on larger values for liberal policy sentiments and lower values for conservative policy sentiments. Fetzer (2018) finds that the support for the UK’s Independence Party is significantly increasing in individuals’ exposure to fiscal austerity.

A similar assumption is made by Benhabib and Przeworski (2006). An exogenous probability of returning to the moderate regime would not affect any of our comparative static results.

See Konrad (2004) for an analysis of political competition with costly signals to voters when voters do not know their own group status.

Note that if unstigmatized individuals can be sure that they will not be stigmatized in the future \(\left( p=0\right)\), we get that \(\hat{y}=\frac{\tau +\frac{s}{1-s}}{\tau }\bar{y}\), which exceeds \(\bar{y}\) if \(\tau >0\). The threshold exceeds average income because of the uniform gains from stigmatization. Obviously, that result may change if the probability p that an individual becomes stigmatized in the future is positive.

In “Appendix” we permit that \(s_{1}\ne s_{2}\) and we explore the conditions under which the negative relationship prevails.

In the present model we assume away any incentive effects of redistribution, permitting us to obtain a full analytical solution for voters’ income thresholds. In our simple model, fully endogenizing the choice of tax rates is not very interesting. With median below mean income, moderate two-party competition leads to an equilibrium with full redistribution. When an extreme party offers the same tax rate, the threshold simplifies to \(\hat{y}=\frac{1+\frac{s}{1-s}}{\left( \frac{1+\delta g^{\alpha }}{1+\left( 1-p\right) \delta g^{\alpha }}\right) ^{\frac{1}{\alpha }}}\bar{y}\), thus leading to the same result on the effects of growth. However, even in the present simple setup, studying a subgame perfect equilibrium is more involved. That complexity holds, in particular, if incentive effects actually must be taken into account. Although interesting, we prefer to leave that analysis for future research.

The countries (time-period) covered in our dataset are: Austria (1994–2002), Belgium (1970–2002), Denmark (1973–2002), Finland (1993–2002), France (1970–2002), West-Germany (1970–2002), Great Britain (1973–2002), Greece (1980–2002), Ireland (1973–2002), Italy (1970–2002), Luxembourg (1973–2002), Netherlands (1970–2002), Norway (1990–1995), Portugal (1985–2002), Spain (1985–2002) and Sweden (1994–2002). That is the largest possible sample given the availability of data from Eurobarometer. See Appendix Table 1 for summary statistics. For a list of the extreme right-wing and left-wing parties, see Table 12.

The data are publicly available at http://​zacat.​gesis.​org/​webview/​index.​jsp.​

The average survey size was 1088, with an interquantile range of [1000, 1049].

One can model actual vote shares as the survey vote shares plus measurement error. Larger measurement errors in the dependent variable has no effect asymptotically on the estimated coefficients of the right-hand-side variables, but it does inflate their standard errors.

The coefficient of 0.07 on the immigration variable in column (2) can be interpreted as follows: for each 0.1 million immigrants, the support for extreme right-wing parties increases by about 0.7 percentage points. The coefficient on GDP per capita growth is around \(-\) 0.3, which means that, conditional on the four control variables in column (2), for each 1 percentage point decline in annual GDP per capita growth, the support for extreme right-wing parties increases by about 0.3 percentage points.

If a semi-elasticity effect is estimated, i.e., by entering the natural logarithm of immigration on the right-hand side of the estimating equation, then, for the specification in column (2), the coefficient on log immigration is insignificant and negative; the coefficient on growth remains negative and significantly different from zero at the 1% level (estimates not reported).

Brückner et al. (2012) use the same instrument to estimate the effects of persistent income on democracy in a world panel. Weather shocks, used by Brückner and Ciccone (2011) to estimate the effects of transitory income on democracy in sub-Saharan African countries, are not suitable for the present paper’s analysis. Leigh (2009) finds that faster world growth increases the likelihood that incumbents are reelected significantly.

The same result can be seen from Fig. 4, which displays a scatter plot between lagged GDP per capita growth and the support for left-wing extremism.

The median median-to-mean after-tax income ratio in Panel A is 0.92; in Panel B the median median-to-mean after-tax income ratio is 0.83.