Perils of parliamentarism? Political systems and the stability of democracy revisited

Abstract

Parliamentary systems are generally regarded as superior to presidential ones in democratic sustenance. This article contributes to the debate on the relationship between systems of government and the survival of democracy by bringing in a new perspective and analysing the experiences of 131 democracies during 1960–2006. We argue that systems of government do matter, but their effects are indirect; they exert their influence through societies' prior democratic records. Confirming the conventional argument, our data analysis shows that uninterrupted parliamentary democracies face significantly lower risks of a first breakdown than their presidential counterparts. Contrary to the common understanding, however, we find that the risk of a democratic breakdown can be higher for parliamentary regimes than for presidential regimes among the countries whose democracy has collapsed in the past. Furthermore, the risk of a previously failed democracy falling again grows as (the risk of) government crises increase(s). Hence our study questions the common belief that parliamentary systems are categorically more conducive to democratic stability than presidential ones.

Keywords:

• democratization
• presidentialism
• parliamentarism
• democratic breakdown

Acknowledgements

The authors would like to thank Kevin J. Fandl, Gaspare Genna, Aníbal Pérez-Liñán, the panel participants of the APSA, MPSA, and ISA, and the anonymous reviewers meetings for insightful comments and suggestions. They acknowledge assistance provided through work supported by the National Science Foundation under Grant No. 0245071.

Notes

Earlier versions of this article were presented at annual meetings of the American Political Science Association, Midwest Political Science Association, and International Studies Association.

Linz, ‘Presidential or Parliamentary Democracy’.

For example, Stepan and Skach, ‘Constitutional Frameworks’; Mainwaring, ‘Presidentialism, Multipartism, and Democracy’; Linz, ‘Presidential or Parliamentary Democracy’; Valenzuela, ‘Party Politics’; Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development.

Cheibub, ‘Minority Governments’; Cheibub, Przeworski, and Saiegh, ‘Government Coalitions’.

Shugart, ‘Presidentialism’; Mainwaring and Shugart, ‘Juan Linz’.

For example, Ulfelder and Lustik, ‘Modelling Transitions’; Svolik, ‘Authoritarian Reversals’.

Stepan and Skach, ‘Constitutional Frameworks’; Linz,‘Presidential or Parliamentary Democracy’; Valenzuela, ‘Party Politics’.

Linz,‘Presidential or Parliamentary Democracy’.

Linz,‘Presidential or Parliamentary Democracy’; Stepan and Skach, ‘Constitutional Frameworks’.

Stepan and Skach, ‘Constitutional Frameworks’, 5.

Ibid., 10.

Linz,‘Presidential or Parliamentary Democracy’.

Shugart and Carey, Presidents and Assemblies.

Mainwaring, ‘Presidentialism, Multipartism, and Democracy’.

Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development.

Cheibub, ‘Minority Governments’; Cheibub, Przeworski, and Saiegh, ‘Government Coalitions’.

Cheibub, Presidentialism, Parliamentarism, and Democracy.

For example, Ulfelder and Lustik, ‘Modelling Transitions’; Svolik, ‘Authoritarian Reversals’.

Shugart, ‘Presidentialism, Parliamentarism’; Mainwaring and Shugart, ‘Juan Linz’.

Shugart, ‘Presidentialism, Parliamentarism’.

Mainwaring and Shugart, ‘Juan Linz’, 452.

Shugart, ‘Presidentialism, Parliamentarism’.

Valenzuela, ‘Party politics’, 216.

Laver and Shepsle, Making and Breaking Governments.

Ibid., 61.

This is not an exhaustive list of assumptions found in Laver and Shepsle's Making and Breaking Governments. We enumerate here only those relevant to our discussion.

See Mezey, ‘Coup in Thailand’.

Martin and Vanberg, ‘Policing the Bargain’.

Mainwaring and Shugart, ‘Juan Linz’.

See Mezey ‘Coup in Thailand’ for Thailand's constitutional crises in the twentieth century.

For instance, according to one source, opposition parties joined an effort to remove Sharif shortly before the coup (BBC News Online, ‘Opposition Happy’).

Barracca, ‘Military Coups’, 145.

This is not to say that countries with no prior history of democratic collapse will be immune to democratic instability. Political actors in these countries can also ‘learn’ to use extra-constitutional measures to resolve political conflict from other countries whose democracies collapsed. There is an extensive literature on the neighbourhood effects and diffusion models. However, this article focuses on the impact of societies' own experiences (that is, their previous democratic failures) on the subsequent risks of another breakdown, and first learning and diffusion effects are outside the scope of the paper.

Schelling, Strategy of Conflict, 80.

Foss and Lorenzen, Cognitive Coordination.

This proposition relates to how actors conceive available options in the choice process. Schelling (Strategy of Conflict, 60) stated that ‘The choices that cannot coordinate expectations are not really “available” without communication’. We would argue that the options that actors cannot even conceive of are not really available. See, for example, the literature on bounded rationality (Simon, ‘Human Nature in Politics’).

We used World Development Indicators for economic variables, which start in 1960.

See, for example, Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development – abbreviated as ACLP hereafter when referring to their data set. Also Freedom House classification of electoral democracy at http://www.freedomhouse.org/template.cfm?page=351&ana_page=341&year=2008

Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development, 20–9.

Ibid., 23–8.

Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development, 23.

Neher, Southeast Asia.

Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development.

The following classifications in our data set differ from the original ACLP data set. Botswana (1966–2006), Fiji (1970–1986), and Samoa (1989–2006) are reclassified as democracies as a result of the non-adoption of the Botswana rule. Central African Republic (1961), Rwanda (1962–1964), and South Korea (1963–1971) are reclassified as dictatorships. Rwanda was in a civil war, Central African Republic's first president dissolved the main opposition party already in 1960, and elections held in South Korea during 1963–1971 were nominal, marked by massive irregularities, and held under the auspices of General Park. Park led the 1961 coup, assumed executive power that year, and elected and re-elected himself during 1963–1971 before he declared the state of national emergency. Therefore, this is a condition that is quite different from a case in which a democratically elected president closes down the legislature. Brazil (1979–1984) was reclassified as a dictatorship due to the occupation of the presidency by an unelected military general, which violates the first condition (elected chief executive). The correlation between ACLP's INST2 variable and our regime variable is 0.94.

See Freedom House, ‘Freedom in World Survey Methodology’. The data on electoral democracy are available from the Freedom House website at http://www.freedomhouse.org

Cheibub, Presidentialism, Parliamentarism, and Democracy. Cheibub's extension of the ACLP data is based on the criteria that include the incumbency and Botswana rules.

The following cases differ from Freedom House's electoral democracy classification. We classified Gambia as a dictatorship since its independence per ACLP data, although Freedom House considers the country to be democratic for 30 years since independence. Fiji's democracy collapses in 2006 rather than 2003. There was no major political event in 2003 that would disqualify the country as a democracy but there was a military coup in 2006. Bosnia-Herzegovina is coded as a dictatorship during the 1990s due to its civil war, although Freedom House classifies the country as an electoral democracy in 1992 and 1996–1998. These coding changes are consistent with Cheibub's (Presidentialism, Parliamentarism, and Democracy) update of the ACLP data except for Fiji (which Cheibub does not list as a democracy, perhaps due to the Botswana rule).

Londregan and Poole, ‘Poverty’, 152.

Beck, Clarke, Groff, Keefer, and Walsh, ‘New Tools’.

For example, Lipset, ‘Some Social Requisites’; Przeworski, Alvarez, Cheibub, and Limongi, Democracy and Development; Epstein, Bates, Goldstone, Kristensen, and O'Halloran, ‘Democratic Transitions’.

Gasiorowski, ‘Economic Crisis’; Svolik, ‘Authoritarian Reversals’.

We also estimated our models using logged inflation as a measure of economic performance. However, it was not statistically significant. Information on economic variables was retrieved from World Development Indicators (World Bank 2007).

Beck, Katz, and Tucker, ‘Taking Time Seriously’.

Ibid., 1271.

Cheibub, Presidentialism, Parliamentarism, and Democracy. In all four models, the coefficients of prior breakdown and number of breakdowns are positive under parliamentary systems as predicted by the theory. The coefficients using two-tailed tests are statistically significant at p = 0.055 (Model 3) and p = 0.063 (Model 4) when estimating with the number of prior breakdowns. The coefficients for the models estimated with prior breakdown are close to achieving statistical significance at p = 0.103 (Model 1) and p = 0.117 (Model 2). Given that the Cheibub extension covers a shorter time period and includes less breakdown cases due to the incumbency and Botswana rules, the replicability of our findings using his data demonstrates the robustness of these findings.

We could not estimate the models using subsamples based on the three-way regime classification because certain key variables were dropped due to perfect collinearity. Since our primary focus is on the presidential–parliamentary distinction, this issue does not invalidate our model specifications or findings.

We did not create a similar graph for presidential countries because the coefficients were not significant at any level of government fractionalization.