Can we draw a sharp distinction between regimes that are democratic and those that are not? If so, what are the criteria? If not, why not?

First published: ; Last modified:

This content was originally written in .
Footnotes were present in the original; sidenotes are used for comments.

Essay

Although it is possible to draw a sharp distinction between regimes that are considered democratic and those that are not – say, by defining democracy as simply when a regime holds regular elections – the fact that an artificial boundary can be drawn sharply does not mean that doing so is desirable or meaningful. In general, taking a dichotomous approach to classifying democracies and non-democracies is more prone to bias and definitional sensitivity than when continuous measures are used. Although sharp boundaries between the categories may be demanded for the purposes of government policy, scales are much more valuable for political scientists when analysing the effect of democracy on actual outcomes, and, moreover, such approaches capture the continuity found within the concepts of democracy and dictatorship. In this essay, I first outline the reasons that a classificatory distinction between democratic and non-democratic regimes is useful within political science at all, and the different forms such a division could take. I will then evaluate attempts to establish criteria to cleanly distinguish between democratic and non-democratic regimes, and identify their conceptual and empirical flaws. Finally, I conclude that whilst it is possible to construct classifications of democracy which are dichotomous, they lack validity

There’s the idea of face validity (does something measure what it purports to) as compared with content validity (does it relate to other measures we would expect it to). Not totally clear on the distinction in practice, or how it applies here, though.
, and therefore graduated measures are generally the most appropriate.

It is important to emphasise that the desire to classify regimes does not stem from a belief that democracy is inherently superior to autocracy, or even that there is some external objective truth about whether a given country is democratic; those would be unwarranted normative claims. Rather, distinguishing between democratic and non-democratic regimes is helpful as it allows academics to test hypotheses about the effects that having characteristics associated with democracy has on a country, and then to make predictions and recommendations from the evidence-based models which emerge. There are, for example, competing theories around which features of democracies and dictatorships are dominant in determining macroeconomic outcomes (Clark et al. 2012, 326): do democracies perform better thanks to rule of law encouraging higher levels of investment, or does the larger inequality found in dictatorships cause investment there to be greater investment? Addressing such questions is beyond the scope of this essay, but keeping in mind the sorts of hypotheses which a distinction between democratic and non-democratic regimes might be used to investigate is crucial in order to establish whether a given classificatory approach is appropriate or not

If I wanted to draw on Yudkowsky, I’d say something like « Democracy is a bundle of characteristics which covary with each other a lot. We don’t care much about democracy in itself, but instead as this bundle of features which might be upstream of other outcomes that are important to us (like life expectancy, self-reported wellbeing, etc.). But this means that we should try to tease apart which particular features are relevant for the causal chain rather than gesturing at the group. Doing so lets us make more meaningful predictions later on, since it’s unlikely likely that other countries will have exactly the same mix of features as the ones in our sample, and we thus need to isolate which features matter. »
.

Some approaches are minimalist, focussing on de jure features of the procedure by which governments are selected in a given country. Others are more substantive

Clark et al. (2012, 158) raise the objection that I have to this kind of approach: “many substantive measures of democracy conflate institutional and procedural factors with the outcomes they are thought to produce.”
, and take into account the de facto context in a nation, such as citizens' freedom of speech and assembly (Boese 2019). In general, it is easier to draw a sharp distinction between regimes that are democratic and those that are not when using a minimalist approach, since (for example) nations either do hold elections or don’t, but the level of civil liberties that residents enjoy lies somewhere on a spectrum. The choice of approach matters: different methods of classification lead to different conclusions when testing the effect that democracy has on outcomes (Clark et al. 2012, 158).

One common minimal, dichotomous approach to classifying regimes is the Democracy-Dictatorship index (DD). Under this formulation, a country is democratic only if it has a multiparty system with an elected legislature and executive, and in which there has been a previous transfer of power (Cheibub et al. 2010).1 DD trivially allows for a sharp distinction to be drawn: either a country meets those requirements or it does not, with very little ambiguity. However, the coherence and usefulness of the categorisations it yields coherent are questionable. The transfer of power condition is included to avoid Type I errors (i.e. calling countries where the incumbent would refuse to give up power if they were to lose an election democratic), yet this means that some pluralist nations such as Singapore are categorised as dictatorships. Cheibub et al. (2010) justify their approach by arguing that we cannot tell whether such nations are democratic or not, and they would prefer to make Type II errors, but this wrinkle is one initial indication of the arbitrariness of binary classification.

Some scholars may acknowledge that the binary approach has imperfections but maintain that it is nonetheless the only appropriate method of classifying democracies. Their arguments generally appeal to the idea that the components of a political system are jointly interdependent in such a way that democracy and dictatorship are differences in kind (Sartori 1987, 183), or that there is no neutral zero point between dictatorship and democracy (Alvarez et al. 1996, 21). From this they conclude that a continuous axis from dictatorship to democracy would be nonsensical. However, these assertions appear to be conceptually confused. Consider an analogy with age and maturity. Like democracy, adulthood is made up of a number of interrelated characteristics (emotional, physical, intellectual), and society is able to draw sharp legal distinctions between adulthood and childhood wherever it likes. This is useful for implementing government policy, since the right to, for instance, drink alcohol and buy cigarettes is a binary one. However, this does not mean that there is a difference in kind between the maturity of adults and of children, or that employing a continuous measure would imply the existence of a zero point between “negative” and “positive” maturity. In the political context, Dahl’s conception of polyarchy (1971, 2) demonstrates that there need not be a purported zero point on a continuous measure if we instead consider it as being over the open interval (0, ∞)(0,\ \infty), where the two extremes are the idealised, but non-actual, forms of dictatorship and polyarchy respectively. So, the conceptual arguments that we must adopt a dichotomous approach seem shaky

And from a practical perspective, a continuous approach might be much more appropriate. If one were trying to predict, say, situational awareness based on age, a continuous measurement of the independent variable would be much more helpful.
.

Moreover, Sartori’s position that political systems function as “bounded wholes” (1987, 184) is dubious in light of the proliferation of electoral authoritarianism. Empirical work by Levitsky & Way (2002) highlighted how several African and Eastern European states held regular elections featuring genuine political competition, but with incumbents distorting the playing field in their favour, often by intimidating independent media and harassing opponents. The persistence of the semi-democratic status quo in many of these countries (Levitsky & Way 2020) suggests that they are not merely undergoing a temporary transition between the two “bounded wholes” of democracy and dictatorship, contrary to Sartori and others. This demonstrates why other attempts to dichotomously define democracy, such as in terms of holding free and fair elections, will falter as DD does. In order to apply such a definition, one must operationalise the meaning of “free” and “fair” in an electoral context, an inevitably subjective endeavour – and more pertinently, one which introduces additional continuous concepts of civil rights and liberties (Bollen 1990). Even if you believe that the components of democracy are complementary and not substitutive, it is simply implausible to suggest, as Collier & Adcock (1999) do, that this means there is a valid sharp distinction between democracies and dictatorships. There being one would imply that there is some threshold of civil liberties at which the electoral arena suddenly switches from being uncompetitive to competitive, a discontinuity which is both conceptually unsatisfactory and empirically refuted by the examples of electoral authoritarian states.

In addition to the conceptual flaws with drawing a sharp distinction between democratic and non-democratic states, taking a binary indicator approach prevents precise explanations and predictions from being made. When a variety of measures of welfare are plotted against level of democracy in a country, the main relationship which emerges is that dictatorships tend to have a larger amount of variation within outcomes, but are not on average substantially different to the typical democracy (Clark et al. 2012, 342). Not only are there countries across the full range of democracy scores (as noted in the previous paragraph), but there are dictatorships across the full range of wellbeing indicators. We can think of producing a dichotomous classification as an exercise in choosing where to position the vertical line dividing democracies and dictatorships. The dispersion and variation of the within the independent variable and all outcome variables means that there is no natural place to place this line – the choice of boundary will be arbitrary because the data do not clump together in two linearly separable clusters. Furthermore, no matter where the boundary is placed, it will not be possible to make reliable inferences about outcomes from the autocracy/democracy classification of a given country, because of the lack of a clear association between the two variables and the enormous amount of variation within the autocracy group. If, as Bueno de Mesquita et al. (2003, 38) do, we instead establish the contours of democracy and dictatorship in a two-dimensional continuous space with axes for inclusion and contestation quantitatively defined, our classifications become far more meaningful. An appreciation of the non-binary nature of democracy enables us to make statistically sound predictions about measures of welfare in a way that is simply impossible if a sharp distinction is insisted upon (Clark et al. 2012, Table 10.4). Separating democratic and non-democratic nations dichotomously is therefore a considerably less valuable and powerful approach to take than using minimal, continuous measures modelled on Dahl’s ideal of polyarchy.

So, to conclude, whilst it is of course possible to draw any number of sharp distinctions between countries which are democratic and those which are not, such distinctions lack coherence and authority. Democracy is not a discrete quality that countries either possess or do not, but rather an ideal which states fulfil to varying extents. As a result, employing a binary indicator to quantify this continuous concept will thus inevitably reduce the resolution of our measurements and demand the introduction of subjective cut-off points. When categorisations are used for implementing government policy, such as whether or not to give development aid to a nation, a dichotomy may be desired in spite of its arbitrariness. However, drawing a sharp distinction between democracies and non-democracies is not a useful approach when investigating the effects that regime type has on outcomes as the continuous nature of democracy means that this results in classifications which are unreliable and reductive.

References

Bollen, K. (1990). “Political Democracy: Conceptual and Measurement Traps.” Studies in Comparative International Development 25(1): 7-24.

Boese, V.A. (2019). How (not) to measure democracy. International Area Studies Review, 22(2), 95-127. https://doi.org/10.1177/2233865918815571

Bueno de Mesquita, B., Smith, A., Siverson, R.M. & Morrow, J.D. (2003). The Logic of Political Survival. The MIT Press.

Cheibub, J.A., Gandhi, J., & Vreeland, J.R. (2010). “Democracy and dictatorship revisited.” Public Choice, 143(1/2), 67–101. https://doi.org/10.1007/s11127-009-9491-2

Clark, W.R., Golder, M. & Golder, S.N. (2012). “Chapter 9: Democracy or Dictatorship: Does It Make a Difference?”, in Principles of Comparative Politics (2nd ed.). London: CQ Press.

Dahl, R.A. (1971). Polyarchy: Participation and Opposition. New Haven, CT: Yale Univ. Press

Levitsky, S., & Way, L.A. (2002). “Elections Without Democracy: The Rise of Competitive Authoritarianism.” Journal of Democracy 13(2), 51-65. https://doi.org/10.1353/jod.2002.0026.

——— (2020). “The New Competitive Authoritarianism.” Journal of Democracy 31(1), 51-65. https://doi.org/10.1353/jod.2020.0004.

Sartori, G. (1987). The Theory of Democracy Revisited. Chatham, NJ: Chatham House

Tutor’s comments

  1. The transfer of power must have happened under the same rules as those by which the current government was elected. ↩︎