The Half-life of Leaders and the Half-life of Regimes

Thinking back on the last couple of posts, a couple of questions arise naturally. First, there is the question of the survival of regimes in general, not just democracy: if most democracies die within 15 years or so, what is the median duration (the “half-life,” if you will: the time it takes for half of them to be gone) of other regimes? And second, there is the question of the relationship between the half-life of regimes and the half-life of leaders: do regimes whose leaders tend to have longer half-lives also have longer half-lives? My interest in these questions stems from my current research on the question of legitimacy: my sense is that legitimacy matters much less than people usually think to the survival of large-scale patterns of political power and authority, so I’m interested in trying to figure out if there are systematic differences in survival between more and less “legitimate” regimes and other political structures. So this is another exploratory post, with lots of graphs.

How do we measure the duration of non-democratic regimes relative to democratic regimes? Though democratic regimes are not always straightforward to identify, non-democratic regimes come in a much wider variety of forms – from hereditary, absolute monarchies to single party regimes and multiparty hybrids, and some of these forms shade gradually into one another over the course of many years. (For a sense of this variety, consider the differences between Mexico before the 1990s under the PRI, whose presidents succeeded each other with clockwork regularity every six years and a lively opposition existed but could never win the presidency, North Korea today, where opposition is non-existent and succession is controlled by a tiny clique, and Mubarak’s Egypt.) To get a handle on this question, I’m going to use the Polity IV dataset, which codes “authority characteristics” in all independent countries (with population greater than 500,000 people) from 1800 to 2010. (I’ve been convinced by Jay Ulfelder’s work that the DD dataset I used in my earlier post is not appropriate to study comparative regime survival due to the way it codes certain democracies where alternation in power has not occurred as dictatorships, which systematically biases the survival estimates of democracies upwards).

The Polity dataset is fairly rich. Most researchers seem to use only the composite indexes of democracy and dictatorship it offers, but these indexes, while useful, do not have a strong theoretical motivation, as Cheibub, Gandhi, and Vreeland argue here. For my purposes, it is best to use the dataset to extract those authority characteristics of political regimes it purports to measure: the mechanisms of executive recruitment, the type of political competition, and the degree of executive constraint. Mechanisms of executive recruitment include hereditary selection, hybrid forms combining hereditary and electoral mechanisms, selection by small elites, rigged elections, irregular forms of seizing power, and competitive elections; types of political competition range from the repressed (all opposition banned, as in North Korea) to the open (typical of thriving democracies); and executive constraints range from unlimited to “parity” with the legislature. (See the Polity IV codebook for a full discussion). In theory, the dataset distinguishes eight kinds of executive recruitment mechanisms, ten types of political competition, and seven degrees of executive constraint, plus three different kinds of “interruption” (including breakdowns of state authority, loss of independence, and foreign invasion and occupation), leading to a possible 563 possible patterns of political authority, but these dimensions are all highly correlated (over .99); indeed, only 212 combinations of executive recruitment, political competition, and executive constraint actually appear in the date, most of them only once and for short periods of time, and it is obvious that some combinations do not even make sense. (And those that do make sense do not always capture all the information we would normally want about a political regime: Polity has no good measure for the extent of suffrage in competitive regimes, for example). But the dataset helpfully indicates how long each of these patterns last, so we can attempt a first cut at the question of the half life of regimes using a Kaplan-Meier graph:

The half-life of an “authority pattern” – a combination of an executive recruitment mechanism, a type of political competition, and a specific form of executive constraint – is 6.6 years, though the tail of the distribution is very long: some of them have lasted for upwards of a century. Switzerland, for example, has had the same authority pattern for 162 years, and Afghanistan retained the same authority pattern from 1800 to 1935 (a hereditary monarchy). As it happens, social and political life comes to be mostly structured in most places by the long-lasting patterns, but most patterns of authority do not last that long. Incidentally, at this level of abstraction there are no great regional differences in the half-lives of authority patterns, though it does seem as if authority patterns last slightly longer in Europe and the Americas than in Africa and Asia:

Yet an “authority pattern” is too amorphous a unit of analysis. We might get a better handle on the question of comparative regime survival by looking specifically at the mechanism of executive selection, since the manner in which the chief power in the state is selected is normally thought to be quite important and to have far-reaching consequences: whether supreme power is achievable by hereditary succession only or through designation within a closed elite or via competitive elections or some other means seems to have important consequences.

Of all the mechanisms of executive selection identified in the Polity IV dataset, only one, “Competitive Elections,” is unambiguously democratic by most people’s lights. Though within the dataset the fact that a regime has competitive elections is no guarantee that it will also have universal suffrage, for the most part “competitive elections” identifies most countries that most people think are democratic. We can thus calculate the duration of all periods of “competitive elections” and compare them to the duration of all “non-democratic” periods – those periods where executive selection happened through some other means. The details are somewhat tricky (see the code), but here are the results:

Some notes. As we might have expected from the discussion in the previous post, full hereditary monarchies (Russia under the Tsars, Saudi Arabia, Iran under the Shah, Portugal and Romania in the 19^th century, Nepal in the 19^th century, among others; there are 65 episodes in 40 countries in the dataset) have the longest half-lives (nearly 32 years; this increases if we collapse the two hereditary monarchy categories. Note these are not “constitutional” monarchies like the British one). But competitive electoral regimes are no slouches, with a half-life of about 17 years (in keeping with Jay’s numbers in this post, though he uses a different dataset), and as time goes on their survival rates seem to converge with those of monarchies. Similarly, “limited elite selection regimes” (e.g., single party-communist regimes, where a narrow clique selects the leader without open competition) have a half-life comparable to that of democracies, but as time goes on they tend to break down more; their survival rates seem to diverge from those of competitive electoral and monarchical regimes. Low survival rates are found especially among political forms that appear to have internal tensions, such as competitive authoritarian regimes, where elections exist and are contested by an opposition, but it is very hard for the opposition to attain real power (e.g., Zimbabwe today). I confess I don’t really understand Polity’s “Executive-guided transition” category, but it’s obviously a regime that is turning into something else (the Pinochet regime in Chile after the 1980 referendum but before the return of competitive elections counts, for example), and “ascription plus election” includes regimes where the monarch retains some real power but the legislature and other executive offices are no longer under its thumb  (only a few are recorded in the data, including Belgium in the late 19^th century and Nepal in the 1980s and 90s); it makes sense that such regimes, halfway between “real” monarchies and purely constitutional monarchies like the British, should have short half-lives as the conflict plays out and either turn into competitive electoral regimes or into more absolute monarchies.

It is also interesting to compare the relative survival rates of competitive electoral patterns of authority vis a vis periods where selection happens by non-competitive electoral means (regardless of whether the selection means stay the same):

Though the difference seems to narrow as time passes, the half-life of non-democracy since the 19^th century has been a bit longer than the half-life of competitive electoral regimes (23 vs. 17 years). In sum, political regimes do not last much more than a generation.

(For those still following, the regional breakdown indicates that competitive electoral periods have had the longest half-lives in Europe and the Americas, whereas non-democracy has had the longest half-lives in Africa and Asia; no special surprises there, though I am not sure about the reason).  

How does this relate to the half-life of leaders? For that, we turn to the ARCHIGOS dataset by Goemans, Gleditsch, and Chiozza, which contains information about the entry and exit date of almost all political leaders of independent countries in the period 1840-2010. It’s a fantastic resource – more than 3000 leader episodes, and information on their manner of exit and entry. And the conclusion one must draw from examining it is that power is extremely hard to hold on to; a ruler’s hold on power seems to decay in an exponential manner (note I haven’t checked that the decay really is exponential in the technical sense, though I'm thinking of doing that). Over this vast span of time, covering all kinds of political regimes, the half-life of leaders is only about 2 years, or a third of the median authority pattern, as we might have expected from the previous post (though the half-life of leaders is even smaller here):

Yet of course it is the people who beat the odds – those who last much longer than the average leader – the ones who shape social and political life. (There’s an endless parade of mediocrities in the dataset, two-bit prime ministers gone after a few months of ineffectual dabbling and the like).

(But don’t some leaders come back to power after losing it? In fact, the vast majority of leaders only attain power once, and never return to power, though about 100 did manage the feat three or more times. In fact, practice does not help; survival in power only appears to decrease the more previous times the leader had been in power, though note that the uncertainty of the estimates also increases, and one might expect that age would take its toll too).

We are now in a position to extend the analysis in the post below by merging the Archigos and the Polity dataset to calculate the survival curves for leaders conditional on the pattern of executive recruitment. Though I would take these curves with a grain of salt, here are the results:

As expected from the previous post, it’s good to be king – the half-life of absolute kings is about 12 years (and it’s almost always king: there are only 41 female leaders in a 3000 case dataset). Interestingly, a similar result for the half-lives of Chinese emperors is reported here (10 years: Khmaladze,  Brownrigg, and Haywood 2010, ungated) as well as for the half-lives of Roman emperors (11 years: Khmaladze,  Brownrigg, and Haywood 2007, ungated). There is something about the deep structure of monarchies in many different periods and societies, it seems, that points to a half-life in power of about 10-13 years for monarchs. 

More generally, authoritarianism pays in terms of leader tenure, despite the fact that non-competitive regimes do not always last longer than competitive ones. The highest half-lives of leaders beyond monarchs are found in limited elite selection regimes, executive-guided transitions (where non-democratic leaders are changing the rules), and competitive authoritarian regimes; but democracies are more lasting than most of these regimes (except for monarchies; see above).

Another way of looking at this is to calculate what we might call the “personalization quotient” of a regime: divide the half-life of the leader (for a given regime) by the half-life of the regime to get an idea of the percentage of the regime half-life that a leader is expected to last. So a monarch is expected to last about 37% of the half-life his regime (31.86 / 12); this is the most intensely personalized of regimes, as one might have expected given that it is devoted to the maintenance of a family line. The next most personalized regimes are competitive authoritarian regimes (28%), “self-selection” regimes (15%), limited elite selection regimes (16%), and “executive-guided transitions” (40%; this is pretty much by definition, however, so I don't make much of them). A competitive electoral regime has a personalization quotient of 8% - an expected leader half-life of about 2, divided by an expected regime duration of about 17 years. From the point of view of such a leader, it pays to try to move towards a competitive authoritarian regime, and it pays for the leader of a limited elite selection regime to move towards a formal hereditary monarchy (as is happening, in a sense, in North Korea right now, and almost happened in Egypt and Libya). 

But are authoritarian regimes more risky, so that leaders will try to hang on to power more? We can also look at that using the archigos dataset. Though leaders in non-democratic regimes have a slightly higher risk of leaving office with their heads on pitchforks or hanging from lampposts, the vast majority leave by "regular" procedures.  

More, perhaps, could be said. I’ve been wondering, for example, about whether there is a relationship between the breakdown of particular regimes and the tenure of leaders, though I’m not sure how to go about tackling that question. From the point of view of the study of legitimacy, however, what strikes me is the general fragility of patterns of authority and rule: few patterns of authority are expected have half-lives that exceed a single generation, and most don’t last nearly as long, regardless of their “legitimation formula” – heredity, competitive elections, ideology, whatever. Of course, some beat the odds, especially some competitive regimes and some monarchies, and these shape history. But the historical evidence suggests that they are in a sense the exception rather than the rule.

Code necessary for replicating the graphs in this post, plus further ideas for analysis, here and here. You will need to download the Polity IV and ARCHIGOS datasets directly, and this file of codes from my repository.

[Update: fixed some typos,  9 Feb 2012]

Comparative Political Leader Survival, 1946-2008

After playing around with Jay Ulfelder's data on the survival of democracy in the previous post, it occurred to me that I have not seen survival estimates for leaders in different kinds of regimes like the ones he discusses for democracies. So, in the spirit of exploratory data analysis, here are some graphs using data from the DD dataset of political regimes by Cheibub, Gandhi, and Vreeland, which provides information about regime type, effective heads of government, and leadership tenure for most countries in the world for the period 1946-2008. (Fuller data and methods note at the end of the post).

First, let's look at a simple estimate of leader survival for all (effective) political leaders in all regimes in the post WWII era:

The figure shows an estimate of the proportion of leaders who are expected to still be in power after n years. So, for example, after four years in power, less than half of all leaders are expected to still be in power, and after 20 years less than 10% of all leaders are expected to still be in power; the majority of all leaders last less than 4 years in power, and the vast majority less than 5. [Update: of course, some of these leaders come back to power after a shorter or longer period out of power.] This may be easier to see if we draw the plot on a logarithmic scale:

This looks like a classic "long tail" distribution of a kind often produced by "rich get richer" processes: most leaders don't last in power very long, but those who beat the odds can do very well indeed, as power feeds on itself and leaders become increasingly difficult to dislodge. (I won't say anything about power laws for fear of attracting the ire of the statistical gods). 

Nevertheless, democratic leaders and non-democratic leaders aren't equally successful at hanging on to power:

While the median democratic leader can expect less than 3 years in power, the median autocrat can expect a bit less than 7. And the gap widens with time: less than 8% of all democratic leaders can expect to hang on to power for more than 10 years, but more than 40% of autocrats do, and no democratic leader in the sample has lasted more than 25 years in power (Lynden Pindling of the Bahamas and Eric Williams of Trinidad and Tobago; your mileage may vary as to how democratic you think they were, but that's how they are coded), whereas nearly 20% of autocrats do. This may seem obvious (after all, autocrats typically impose larger barriers to political competition than democratic leaders, and ordinary people face larger obstacles in trying to get rid of them) but it also presents a bit of a puzzle, for democracies are supposed to be more responsive to popular wishes and more legitimate, and dictators are always at risk of being overthrown by their close associates. (For one influential explanation of the observed pattern of survival by Bueno de Mesquita, Smith, Siverson, and Morrow, see here and here). The greater legitimacy of democratic leaders, and their closer connection to popular opinion (to whatever degree: let's not exaggerate, either), does not seem to translate into a surer hold on power. 

Not all autocrats do equally well; absolute monarchs are especially successful at holding on to power:

Though the uncertainty of the survival estimate is larger for monarchs than for other regimes (there are just fewer monarchs in the sample) their advantage is large enough to be noticeable above the noise: nearly 60% of all monarchs can expect to last 20 or more years in power, while only 20% of other autocratic rulers can expect to survive that long, and less than 1% of democratic leaders can hope for such a career. This is another reason to think the Middle Eastern monarchs are probably safer from being overthrown than the leaders of the "republican" regimes, as Victor Menaldo has recently argued. His argument points to specific features of the political culture of these monarchies that enable elites to better monitor and discipline leaders; but other things may be going on as well (monarchs elsewhere in the world also appear to have done well, so whatever enables monarchs in the Middle East to survive appears to also work elsewhere, though admittedly most of the world's absolute monarchs since 1946 have been concentrated in the Middle East). It is also interesting to note that military and civilian dictators do not differ (much) in terms of their survival expectations (the estimates fall within each other's 95% confidence intervals), despite theoretical and empirical work that suggests that military regimes are less stable than civilian dictatorships. (Of course, this could be due to any number of things, including problems with the coding of the data and the fact that the stability of regimes is a different thing from the stability of any given leader's grip on power).

I was also curious to see whether the survival of leaders differs across regions of the world. And at least for non-democratic leaders, that seems to be the case:

There's a lot of uncertainty in these estimates (and I could have made a mistake), but in general it seems to be the case that autocratic leaders have had less success hanging on to power in Latin America, despite the USA's not always benevolent influence in the region. That was surprising to me, so perhaps someone will tell me why this is wrong. By contrast, democratic leaders all have very similar survival expectations all over the world; no evidence of "regional" effects seems evident:

Now that I've mentioned the USA's influence, we might as well look into whether autocrats (or democratic leaders) have had more trouble hanging on to power during or after the cold war. Surprisingly, it seems they have not: leaders in both regimes had the same survival expectations in both periods. But this was tricky to figure out how to calculate, and it is the most likely spot where I might have made a mistake (see sources and methods note below):

Sources and methods. A full description of the DD dataset can be found here, including the criteria it uses for categorizing regimes as democratic or non-democratic and a general defense of its methodological approach. (It used to be possible to download it as well from that page, but the form no longer seems to be working. I've animated the dataset here.) These criteria have been criticized for a variety of reasons, but in general DD does not suffer from worse problems than many of the other common datasets of political regimes (like Polity IV or Freedom House). It is possible that some of the coding decisions they make might influence the estimates of survival presented above, e.g. because they err on the side of classifying some regimes as dictatorships that could have been considered democratic (when there has been no alternation in power). This would tend to bias downward the survival estimates of democratic leaders. At any rate, DD includes information about leaders and their tenure, which is missing in other datasets and makes the data-wrangling easier, though this information is not always complete (there is sometimes more than one leader in a year for a given country, a fact that the dataset must omit, given its country-year resolution) and is not quite in the right format for survival analysis. I thus had to reshape  it (R code and a general description of the process; rank amateurism on display). I created three data files: one for the plots of survival for all leaders and leaders by democracy/non-democracy (ddsurvival.csv); one for the plots of survival by autocratic regime type (ddsurvival2.csv); and one for the plots of survival during and after the cold war (ddcoldwar.csv). (R code for generating all plots is here). These files treat the leader spell as a case; "right censoring" occurs when the leader dies or if the leader is still in power by 2008 (see the DD codebook; the files use a variable called ecens2). Since DD does not distinguish between deaths by natural causes and political assassinations or death in revolution, this introduces a certain amount of bias; in theory, "political" deaths should not result in"censoring" of the data. I should note that the plot of survival by autocratic regime  type does not take into account some cases where "left censoring" occurs (i.e., when a regime starts before 1946), though the number of cases where that is a problem is very small. Finally, there are a small number of repeated cases in ddsurvival.csv and ddcoldwar.csv due to problems guessing the right "entry date" for the leader; these must introduce some small amount of error, though I couldn't possibly say how much or in what direction the bias would work.

[Update, 1/31/2012: Fixed minor typos]
[Update, 1/02/2012: Changed location of code and data files]

Exit, Voice, and Democracy

Speaking of exit, voice, and domination, here's a very interesting paper by Mark E. Warren in the latest APSR: Voting with Your Feet: Exit-based Empowerment in Democratic Theory (gated, ungated earlier version), Volume 105, Issue 04, November 2011 pp 683-701. Abstract:

Democracy is about including those who are potentially affected by collective decisions in making those decisions. For this reason, contemporary democratic theory primarily assumes membership combined with effective voice. An alternative to voice is exit: Dissatisfied members may choose to leave a group rather than voice their displeasure. Rights and capacities for exit can function as low-cost, effective empowerments, particularly for those without voice. But because contemporary democratic theory often dismisses exit as appropriate only for economic markets, the democratic potentials of exit have rarely been theorized. Exit-based empowerments should be as central to the design and integrity of democracy as distributions of votes and voice, long considered its key structural features. When they are integrated into other democratic devices, exit-based empowerments should generate and widely distribute usable powers for those who need them most, evoke responsiveness from elites, induce voice, discipline monopoly, and underwrite vibrant and pluralistic societies.

Warren explicitly argues for a connection between mechanisms of exit and the promotion of nondomination, something which I had idly wondered about, and rightly argues that exit has often been neglected in democratic theory, even though modern democracies obviously depend at a basic level on certain forms of exit (e.g., from one political party to another). I also found Warren's discussion of the varieties of exit and their interaction with voice mechanisms (e.g., exit as signalling vs. exit as silence, and exit as free-riding vs. exit as empowerment) insightful, and his discussion of the ways in which exit mechanisms can be incorporated into modern democracies provocative.  (I should note that my first reaction to his argument was "I wish I'd written this paper!").

I have some quibbles, however. Warren notes that democracy is typically understood in terms of a voice-monopoly model in which collective voice is required to discipline the  potentially problematic effects of the state monopoly on violence:

The democratic case for voice usually assumes monopoly organizations. It does so normatively—voice is most important within the context of monopolistic organizations. And it does so structurally—monopoly induces voice by restricting exit. 

In these two respects, Hirschman's analysis tracks the fact that modern democracy was born of a specific kind of monopoly—that of states. In its eighteenth- and nineteenth-century origins, the democratic project focused on increasing inclusions within states that had effectively consolidated power by controlling territory, developing administrative capacities, and regularizing sovereignty through constitutional means (Poggi 1990, chaps. 1–2). The justifications for voice are closely related to these elements of monopoly in two ways. First, when a collectivity controls key features of livelihood, such as security, and solves collective action problems through coercion, then individuals subject to that coercion should have a say in how it is deployed. Second, the greater the costs of exit to individuals, the greater the need for voice. Though liberal-democratic states do not legally restrict exit from their territories, they recognize that exit is costly: It is disruptive of family, social support networks, careers, language, and culture, and can mean giving up the protections and welfare entitlements of citizenship. These monopoly-like effects are well recognized and justified by the existence of voice mechanisms—that is, democratic processes that legitimate the monopoly-like properties of the state. Thus it is appropriate that democratic theory has focused on equalities of political resources, secured by positive political rights (voting, speech, association) and related welfare rights (education and income security), as well as on the mechanisms such as electoral systems, judicial systems, public sphere discourse, and civil society activism through which citizens’ voice is translated into influence over law and policy (see, e.g., Habermas 1996, chap. 8).

The depth of attachment to monopoly within democratic theory stems from the fact that it is structurally necessary for the provision of common goods. As Hirschman's analysis suggests, democracies are sensitive to problems of collective action: Defectors from collectivities undermine democracy by undermining the possibility of common choice (Barry 1974). Union organizing is the archetypal case: The worker who breaks with the solidarity of the bargaining unit also undermines the capacity of the union to serve its members. More generally, as Olson (1971) famously detailed, when individuals are left to weigh the costs and benefits of collective action, larger groups tend to return fewer benefits, causing individuals to exit the collectivity, which in turn undermines the provision of public goods. As Hobbes understood, monopoly removes the threats to common security and provision posed by defection. Similarly, democratic theorists—particularly those focused on the important relationship between solidarity and collective choice—view exit opportunities as harmful, indeed, so much so that, as Hirschman (1970) observes more generally, exit is often branded criminal or treasonous (17).

Warren then rightly argues that voice is insufficient to the task of disciplining monopoly, especially given the scale of and the dispersion of power in modern states, and also that forms of exit can also play a role in ensuring that people affected by collective decisions are not unjustly dominated. Yet he does not discuss the possibility of exit from the state monopoly (aside from the brief mention of migration quoted above) except in terms that assimilate these forms of exit to "free-riding" (e.g., capital flight that hollows out public services). And the forms of exit he does discuss (what he calls "enabled" and "institutionalized" exit) are more or less dependent on the state insofar as they require the state to provide resources to make effective the ability of individuals to leave dominating relations, or to transform relations of domination into relations of choice. For example, a policy of full employment can be understood to enable exit from oppressive employment relations by reducing the costs of unemployment; and similarly extensive social safety nets, or a guaranteed minimum income can enable exit from such relations by making formal options (like quitting a job) much more easily taken. But forms of enabled exit are presented as dependent on the state in ways that suggest that Warren implicitly values voice more than exit, or at least thinks that voice is normatively or structurally prior to exit, and sets limits to its exercise, a view that seems to me to be unwarranted. (So Warren is fairly critical of the market as a mechanism for exit, in part because he thinks that the market tends to be biased against those with fewer resources). Yet it is by no means clear that all forms of exit from the state monopoly should be understood as forms of free-riding (see, for example, James C. Scott's work), or that enabled exit (making effective formal opportunities for exit) should be understood as something that only states can (or should) structure and provide, even if enabling exit may on occasion require large-scale collective action.

Perhaps this is a result of trying to fit a discussion of exit within democratic theory rather than simply liberal theory. It seems to me that there is something like a liberalism of voice that incorporates exit to a greater or lesser extent in its basic structure, and a liberalism of exit that similarly incorporates voice to some greater or lesser extent in its structure. Both forms of liberalism are concerned with nondomination, but they differ in their normative evaluations of the relative importance of exit and voice, in part due to different understandings of the relationships between, and the value of, the individual and the community. Warren's argument pushes a liberalism of voice closer to a liberalism of exit, but his position remains, in important respects, a liberalism of voice.

Crowdsourcing a Democracy Index: An Update

(Part 1 of possibly several, depending on time and mood)

A couple of months ago, I set up a democracy ranking website using the Allourideas software as part of a class project to crowdsource a democracy index (which has now been completed; more on that project in an upcoming post). The site works by presenting the user with a random comparison between two countries, and asking them to vote on which of these countries was more democratic in 2010 (click here if you can't see the widget below):



The 100 or so students in my class started the ball rolling, and their responses generated an initial democracy index that had a correlation of about 0.62 with the Freedom in the World index produced by Freedom House: respectable but not great. The post describing the initial results got some links from Mark Belinsky, the Allourideas blog, and Jonathan Bernstein, which increased the number of votes substantially. In fact, as of this writing, the website has registered 4402 (valid) votes, from about 203 different IP addresses, mostly in the USA, New Zealand, and Australia:

4,402 valid votes means at most 4,402 distinct comparisons out of a possible 36,672 potential comparisons of 192 countries (most comparisons have appeared only once, but a few have appeared a couple of times), or about 12% of all possible comparisons. How has the increase in the number of voters changed the generated index? And how does it compare to the current Freedom House index for 2010? As we shall see, the extra votes appear to have improved the crowdsourced index considerably.

Here is a map of the scores generated by the "crowd" - i.e., voters in the exercise (darker is more democratic, all data here):




And here's a scatterplot comparing the generated scores to Freedom House's scores for 2010 (click here for a proper large interactive version):

Powered by Tableau

The Y axis represents the score generated by the Allourideas software: basically, the probability that the country would prevail in a comparison with a randomly selected country. For example, the Allourideas software predicts that Denmark (the highest ranked country) has a 96% chance, given previous votes, of prevailing in a “more democratic” comparison with another randomly selected country for 2010, whereas North Korea (the lowest ranked country) only has a 5% chance of prevailing in this comparison. The X axis represents the sum of the Freedom House Political Rights and Civil Liberties scores for last year (from the “Freedom in the World 2011” report), reversed and shifted so that 0 is least democratic and 12 is most democratic (i.e., 14-PR+CL). The correlation between Freedom House and the crowdsourced index is a fairly high 0.84 (which is about as high as the correlation between the combined Freedom House score and the Polity2 score for 2008: 0.87). But how good is this, really? What do these scores really represent?

At the extremes, judgments of democracy appear to be “easy”: Freedom House and the crowd converge. For example, among countries that Freedom House classifies as “Free,” only six countries (Benin, Israel, Mongolia, Sao Tome and Principe, and Suriname) receive a score of 40 or below from the “crowd,” which is the highest score that any country Freedom House classifies as “Not Free” receives (Russia). But in the middle there is a fair amount of overlap (just as with expert-coded indexes, whose high levels of correlation are driven by the “extreme” cases – clear democracies or clear dictatorships). Some of these disagreements could further be attributed to the relative obscurity of some of the countries involved, given the location of the voters in this exercise (few people know much about Benin, and anyway the index got no votes from Africa), but some of the disagreements seem to have more to do with the average conceptual model used by the crowd (e.g., the case of Israel). The crowd would seem to weigh the treatment of Palestinians more heavily than Freedom House in its (implicit) judgment of Israel’s democracy. This is unsurprising, since the website does not ask participants to stick to a particular “model” of democracy; the average model or concept of democracy to which the crowd appears to be converging seems to be slightly different than the model used by Freedom House.

We can try to figure out where the crowd differs the most from Freedom House by running a simple regression of Freedom House’s score on the score produced by the crowd, and looking at the residuals from the model as a measure of “lack of fit.” This extremely simple model can account for about 69% of the variance in the crowdsourced scores on the basis of the Freedom House score (all data available here); we can improve the fit (to 72%) by adding a measure of “uncertainy” as a control (the number of times a country appeared in an “I don’t know” event, divided by the total number of times it appeared in any comparison). What (I think) we’re doing here is basically trying to predict Freedom House’s index on the basis of the crowdsourced judgment plus a measure of the subjective uncertainty of the participants. The results are of some interest: for example, participants in the exercise appear to think Venezuela, Honduras, and Papua New Guinea have higher levels of democracy than Freedom House thinks, and they also appear to think that Sierra Leone, Lithuania, Israel, Mongolia, Kuwait, Kiribati, Benin, and Mauritius have lower levels of democracy than Freedom House thinks.

A more interesting test, however, would be to do what Pemstein, Meserve, and Melton do here with existing measures of democracy. Their work takes existing indexes of democracy as (noisy) measurements of the true level of democracy and attempts to estimate their error bounds by aggregating their information in a specific way. I might try do this later (I need to learn to use their software, and might only have time in a few weeks), though it is worth noting that a simple correlation of the crowdsourced score for 2010 with the “Unified Democracy Scores” Pemstein et. al. produce for 2008 by aggregating the information from all available indexes is an amazing 0.87, and a simple regression of one on the other has an R² of .76. So the crowdsourced index seems to be doing something much like what the Unified Democracy Scores are doing: averaging different models of democracy and different "perspectives" on each country.

This all assumes, however, that there is something to be measured – a true level of democracy, which is only loosely captured by existing models. On this view, existing indexes of democracy reflect different interpretations of the concept of democracy, plus some noise due to imperfect information and the vagaries of judgment; they each involve a “fixed” bias due to potential misinterpretation of the concept, plus the uncertainty involved in trying to apply the concept to a messy reality whose features are not always easy to discern (try figuring out the level of civil rights violations in the Central African Republic compared with Peru in 2010, quick!). The crowdsourced index actually goes further and averages the different interpretations of democracy of every participant, just as the Unified Democracy Scores aggregate the different “models” of democracy used by different existing indexes. To the extent that the crowd’s models converge to the true model of democracy, then the crowdsourced index should also eliminate that “bias” due to misinterpretation. But it is not clear that there is a true model, or that the crowd will converge to it even if it existed: the crowdsourced index may have a higher bias (total amount of misinterpretation of the concept) than the indexes created by professional organizations. (And this conceptual bias might shift if more people from other countries voted; I’d really love to get more votes from Africa and Asia).

Even if there is no true model of democracy, it would be interesting to “reverse-engineer” the crowd’s implicit model by trying to figure out its components. (What do people weigh most, when thinking about democracy? Violations of civil liberties? Elections? Opportunities for participation? Economic opportunities?). One could do this, I suppose, by trying to predict the crowdsourced scores from linear combinations of independently gathered measures of elections, civil liberties, etc.; some form of factor analysis might help here? My feeling is that the crowd weighs economic “outcomes” more than experts do (so that crowdsourced assessments of democracy will be correlated with perceptions of how well a country is doing, like GDP growth), but I haven’t tried to investigate that possibility.

It would also be interesting to repeat the exercise by asking people to stick to a particular model of democracy (e.g., Freedom House’s checklist, or the checklist developed by my students – more on that later). It would also be great if the allourideas software had an option that allowed a voter to indicate that two countries are equal in their level of democracy (I think one could do this, but then I would have to modify the client; right now, the only way of signalling this is to click on the “I don’t know” button). Perhaps next year I will try some of these possibilities. All in all, it seems that crowdsourcing a democracy index produces reasonable results, and might produce even better results if the crowdsourcing is done with slightly more controls. (E.g., one could imagine using Amazon's "Mechanical Turk" and a specific model of democracy for generating data on particular years). I would nevertheless be interested in thoughts/further analysis from my more statistically sophisticated readers.

In an upcoming post I will explain how my students produced an index of democracy for 2010, 1995, and 1980, and how that crowdsourced effort compares with other existing indexes. (Short version: pretty well).

[Update 8:40pm: Made some minor changes in wording, added a couple of links]

More on Inequality, Democracy, and Dictatorship: Is there a “Natural Rate of Inequality”?

(Continues the discussion in this post, with more graphs, more data, more theory, and more verbiage. Mostly exploratory, considering further research. Statistician General’s warning: all statistical analyses in this post should be taken with large heapings of salt, since they have not been produced by a trained and licensed statistician and do not provide appropriate guidance regarding the uncertainty of any estimates. If you don’t mind a spot of quantitative social science from someone who was not trained in these dark arts but who is overly excited about learning to produce pretty graphs, go on.).

In an earlier post, I discussed some recent models of the relationship between inequality, political regime types, and democratization (e.g., Acemoglu and Robinson or Boix). The basic ideas in these models are pretty simple, even simplistic. In democracies, governments are (ideally, at least) responsive to the interests of the majority of the population, and in particular to the interests of the “median” voter (the voter in the middle of the distribution of income among voters), whereas in dictatorships governments are more responsive to the interests of smaller – sometimes much smaller – groups. To the extent that dictators are responsive to the interests of constituencies where the median income is higher than the median income in society (the typical case), we should expect that dictatorships will tend to redistribute less (to lower income groups) and have higher levels of income inequality than democracies, other things being equal (and other things are not always equal!). Moreover, these models indicate, the higher the level of inequality, the higher the degree of social conflict over the level of redistribution and ultimately over the type of regime, since “one off” redistribution in the face of occasional protest or other contentious action is not sufficiently “credible.” Hence we should expect that in the long run, democracy should be unsustainable at very high levels of inequality, and the only stable regime outcomes should be forms of dictatorship: “leftist” dictatorships where the poor (or rather, people claiming to act in their name) expropriate the rich, and “rightist” dictatorships where richer elites restrain redistributive demands by non-elites through coercive means. Finally, we should observe more regime change at higher rather than lower levels of inequality, more stable regimes at lower rather than higher levels of inequality, and more transitions to stable democracy at middle levels of inequality.

Using data on inequality from the University of Texas Inequality Project (1960-1996) plus data from the World Bank (which I didn’t use in my previous post but extends the UTIP data to 2008 for some countries), and data on political regimes by Cheibub, Gandhi, and Vreeland (1956-2008), we can see that some of these theoretical expectations appear to be reasonably well validated.[1] Here is a plot of the distribution of inequality, as measured by the gini index, in democracies and dictatorships (reproducing the first plot in my earlier post, but with World Bank inequality data added):

The median gini for democracies is 38.6; the median gini for dictatorships is 45.6. (means 40.1 and 44, respectively, with N= 3,321 observations between 1963 to 2008; similar patterns appear if we look only at particular periods, like the post cold war era). If we could add the vast majority of non-democratic systems in human history, the pattern would be even more obvious; as Lindert, Milanovic, and Williamson have argued, ancient non-democratic societies (i.e., the vast majority of all ancient agricultural societies) were at the “inequality possibility frontier” – elites extracted the maximum surplus from society.

But as I mentioned in my earlier post, it is obvious that the distribution of inequality in both democracies and dictatorships is very wide: lots of democracies have high gini values, and lots of dictatorships have low gini values. We do not see two clearly defined “peaks” in the distribution; rather, the distribution of inequality in both democracies and dictatorships appears to be “bimodal” – with distinct high inequality and low inequality peaks.

This is relatively easy to explain in the case of dictatorships: most of the dictatorships with low inequality appear to be communist countries, though there are fewer of these – exactly what we would expect from the theoretical models. (It is, after all, easier to organize a coup than a social revolution). Here’s a picture of the distribution of inequality in dictatorships only, split among communist and non-communist dictatorships:

The communist dictatorships are clustered narrowly at a low level of measured inequality (median gini 28.9; more on the “measured” bit in a minute), while the non-communist dictatorships have a somewhat broader distribution centered around a larger level of inequality (median gini 45.9).

The bimodal distribution of inequality in democracies is harder to explain; like dictatorships, democracies appear to have both a low inequality and a high inequality equilibrium. Why?

One factor that might seem to matter is simply the length of time a democratic regime has been in existence: redistribution capable of affecting the level of inequality in a society seems to take time. Here’s a plot of the distribution of inequality in democracies, split between those democracies that have endured for less than 10 years and those democracies that have endured for more than 10 years:

Younger democracies appear to have larger levels of inequality (median gini for established democracies: 36.2; median gini for new democracies: 44.4). In fact, while democracies appear to become more equal with time, dictatorships appear to become less equal:

 

And while it seems that democracies become more equal as they become richer, dictatorships appear to remain as unequal as before:

I won’t put too much stress on these graphs; the patterns in individual countries do not always or even often bear out the apparent overall pattern, and it is possible that this is just an artefact of the sparseness of the inequality datasets and the general badness of the data from dictatorships. Most “old” democracies in the dataset start at low levels measured inequality but appear to increase their level of inequality over time (e.g., the USA, France), whereas most “new” democracies start at high levels of inequality and appear to decrease these levels of inequality over time. Since there are more “new” democracies than “old” democracies here, it is possible that we are merely seeing is a kind of cohort effect, though one that is consistent with the basic theory: new democracies start at high levels of inequality, and many don’t last long (because of opposition to redistribution by elites), which skews the right hand panel so that it looks as if democracies become less unequal over time. Most dictatorships transition to democracy at a gini of around 45 (which is high for democracies), but that’s because that’s the median gini for dictatorships; similarly, most democracies transition to dictatorship at a gini of around 45 (perhaps because most new democracies are less stable, and they transition to dictatorship before engaging in significant redistribution?).

Moreover, it is obvious that democracies do become quite unequal sometimes. The USA is an obvious case. Here it is interesting to note that the USA is not a new democracy and is clearly quite rich, so (given the previous graphs) we would predict inequality to go down, but it seems to have been on an upward trend even looking at the long run (not just at the last decade):

(Similar patterns are visible in many rich democracies – France, for example). More on why this might be the case in a minute. But let’s think of different possibilities for why democracies might (or might not) decrease inequality. Consider what happened in Poland after the collapse of communism (similar trends are visible in Hungary, Bulgaria, and other communist countries that transitioned to democracy):

In this case, it seems that the transition to communism triggered the wholesale conversion of political access (the main inequality in these societies) into monetary assets, leading to a higher equilibrium level of measured income inequality. Measured income inequality was actually misleading about the distribution of power in communist countries; just because they were “equal” societies in income terms did not mean they were “equal” societies in the things that income can buy elsewhere, and when the basis of the regime changed, the “true” inequality in society reasserted itself in income terms, though it still remained relatively low in comparative terms. (An alternative story: perhaps with the to a market economy, people in Poland and other communist countries had the opportunity to trade off more income against increased inequality, and they took it. This is also plausible, but not my focus here; it is less plausible in places where wholesale conversion of communist apparatchiks to well-connected biznesmeni took place, as in Russia).

   

Sometimes democracy is, in a sense, too successful at an earlier time in redistributing income, prompting a reaction from elites. Consider Chile:

Inequality decreases fast until the 1973 coup, partly because of redistributive policies pushed by the left, at which point it increases again greatly. The interpretation is obvious: the elite could not stomach so much redistribution, and returns Chile to a higher level of inequality by coercive means. (The military Junta led by Pinochet made this point rather explicitly: their mission was to destroy communism and its leftist sympathizers in Chile. So they arrested and sometimes killed the leaders of leftist parties and coercively defanged or banned labor unions). After the transition to democracy in the late 1980s, inequality seems to stabilize at a higher level: the new democratic governments are constrained in the amount of redistribution they can undertake, both constitutionally and prudentially, and at any rate, the structure of Chilean society changes – labour unions have less power, elite assets are more mobile, etc. So elites are willing to transition to democracy, without fearing Allende-style redistribution. It’s like Chile’s long-term “natural” rate of inequality – the rate that is consistent with the maintenance of a democratic regime is somewhere around a gini of 45.

South Korea presents yet another possibility:

This pattern is also nicely consistent with the basic theory, though in a different way. Here we have a right-wing dictatorship facing a communist neighbour that presented a credible but far more redistributive model. (After the Korean War and until the mid 70s, most people thought the North was doing better than the South). In these circumstances, democracy was too threatening to elites: it was too easy to imagine a communist takeover by electoral means. It was still necessary to defuse the threat of social revolution through some redistributive measures (there were some fairly extensive land reforms, if I am remembering correctly), but not through institutionalized democracy, which was too risky. Promises of redistribution were made credible by the communist threat to the north, and in fact carried out to some extent. Eventually, however, inequality decreased sufficiently and the Northern model became sufficiently unattractive that democracy became much less costly to elites, leading to a transition in the late 1980s. Inequality again appears to stabilize after the transition.

A fourth pattern is found in Thailand:

(The highlighted points are years of successful coups; there were more unsuccessful ones, and the 2006 coup is not shown – no gini data for that year). We might interpret this as follows. Democracy is introduced in the late seventies (though resisted, as shown by coups in 1976) and is immediately associated with a decline in inequality, presumably through redistributive policies, but remains plagued with coups (more unsuccessful ones not shown), mostly supported by the elite. Yet the elites do not have enough power to sustain a military regime indefinitely; military coups merely postpone the resumption of redistributive politics. (I don’t know of Thailand inequality data for 2006 and after, but it would be interested to see what it looks like).   

The individual patterns are not always so clear. In fact, in most cases where there is data, no pattern is readily discernible: the aggregate pattern is clear, but the level of inequality in individual countries sometimes appears to fluctuate without any apparent connection to regime type. In some countries, transitions to democracy occur as inequality is going down (the South Korean pattern), in others, as it is going up; and in others the trend appears flat, at least given the available data. In some cases, periods of dictatorship are associated with increases in inequality, in others with decreases in inequality, and similarly for periods of democracy.

I suppose someone with knowledge of individual country histories could make sense of any given country pattern, and someone with better statistical skills could design an appropriate measure of whether inequality goes up or down, on average, during democratic or dictatorial periods. My best guess is that you would need to do a fixed effects model – controlling both for factors that affect the level of inequality at a global scale, and for country-specific factors that affect the level of inequality in a specific country. For example, inequality appears to have increased in many countries in the world in the last decade, presumably due to changes in the global economy – but more in some countries than others, presumably due to particularities of each country, including the influence of the political regime. So we would need to distinguish those global effects that are not subject to political control from the effects of political regime on inequality. Fiddling around a bit with something along these lines (using some advice provided by Eric Crampton, though all errors are mine), I cannot find any specific pattern when controlling for obvious things; if anything, democracy seems to increase inequality over time relative to dictatorship when using a full time series model (though it may be that I am not very good at interpreting the coefficients I’m getting, or that I’m mispecifying the thing somehow; for example, maybe one needs specific kinds of lags, etc. If you have expertise and would like to collaborate on figuring this out, please let me know.).

But what does come out of that fiddling as an important factor determining the level of inequality over time is the number of previous transitions to authoritarianism, which we might interpret as a proxy for the power of the elite to prevent redistribution. In various simple regressions, an additional transition to authoritarianism seems to increase the level of inequality by a unit in the gini index, and democracies with higher numbers of transitions to authoritarianism in the past seem to exhibit higher long-term levels of inequality. Consider this scatterplot:

When looking at the average gini of democracies that have existed for more than ten years, we see that democracies with fewer transitions to authoritarianism in the 58 year span (1962-2008) of the dataset seem to be scattered all over the place, though the upper range is less populated. But inequality clearly increases with previous transitions to authoritarianism, and the range of “permissible” inequality seems to narrow – with more transitions to authoritarianism, the narrow the range of inequality and the higher the mean. (The pattern is visible when looking not just at the mean gini of democracies existing for more than 10 years, but also at the mean gini of all democracies).  

How can we understand this? Here’s one possibility, and (finally) the justification for the title of this post. With fewer experiences of dictatorship in the recent past, democracies can enact a wide range of redistributive policies, depending on beliefs about luck and hard work, “ideological investment” by various interested parties, reasonable arguments, the organizational ability of various actors like labor unions and business associations, the visibility of wealth and the economic structure of society, etc. Such policies have a wide range of consequences, and so inequality may fluctuate quite drastically over time in established democracies, though it will in general experience some downward pressure relative to dictatorships. Perhaps there are “arms races” in which organizational and ideological innovations by representatives of the “left” (labor unions and particular forms of contentious politics like strikes, Marxism, etc.) are in time neutralized by organizational innovations by representatives of the “right” (think-tanks and particular business organizations, legal ways of constraining the power of unions, “neoliberal” ideas, etc.) and vice-versa, so that in the long run, inequality follows a trend determined by structural factors in the economy: the long-run “natural rate of inequality” for the society, kind of like what economists sometimes refer to as the “natural rate of unemployment.” Something like this has perhaps happened in the USA (at least if we credit people like Hacker and Pierson), where “right-wing” actors developed organizational and ideological innovations to counter the organizational and ideological tools of the “left;” but in the long run, what we should see, short of a change in regime, is a reversion to the mean trend, itself determined by economic factors that are not easily susceptible of political control.

 In some societies, however, some of these redistributive policies prove intolerable to elites, who then stage coups. When the society returns to democratic rule, representatives of non-elites know something about the ability of elites to credibly commit to extreme measures in the face of redistributive policies, so the range of distributive outcomes that appears acceptable to all narrows. (Something like this appears to have happened in Chile, for example, where successive “left” governments have “learned” from the past not to engage in certain kinds of redistributive policies). The more successful coups there are in a society, the narrower this range. Over time, you see the development of a bimodal distribution of inequality in democracies: societies where people have learned about the “red lines” that ought not to be crossed have a higher long-run level of inequality given the structure of their economy.

As noted earlier, take all of this analysis with a grain of salt; this is more exploratory than anything else, and I am certainly not a properly trained statistician. I am just curious, and considering further research on these topics. (Collaboration possibilities are also welcome).

---

[1] Data on inequality is patchy and often of poor quality. The UTIP dataset is basically the best there is for cross country comparisons over time (going back as far as 1963 for some countries); the World Bank adds some more data points, especially after 1996 (when the UTIP data ends). I mentioned in a previous post why I like the Cheibub, Gandhi, and Vreeland dataset on political regimes so much - in particular, it operationalizes a clear and theoretically justified distinction between democracies and non-democracies - but more perhaps later.

Crowdsourcing a Democracy Index

(Sorry for the recent neglect of the blog. I just started teaching again, and that tends to absorb all my energy. So here’s a teaching-related post on something I’ve been doing in one of my classes).

One of the things my students are doing in my “Dictatorships and Revolutions” class this term is constructing a democracy index/regime classification like those produced by Freedom House, the Polity project, or the DD dataset of political regimes I’ve used in this blog in the past (see, e.g., here and here).[1] We are looking at examples of how different regime classifications can be constructed, discussing some of their problems, and then collectively constructing a set of criteria for classification, which we will ultimately use to actually code all 192 or so countries in the world at intervals of about five years for a couple of decades. (If you are interested in the actual details of how the exercise is organized, e-mail me; this whole thing is still quite experimental, so I would not mind some feedback. It’s turning out to be a bit complex). Since there are over 100 students in the class (around 120, in fact), this means that we can achieve full coverage (and even some overlap) if each student codes just 2 countries (at various points in time), and I am planning to assign 4-5 countries to each student (so each country gets at least 2 coders).  We will then examine how our crowdsourced index or regime classification compares to some of the other indexes and regime classifications.

As a warm-up exercise, I set up a democracy ranking website using allourideas.org, which I learned about some time ago via the good orgtheory people. Basically, this is a webpage where you are presented with a comparison between two countries, and asked which one is more democratic (you can answer “I don’t know,” and give a reason). The results of the pairwise comparisons can be used to generate a ranking, which represents something like the probability that a given country would be more democratic than a randomly selected country. (But rather than read this explanation, why not go play with it? It can be addictive, and it’s basically self-explanatory once you see it). I asked the students to go to this website in the first class of the term, and to vote; a lot of them voted (an average of about 14 times, i.e., 14 comparisons). I didn’t know exactly what to expect, but I was sort of hoping for a “wisdom of crowds” effect. And there is, indeed, something like that, but the effect is small. Here’s a graph (link for full screen):

Powered by Tableau

The y axis represents the sum of Freedom House’s political rights and civil liberties scores: 2 is most free, 14 least free. The x axis represents the “ranking” of the countries as calculated by the Allourideas software, ranging from 4 (North Korea has only a 4% chance of prevailing in a “more democratic” comparison against a randomly selected country) to 93 (Australia; New Zealand scored 92, and was for a time in first position, which is to be expected from a group from New Zealand; see the complete ranking here). Note that these numbers do not reflect the judgments of “individual” students, but the calculated probability of prevailing in a comparison against a randomly selected country, given the information available from previous pairwise comparisons. (No student or set of students actually “ranked” North Korea last or Australia first). The size of the bubbles is proportional to the class’ subjective “uncertainty”: basically, the number of times a country was involved in an “I don’t know” answer divided by the total number of times the country appeared in any comparisons. There were 1250 votes submitted, but since there are 192 countries, the number of possible comparisons is 36,672, which means that a relatively large number of potential comparisons never appeared. (Which is part of the reason I am posting this here – I want to see what happens if lots of people engage in this informal ranking exercise).

There’s clearly a correlation between the rating by Freedom House and the informal rankings generated by the pairwise comparisons produced by the students – about -0.62, which is pretty respectable. (Some of the correlations between Freedom House and other measures of democracy are not much higher than this). A simple regression of the Freedom House ratings on the rankings generated by the students gives a coefficient of -0.11 (highly significant, not that that matters much in this context), which means that an increase of 10 points in the student-generated ranking is associated with a decrease of about 1 point in the combined Freedom House PR+CL score. (A more thorough analysis could be undertaken, but I don’t feel qualified to do it; I’ve put up the data here for anyone who is interested in doing some more exploration, and will update it later if enough other people participate in the ranking exercise).

Most of the “obvious” cases appear at the extremes – developed, well-known democracies get a high ranking, while obvious dictatorships mostly get a low ranking. Many of the countries that seem to be misplaced, however, appear to be either small and little talked about in the news or not especially well-known to students; see, for example, Ghana (which is ranked lower than it should be, if Freedom House is right) and Armenia (which is ranked higher than it should be, if Freedom House is right). Would this change if more people contributed to the ranking, especially people from a variety of countries around the world (I know this blog gets a small readership from a number of unlikely countries –could my kind readers send this link around to people who might be interested, e.g., students?). Here's a heatmap of the student-generated rankings (darker is more democratic):

The map seems reasonable enough to the naked eye. It seems that even a simple informal ranking exercise can be a reasonable approximation to a professional ranking (like that generated by Freedom House) if the people doing the ranking have some knowledge of the countries being compared, so I would expect that more people participating would probably move the informal ranking closer to Freedom House’s measure. (Maybe this is a most cost-effective method of generating a democracy index – “the people’s democracy index,” as it were). But it could also be the case that the ranking would diverge more from the Freedom House ranking as people from diverse countries participated, with different understandings of democracy. Perhaps global opinion about which countries count as most democratic would diverge sharply from the opinions of Freedom House’s expert coders. Or perhaps it would be affected by national biases – people from particular countries would have a tendency to rank it higher/lower than a more “objective” ranking would. It would be interesting to know – so it would be great if you could spread the word by sending  this link around!

(I have also wondered whether this method would work for generating “historic” data on democracy. But the obvious way of doing this would introduce many very unlikely or difficult comparisons– e.g., could we meaningfully compare democracy levels in 1964 Gambia vs. 1980 Angola using this method? – and the less obvious way would require one to set up a website for each distinct year).

---

[1] Technically, an index of democracy and a regime classification are two different things. The Economist and Freedom House produce indexes of democracy/freedom – an aggregated measure of the degree of democracy in a given country at any particular point in time, ranging from 0 to 100. A regime classification instead takes regimes as types, and attempts to determine whether a given country should be categorized as one kind or another.