That seemed strange to me, but that the time there were simply too many other issues to look at. Still, the assertion nagged me and I wanted to come back and look at the data some more. Of course the authors are right – the folks at the Guardian can count as well as the next journalist – but there are a number of caveats that should have been mentioned in the article that paint a very different picture. The obvious conclusion would seem be that the “NO” vote has only a 1 in 10 chance of coming out on top based on history – and this point was actually tweeted quite a bit before the 19th of September. The truth is far different.
First of all, it is not very useful to go back to 1846 for results, any more than it would be for me to investigate referendum results in the Byzantine Empire or Ming China. The world was simply too different, and nationalism was only just beginning its take-over of the political architecture of Europe. The historical period must therefore be shorter for this survey to be useful as a predictive tool.
Another problem is that the article doesn’t consider any other factors. It treats the Lithuanian referendum on independence from the Soviet Union in the exact same manner as it treats the Quebec referendum on independence from Canada. Yet intuitively, there should be a significant difference between people voting to escape from a dysfunctional autocracy and those voting to separate from a liberal democracy.
This is what I set out to test.
I looked at referendums on independence since 1946. I picked that date as being within the confines of the “modern world” which we might define as the post-World War 2 international order imposed by the victorious Allies and the Soviet Union. Anything prior to that belonged to a different world and we would not be comparing apples to apples. I applied some additional criteria to try to keep the comparison relevant:
- I eliminated referendumswherethe presence of a significant foreign military force seemedtopreclude any possibility of impartiality of turnout or results. These included:
- Transnistria from Moldova, 1991
- Nagorno-Karabakh from Azerbaijan, 1991
- South Ossetia from Georgia, 1992
- Kurdistan from Iraq, 2005
- Crimea from Ukraine, 2014
- Donetsk from Ukraine, 2014
- Lugansk from Ukraine, 2014
- I excluded referendums where the question was not strictly of independence (such as a referendum on maintaining an independence status already proclaimed). These included:
- Moldova, 1994
- Transnistria, 2006
- South Ossetia, 2006
- I excluded any referendum that did not have published voting data:
- United Nations Bahraini Independence Survey, 1970
This is not to say that the excluded referendums were not representative of the will of the people; it is possible that they were. I excluded them because conditions existed that make it impossible to judge if the referendum was valid or not.
That leaves us with 39 referendums.
To determine the degree of authoritarianism vs. democracy of the pre-referendum state, I used the Polity Project historical data series provided by the Center for Systemic Peace which covers all major countries from 1800 to the present. The composite scale provided by the project goes from pure autocracy (-10) to pure democracy (+10) and can be sub-divided into the different components of governance and inclusiveness. For this analysis, I looked only at the composite score. Because a referendum is not organized overnight, nor does it take place in a vacuum, I took the average polity score for the two years preceding the referendum to reflect the institutional conditions that led up to the vote in the first place.
What comes out is this chart:
The horizontal axis is the degree of autocracy or democracy in governmental structures at the time of the referendum and two years prior. This value is for the existing nation, not the proposed one. The vertical axis is the percentage of voters in favor of independence; it should not be confused with the percentage of eligible voters and does not reflect differences in turnout. It should be noted that a majority of 50%+1 has not always been sufficient to secure independence: for example, in the two referendums held by Tokelau for separation from New Zealand, a supermajority of 66.6% was required (both failed).
The chart reveals many things:
- Democracies (Polity Score = 10) have far more referendums than autocracies. That should hardly come as a surprise, since the basic organizing principle of democracy isthe recognition ofindividual rights and the sovereignty of the people;whereas autocraciesare run by elites of one kind or another and are as often working against the interest of their own people as for them. In any case, there is little or no recognition of individual rights and citizens are discouraged, often violently, from organizing themselves, much less voting in plebiscites.
Conclusion: If your national government is categorically opposed to your organization of a referendum, you may want to question the definition of your governance structure.
- When people in autocracies do get to vote on independence, they do so overwhelmingly in favor of it. Although we have a fair number of independence referendums from autocracies in our time period (n=15) they are all clustered around two events: the fall of the Soviet Union and the disintegration of Yugoslavia. Both events occurred over the period 1991 and 1992, and can’t really be treated as individual events.
Conclusion: people who live under despotic governments will ALWAYS vote to get out of it, if given the opportunity.
- Democratic reforms are an effective way of reducing the desire for independence, if they aresubstantial. Conversely, small reforms only increase the desire for big reforms andmay not be enough to satisfy the citizens. We have a very small sample size here:
- The introduction of glasnost and perestroika in the USSR under Gorbachev reduced the Soviet Union’s polity ranking from -7 (highly autocratic) to 0 (neutral) in the space of 3 years, yet the Soviet Republics voted overwhelmingly for independence. It is likely that the reforms were too little, too late, and only whetted people’s appetite for more freedom;
- On the other hand, in the 15 years subsequent to the break-up of Yugoslavia, the inheritor state of Serbia and Montenegro had achieved considerable progress towards democratization (from -4 to +6). At that point, the people of Montenegro nonetheless voted for independence, but only by 55% to 45% (the average secession vote from Yugoslavia was 95%).
Conclusion: Don’t do partial reforms. If you are going to become a democracy, go whole hog. Otherwise, you reap none of the benefits, but you keep all of the risks.
- Democraciesprovide more than half of the instances of independence referendums since the Second World War (n=25). Of these, 12 referendums resulted in a “NO” vote, while the other 13 resulted in a victory for “YES”. Of these “YES” victories, 2were unrecognized (Faroe Islands, Rhodesia), 2 were non-binding (Greenland, Veneto – the latter was also unrecognized), and 2 weremajority votes that nevertheless didn’t pass the hurdle for independence (Tokelau).Still, that would seem toindicate that in more than 50% of cases involving democracies, the people voted to get out. But even that percentage is too high:
- 9 of the 13 “YES” votes were held in colonial territories administered by a government that was on a separate continent (Guinea, Samoa, Algeria, Rhodesia, Djibouti, Aruba, Micronesia, Tokelau x2). It is understandable that the colonists would want to be free of their imperial masters, no matter how democratic it was in Lyons or Leicester.
- Imperialism is not an absolute guarantor of a “YES” vote, unlike despotism. In 6 cases, a colonial territory voted to remain with the imperial power (Puerto Rico x 4, New Caledonia, Bermuda). We could even add Aruba, which after voting “YES” for full independence, only partially implemented it and remains happily a state within the Kingdom of the Netherlands.
- Of the remaining 16 cases, only 6 (37%) involved a non-colonial territory separating from a nation on the same continent (Faroe Islands, Malta, Gibraltar x2, Greenland, Veneto). The latter two referendums were non-binding. Malta was a late addition to the British Empire (1814) and always maintained its unique language, traditions, and local government.
- Of the 5 referendums held in integral parts of a national territory (Northern Ireland, Quebec x2, Scotland, Veneto) only the non-binding and unrecognized Veneto referendum of 2014 has prospered (20%).
Conclusion: Full democracies are relatively resilient (small sample) against internal secession threats, but not against independence threats in overseas territories. The greater the distance in geography, culture, language, religion and ethnicity, the more likely a people will vote to govern themselves.
I feel comfortable with that conclusion, especially when compared to the universal rejection of authoritarian systems. If nationalism is a “universal solvent”, democracy seems to be one of the few glues that can nonetheless hold people together. But there is still the nagging doubt created by the massive vote for in Veneto for independence from Italy; Scotland and Quebec have come very close to voting yes; and polls in Catalonia indicate that a “YES” vote would win if a referendum were held today. There must be some additional factors that separate the “at risk” from the “not at risk” democracies in the world.
The sample size of referendums is too small to do a statistical analysis, but we can look at some likely metrics to see if they directionally, intuitively tell us anything. This is what a statistician would call an “experiential model” – one built on the designer’s experience rather than on a mathematical basis.
Not all democracies are alike, and not surprisingly, it would seem to that governance is highly relevant in identifying “at risk” democracies. Italy, Spain, the USA, the UK and Canada are all “absolute democracies” (polity score = 10) and all either have had or are at imminent risk of having an independence referendum.
In the cases of Italy and Spain, where in the former the “YES” vote won a non-binding referendum and where in the latter some polls show a majority of Catalans supporting “YES”, key governance metrics are worse than in countries that have survived similar challenges intact.
Italy and Spain have greater corruption, less judicial independence, less enforcement of the rule of law and less press freedom than the three democracies that have survived their own separatist challenges in recent years. That would appear intuitive (there is a reason my website is called “Common Sense”): when people are frustrated, outraged, demoralized, disengaged and politically marginalized, they begin to view the system itself as the enemy. Reformation requires an element of faith and trust; when all reformation efforts seem doomed to failure, people start thinking of more radical alternatives and those are typically either revolution or secession.
Let me be clear: I’m not accusing Italy or Spain of being autocracies. They are still full democracies. But within the ranks of full democracies, they appear vulnerable due to their lower governance scores. Spain also suffers from a greater “official” heterogeneity, though Italy is not as “full of Italians” as a foreigner might believe. The existence of an alternative, minority language is also a risk factor: modern nationalism is founded on the myth of ethno-linguistic purity after all and it works best as a solvent on those seams. The higher the percentage of alternative language speakers, the greater likelihood that popular discontent will emerge as a bid for independence rather than a bid for internal revolution.
That is why you have the Catalans wanting independence while the Castilians support Podemos. That is why you have the Liga Nord in Veneto while you have the Five Star Movement in Rome.
Based on these metrics, Italy and Spain are “high risk” democracies, though by no means comparable to even moderately unfree states. Even the US should raise eyebrows: our democracy is by no means all it should be. If America doesn’t have a strong secessionist movement (at this time) it has the Tea Party and plenty of other fringe groups that seek a radical change to the existing governance structure.
What is the solution for “at risk” states? The solution is evidently not repression: we’ve already established that people ALWAYS vote to leave the autocratic state.
Clearly it is more democracy. Improving governance is always a good idea, even when a sizable part of your citizenry isn’t threatening to send the central government packing. More transparency, more judicial independence, less corruption, more press freedom: all of these measures translate into greater accountability of politicians and elites to the people who elect them. That is always a good thing.
 Alberto Nardelli and George Arnett, “When given the chance, countries tend to say yes to Independence,” The Guardian, 16 September 2014
 The Polity Project User Guide notes that authoritarianism and democracy are not mutually exclusive means of organization, they do not exist on a single sliding scale. Elements of each can coexist in the same governance structures: it is possible to have a government with a great deal of citizen participation (+ democracy) but with a powerful elected executive that has a low degree of accountability until the next election (+ autocracy). Countries with high scores in both categories would look something like the “illiberal democracies” of Erdogan in Turkey and Orban in Hungary. However, for the purposes of this analysis, the single composite index is still more useful.
 When I worked in Padua in 2005 (a city in Veneto), I was surprised by how strong the feeling was among the locals that anyone who lived south of Rome was a foreigner. Some might have drawn the line at the Po River.