Fractionalization and Democratization in Turkey, Part I.

May 12, 2019
Part one of a paper written for a course on Comparative Political Economy.

“Demokrasi bizim için bir tramvaydır. İstediğimiz durağa gelince ineriz.”
—Recep Tayyip Erdoğan

"For us democracy is like a tram car. When we get to our stop, we get off."
—Recep Tayyip Erdoğan

Much of the current political and economic analysis on Turkey focuses on the wavering commitment to democracy shown by the Justice and Development Party (Adalet ve Kalkınma Partisi, AKP) and its leader, Recep Tayyip Erdoğan. The AKP-led government’s assault on other institutions—the military, the judiciary, professional associations, journalists—seems a dramatic reversal of fortune for Turkish democracy, and for the party held up by the West as a positive example for the Muslim world. It is, however, simply the latest stage in a long-term change in elites caused by economic development in a nation with a highly fractionalized ethnolinguistic composition. The primary symptoms of this change in elites are political fractionalization and economic inflation.

Since the founding of the Republic of Turkey, Anatolia’s diverse population has been held together by a military-bureaucratic elite which arose in the late-nineteenth century amidst the collapse of the Ottoman Empire. This elite, traumatized by several hundred years of Western (predominantly British and Russian) manipulation and predation, has traditionally sought to de-emphasize ethnolinguistic and religious differences in favor of a centralizing nationalist project through emphasizing unity in the face of external threats (Soviet or US domination), common identities (pan-Turkism or Islam), or through repression (establishing official languages and religions). At the same time, paradoxically, defense against external threats required economic development and industrialization from what was, at the beginning of the twentieth century, an essentially feudal society. This development, in turn, created countervailing demands for democracy and representation for all sections of Turkey’s highly fragmented population.

Defining fractionalization

Fractionalization as a measure of ethnolinguistic diversity is commonly calculated using one minus the Herfindahl index of group shares. This is the sum of the share of total population made up by each ethnic group squared. For Turkey, an accurate calculation of this index is quite difficult, and indeed has led to the use and propagation of several incorrect or misleading values. Anthony Annett, for example, provides an ethnolinguistic fractionalization value of 0.19 and a religious fractionalization value of 0.02 for a combined value of 0.10.1 Alesina, Devleeschauwer, Easterly, Kurlat, and Wacziarg, by contrast, provide an ethnic value of 0.32, a linguistic value of .2216, and a religious value of .0049. Compare this with the respective values for the US: .4901, .2514, and .8241.2 Turkey’s values seem particularly low, especially when one considers the country’s true diversity. According to a recent estimate by Ethnologue, the following languages are spoken by over 500,000 people: Turkish, Kurmanji Kurdish, Zazaki Kurdish, Levantine Arabic, Mesopotamian Arabic, Circassian, Persian, Azeri, and Romani. The extremely low religious diversity figures provided by both Annett and Alesina et al. must be based on the assumption that Turkey is 99% Muslim; but these figures fail to take into account the Shia-Sunni split—approximately 20-30% of the Turkish population, including both Turkish and Kurdish language speakers, follows Alevi Shiism, with profound implications for political allegiances.3 Calculating based on the Ethnologue data (the Turkish census has not examined ethnicity or language since 1965), I have generated a combined Turkish fractionalization value of .6793, considerably higher than previous estimates.4

This method can also be applied to calculate political fractionalization by replacing ethnolinguistic group share with vote share per party or party-held seats in parliament. In the Turkish case, both are helpful as there can be quite a large difference between vote share and seat share, caused first by district size and d’Hondt method parliamentary seat allocation and, after 1980 (more importantly), a 10% national threshold for any party entering parliament.

Parliamentary fractionalization can also be measured by the number of cabinets formed in a given period with the idea that hung parliaments (highly fractionalized) will have less stable governments and more frequent dissolving and reforming of coalitions. This holds true of Turkey in the late 1970s as well as the mid 1990s. Similarly, one could also measure the number of parties with ministerial portfolios per cabinet, an index of the difficulty of coalition formation.

For the purposes of this article, fractionalization has been calculated based on vote totals per general election, as well as parliamentary seat totals at the start of a term. This data shows that parliamentary fractionalization is almost always well below vote share fractionalization. Meanwhile vote share fractionalization for every general election since 1950 averages .699, very close to the estimate of ethnolinguistic fractionalization in Turkey.

The 1970s and the 1990s in Turkey

A visual comparison of political and economic data can provide immediate insight into potential relationships. If political fragmentation is one of the main stories of the Turkish 1970s and 1990s, inflation is the predominant economic one. Starting in 1977, Turkey’s import substitution industrialization economy entered a major crisis with triple-digit inflation, high unemployment, and factories stopped due to lack of imports. The spike in the misery index triggered street violence and finally the 12 September 1980 coup. Following the restoration of democracy in 1983, inflation rates slowly rose again as the economy reoriented toward trade and exports. Throughout the 1980s and 1990s the Turkish government ran large deficits, covered primarily by foreign investment. A crisis unfolded in the late 1990s, with a “postmodern coup” in 1997, culminating in 2001 amidst ineffective coalition government and an approximate exchange rate of 1,500,000 Turkish Lira to one dollar.

Given the recurring simultaneity of inflationary crisis and extreme political fragmentation, it might be useful to test if there’s any causal relationship between the two.

To that end, I calculated R-values between both inflation and vote share fractionalization as well as inflation and parliamentary fractionalization. Interestingly, there seems to be a relatively high correlation between inflation and electoral fractionalization with an R-value of .6831. However, there is a much weaker correlation between inflation and parliamentary fractionalization, with an R-value of .3917.

Persson, Roland, and Tabellini propose a possible explanation linking political fragmentation with inflation: opportunistic government consumption with reelection motives in mind.5 Indeed, Turkey’s late twentieth century coalition governments are widely thought to have pursued patronage policies on a large scale.6 A quick visual comparison of election dates and percent GDP growth shows spikes of growth following about one year after elections. But if electoral rules artificially reduce electoral fractionalization and result in parliamentary fractionalizations only moderately correlated to inflation, there must be other causal factors at play.

This is not to say that unwieldy coalition governments have no effect on inflation. Indeed, the military interventions of 1960 and 1980 both drastically reduce inflation over the course of non-democratic rule acting almost as the graphite rods in a nuclear reactor. But there could be other factors as well.

Perhaps instead, it is inflation that causes fractionalization—voters upset with rising costs and unemployment vote against the governing coalition, but cannot coordinate on a single alternative. This results in another hung parliament, only with different parties this time. However, eventually one would expect some degree of coordination to develop and push the system toward two-party alternation. That this does not happen is a sign that political fractionalization is not a result of voter punishment strategies following economic difficulty but rather that it is a constant background factor that cannot be overcome, even in the face of crisis.

Taking a cue from the common assumption that the 1970s inflationary crisis was a result of the failure of import substitution industrialization, I tested the correlation between inflation and industry as percent of GDP, arriving at an R-value of .6354.7 On the other hand, industrial development has a much weaker correlation with vote share fractionalization at .4747 and almost no correlation with parliamentary fractionalization at .1755.

The relatively high correlation between inflation and vote share fractionalization as well as between inflation and industrialization, but the lack of corresponding correlation between industrialization and vote share fractionalization indicates that both fractionalization and industrialization may be parallel developments flowing from the same cause, both with constructive effects on inflation.

1. Annett, Anthony. “Social Fractionalization, Political Instability, and the Size of Government.” IMF Staff Papers, Vol. 48, No. 3 (2001), pp. 561-592.
2. Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg. “Fractionalization.” Journal of Economic Growth, 8 (2003), pp. 155-194.
3. Alevis tend to vote against Islamist parties.
4. To view the Ethnologue data, visit Ethnologue
5. Torsten Persson, Gerard Roland, and Guido Tabellini, “Electoral Rules and Government Spending in Parliamentary Democracies,” Quarterly Journal of Political Science (2007), pp. 1-34.
6. See, for example, Ilter Turan, Turkey’s Difficult Journey to Democracy, Oxford University Press (2015), pp. 93, 184.
7. Both inflation and industrial sector percentage data are pulled from Worldbank datasets and only go back to 1960.

Subscribe

Index Atlas Weekly Mailing List