Conflict dynamics and costs in the Greek Civil War 1946–1949

Abstract

Using a new set of data from Greek Army sources, US military archives, and Communist Party documents, the paper provides a quantitative analysis of the armed confrontation that took place in Greece during 1946–1949. A dynamic Lotka–Volterra model is estimated, pointing to the existence of a conflict trap that explains the prolongation of the civil war and its dire consequences for the country. A regional analysis finds that the mobilization of guerrilla forces was crucially affected by morphology and the local persecutions of political rivals. Using neoclassical growth-accounting, the economic cost of the conflict is estimated to surpass an annual GDP, in line with similar findings in contemporary civil wars. The same framework is employed to assess the outcome in counterfactual situations discussed in this paper.

Keywords:

• Civil war
• Greece
• Equilibrium and stability conditions
• Production function

JEL Classification:

• C62
• E23
• N44
• O52

Acknowledgments

The author is thankful to D. Varakis and Ch. Triantopoulos for compiling data from military archives, and to participants in seminars held in the Bank of Greece, the LSE and at Yale University for useful remarks on an earlier version. Suggestions and comments by two anonymous referees were particularly helpful. The usual disclaimer applies.

Notes

¹ This is frequently called the ‘survivor’ function, but here the term is unsuitable for describing battle deaths.

² The Herostratian decision was taken in 1989 by a short-lived Government of the conservative and the communist parties on the naïve expectation that national reconciliation is better achieved by eliminating historical records rather than by studying and trying to understand the course of history.

³ Battle deaths and casualties include all people, combatant and civilian, that are killed or injured in armed engagements.

⁴ The decision was initially taken in June 1945, by the 12th Plenary of KKE; see Rizospastis (2011, 149).

⁵ The Third Decree of the State was issued in July 1946 and initially established eleven martial courts in key cities. A few months later the number rose to 30, covering most of the country.

⁶ For an account of town sieges by the guerrilla army see Marantzidis (2010, 192).

⁷ The main army operations were ‘Falcon-Ierax’ and ‘Stork-Pelargos’ (4/1947), ‘Eagle-Aetos’ (5/1947), ‘Swan-Kyknos’ (6/1947) and ‘Crow-Korax’ (5–8/1947). Casualties are given by GES (1976) and GES (1980) as-described in Appendix 1.

⁸ The hypothesis of no breakpoint between January and March 1948 is rejected at a range of levels from 1% to 10% for total casualties and battle-deaths as shown in Table I.

⁹ The most notorious concentration camp was at Makronisos where 28,800 soldiers and officers were kept during 1947–1950. Though the Government hailed the camp as a ‘moral transforming institution’, several of the interns perished out of torture and starvation; for an account see Kaltsogia-Tournaviti (2001, 72).

¹⁰ Details of the new structure in DAG are given by Kyritsis (2006, 28).

¹¹ Rizospastis (2011, 457).

¹² Rizospastis (2011, 290). That was the first open disagreement about strategy, with the DAG leader supporting the partisan fighting and the Secretary General of KKE insisting on urban struggle.

¹³ For a description see Rizospastis (2011, 292).

¹⁴ The characteristic equation takes the form z² − bz + c = 0, where b is calculated as the trace and c as the determinant of the Jacobian. A necessary and sufficient condition for two stable roots is |b| < 1 + c < 2, no matter if they are real or complex.

¹⁵ For example, Din (2013) considers a non-linear autonomous discrete-time model in fractional form and shows that the non-trivial equilibrium is asymptotically stable only if several complicated conditions are imposed upon the parameters.

¹⁶ Pointedly, von Clausewitz (1976, 480) was advising that a guerrilla war should not be conceived as an isolated process but ‘in the framework of a war conducted by the regular army’.

¹⁷ Estimation was also carried out for battle casualties and results are similar to those reported for deaths. This is somehow embedded in the data as the figures for DAG fighters wounded in 1948–1949 were approximately set in the Government records to be three times that of deaths; estimation details are available by the author.

¹⁸ This looked like a ‘situation being pregnant with disaster’, as figuratively is described by von Clausewitz (1976, 596). It arises when the army is ‘taking things the easy way – using superior force to filch some provinces, preferring the security of the minor conquest to a major success’.

¹⁹ That outcomes are not necessarily determined by army numbers was indignantly expressed by a Government supporter who was skeptical about ‘the alleged mathematical assertions … on so many more armies than bandits … How then it happens that the former do not snatch the latter from the neck, to finish them off?’; daily Kathimerini 30/1/1049, reprinted in Rizospastis (2011, 397–398).

²⁰ Averof-Tositsas (2010, 323–324) claims too that high ranking officials in the US were considering to opt out, while the Government was seriously contemplating defeat.

²¹ According to his own testimony, the new DAG leader was notified in 20/4/1949 that ‘Stalin put forward the case for retreating, for ending the armed struggle’; quoted in Rizospastis (2011, 449).

²² The sweeping victory in Grammos was seen by many as vindicating the supremacy of military professionals over self-trained communist leaders who ignore fundamental principles of tactical war; see, for example, Tsakalotos (1971, 317). Though factually true, the assertion should also include the huge material superiority as another critical factor.

²³ As reported by Stathakis (2002, 66).

²⁴ Similar estimates are found for the destruction in infrastructure; for example, Babanassis (2001) reports that 15% of rail lines were destroyed because of the conflict.

²⁵ Averof-Tositsas (2010, 385) asserts that material destruction due to the civil war amounted to USD 250 million in 1948 prices. By further adding damages in dwellings, refugees’ costs and labor time forgone, the loss is raised to USD one billion in 1948 prices. Greek GDP in 1948 was Drs 63,706 million or 1319 million in 1948 prices; (the 1970 exchange rate was 30 Drs/ and US CPI 1970/1948 was 1.61). Thus, the estimate represents 75.82% of annual GDP in 1948, roughly two thirds of the present estimate. The difference is probably due to the fact that Averof leaves out of calculation the losses in human capital due to forced expatriation and imprisonment.

²⁶ Applebaum (2012, 49) describes that the Communist International in Moscow was training key Italian communists to seize power in postwar Italy. According to Rizas (2001), the leadership of the Italian Communist Party was severely reprimanded by Moscow for its failure to act according to Stalin’s expectations.

²⁷ Past casualties lead to more losses in the future if the army is either constantly depleted from critical support units or is panic stricken after a major defeat. However, none of them is compatible with a prolonged conflict. Losses can also be self-multiplied in suicidal insurgencies where a new wave of martyrs follows those previously perished but, again, this cannot last for very long.