Navigation – Plan du site

AccueilNumérosXXXIII-2The 1946 Allied Mission to Observ...

The 1946 Allied Mission to Observe Greek Elections

An Experiment in Quantitative Political Science
La Mission alliée d’observation des élections grecques de 1946. Une expérience de science politique quantitative
Jean-Guy Prévost
p. 163-192

Résumés

La Mission d’observation alliée des élections grecques de 1946 fut mise sur pied afin de garantir l’équité des élections et du plébiscite sur la monarchie que les factions politiques rivales s’étaient engagées à tenir à la suite des affrontements armés de décembre 1944 et de janvier 1945. À cette fin, on mit à profit les développements les plus récents dans le domaine de l’échantillonnage probabiliste et une équipe d’experts fut chargée d’effectuer une série d’enquêtes sur les conditions dans lesquelles ces consultations se tenaient. Le but de cet article est d’examiner cet épisode controversé mais largement oublié et qui se situe au carrefour de l’histoire de la guerre froide et de l’histoire de l’application des probabilités.

Haut de page

Notes de l’auteur

The author kindly thanks Catherine Michalopoulou and the two anonymous referees for their helpful comments.

Texte intégral

  • 1 Documents relating to AMFOGE are filed under: National Archives, Department of State, Record Group (...)

1The 1946 Allied Mission to Observe Greek Elections (or AMFOGE, from its unofficial title: Allied Mission for Observing Greek Elections) is remembered today mostly as a minor and, given the course of events, largely inconsequential episode of the Greek civil war (1943-1949). It was hailed by its main proponents as a robust scientific undertaking, while the contemporary press and later historians have almost unanimously questioned its results if not its motives. The purpose of this paper is basically to give an overview of what AMFOGE experts were doing, and of some of the political and scientific entanglements they ran into. In order to do this, we must start by setting AMFOGE in context and highlighting its importance in terms of our social, historical and political understanding of statistical expertise. Sections 2 and 3 will analyze the contrasting views of statisticians who were directly involved in this episode and of later historians who studied the Greek civil war. From there, we will move on to examine the specific surveys conducted by AMFOGE experts during their time in Greece and some of the political debates they were involved in at that time. The present paper is just part of an ongoing research project aimed at giving a full, contextualized account of the AMFOGE mission, based on published sources and archival evidence.1

1. AMFOGE in context

  • 2 “Agreement between the Greek Government and EAM”, Foreign Relations of the United States: Diplomat (...)
  • 3 “The Soviet Foreign Commissar (Molotov) to the Secretary of State”, Foreign Relations of the Unite (...)
  • 4 Report of the Allied Mission to Observe the Greek Elections, Washington, United States Government (...)

2AMFOGE was set up as a result of the Varkiza Agreement of February 12, 1945 in which opposing Greek political factions that had been engaged in internecine conflict for some two years consented to the rapid organization of a plebiscite on the future of the monarchy, to be followed by elections for a Constituent Assembly. This agreement, concluded the day after the Yalta Declaration of Liberated Europe was made public, was first and foremost a result of the relative defeat in Athens of the left-wing, Communist-dominated National Popular Liberation Army (ELAS) at the hands of the Greek government police and army, strongly backed by British troops. To guarantee the fairness of the upcoming plebiscite and elections, it was also agreed that “for the verification of the genuineness of the expression of the popular will the great Allied Powers shall be requested to send observers”.2 The Soviet Union soon let it be known that it objected to “the practice of control by foreign governments over national elections” and declined the invitation.3 But France agreed to join Great Britain and the United States in this undertaking, and, with the consent of the Greek government, the Allied Mission to Observe Greek Elections was thus formed. It was to be headed by Ambassador Henry F. Grady for the United States, Mr. R. T. Windle for Great Britain, and General Arnaud Laparra for France.4 AMFOGE was officially a civil mission, but given the precarious housing and road conditions of post-war Greece, most of its personnel were in fact military officers. In the meantime, the order of the upcoming plebiscite and elections were reversed. While the elections were to be held on March 31, 1946, only after the electoral victory of right-wing forces was the plebiscite finally scheduled for September 1st of that same year. A second, less well-known mission designated as AMFOGE II (its full name was Allied Mission to Observe the Revision of Greek Electoral Lists), to be headed by ambassador Leland Morris for the United States and R. T. Windle again for Britain, was entrusted with checking the electoral rolls to be used for the plebiscite and observing its progress.

  • 5 Z. Karabell, 1999, p. 17.

3The prospect of holding fair and free elections under the volatile political conditions prevailing in Greece was an unlikely one. At the start of 1946, tensions between left- and right-wing factions ran high. Despite the promise made at Varkiza, no general amnesty was implemented and left-wing militants faced arrest and imprisonment. Accusations of fraud and of intimidation by the police and gendarmerie during revision of the electoral lists were widespread, originating mainly from the Left. And even though armed gangs had been officially dissolved and disarmed, political violence was rampant throughout 1945. On an international level, Greece’s Balkan neighbors (Albania, Yugoslavia and Bulgaria) had by then fallen under Communist rule and, as “the key to the eastern Mediterranean […], Greece was of critical importance to the Allies”.5 But it was hoped that the inauspicious political conditions would be made more favorable by the presence of foreign observers and, notably, by AMFOGE’s unprecedented scientific expertise in assessing the fairness of elections.

  • 6 “A Chapter in Quantitative Political Science. Observations on the Elections in Greece”, draft pape (...)

4Overall, AMFOGE I was composed of 1,155 people, with a little over 100 civilians, most of them Americans. The large majority were either observers – essentially (unarmed) army officers in uniform – or support personnel, but they included a small contingent of statistical experts entrusted with making the best use of recently developed scientific sampling survey methods in order to fulfill the Mission’s tasks. The three most prominent of these statisticians were Jerzy Neyman, a world-renowned mathematician of Polish-Russian origin who had become a professor at Berkeley University in the late 1930s, W. Edwards Deming, a physicist by training who had worked as a mathematical and sampling expert with the United States Bureau of the Census and the Bureau of the Budget from 1940 to 1945, and Raymond J. Jessen, an authority on agricultural sampling from Iowa State University. While Neyman soon raised doubts about the Mission’s preliminary conclusions on the validity of the electoral lists and was thereafter dismissed for allegedly violating the Mission’s political neutrality, Jessen, Deming and all the other experts remained apparently convinced, throughout their stay and also after their return, both of the practical feasibility of assessing the fairness of elections held in such dire circumstances and of the enduring scientific value of the conclusions reached. As some of them stated in the summer of 1946, the Mission was there “to observe the election, obtaining quantitative information concerning compliance with the election laws of Greece”. They claimed that “there was no intention on [its] part […] to enforce the election laws or to change them or pass judgment on them; neither was this in any sense an opinion poll predicting the election”.6 The task at hand was to assess the fairness of elections and the validity of results, and this AMFOGE did, or at least attempted to do.

5As its “unanimous” report, made public just ten days after the election, stated in its “Final Judgment”:

  • 7 AMFOGE I Report, pp. 26-27.

“The Mission […] concludes that notwithstanding the present intensity of political emotions in Greece conditions were such as to warrant the holding of elections, that the election proceedings were on the whole free and fair, and that the general outcome represents a true and valid verdict of the Greek people.”7

  • 8 “The Mission is satisfied that the revision and recompilations of the electoral lists as observed (...)
  • 9 “Report by Messrs. Richard T. Windle and Leland Morris, Chiefs of the Allied Mission to Observe Gr (...)

6This judgment was, of course, immediately and fiercely challenged by Greek left-wing forces, which had finally called for voters to abstain in protest against alleged widespread fraud and violence in the months preceding the election. A number of international press correspondents were likewise doubtful of AMFOGE’s claims and later historians of the Greek conflict have almost unanimously echoed this skepticism. A similarly favorable judgment was nonetheless expressed on August 19 by AMFOGE II with regard to the revision of the electoral lists.8 Regarding the plebiscite itself, whose results had elicited comparable, if not greater disbelief than those of the elections on the part of left-wing parties and journalists, the Mission released a more ambivalent report on September 7 stating that “there is no doubt in our minds that the party representing the government exercised undue influence in securing votes in support of the return of the King, but without that influence we are satisfied that a majority of votes for the King’s return could have been obtained”.9 A week later, a supplementary report, which was never published, struck a much more skeptical note:

  • 10 “Supplement to Report on the Observation of the Greek Plebiscite”, September 13, 1946, quoted in G (...)

“The information thus gathered clearly shows that conditions necessary for the holding of a fair Plebiscite did not exist in several sections of Greece; and that the final count is not true. Whether the absence of these conditions would have led to a majority vote against the King is not a question which this Mission can answer with certainty. Its opinion is, however, that the King would in all events have received a small majority”.10

  • 11 “Allied Mission to Remain in Greece for Informal Observation”, Department of State Bulletin, vol.  (...)

7But whereas the revision of electoral lists was conducted under the guidance of sampling methods, judgment on the plebiscite itself rested upon “informal” observation carried on a “smaller scale”, as there was neither the time nor the resources to apply the more elaborate methods implemented for the March elections.11

  • 12 “Miniature sample census” was an expression used by Jessen and his co-authors in “Observations”, p (...)

8One reason for AMFOGE I’s apparent confidence in the results of its surveys and observations was their rigorous compliance with the latest scientific, mathematically based sampling methods, directly supervised in the field by the world’s foremost experts. A whole series of sample surveys were conducted in Greece under the authority of AMFOGE I during the months preceding the March election, on election day and in its immediate aftermath. They were very practical in nature and dealt with issues such as the validity of electoral rolls, their printing and posting, voter registration, complaints of fraud, intimidation and violence, irregularities on election day, as well as electoral participation or abstention and illegal voting. AMFOGE II, on the other hand, limited itself to (1) estimating the number of surplus names on the lists by taking a sample of names and then tracing these individuals, and (2) estimating the number of individuals that should have been on the list but were not. This led, however, to what was probably the most technically elaborate of all the AMFOGE surveys, the so-called “miniature sample census” based on a carefully constructed household sample; it also had a distinctly scientific purpose, that of estimating the size of Greek population and some of its basic characteristics, given the absence of any recent census and the catastrophic, but at that time undocumented, demographic consequences of the war.12

  • 13 See note 6.

9Notwithstanding the Mission’s claim that its methods were well-tested, random sampling surveys of human populations were still a fairly recent development (the 1940 census of the United States, held in admittedly more peaceful conditions than those of Greece six years later, was the first to use a random sample). All of AMFOGE’s surveys may thus be described as somewhat experimental. In fact, it may be argued that these experiments provided feedback on a still largely untried methodology, in a context where information collection and recording met with significant obstacles. As a whole – and borrowing the title of a draft paper written by three of the experts in the summer of 1946 – the AMFOGE surveys may be described as an experiment in “quantitative political science”.13 Given the highly conflictual context and the inevitable political exploitation to which they would be put, they may also be described as a large-scale mobilization of random sampling theory and methods for the purpose of interpreting and, given the conclusions reached, conferring legitimacy on controversial election results. The confidence of AMFOGE’s experts in the scientific validity of their methods, and in the results as expressed in the Mission’s report and in the scientific papers subsequently published by some of those experts, is indeed remarkable, given the extremely tense political situation on the ground and the severity of the impediments to any form of objective observation or inquiry, on which most historians of the period readily agree.

  • 14 On the rise of sampling as seen from a socio-historical perspective, see A. Desrosières, 2002 (spe (...)
  • 15 T. M. Porter, 1995, p. ix.
  • 16 T. M. Porter, 1995, p. 8.

10It is to this controversial yet largely forgotten episode – at the intersection between the history of Anglo-American foreign intervention during the early Cold War and that of survey sampling as a scientific tool for gathering information and estimating its value – that this paper will be devoted. Historians' largely negative assessment of the 1946 Greek elections and plebiscite; the escalation of violence that occurred just after the elections and the resurgence of full-scale civil war; contrasting evaluations of the respective roles played by Britain, the United States and the USSR in Greece’s predicament; the latter’s difficulty in securing a peaceful and democratic settlement for decades after the Communist guerrilla was defeated, have all competed to push any detailed examination of AMFOGE, and of its scientific dimension especially, into near oblivion. From the perspective of political history, the fate of the observation mission indeed appeared secondary, and if not anecdotal, at any rate merely an element in Western Powers’ maneuvers, as the obvious entanglement between AMFOGE and British and American political personnel in Greece seemed to prove. Drawing on new insights developed over the last decades in the history and sociology of science, this episode may, however, be examined more closely as a remarkable instance of scientific knowledge, expertise and technique deployed with the intent or the hope of solving a contentious political issue. Statistics, it is true, has never been far from politics, as its somewhat direct ancestor, seventeenth-century political arithmetic, and its etymology – State-istics – both suggest. But random sampling, which was born out of the encounter between traditional statistics and probability calculus in the early 1930s, redefined the discipline and its practice along more esoteric, mathematical lines, and allowed for the emergence of a new type of statistician, the sampling expert. And since the art of sampling made it possible to study human populations using the same methods as those applied in industry or in agriculture, it represented a revolutionary step in the detachment that quantification – which can be defined as the representation of reality through the language of numbers – already provided.14 With the advent of sampling, whatever its other advantages for measuring reliability or cost-effectiveness, statistics became more than ever a “technology of distance”.15 The AMFOGE experiment, the Mission’s specifically statistical or sampling dimension, thus deserves to be re-examined, as it provides an especially telling example of statistics’ attempt at reaching fairness, impartiality and objectivity through the impersonality of numbers.16 A review of the assessments by its own members in their subsequent writings and memoirs indeed shows a profound belief in the possibilities of statistics, and especially sampling, in this regard, as well as a lasting commitment to such views.

2. Quantification as de-politicization: the statisticians’ view

  • 17 L. R. Frankel, 1951, p. 129.

11As soon as AMFOGE’s report on the March elections was published, three of the mission’s statistical experts started working on a scientific paper summarizing the nature of the surveys they had undertaken. The above-mentioned draft, entitled “A Chapter in Quantitative Political Science”, signed by Jessen, Deming and Joseph F. Daly, and dated July 1946, while AMFOGE II was still in operation, was circulated for comments. First submitted to The Journal of Political Science in 1947, it was finally published in the American Sociological Review in 1949, as “Observations on the 1946 Elections in Greece”, with Oscar Kempthorne added as co-author. A second paper, entitled “On a Population Sample for Greece” and signed by Jessen, Deming and Richard H. Blythe, was circulated for comments in early November 1946, and published in the Journal of the American Statistical Association in September 1947, with Kempthorne again as co-author. This second article was later reprinted, almost without changes, as a chapter of Deming’s 1950 textbook Some Theory of Sampling, which one contemporary reviewer described as likely to become “for some time […] the ‘bible’ of sampling statisticians”.17 Together, these papers may be described as the truly scientific legacy of AMFOGE – in contrast to its official or political output.

  • 18 R. J. Jessen, W. E. Deming, J. F. Daly & O. Kempthorne, 1949, pp. 12-13.
  • 19 R. J. Jessen, W. E. Deming, J. F. Daly & O. Kempthorne, 1949, p. 16.
  • 20 R. J. Jessen, W. E. Deming, J. F. Daly & O. Kempthorne, 1949, p. 13.

12“Observations” was basically a defense of quantification as “an adequate basis for action when important decisions are pending”. As the authors stated at the outset, “the mere existence” of intimidation and irregularities were not sufficient grounds for declaring an election valid or invalid. It was first necessary to address a series of questions of a quantitative nature, namely: “how extensive are these irregularities? Are they so numerous as to substantially affect the outcome of the election? And to what extent are they deliberate?” To make a decision, “quantitative results” rather than “rumours, assertions, pressure and impressions” were needed. It was therefore decided to use sampling procedures based on “the theory and practice of mathematical statistics”, well defined in advance, and which left no place for “the convenience of the observer nor anyone’s judgment regarding ‘representative’ or ‘typical’ areas nor anyone’s desires to be in or out the sample […]”. In other words, quantification provided objectivity through impersonality. But quantitative information was not sufficient – it could have been obtained by simply summing observations; “a measure of its reliability was (also) required”. It was here that “the use of appropriate statistical methods” in order to obtain “the desired information […] with any desired accuracy […]” became relevant. The purpose of the paper was therefore “to distinguish between (i) quantitative results obtained from a proper statistical plan; and (ii) information furnished by general impressions without a proper statistical plan”. This distinction was highlighted by contrasting a series of questions amenable to quantification (“[…] how many men on an election list are actually living at the addresses shown? How many have died? How many men over 21 are registered? How many are not registered?”), which were those asked in the Mission’s surveys, with a series of qualitative assertions (“widespread intimidation”, “wholesale illegal registration”, “complete lack of law and order”), which could be understood as expressing the viewpoint of those who criticized the election – and the Mission.18 The bulk of the article was a general, non-technical description of the statistical procedure and of the content of the six surveys conducted under the aegis of AMFOGE I. The final summary reasserted the superiority of the Mission’s findings based on the statistical sampling method (defined as the “systematic collection of first-hand information in accordance with a plan which gave every potential voter and every polling place in Greece a preassigned chance of being investigated”), over “the numerous charges and counter-charges that marked every stage of the political campaign”, which were based, for their part, on the obvious fallacy of “generalizing from selected instances”.19 Overall, “Observations” summarized the manner in which statisticians envisioned the problem that had brought the Mission into existence and the solution it required. Managing and passing judgment on a sensitive political issue such as the validity of elections held in difficult conditions meant replacing qualitative, general and unverifiable statements emanating from contending parties with quantitative, specific and verifiable information obtained by trained observers. And this called for a method capable of providing a desired and measurable degree of accuracy, i.e., the method of probabilistic sampling that had been developed and proven “in the United States, Great Britain, India and elsewhere” over the course of the preceding decade.20 In other words, the political problem could be solved in so far as it could be translated or converted into a technical problem: from there, as the Mission’s final judgment about “charges and counter-charges” suggests, it was possible to draw what we may describe as de-politicized political conclusions.

  • 21 In a letter to William Baxter of the State Department in which he sought permission to publish the (...)
  • 22 C.-A. Chiang, 1949; Y. P. Seng, 1949. The first scholarly paper quoting “On a Population Sample” ( (...)
  • 23 C. Michalopoulou, 2014, p. 201.

13“On a Population Sample”, which focused on the inquiries made during AMFOGE II, was a much longer and altogether much more ambitious paper from the statistical point of view. It described in painstaking detail the procedures used to conduct a sample survey of households during the summer of 1946, and to estimate the size of the Greek population, as well as some of its characteristics. The first section was concerned mainly with the sample design while the second presented results for the estimates of population characteristics. Compared with “Observations”, this was a more elaborate and more technical paper, and it was accordingly published in a statistical journal. But it also represented a further and decisive step in the de-politicization of the problem at hand: even though AMFOGE II’s mandate was cited as the background against which the survey was conducted, no mention was made here of the election or plebiscite results or of AMFOGE’s assessment of these events.21 The detachment from Greek politics achieved by the paper can be measured in the way it was received by other statisticians around the world: the “Greek sample design” was soon hailed as a model of how the impossibility of conducting a complete census could be addressed, without any reference to the political conditions of the survey or how these might have influenced the results.22 It was not until 2014 that a statistician linked an assessment of AMFOGE’s statistical legacy (“[the statisticians] made no contribution to theory [but] they did demonstrate in an exemplary fashion the advances in sample survey theory and practice”) to an examination of the context in which the surveys were conducted.23

  • 24 C. M. Woodhouse, 1948, p. 264.

14Alongside this summary of the AMFOGE statisticians’ scientific output, the assessment presented in Christopher M. Woodhouse’s 1948 Apple of Discord, the first historical account of the Greek civil war, deserves to be mentioned. Woodhouse was a former operative of the Special Operations Executive (SOE) who acted as Commander of the British Military Mission (later Allied Military Mission) to help the Greek resistance during German occupation. He then served briefly as secretary general of the British section of AMFOGE. In his book, he fully endorsed the Mission’s “scientific methods”, which he described as based “on a wealth of statistical accuracy”, as well as its conclusions, which were reached “by way of scientific proof, not by way of subjective impressions”.24 The same language can be found in testimonies published by the Mission’s statisticians in later years, which generally combine this de-politicized, technical view of their work with an undeniable degree of political naïveté. For instance, Joseph Daly, who had spent the war working in quality control in the Navy and went on to a long career in the U.S. Bureau of the Census after his Greek mission (he was there only for AMFOGE I), summarized the episode as follows:

  • 25 J. F. Daly, no date, p. 2.

“About 1946, I was sitting there waiting to be ‘mustered out’ when Morris Hansen and Ray Jessen (Statistical Laboratory, Iowa State College) came along and said that they needed somebody to go to Greece to help run a survey to monitor the Greek election. President Harry Truman was still working on the Marshall Plan at this time and the Communists were in the underground in Greece and seemed to have quite a following. In addition, there was a question whether the Marshall Plan would work – whether we [the United States] dared to try to assist the southern European countries to try to get them out of the Communist orbit. So there was a gang of us that went over there, Jessen and Arnold King from Iowa State, Deming, and Jerzy Neyman. We were suppose (sic) to conduct a survey to find out how many people there really were that were eligible to vote and what proportion of them voted; if possible, we would try to find out something about the people abstaining from the election. When it was all over, it became pretty clear that the Communists would not carry more than about 15 percent of the vote, and President Truman went ahead with the Marshall Plan in Greece, and it worked out all right. It was a very interesting job, very interesting group of seven or eight of us working over there. We stayed in Greece for about 3 months and that’s what really got me interested in survey sampling.”25

15Oscar Kempthorne, who had worked at the Rothamsted Experimental Station in Britain before joining the Mission in Greece, described it in those words:

  • 26 J. L. Folks, 1995, p. 324.

“During WWII the Greeks worked hard to stop the Nazis – the Germans – and a left wing party developed. Now this was in conflict with the right wing party, and after the conclusion of WWII, there was a civil war – a fight between the right and the left. The allied governments (that is, the United States, Great Britain and France) decided that there should be an election and that this election should be monitored by them, which is really a rather interesting political move. The United States State Department decided to form a technical mission to aid in the observation of the Greek elections. This technical mission consisted of political people, but critically it depended on a technical staff of statisticians. Now, for some reason that I do not understand, the United States pushed the idea of having a sampling group. So a technical statistical group was made up to assist in the observations and the evaluation of the election. It was led by Ray Jessen (a leader in real work in sampling) of what was then Iowa State College, and it included on the staff W. Edwards Deming, Jerzy Neyman, Joe Daly and Arnold King from the United States and me from Great Britain. There should have been also a person from France, but I believe the conclusion, erroneous or not, was that there wasn’t anyone in France who could really lend technical support. So off we went to Greece. We – Jessen and the team – set up a sampling scheme in which a sample of precincts was drawn and observations were made on the sample by army officers. The survey was done, the report was made and it was concluded that the election result was fair. The result of the election was that the right wing group won the election and the left wing group lost. An outcome of this, which I heard about only recently, was that the left wing people were badly treated by the right wing people, and that many of them had to leave Greece in order to survive. The election was a very important part of the development of politics in Greece. I spent perhaps three months in Greece working on this mission. A paper on this whole plan was written by Jessen, R. H. Blythe, Deming, and me [Jessen, Blythe, Deming and Kempthorne (1947)]. It is of interest that Neyman was dismissed by the mission for actions contrary to its policies.”26

  • 27 R. J. Jessen, 1978, p. 319.
  • 28 “Here there were 60 British, 60 American, and 40 French investigators. Assignments of these nation (...)
  • 29 R. J. Jessen, 1978, p. 496.

16Neither Deming nor Jessen seem to have published testimonies of their experience, but the 1947 and 1949 papers they co-authored were frequently cited in the latter’s own 1978 Statistical Survey Techniques. The “miniature sample census” was thus presented as an “example of using measure-of-size information in various ways”.27 In a section on investigator errors and bias, Jessen also referred to the Greek Election Survey as a case where “the interpenetrating-samples method was used to determine whether nationality of investigators had any effect”.28 Finally, as an example of a quality check for internal consistency, Jessen cited the fact that two of the AMFOGE surveys “provided two independent estimates of total registered voters (that) were within reasonable sampling error of each other”. He then commented: “This result was reassuring, because so many things could have gone wrong due to circumstances”.29

  • 30 C. Reid, 1998, p. 203.

17Neyman’s own recollection of the episode was quite different. In interviews he gave during the late 1970s for the biographical account produced by Constance Reid, Neyman gave a detailed summary of the events that led to his dismissal. Two points deserve mention here. First, he declared that the AMFOGE surveys represented “a difficult assignment” but “not one of any great theoretical interest”, a view that may stem partly from the abrupt termination of his involvement, but also reflects his own prominent place in the more general history of random sampling. Regarding the conclusion of the Greek inquiries, his “considered stand” was based on a purely political argument, and he did not discuss statistical issues. The elections were “a miscarriage of justice and […] the American mission played a role in it that I regret”.30 In a radio talk he gave on his return to Berkeley in 1946, Neyman had declared:

  • 31 C. Reid, 1998, p. 208.

“The key to the question whether the elections were fair and whether their result represents the popular will of the country, seems to be provided by the fifth news item, published only two days ago. This item reports that the victorious Royalists favor the continued presence of the British army in Greece.
To my mind, if elections were really fair and if the Royalists did have behind them a large majority of the population, there would be no need for them to seek the continued support of a foreign army.”31

18This judgment, as we shall now see, was not very different from that expressed by most of the historians who have dealt with that period.

3. Quantification as charade: the historians’ view

  • 32 R. Clogg, 1987, p. 21.
  • 33 C. M. Woodhouse, 1976, p. 175. In his 1982 memoirs, Something Ventured, Woodhouse makes only one r (...)
  • 34 J. K. Campbell, 1972, p. 46.
  • 35 R. Clogg, 1987, p. 18. “These estimates are essentially guesswork […]”, T. D. Sfikas, 1994, p. 87.

19The historiography of the Greek civil war has been affected by a number of major issues. Was the civil war basically a protracted attempt by the Greek Communists to seize power? Or was it, on the contrary, the unpremeditated result of a series of compromises, misunderstandings, and defensive moves on the part of both Left, Center and Right? What were the respective roles of foreign intervention and internal politics in the unfolding of the conflict? In retrospect, the 1946 elections and plebiscite were generally depicted as “critical turning points in the progress towards the civil war that wracked Greece between 1946 and 1949”, and the two observation missions were often treated as not much more than a footnote.32 An interesting example of this attitude is provided by C. M. Woodhouse’s second and more documented account of the Greek civil war (1976), in which he revised his earlier judgment on the mission’s scientific accomplishments. Instead of a full page, he now devoted only a single sentence to AMFOGE, the term sampling was used in quotes, and the author described the Mission’s conclusions on abstention as “far from unquestionable”.33 In an endnote, Woodhouse quoted John Campbell, himself a liaison officer during the armed clash in Athens in December 1944, who had, in a 1972 account, characterized AMFOGE’s abstention figures as “wishful speculation”.34 Likewise, Richard Clogg described these estimates as “essentially guesswork”, a judgment echoed by T. D. Sfikas.35 David H. Close, for his part, wrote:

  • 36 D. H. Close, 1995, p. 174.

“The members of AMFOGE were influenced by the British and committed to their governments’ desire that elections be held soon. Their ignorance of Greek conditions made them figures of ridicule, and they could do nothing to eliminate the effect on voters of the coercion that preceded their arrival and was expected to follow their departure.”36

  • 37 G. T. Mavrogordatos, 1981, p. 184. Mavrogordatos, whose analysis also covered the September plebis (...)
  • 38 G. T. Mavrogordatos, 1981, p. 182.

20The most detailed analyses of the March 1946 election are those of George Th. Mavrogordatos and Heinz Richter. Whereas some authors tend to see the election and plebiscite as simply coinciding with the resurgence of the civil war and consider that the spiral of violence had already been set in motion by that time, Mavrogordatos emphasized their importance and described them as “the last lost chance of peaceful political evolution”.37 Given the fact that the post-Varkiza “White Terror” amounted to “a one-sided civil war waged by the monarchist right against its defenseless opponents”, he argued that the conditions for a fair and valid consultation were by no means satisfied and that it was futile for AMFOGE to try to accomplish its mission.38 His severe overall judgment of AMFOGE I and II deserves to be quoted in full:

  • 39 G. T. Mavrogordatos, 1981, p. 187.

“In retrospect, the two Allied Missions can only be described as elaborate charades, certainly a sad conclusion given the extraordinary human resources, technical skill, goodwill, and plain honesty particularly, but not exclusively, on the American side. Ample evidence of the bias and manifold inadequacies that the left denounced at that time is to be found in the missions’ own documents, and in their own files. No matter how honorable the intentions of their members may have been, it should have been obvious to all from the very beginning […], that the two missions could not possibly have subverted and contradicted the established policy of their governments. No matter what the actual findings might be, it was therefore inconceivable that they would have challenged the validity of the election or the plebiscite.”39

21Speaking more specifically of AMFOGE I and its restricted mandate (“[it] was not supposed to concern itself with public order, nor was it empowered to enforce fairness”), Mavrogordatos also wrote that:

  • 40 G. T. Mavrogordatos, 1981, p. 192.

“If it was theoretically intended to provide information upon which a subsequent decision could be based, it was apparent that by that time the two governments [of Britain and the United States] had implicitly abandoned any other option, except that of accepting the electoral outcome in any case. The mission was thus eventually only expected to furnish a reasonably convincing ex post facto justification for holding the election on March 31.”40

  • 41 G. T. Mavrogordatos, 1981, p. 187.
  • 42 G. T. Mavrogordatos, 1981, p. 379. In his 1965 survey of Greek political forces, the French politi (...)
  • 43 H. Richter, 1986, pp. 444-445.

22Being a creation of the wartime Allies (plus France and minus the Soviet Union), AMFOGE was, from the outset, deeply entangled with the – not necessarily identical – policies of the British and American governments and its leading exponents themselves held quite definite views about Greek politics. According to Mavrogordatos, the Mission soon became “an active exponent” of “the British and American policy against the postponement of the election, regardless of the conditions prevailing in Greece” by claiming that it could not maintain its presence far beyond the scheduled date.41 In an extended endnote, he even spoke of the “total subordination of the two missions to Anglo-American policy”.42 Richter, whose later book provided a careful and detailed narrative of Britain’s intervention in Greece in 1945 and 1946, also placed heavy emphasis on the Mission’s political bias. As proof of this, he mentioned former SOE commander Woodhouse’s appointment to AMFOGE, the content of the documents prepared in view of the Mission members’ “indoctrination”, but also the massive presence of officers among observers and that of upper-class Greeks among their interpreters, both groups being likely unsympathetic to the Left and to the plight of its supporters.43

  • 44 H. Richter, 1986.
  • 45 H. Richter, 1986, p. 447.
  • 46 G. T. Mavrogordatos, 1981, p. 378. Already in 1948, American political scientist Earl Latham had i (...)

23Concerning the Mission’s statistical findings, Richter wrote that “the data on which AMFOGE’s calculations were to be based were either wholly lacking or were grounded in arbitrary assessments”.44 He successively challenged and analyzed AMFOGE’s estimates of the total Greek population, of the number of people entitled to vote and of those who abstained. However, since his analysis was based on the Mission’s published report, he did not closely examine the specifically statistical aspects of the surveys and the nature of the choices made by the experts. Typical of this approach is his description of the Mission’s estimated abstention rate, which he described as being the result of a “sample opinion poll”, which the ambient political conditions rendered unreliable.45 Mavrogordatos’ analysis also focused on this estimate of the abstention rate, which was met with disbelief and sarcasm by the leftist parties that had advocated abstention and by many foreign observers. According to Mavrogordatos, even though the abstention figures were “reportedly derived from a sample survey of voters”, the political conditions in which the election had been held and the fact that “admission of political abstention involved grave dangers for the individual concerned” made them unreliable and “deal[t] a death blow to the pretenses of ‘scientific validity’ that the mission made a great deal of”.46 Here again, his remarks deserve to be quoted at length:

  • 47 E. Latham, 1948, p. 379.

“The arbitrary estimate of political abstention by AMFOGE I, concealed behind a transparent cloak of ‘scientific sampling’, amply illustrates the two major sources of bias: political expediency, and political naïveté. In this case, the former required that republican abstention and its consequences be scaled down by all means. […] On the other hand, political naïveté allowed a choice of means and arguments that could hardly persuade Greek and foreign public opinion, while they discredited the mission. […] Naturally, not all participants were equally guilty of both sins: political expediency was foremost in the minds of leading participants, while naïveté and inexperience plagued the mass of observers, mostly junior officers.”47

24Indeed, one may well wonder if the statistical experts we have encountered so far belonged to the group of “leading participants”, in which case they were consciously engaged in manipulation, or to that of the naïve “observers”, as Daly’s and Kempthorne’s above-mentioned comments may suggest.

  • 48 E. Latham, 1948, p. 379.

25Besides, while giving AMFOGE I and II considerably more attention than any other author, Mavrogordatos and Richter did not really describe or discuss the concrete nature of the work undertaken by the statistical experts during their stay. For instance, one of Mavrogordatos’ examples of the two Missions’ “total subordination […] to Anglo-American policy” concerned the fact that “AMFOGE I had recommended that ‘a new census should be taken and a complete revision of the registration lists made before the opinion of the Greek people is sought again on matters of national import’”.48 It is of course a fact, as he points out, that no such census could be conducted once it was decided that the plebiscite would be held soon after the election, and it is also a fact that the revision of the registration lists was in no way exhaustive. But this remark does not take into account the fact that, whatever its shortcomings, AMFOGE II applied a sampling method for their revision of the registration lists. This was the preferred method from the outset: not only did the Mission’s experts consider sampling to be more expedient than complete enumeration, but they also judged it far superior from a scientific point of view precisely because, contrary to the latter, it provided, in statistical parlance, a reliable estimate of error. The incomplete character of the revision was in this sense a strength rather than a weakness. Neither did Mavrogordatos make any mention whatsoever of the “miniature sample census” conducted by AMFOGE II, which provided an estimate of the total population against which participation could be assessed.

4. AMFOGE as an experiment and as a battlefield

26That the Mission’s presence and activities soon became instrumental in conferring legitimacy to the Greek election and thereby in condoning British and American policies on Greece’s future, is not open to dispute. But this means considering the AMFOGE episode first and foremost from the standpoint of politics, downgrading the scientific and technical dimensions to the status of “guesswork,” “speculation,” or “charade,” and interpreting the role of experts through the lens of manipulation or, at best, self-deception. But the interplay between politics on the one hand, and science and technical expertise on the other can, in many regards, be seen as more complex. It may also provide useful insight into the role of expertise in politics – for instance: what types of knowledge and know-how were mobilized to address issues and problems such as those raised by the Greek election and plebiscite? – as well as into the politics of expertise – for instance: what kinds of arguments allowed for the settlement of an issue and who was empowered to take part in these discussions? In the present case, statistics, and sampling as a specific kind of statistics, must be placed at the center of attention. In this section, however, we will simply attempt to give an overall idea of what the AMFOGE experts were doing by first summarizing the specific surveys that were conducted and the methods used, and then commenting on two episodes of a more political character, namely the “Neyman affair” and the interactions between the sampling experts and representatives of the EAM (National Liberation Front) coalition.

The AMFOGE I surveys

  • 49 C. Michalopoulou, 2014, p. 199 (Table 1).

27Six different surveys – or inquiries, as they were designated by those who conducted them – were conducted by AMFOGE I. Four took place before election day, one on election day, and another in the days that followed. It should be recalled that most of the observers chosen to conduct these surveys were military personnel in uniform (though unarmed and identified by the Mission’s insignia). The majority did not speak Greek and most interviews were conducted by Greek interpreters chosen by the Mission. We will briefly summarize the purpose of each survey and the nature of the questions asked. Michalopoulou’s 2014 paper provides a detailed description of the sample designs of each of survey.49

  • 50 “Report of the Planning and Sampling Staff, Central Office, AMFOGE”, April 9, 1946, “Statistical S (...)
  • 51 “Form A”, February 26, 1946 (Box 14).

28Inquiry A included a series of questions on the printing and posting of election lists and it covered approximately 50% of the country’s 3,000 registration offices. It was designed to provide “information about the registration procedure and the sentiment of party leaders” and to “bring to the notice of the Mission the conditions which might be expected to prevail on election day”.50 Some of these questions were addressed to the Mission’s observers, whose answers were necessarily based on their own observations and on information they could collect from registration officials; other questions were addressed to the representatives of political parties. For instance, observers were asked: “Is the electoral list recompiled (or entirely new for 1945 or 1946?)”; “Or, does it consist of old lists plus supplements?”; “What credentials did an individual require in order to obtain registration?”; “If lists were revised rather than recompiled, have dead person’s (sic) names been marked from the positive (engegramenci) lists?”; “If NOT, explain why not?”; etc. Party representatives, for their part, were asked questions such as: “Have you (the party representative) examined the registration lists of this precinct (or parish, or enoria)?”; “If NO, why didn’t you examine the lists?”; “Did you examine the lists to see if all members of your party were registered in your precinct?”; etc.51

29Inquiry B “was based upon a cross-section of the Greek voting population”:

  • 52 “Report of the Planning and Sampling Staff, Central Office, AMFOGE”, April 9, 1946, “Statistical S (...)

“The purpose of this survey was to estimate (i) the total number of eligible voters in Greece, (ii) the ratio of that number to the total population, (iii) the proportion of eligible voters who were not registered.”52

  • 53 “Household interview regarding registration. Form B”, February 28, 1946 (Box 14).

30It consisted of interviews with persons inhabiting the same household and concerned individual registration. Each adult male (women were not entitled to vote) was asked to specify his occupation; the date he had moved to his present location and the place from where he had moved; and if he had registered for the election. If he had not registered, he was asked why not; his registry status was then verified and he was asked if he intended to vote or not, etc. Respondents were also asked if they thought that “the voting in your parish (precinct) (enoria) will be conducted properly or improperly” or that “anyone in this immediate neighbourhood may be afraid to go and cast his ballot for his party”.53 Survey B was based on a random sample of 2,000 households, representing large cities, small towns and villages throughout the country.

  • 54 “Instructions for Form C-2. Verification of residence of registered voters”, February 27, 1946, p. (...)
  • 55 “Form C-2”, appended to the preceding document (Box 14).

31Inquiry C sought to assess the validity of the electoral lists, i.e. “to estimate whether and, if so, to what extent they were ‘padded’”.54 Covering the same areas as B, it drew a systematic sample of names from the electoral lists and asked neighbors a series of questions to establish the validity of each entry (“1. Do you know the registrant? 2. What is his occupation? 3. How old is he? 4. How long has he lived at that address?” etc.).55 Names present on the list as well as names that had been crossed out during the revision by the authorities were included in the sample.

  • 56 “Form D-1” (Box 14).

32Inquiry D, carried out on election day, was concerned with voting irregularities. Observers had to answer a series of questions on the organization and progress of the vote as well as on electoral results in the polling place concerned. It included, for instance, questions about the physical layout of the polling place (“Was there a place where the voter could withdraw and put his ballot in the envelope without being seen by anybody?”), the presence of party representatives, their satisfaction “with the way the balloting was going on”, or the presence of “drunken men” or “armed individuals […] except from the regular police”.56 Two distinct samples were created: first, a systematic sample of polling places (105 out of a total of 3,398), where observers were present all day long; and a second sample of 120 routes that enabled observers to move from one place to another during the day, so as to cover a maximum number of polling places and also to be seen by a maximum number of voters. No fewer than 708 polling places were covered in this manner. Overall, nearly 25% of polling places were observed all day long or during a part of the day.

33Inquiry F, conducted on the immediate aftermath, sought to assess electoral abstention and its motives. Since EAM, the coalition of Left parties, had finally called upon voters to boycott the election and about 1,107,000 people had voted out of an electorate estimated by AMFOGE at 1,850,000, it could be said that the way the election was judged depended to a large extent on this question. Inquiry F comprised a systematic sample of 1,345 names drawn from the electoral lists of 113 polling places. The chosen names were matched against electoral committee reports indicating who had voted and who had not. The persons concerned were then traced and asked if they had indeed voted or not, and why.

  • 57 Inquiry E is also sometimes mentioned. This one was different, since it covered a variety of inves (...)

34From all this,57 the Mission drew a series of carefully weighted conclusions that were quantitative in nature yet expressed in qualitative terms:

  • 58 AMFOGE I Report, pp. 26-27. Our emphasis. A full report, to be “published in volume form”, was ann (...)

“The registration lists in large areas contained irregularities but there was no significant amount of illegal voting. Intimidation existed, in some degree, from both extremes and was even on occasion given countenance by members of the gendarmerie but it was not extensive enough to affect seriously the election. The practice of deliberate abstention did not reach large proportions.”58

35But its report also presented the “main findings of the planning and sampling staff” in an appendix that also included a summary of the survey results. As mentioned earlier, these conclusions were immediately challenged. As will be obvious later, when we discuss Neyman’s and then EAM’s criticisms of the Mission’s work, conclusions drawn from these sampling surveys were based on a number of fundamental, yet controversial, assumptions, and, in certain cases, on pure guesswork.

The AMFOGE II surveys

  • 59 The results of these inquiries were compiled in two tables featured in “Report of the Allied Missi (...)

36AMFOGE II’s purpose was to determine if the lists used for the March election could also be used for the September plebiscite on the monarchy. To that end, two kinds of samples were devised. The first comprised names included in the electoral list. All these names were checked, so as to estimate the number of listed names that should not have been present (dead persons, fictitious names, duplicates, etc.).59 The second sample comprised 3,052 households to which a series of questions were asked. It was described as a “miniature sample census” that made it possible to: (a) estimate the number of names that should have been on the list but were absent; (b) verify some of the AMFOGE I survey results; (c) to provide, given the absence of any recent countrywide census, an overall picture of the Greek population. To this end, alongside the questions specifically designed for the Mission’s purpose, others dealing with characteristics such as sex, age, literacy, and occupation were included with a view to conducting a more scientific survey.

  • 60 Michalopoulou has described the population sample as “exemplarily designed, conducted, and present (...)
  • 61 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 359.
  • 62 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, pp. 363-364. This method was prese (...)
  • 63 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 360.
  • 64 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, pp. 360-367.

37As mentioned earlier, the miniature sample census was by far the most elaborate survey undertaken by the Mission’s statisticians and it is worth describing the procedures it followed in some detail.60 The sample was planned in such a way that “every household had a known probability of being drawn”, that any “bias of selection” was eliminated, that “the standard errors (were) computable”.61 No substitution was allowed, notwithstanding the difficulty of reaching a selected household. Material constraints had to be taken into account, namely the distance that could be covered in a single day by interviewers, the time devoted to sample selection, the speed of interviews, and the time needed to find and examine the electoral rolls. It was estimated that 65 teams in the field for three weeks would be able to interview 15 households per day. On the basis of the AMFOGE I surveys, it was decided that the sample would be selected from “about 200 cities, towns and villages drawn from strata suitably created on the basis of geographical divisions and the population of 1940”. The overall sampling ratio was established as 1 in 500, in line with the principle of optimum allocation, as developed by Neyman in 1934.62 Villages, towns or cities, along with the satellite villages and any surrounding households constituted the “primary sampling unit”. Information on these units came from the 1940 census and from geographical maps provided by the Greek government.63 From these primary units, 200 subsampling zones were selected, and then a group of 15 households within each one. Given the hypothesis of a total population of 7.5 million and of a household size of 5 persons on average, a sample of 3,000 (200 × 15) households represented 1/500 of all households. Final selection of the households to be interviewed depended on the information available: if a detailed and up-to-date list of households living in a village or parish existed, systematic sampling could be applied; if only a map representing dwellings was available, these were numbered and then systematically sampled; if the map only showed housing blocks, the procedure depended on information that could be gathered about the number of households per block. The final result was a sample of 3,052 households, of which no fewer than 3,042 were actually reached.64

  • 65 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 367.
  • 66 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 369.

38Two estimates of the total population of Greece were then compared: one obtained by simply multiplying the sample total by 500, the other by using information collected for each sampling area in the course of the 1940 census and applying a regression method.65 The former method gave a total of 6,975,000 with a standard error of 3.7%, while the latter produced a total of 7,257,000, with a standard error of 2.1%.66

The Neyman affair

39By 1946, Jerzy Neyman was a major figure in the world of mathematical statistics and his presence among AMFOGE’s scientific staff reinforced the Mission’s credibility. However, he was sent home after only six weeks in the field, after becoming involved in two specific controversies.

a. Dissent

  • 67 “Letter from R. Jessen to F. Kohler”, February 15, 1946 (Box 4).
  • 68 “Letter from J. Neyman to F. Kohler”, February 15, 1946 (Box 4). Neyman’s emphasis.

40To check the validity of electoral rolls, each expert was assigned a number of districts: Jessen and Kempthorne were in charge of Athens, Neyman of Salonika, and so on. When Jessen (who was, it should be remembered, chief of the sampling staff) submitted a report in which he stated that “any claim of extensive interference with registration in Athens is in contradiction to the facts as observed by eliciting information from households”, Neyman immediately wrote to Foy Kohler, the Mission’s secretary general,67 stating that such a conclusion was open to five criticisms: (1) the survey “covered a sample of residents”, even though a significant number of refugees were present in Athens; (2) in some cases, “the samplers got the impression that persons […] were reluctant to enumerate all the people residing at a given address”; (3) “complaints of extensive interference with registration usually mention intimidation by police”; (4) a number of people, who were “more or less in hiding […] were not covered by the sampling survey”; and (5) the survey results are “perfectly consistent” with the hypothesis of large-scale intimidation, leading to non-registration. No conclusion should therefore be drawn from the Athens survey.68

  • 69 These were: Ioannis (John) Mikhalopoulos, director of the General Statistical Service of Greece (G (...)
  • 70 “Discussion with 1. Mr. John Mikhalopoulos 2. Mr. Demetrios Calitsounskis 3. Mr. Angelos Angelopou (...)

41A few days earlier, on February 11, Neyman had held a meeting with three Greek statisticians, in the presence of Kohler and Jessen.69 Although Kohler stated from the outset that the discussion was to be of a “technical” and “professional” nature,70 the invitation sent by Neyman to the Greek statisticians on February 9 included three questions that had an indisputable political dimension:

  • 71 “Letter from J. Neyman to J. Mikhalopoulos”, February 9, 1946 (Box 4).

“1. How could one account for the, apparently, very high proportion of the population which is entitled to vote? 2. What is the economic background of the differences in the political outlook of the various groups of the population? 3. What are the economic and political differences of different geographic regions of Greece?”71

  • 72 Michalopoulou observes that Neyman’s estimate, which was not based on any survey, was finally not (...)
  • 73 “Discussion with 1. Mr. John Mikhalopoulos 2. Mr. Demetrios Calitsounskis 3. Mr. Angelos Angelopou (...)

42The minutes of the meeting report a discussion that focused on the total population size and on the number of males over 21 (who were entitled to vote). Estimates of the latter ranged from a high of 2,100,000 (proposed by Neyman, but which he admitted was probably too high) to a low of 1,600,000.72 Interestingly, the number chosen by the Mission in its final report (1,850,000) was that proposed by the Greek statisticians. It was not based on any sampling procedure but on a prior assumption, namely that the total population size in 1946 was basically identical to that of 1940. The issues of possible inaccuracies and of refugees who were not registered or registered at their former address were discussed. One Greek participant kept “repeating that normal political conditions are essential for conducting an election and cannot be ignored”.73 The criticisms raised in Neyman’s February 15 letter may reasonably be interpreted as an echo of that discussion.

b. Faux pas

  • 74 Georges Tenekides (1910-1990) was a professor of international law in Athens from 1938 to 1941, an (...)
  • 75 “Letter from J. Neyman to Ambassador H. F. Grady”, March 11, 1946 (Box 4).
  • 76 “The Enemies of democracy” Nea Alithia (Salonica), translation, February 27, 1946 (Box 4).

43The events that led directly to Neyman’s dismissal occurred a few days later. From February 20 to 24, Neyman was sent to conduct the survey in Salonika. There, he met Professor Georges Tenekides.74 On his invitation, Neyman went to a village (Khortiatis) where, according to the former, the electoral rolls deserved investigation. Neyman reported that he had found “practically no irregularities” in Salonika, a “few more” in Gefiras and a “substantial number” in Khortiatis.75 On February 27, a right-wing newspaper in Salonika denounced Tenekides for being a Communist and Neyman for being a dupe whose report had been dictated to him.76 When Tenekides learned of the article’s existence, he asked Neyman to reply publicly.

  • 77 “Memorandum. Subject: Termination of the Services of Dr. J. Neyman”, March 20, 1946 (Box 4).
  • 78 “Letter from J. Neyman to the Rector of the University of Salonica”, March 13, 1946 (Box 4).
  • 79 “An Honest Observer”, Eleftherie, translation, March 29, 1946 (Box 4).

44When Neyman discussed the issue with officials, he was advised to drop the matter, since it would embroil the Mission in political controversy. As he continued to press for a public response, he was “informed that if he persisted in that line of action, his services would have to be terminated in the interests of the Mission”.77 Despite this warning, Neyman decide to write to the Rector of the University of Salonika and point out the errors he had found in the article.78 When this became known, he was reproached for infringing the Mission’s strict neutrality policy and for having disobeyed an express order from Ambassador Grady. He was immediately sent back to the United States. On March 29, EAM’s daily paper presented Neyman’s departure as a resignation, motivated by the fact that he had allegedly discovered numerous falsifications that his colleagues did not wish to investigate.79

EAM’s viewpoint

  • 80 EAM (Greek National Liberation Front) was the main resistance organization during the Nazi occupat (...)
  • 81 “To AMFOGE” (otherwise untitled document), April 7, 1946 (Box 6).

45Two documents, respectively dated December 9, 1945 and April 7, 1946, present the viewpoint of EAM, the Communist-led coalition, on the Mission’s mandate and its report.80 Both report in detail the alleged fraud and violence that preceded or accompanied the elections. The second one, for instance, contained a set of “statistics covering [the] criminal action” of “armed royalist organizations [and] the security and military forces [that] cooperated with them”.81 It mentioned notably:

“1. 1,252 verified murders of democrats

2. 6,950 verified cases of wounded democrats

3. 38,400 verified molestations and torturing

4. 79,169 arrests

5. 185 criminal assaults on women

6. 2 organized coup d’état

7. 895 clipping the hair, pillorying

8. 20,000 political refugees who took refuge in Athens

9. 6,597 robberies, destruction of houses, etc.

10. 7,576 armed gangsters

11. 602 raids and blowing up of newspapers offices and printing works

  • 82 “To AMFOGE” (otherwise untitled document), April 7, 1946 (Box 6).

12. 185 monarchist bands in operation”.82

46Appended to the document were four enclosures, which, according to EAM, provided evidence of large-scale fraud. But both documents also contained a rather elaborate statistical discussion.

a. The number of legal voters

  • 83 V. G. Valaoras, 1946, p. 215.
  • 84 “Of the C.C. of the Political Parties’ Coalition of EAM to the Allied Observers for the elections (...)

47Nearly everyone conceded that the total number of officially registered voters as of December 1945 (2,211,791, according to the government) was much higher than the (unknown) total number of voters who were entitled to vote. In a section significantly titled “Figures speak by themselves”, EAM’s December document recalled that in 1936, the total number of persons who had exercised their right to vote was only 1,278,000. Given an abstention rate of 20% – described as the “usual percentage” –, this meant that the number of people entitled to vote totaled around 1,500,000. The document then introduced an obvious error, stating that Greece’s population at that time stood at 6,127,595: this meant for EAM that the proportion of voters in the population was close to 25%, but no source was given here for a number that was lower than the result (6,204,684) arrived at by the 1928 census. Since then, the 1940 census had counted a total population of 7,344,642, but the war years had caused a net loss that had escaped measurement, since, as a contemporary analyst noted, “civil registration of vital statistics was the first administrative procedure to be paralyzed immediately after the occupation”.83 EAM estimated that, in 1946, a maximum of 6,500,000 was likely. If the 25% proportion was constant, this meant that the number of lawful voters was around 1,600,000. EAM concluded from this that the difference between this number and the one published by the government – more than 2.2 million – represented the extent of the fraud that was planned.84 A correct estimate of the number of potential voters was indeed one of the challenges to be addressed by the Mission’s statistical experts.

  • 85 “Statistical supplement to Report of the Planning and Sampling Staff”, Central Office, AMFOGE, Apr (...)

48During the above-mentioned meeting between the Mission’s experts and the three Greek statisticians, estimates for this parameter varied between 1,808,000 and 1,993,000. From inquiry C, the Mission’s statisticians concluded that 13% of registered names were invalid, which led to a downward revision of the government estimate (2.2 million) to 1,920,000. Inquiry B, on the other hand, led to the conclusion that men over 21 represented 27% of the total population and that 93% of them were registered. But in order to estimate the total number of valid voters, total population size must also be estimated. According to the United Nations Relief and Rehabilitation Administration (UNRRA), the distribution of food rations was based on an estimated total population of 7,500,000; but this was an assumption. Censuses conducted in 1928 and 1940 had counted 6,204,684 and 7,334,642 persons, respectively. Assuming that the 1928-1940 mean population growth rate remained unchanged over the 1940-1946 period, a total of 7,914,621 is obtained. However, taking into account excess mortality during the war years, the Mission’s experts revised this number downward to 7,500,000. Applying the above-mentioned proportions, the number of people entitled to vote came to 1,989,000, of whom 1,850,000 were duly registered.85 The major difference between the Mission’s and EAM’s respective estimates thus resided in their different estimates of mortality due to war and famine: a net loss of 800,000 according to EAM, versus a loss of 400,000 according to the Mission, offset, however, by 600,000 births during that period. Both estimates were based on assumptions that could not be verified at the time.

b. The extent of abstention

  • 86 AMFOGE I Report, p. 26.
  • 87 AMFOGE I Report, p. 32. This “median value” was justified on the basis that half of those who clai (...)
  • 88 “To AMFOGE” (letter from Central Committee of the Political Parties of EAM), April 7, 1946 (Box 6) (...)

49In the document published by EAM to state its contestation of the election results and its disagreement with the Mission’s favorable conclusions, a number of topics are mentioned, namely acts of violence that allegedly occurred during the weeks prior to the election day, and the state of the electoral rolls. But a new problem now became the focus of attention, namely the extent of abstention. Given the estimated number of potential voters (1,850,000) and the numbers of persons who had actually cast their vote (around 1,107,000), abstention totaled 40.3%. On the basis of inquiry F, AMFOGE estimated the rate of politically-motivated abstention via the numbers of respondents who claimed to have “participated in the boycott” (9.3%). This was the number given by AMFOGE in its press release and report.86 In fact, since 17.9% of respondents were recorded as not having voted because they were “not in the neighbourhood on election day”, the report mentioned that abstention may actually have reached 20%, but also put forward 15% as a “median” value.87 But since the Greek government had earlier argued that up to 2.2 million people were potentially entitled to vote, this meant that abstention may have reached 50%. According to EAM, whose member parties had called for a boycott, all those who did not vote were politically motivated to abstain, and all efforts to tamper with these numbers – for instance, by subtracting from the number of potential voters all the dead people whose names were still on the rolls – were designed to lessen the political significance of abstention and were thus politically biased. To a degree, EAM’s argument was a win-win: if the government’s estimate of potential voters was correct, then abstention was huge; if the government’s estimate was wrong, then the electoral rolls, and therefore the elections, were a vast fraud. EAM added that if there had been zero population growth from 1940 to 1946 – it argued that Greece had in fact experienced a net loss – then the share of people who had effectively voted (by contrast with people entitled to vote) should have amounted to 18%. Given that the number of people who voted was higher that the number predicted by the zero-growth hypothesis, this again proved the existence of massive fraud. EAM thus concluded that 250,000 potential voters were not registered – something the Mission’s experts had not noticed in their surveys. The abstention rate thus stood at 50%, “an incontestable majority of the entire number of electors”, according to EAM.88

Conclusion

50The present paper has sought to provide readers with a picture of AMFOGE’s statistical work and of the inescapable political entanglements it ran into. While it may be presumptuous to draw detailed conclusions from such a brief overview, a few observations may be in order.

  • 89 G. Midgley, 2000, p. 125.

51Paying closer attention to the properly scientific dimension of AMFOGE than historians of the period have done may enable us to reinterpret the political category of intervention (as in British or American “intervention in Greece”) through the lens of epistemological and scientific inquiry. It is, for instance, common nowadays to argue that “observation” – as in the “O” of the acronym AMFOGE – always constitutes some form of “intervention”. Basically, this refers to the acknowledgement that any scientific observation or experiment must be organized or controlled and, in that sense, represents intervention on the part of the observer or experimenter.89 Closely examining how observations were organized, controlled, described and then interpreted is one of the aims of our ongoing research on AMFOGE. As such, this partly overlaps with the statistical experts’ own vision of their work: statistical inquiry is, by definition, organized and controlled observation. But the notion of observation as intervention also raises the question of the effect of that intervention on the observed results: in the case of AMFOGE, it has been pointed up from the outset that the manner in which the observation was conducted on the ground – for instance the use of uniformed officers as observers – might have biased the observation results and thus the overall conclusions of the mission.

  • 90 R. J. Jessen, 1978, p. 453.
  • 91 These scales and instruments are, of course, themselves the product of a complicated history. See (...)
  • 92 A. Desrosières, 2008, p. 10.
  • 93 A. Desrosières, 2008, p. 11.

52This theme can be further explored by making use of Desrosières’ distinction between “quantification” and “measurement.” The latter, which, for instance, Jessen defines as “the classifying, counting, or measuring needed on each OU (observation unit) to obtain the basic data required of the survey” or, alternatively, “the activities of obtaining the final observations,”90 belongs to what Desrosières refers to as “realist metrology” applicable to the natural sciences or to applied activities such as engineering. One can indeed “measure” the distance between two points, the height of a building or the pressure of a gas because universally recognized scales of measurement and calibrated instruments already exist.91 To counter the problematic epistemological assumptions associated with the extension of this measurement concept to the domain of social and political sciences, Desrosières proposed the more suitable concept of quantification, which he defined as the process whereby “things that were up to now conveyed through the use of words come to exist and be expressed under the guise of numbers”.92 By contrast with measurement, quantification encompasses all the prior operations through which “a series of conventions of equivalence” emerge, implying “comparisons, negotiations, compromises, translations, inscriptions, encodings, codified and replicable procedures, and computations leading to numerical expression.” Measurement comes into play only as a last stage and as the implementation of all these conventions.93

  • 94 To take one admittedly extreme example of such violence, an attack on Litokhoro – for which the Co (...)

53The work done by the AMFOGE experts in Greece can therefore be seen as a complex process of quantification: from this perspective, the focus can shift from the self-justifying rhetoric of experts, largely derived from the alleged robustness of recent developments in probability-based sampling methods, to the whole series of qualitative judgments that need be made before proper measurement can start. One telling example is the “guesstimation” by AMFOGE experts of the extent of abstention, where the numbers put forward were based on considerations that failed to take into account the problems raised first by Neyman and later by scholars like Latham (see note 45), notably the use of military personnel as interviewers, the abundant recourse to interpreters and the effect of ongoing violence on the reliability of the respondents’ replies.94 As archival evidence shows, discussions and, in the absence of hard data, negotiations about the total number of living Greeks in 1946 – a crucial variable, as we have seen, for further calculations and, ultimately, decision-making about the course of action to be taken – also could not be conducted without references to the conditions in which the electoral laws were applied. From this perspective, assumptions, guesses, rules of thumb, practical advice, bargaining and compromises were as much part of AMFOGE as a scientific experiment as were the notions of optimum allocation or of regression.

Haut de page

Bibliographie

Primary sources

Archives

National Archives, Department of State, Record Group 59, Records Relating to the Allied Missions for the Observation of Greek Elections (AMFOGE I and AMFOGE II) (Lot File No. M-72), HMS Entry Number A1 1376.

Printed sources (Government documents)

“Agreement between the Greek Government and EAM”, Foreign Relations of the United States: Diplomatic papers, The Near East and Africa, 1945, Volume VIII, Document 63.

“The Soviet Foreign Commissar (Molotov) to the Secretary of State”, Foreign Relations of the United States: Diplomatic papers, The Conference of Berlin (Potsdam Conference), 1945, Volume II, Document 1063.

“Revision of Greek Electoral Lists”, Department of State Bulletin, vol. XV, no. 374, September 1, 1946, p. 424.

Report of the Allied Mission to Observe the Greek Elections, Washington, United States Government Printing Service, 1946.

“Allied Mission to Remain in Greece for Informal Observation”, Department of State Bulletin, vol. XV, no. 374, September 1, 1946.

“Report by Messrs. Richard T. Windle and Leland Morris, Chiefs of the Allied Mission to Observe Greek Elections. Report on the Observation of the Greek Plebiscite”, September 7, 1946. Foreign Relations of the United States, 1946, vol. VII, p. 206.

Secondary sources

Adler, Ken, The Measure of All Things: The Seven-Year Odyssey and Hidden Error That Transformed the World, New York, The Free Press, 2002.

Campbell, John Kennedy, “The Greek Civil War”, in Evans Luard (ed.), The International Regulation of Civil Wars, London, Thames and Hudson, 1972, pp. 37-64.

Chiang, Chih-Ang, “Using the Pao as the Primary Sampling Unit. Some Notes and Reflections on the Possibilities of a Census of China by Sampling”, Population Studies, vol. 2, no. 4, March 1949, pp. 444-453.

Clogg, Richard, Parties and Elections in Greece: The Search for Legitimacy, Durham (NC), Duke University Press, 1987.

Close, David H., The Origins of the Greek Civil War, London and New York, Longman, 1995.

Daly, Joseph F., “Oral History – Joseph F. Daly” [online], 9 p.
URL: https://www.census.gov/history/pdf/ Daly_Oral_History.pdf.

Deming, W. Edwards, Some Theory of Sampling, New York, Wiley, 1950.

Desrosières, Alain, “Three Studies on the History of Sampling Surveys: Norway, Russia-USSR, United States”, Science in Context, vol. 15, no. 3, 2002, pp. 377-383.

Desrosières, Alain, Pour une sociologie historique de la quantification. L’argument statistique I, Paris, MinesTech, 2008.

Folks, J. Leroy, “A Conversation with Oscar Kempthorne”, Statistical Science, vol. 10, no. 4, November 1995, pp. 321-366.

Frankel, Lester R., “Some Theory of Sampling, by William Edwards Deming”, Journal of the American Statistical Association, vol. 46, no. 253, March 1951, pp. 127-129.

Henry, Louis, “Résultats d’un sondage sur la population de la Grèce en 1946”, Population, vol. 2, no. 4, 1947, pp. 785-789.

Jessen, Raymond J., Statistical Survey Techniques, New York, Wiley, 1978.

Jessen, Raymond J., Blythe, Richard H., Kempthorne, Oscar & Deming, W. Edwards, “On a Population Sample from Greece”, Journal of the American Statistical Association, vol. 42, no. 239, 1947, pp. 359.

Jessen, Raymond J., Deming, W. Edwards, Daly, Joseph F. & Kempthorne, Oscar, “Observations on the 1946 Elections in Greece”, American Sociological Review, vol. 14, no. 1, February 1949, pp. 11-16.

Karabell, Zachary, Architects of Intervention: The United States, the Third World and the Cold War, 1946-1962, Baton Rouge, Louisiana State University Press, 1999.

Kruskal, William & Mosteller, Frederick, “Ideas of Representative Sampling”, in Donald W. Fiske (ed.), Problems with Language Imprecision, San Francisco, Jossey-Bass Publishers, 1981.

Latham, Earl, “Political Statistics”, The Journal of Politics, vol. 10, no. 4, Nov. 1948, pp. 636-658.

Mavrogordatos, George T., “The 1946 Election and Plebiscite: Prelude to Civil War”, in John O. Iatrides (ed.), Greece in the 1940s: A Nation in Crisis, Hanover and London, University Press of New England, 1981, pp. 181-194.

Meynaud, Jean, Les forces politiques en Grèce, Montreal, Presses de l’Université de Montréal, 1960.

Michalopoulou, Catherine, “A Unique Collaboration: Prominent Statisticians’ Survey Work in Greece in 1946”, The American Statistician, vol. 68, no. 3, 2014, pp. 196-203.

Midgley, Gerard, Systemic Intervention: Philosophy, Methodology, and Practice, New York, Kluwer Academic Publishers, 2000.

Neyman, Jerzy, “On the Two Different Aspects of the Representative Method: The Method of Stratified Sampling and the Method of Purposive Selection”, Journal of the Royal Statistical Society, vol. 97, no. 4, 1934, pp. 557-625.

National Statistical Service of Greece, Résultats du recensement de la population effectué le 7 avril 1951, vol. I, Athens, 1961.

Politakis, George, The Post-War Reconstruction of Greece: A History of Economic Stabilization and Development, 1944-1952, London, Palgrave, 2018.

Porter, Theodore M., Trust in Numbers: The Pursuit of Objectivity in Science and Public Life, Princeton, Princeton University Press, 1995.

Prévost, Jean-Guy & Beaud, Jean-Pierre, Statistics, Public Debate and the State, 1800-1945: A Social, Political and Intellectual History of Numbers, London, Pickering and Chatto, 2012.

Reid, Constance, Neyman, New York, Springer-Verlag, 1998.

Richter, Heinz, British Intervention in Greece. From Varkiza to the Civil War, February 1945 to August 1946, London, Merlin Press, 1986.

Seng, You Po, “Practical Problems in Sampling for Social and Demographic Inquiries in Undeveloped Countries”, Population Studies, vol. 3, no. 2, September 1949, pp. 170-191.

Sfikas, Thanasis D., The British Labour Government and the Greek Civil War 1945-1949: The Imperialism of “Non-Intervention”, Keele, Keele University Press, 1994.

Valaoras, Vasilios G., “Some Effects of Famine on the Population in Greece”, The Milbank Memorial Fund Quarterly, vol. 24, no. 3, July 1946, pp. 215-234.

Woodhouse, Christopher M., Apple of Discord: A Survey of Recent Greek Politics in their International Setting, London, Hutchison & Co., 1948.

Woodhouse, Christopher M., The Struggle for Greece 1941-1949, London, Hart-Davis, MacGibbon, 1976.

Woodhouse, Christopher M., Something Ventured, London, Granada, 1982.

Haut de page

Notes

1 Documents relating to AMFOGE are filed under: National Archives, Department of State, Record Group 59, Records Relating to the Allied Missions for the Observation of Greek Elections (AMFOGE I and AMFOGE II) (Lot File No. M-72), HMS Entry Number A1 1376. This archive consists of 18 boxes, containing a large series of folders filled with memoranda, minutes, incident reports, drafts, correspondence, directives, forms, lists of personnel, press releases, pamphlets, etc. An instrument describing the records in the boxes is provided with Box 1. We will refer to all material from this source by box number. Another relevant data source is the W. Edwards Deming archive at the Library of Congress. Material on AMFOGE is contained in three folders in Box 23. Part of this material can also be found in the AMFOGE archive, but it also includes some correspondence to Deming, and some detailed sampling plans. We have not yet been able to consult Greek sources and we are conscious that our knowledge of the views of AMFOGE’s Greek critics (see section 4) is based solely on the translation of material selected by AMFOGE.

2 “Agreement between the Greek Government and EAM”, Foreign Relations of the United States: Diplomatic papers, The Near East and Africa, 1945, Volume VIII, Document 63.

3 “The Soviet Foreign Commissar (Molotov) to the Secretary of State”, Foreign Relations of the United States: Diplomatic papers, The Conference of Berlin (Potsdam Conference), 1945, Volume II, Document 1063.

4 Report of the Allied Mission to Observe the Greek Elections, Washington, United States Government Printing Service, 1946, p. iii. Hereafter designated as AMFOGE I Report.

5 Z. Karabell, 1999, p. 17.

6 “A Chapter in Quantitative Political Science. Observations on the Elections in Greece”, draft paper by Raymond J. Jessen, Joseph F. Daly, and W. Edwards Deming, dated 26 July 1946, p. 1 (Box 9).

7 AMFOGE I Report, pp. 26-27.

8 “The Mission is satisfied that the revision and recompilations of the electoral lists as observed by it attain a degree of fairness and accuracy which justifies their use in seeking the opinion of the Greek people in matters of national import.” “Revision of Greek Electoral Lists”, Department of State Bulletin, vol. XV, no. 374, September 1, 1946, p. 424.

9 “Report by Messrs. Richard T. Windle and Leland Morris, Chiefs of the Allied Mission to Observe Greek Elections. Report on the Observation of the Greek Plebiscite”, September 7, 1946. Foreign Relations of the United States, 1946, vol. VII, p. 206.

10 “Supplement to Report on the Observation of the Greek Plebiscite”, September 13, 1946, quoted in G. T. Mavrogordatos, 1981, p. 376.

11 “Allied Mission to Remain in Greece for Informal Observation”, Department of State Bulletin, vol. XV, no. 374, September 1, 1946, p. 425.

12 “Miniature sample census” was an expression used by Jessen and his co-authors in “Observations”, p. 12. Two occurrences can also be found in Some Theory of Sampling, where the United States Monthly Report on Labor Force is described as a “miniature monthly sample census” (W. E. Deming, 1950, pp. 158 and 269). Of course, the idea of a “sample census” may seem a contradiction in terms, and viewing a sample as a “miniature” of the population is “rarely appropriate and achievable in statistics” (W. Kruskal & F. Mosteller, 1981, p. 9). Unsurprisingly, the idea has not caught on.

13 See note 6.

14 On the rise of sampling as seen from a socio-historical perspective, see A. Desrosières, 2002 (special issue on survey sampling); J.-G. Prévost & J.-P. Beaud, 2012, ch. 8: “Politics of the Sampling Revolution”, pp. 153-172.

15 T. M. Porter, 1995, p. ix.

16 T. M. Porter, 1995, p. 8.

17 L. R. Frankel, 1951, p. 129.

18 R. J. Jessen, W. E. Deming, J. F. Daly & O. Kempthorne, 1949, pp. 12-13.

19 R. J. Jessen, W. E. Deming, J. F. Daly & O. Kempthorne, 1949, p. 16.

20 R. J. Jessen, W. E. Deming, J. F. Daly & O. Kempthorne, 1949, p. 13.

21 In a letter to William Baxter of the State Department in which he sought permission to publish the paper, Deming wrote for instance: “You will observe that no mention is made of the election results or the condition of the registers.” “Letter from W. Edwards Deming to William Baxter”, dated November 6, 1946 (Box 10).

22 C.-A. Chiang, 1949; Y. P. Seng, 1949. The first scholarly paper quoting “On a Population Sample” (and summarizing its method and findings) was by French demographer Louis Henry: L. Henry, 1947. Interestingly, the AMFOGE team of experts had no French statisticians (they were all American or British), which testifies to the cultural context in which probability-based survey sampling emerged as well as to its novel status in 1946-47.

23 C. Michalopoulou, 2014, p. 201.

24 C. M. Woodhouse, 1948, p. 264.

25 J. F. Daly, no date, p. 2.

26 J. L. Folks, 1995, p. 324.

27 R. J. Jessen, 1978, p. 319.

28 “Here there were 60 British, 60 American, and 40 French investigators. Assignments of these national groups to voting places over Greece were randomized separately for each of the three groups, despite objections from the military organization that this would unduly burden the logistics problems in supplying the investigator teams (consisting of a military officer, his driver, and his interpreter, together with a jeep and trailer stocked to provide the team the means of support wherever it was assigned in Greece). The random assignments assured us, after examining the results, that nationality differences in this case were insignificant.” R. J. Jessen, 1978, p. 475.

29 R. J. Jessen, 1978, p. 496.

30 C. Reid, 1998, p. 203.

31 C. Reid, 1998, p. 208.

32 R. Clogg, 1987, p. 21.

33 C. M. Woodhouse, 1976, p. 175. In his 1982 memoirs, Something Ventured, Woodhouse makes only one reference to the fact that he was the British secretary general of AMFOGE, without passing any judgment on its work (id., 1982, p. 102).

34 J. K. Campbell, 1972, p. 46.

35 R. Clogg, 1987, p. 18. “These estimates are essentially guesswork […]”, T. D. Sfikas, 1994, p. 87.

36 D. H. Close, 1995, p. 174.

37 G. T. Mavrogordatos, 1981, p. 184. Mavrogordatos, whose analysis also covered the September plebiscite, is, to our knowledge, the first author to use AMFOGE archival evidence. Michalopoulou (see note 22) was also able to do so.

38 G. T. Mavrogordatos, 1981, p. 182.

39 G. T. Mavrogordatos, 1981, p. 187.

40 G. T. Mavrogordatos, 1981, p. 192.

41 G. T. Mavrogordatos, 1981, p. 187.

42 G. T. Mavrogordatos, 1981, p. 379. In his 1965 survey of Greek political forces, the French political scientist Jean Meynaud was equally severe in his criticism: “Taking into account the composition and functioning methods of the Mission, it would seem very naive to see in this certificate of good democratic conduct anything beyond an attempt to buttress the international status of a government entirely subservient to the Anglo-Saxon powers.” J. Meynaud, 1960, p. 80.

43 H. Richter, 1986, pp. 444-445.

44 H. Richter, 1986.

45 H. Richter, 1986, p. 447.

46 G. T. Mavrogordatos, 1981, p. 378. Already in 1948, American political scientist Earl Latham had ironically referred to AMFOGE’s results, underlining the fact that, while “the sample of voters was constructed according to the best scientific principles by the highest scientific talent, and rigidly and scrupulously followed in the interviewing”, the fact that these interviews were conducted by soldiers in uniform cast “a reasonable doubt about the reliability of answers to a question about coercion, required to be made in the presence of the military” (E. Latham, 1948, p. 646).

47 E. Latham, 1948, p. 379.

48 E. Latham, 1948, p. 379.

49 C. Michalopoulou, 2014, p. 199 (Table 1).

50 “Report of the Planning and Sampling Staff, Central Office, AMFOGE”, April 9, 1946, “Statistical Supplement”, p. 1 (Box 8).

51 “Form A”, February 26, 1946 (Box 14).

52 “Report of the Planning and Sampling Staff, Central Office, AMFOGE”, April 9, 1946, “Statistical Supplement”, p. 1 (Box 8).

53 “Household interview regarding registration. Form B”, February 28, 1946 (Box 14).

54 “Instructions for Form C-2. Verification of residence of registered voters”, February 27, 1946, p. 1 (Box 14).

55 “Form C-2”, appended to the preceding document (Box 14).

56 “Form D-1” (Box 14).

57 Inquiry E is also sometimes mentioned. This one was different, since it covered a variety of investigations carried out in response to specific allegations of fraud or violence. Given the Mission’s limited means, sampling was used here only to select cases that could be the object of further investigation. Inquiry E thus fell outside the specific competence of the sampling staff.

58 AMFOGE I Report, pp. 26-27. Our emphasis. A full report, to be “published in volume form”, was announced, including a series of appendices, but it was never released (p. 35). A draft version of this full report can, however, be found in the archive (“Report of the Allied Mission to Observe the Greek Elections”, April 10, 1946 [Box 5]).

59 The results of these inquiries were compiled in two tables featured in “Report of the Allied Mission to Observe Revision of the Electoral Lists”, August 19, 1946 (Box 10).

60 Michalopoulou has described the population sample as “exemplarily designed, conducted, and presented” (C. Michalopoulou, 2014, p. 198).

61 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 359.

62 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, pp. 363-364. This method was presented for the first time in “On the Two Different Aspects of the Representative Method: The Method of Stratified Sampling and the Method of Purposive Selection” (J. Neyman, 1934).

63 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 360.

64 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, pp. 360-367.

65 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 367.

66 R. T. Jessen, R. H. Blythe, O. Kempthorne & W. E. Deming, 1947, p. 369.

67 “Letter from R. Jessen to F. Kohler”, February 15, 1946 (Box 4).

68 “Letter from J. Neyman to F. Kohler”, February 15, 1946 (Box 4). Neyman’s emphasis.

69 These were: Ioannis (John) Mikhalopoulos, director of the General Statistical Service of Greece (GSSG) and ex-officio member of the International Statistical Institute (ISI); Angelos Angelopoulos (1904-1995), an economist who had acted as Secretary of State for economics in the EAM so-called “Mountain government” in 1944 and would be forced to leave the country in 1946 (G. Politakis, 2018, p. 9); and Demetrios Calitsounskis (1888-1982), a professor of statistics at the University of Athens.

70 “Discussion with 1. Mr. John Mikhalopoulos 2. Mr. Demetrios Calitsounskis 3. Mr. Angelos Angelopoulos”, February 11, 1946 (Box 4).

71 “Letter from J. Neyman to J. Mikhalopoulos”, February 9, 1946 (Box 4).

72 Michalopoulou observes that Neyman’s estimate, which was not based on any survey, was finally not far from that obtained by the AMFOGE 2 miniature sample census (C. Michalopoulou, 2014, p. 199).

73 “Discussion with 1. Mr. John Mikhalopoulos 2. Mr. Demetrios Calitsounskis 3. Mr. Angelos Angelopoulos”, February 11, 1946 (Box 4).

74 Georges Tenekides (1910-1990) was a professor of international law in Athens from 1938 to 1941, and then in Salonika until 1946. Having played an active role in the resistance during the war, he was ousted from his position in late 1946, as the civil war began.

75 “Letter from J. Neyman to Ambassador H. F. Grady”, March 11, 1946 (Box 4).

76 “The Enemies of democracy” Nea Alithia (Salonica), translation, February 27, 1946 (Box 4).

77 “Memorandum. Subject: Termination of the Services of Dr. J. Neyman”, March 20, 1946 (Box 4).

78 “Letter from J. Neyman to the Rector of the University of Salonica”, March 13, 1946 (Box 4).

79 “An Honest Observer”, Eleftherie, translation, March 29, 1946 (Box 4).

80 EAM (Greek National Liberation Front) was the main resistance organization during the Nazi occupation. By 1946, it consisted of five parties, namely the KKE (Communist Party), the Democratic Radical Party, the Agrarian Party, the Socialist Party of Greece and the Democratic Union; but the Communists were clearly its leading component.

81 “To AMFOGE” (otherwise untitled document), April 7, 1946 (Box 6).

82 “To AMFOGE” (otherwise untitled document), April 7, 1946 (Box 6).

83 V. G. Valaoras, 1946, p. 215.

84 “Of the C.C. of the Political Parties’ Coalition of EAM to the Allied Observers for the elections in Greece”, December 9, 1945 (Box 8).

85 “Statistical supplement to Report of the Planning and Sampling Staff”, Central Office, AMFOGE, April 9, 1946 (Box 8).

86 AMFOGE I Report, p. 26.

87 AMFOGE I Report, p. 32. This “median value” was justified on the basis that half of those who claimed to have not voted because they were “not in the neighbourhood on election day” may in fact have been disguising politically motivated abstention.

88 “To AMFOGE” (letter from Central Committee of the Political Parties of EAM), April 7, 1946 (Box 6).

89 G. Midgley, 2000, p. 125.

90 R. J. Jessen, 1978, p. 453.

91 These scales and instruments are, of course, themselves the product of a complicated history. See for instance Ken Adler’s The Measure of All Things on the advent of the metric system (K. Adler, 2002).

92 A. Desrosières, 2008, p. 10.

93 A. Desrosières, 2008, p. 11.

94 To take one admittedly extreme example of such violence, an attack on Litokhoro – for which the Communists later claimed responsibility – occurred on the night before election day, March 30-31, 1946, when an armed band of 60 men descended from Mount Olympus, resulting in eight deaths and a postponement of the vote until a week later (C. M. Woodhouse, 1976, p. 173). Likothoro was not among the sampled units. AMFOGE of course paid some attention to the conduct of future interviewers, to their relationship with the interpreters and to possible inadequacies of its training sessions. It would be interesting to examine these practical guidelines in more detail, since the validity of the sampling procedure depends on a measurement error that remains modest overall.

Haut de page

Pour citer cet article

Référence papier

Jean-Guy Prévost, « The 1946 Allied Mission to Observe Greek Elections »Histoire & mesure, XXXIII-2 | 2018, 163-192.

Référence électronique

Jean-Guy Prévost, « The 1946 Allied Mission to Observe Greek Elections »Histoire & mesure [En ligne], XXXIII-2 | 2018, mis en ligne le 02 janvier 2022, consulté le 16 mai 2025. URL : http://journals.openedition.org/histoiremesure/8146 ; DOI : https://doi.org/10.4000/histoiremesure.8146

Haut de page

Auteur

Jean-Guy Prévost

Centre interuniversitaire de recherche sur la science et la technologie (CIRST), Université du Québec à Montréal (UQAM), Canada. E-mail: prevost.jean-guy@uqam.ca

Articles du même auteur

Haut de page

Droits d’auteur

Le texte et les autres éléments (illustrations, fichiers annexes importés), sont « Tous droits réservés », sauf mention contraire.

Haut de page
Search OpenEdition Search

You will be redirected to OpenEdition Search