Discussion of phenomena like strategic voting and agenda manipulation has largely occurred at the theoretical level. While many theoreticians of social choice claim that the possibility of a disequilibrium and the possibility of political manipulation fundamentally change our ideas of collective decision processes, and even reduce them to meaninglessness (Riker 1982), there is very little empirical evidence relating to these phenomena. For example, in his study of the Norwegian Storting, B. E. Rasch concludes that "strategic voting. . . is almost non-existent in the Norwegian parliament. . . . Furthermore, we can be quite sure that this marked tendency towards sincere voting hides no deceptive manoeuvring" (Rasch 1987, 63). Most work on strategic voting and agenda manipulation uses only hypothetical examples; the few available real-world examples tend to be highly speculative, and the best examples found are discussed again and again. The situation seems to be typical of many of the social sciences: theoretical and empirical studies are conducted in different worlds.

In this article, I try to bring the study of strategic voting back to the real world. One single event is discussed - the election of Urho Kekkonen as President of Finland in 1956. From a social choice perspective, the case is especially interesting for the following reasons:

(1) The result of the election was extremely important in Finnish politics.
The constitutional position of the President is strong. For the purposes

Resumé

The discussion of strategic voting and agenda manipulation tends to be highly abstract. Most real-world examples are either artificial or politically insignificant; the few paradigmatic examples which do exist are discussed again and again. This is one reason why many empirically minded political scientists often ignore the whole social choice approach. The purpose of this article is to focus on a single event - the election of the Finnish President in 1956 - and to show that almost all theoretically interesting phenomena studied in the social-choice literature were present. The paper is based on empirical material, especially on subjective descriptions provided by the main actors. This kind of approach is recommended as a complement to the deductive-theoretical approach.

of this study, the most relevant constitutional powers are (i) the status of the President as the leader of foreign policy, and (ii) his power to nominate the Cabinet - at least de facto, he has a say even in the nomination of individual ministers. Kekkonen exercised effectively all the power given to him. Moreover, he was able to renew his presidency no less than four times, for a total period of 25 years. It is no exaggeration say that the election in 1956 was the the most important

single political (or at least electoral) event in post-war Finnish history.
(2) Although there are no scientific studies on the subject, the course of

events is well documented. Thus we know a lot about the preferences

of the main actors and about the strategic calculations involved.
(3) In the light of the documentary evidence, there is no question but that

strategic voting and agenda manipulation actually took place, and
determined the outcome.

(4) As far as I know, there is only one single reference to the case in the
literature on social choice (Tseblis 1990). This gives a correct account
of the process, but does not go into detail.

My sources consist of the memoirs of the leading politicians, of journalistic works and of the few works written by professional historians. All important points of view, except those of the Communists, are represented. The material is full of personal biases and speculations (and professional historians to be no better in this respect than old politicians). This is an advantage rather than a disadvantage. For the questions relevant for this study relate to how the actors themselves modelled the strategic situation. Can their reasoning (as they themselves describe it) be translated into the language of the models of social choice?

The Institutional Background

The system pertaining to presidential elections which has been in effect for many years is quite complex.¹ According to this system, a 300-member electoral college is elected on the basis of proportional representation. One month after its election, the electoral college then assembles and elects the President. The system used by the electoral college is a modification of the plurality runoff. In the first round, the college votes upon all the candidates put forward by the parities. If no candidate wins more than 150 votes, the electors vote again. If no absolute majority is reached on the second ballot, a third round is arranged between the two candidates who received the most votes in the second round. In the first and the second rounds, the electors have been free to introduce new candidates. The choice, moreover, did not have to be made among the candidates who "officially" stood for

election. Formally, the electors could choose anyone they liked, but in political practice there has traditionally been a clear difference between the candidates introduced during electoral contests and "dark horses" introduced in the electoral college. The introduction of new candidates has generally been looked upon with suspicion.

The election has been secret. In some cases, party groups have demanded that individual electors show their voting slips either to the party whip or to other electors. The Social Democrats, for example, used the latter method in 1956. However, it has been noted that this system is not watertight. Generally, the party leaders were able to infer the choices of other groups, but not to monitor the behaviour of individual electors.

This system seemed to be especially attractive for political manipulators, for several reasons. First, the plurality runoff does not satisfy the condition of monotonicity, and this property can be utilized by the strategists. As Nurmi (1987, 75) says

... in general, when the plurality runoff method is used, a group supporting a candidate
with almost but not quite 50% of the first ranks is well-advised to think how to distribute
its "surplus" votes in the first round, i.e. the votes that exceed the number needed to get

their favourite into the second round

Another well-known property of the plurality runoff method is that although a candidate who is a Condorcet winner will be elected if he or she survives to the final round, there is no guarantee that he or she will not be eliminated in the earlier rounds (Nurmi 1987, 50-51). Finally, the open agenda, or the possibility of introducing "surprise" candidates without any restrictions, makes the system vulnerable to agenda manipulation.

All these formal properties are relevant for the present study. There
have, in addition, been several non-formal institutional properties which
also made manipulation (in a wide sense) an attractive alternative:

(1) The system of proportional representation used in the election of the
college ensured that several groups were always represented. The runoff
method seemed to have had a similar effect. (Wright & Riker 1989)

(2) After the election of the college, the parties had a whole month to collect information and consider their strategies, and between the rounds there were "negotiation breaks", which gave them time to revise their strategies.

(3) In a system with three rounds, the actors can learn more about each
other's preferences than in a two-round plurality runoff or in a simple
plurality system.

(4) As compared with a direct election, an election in a small college is
much easier to manipulate.

It is no wonder that "the presidential game" - as it is popularly called - is,

in Finland, an art of ts own. Given this situation, it is probable that an
exact analysis of some other Finnish presidential elections would also reveal
interesting strategic situations (see Lagerspetz, forthcoming).

The Story

Dramatis Personae

The Conservatives (Kansallinen Kokoomus = KOK; 57 electors) backed Sakari Tuomioja. Tuomioja himself was not a party member, but an independent liberal. The party had chosen him in order to attract votes from the political centre (Junnila 1980, 145).

The Swedish Party (Svenska Folkpartiet = SFP; 20 electors), a loose coalition uniting Swedish-speaking Finns from the far right to the centre, was originally willing to support Tuomioja's candidacy. When, however, the Swedish-speaking Fagerholm became the Social Democrat candidate, it was forced to put up a candidate of its own, Ralf Torngren (Tuomioja 1986, 263).

The Finnish People's Party (Suomen Kansanpuolue = KP; 7 electors) was a small liberal party. Some of its prominent members were willing to support Tuomioja, but this became impossible when he was nominated as a Conservative candidate. The party tried to attract the same voter groups as the Conservatives. There was, moreover, a family quarrel between the party and the small group of independent liberals led by Tuomioja. Both groups saw themselve:* as the political heirs of the old liberal Progressive Party (Tuomioja 1986, 256-257; Junnila 1980, 147).

The Agrarians (Maalaisliitto = ML; 88 electors) supported Urho Kekkonen. As a long-serving Prime Minister, Kekkonen had already become the most important, and the most controversial, politician in Finland. The controversy around Kekkonen was related to his personality, and to his key position in the struggle on foreign policy (see below).

The Social Democrats (Suomen Sosialidemokraattinen puolue = SDP; 72 electors) were internally divided (the party was to split officially in 1957). Before the election, there was a competition for the candidacy between Kar I-August Fager holm and the old wartime leader, Vaino Tanner. Fagerholm nominated in the party congress by a very narrow margin, and the powerful faction kid by Tanner looked at his candidacy with extreme suspicion.

The Communists were operating under an umbrella organization (SuomenKansan Liitto = SKDL; 56 electors). Nominally, the SKDL was not a Communist movement, but in those days it was under the complete control of the party. The SKDL candidate, Eino Kilpi, was not, however, a Communist party member. The party, one of the strongest in

the Western world, was trying to break out from its parliamentary isolation
and was eager to cooperate over the ideological barricades.

The incumbent President, Juho Kusti Paasikivi, had a Conservative background. However, by his skilful diplomacy, he had managed to create good relations with the Soviet leadership, and his personal authority was unquestioned. Had he not been :;o old (he was 85 years of age), he would certainly have been re-elected.

The Preferences of the Actors

There were at least two relevant political dimensions in the situation. On the one hand, there was the traditional right-left dimension. On the other hand, parties were (sometimes internally) divided over the questions of foreign policy. The Communists;, the majority of the Agrarians (led by Kekkonen) and some "realists" inside the SDP and the bourgeois bloc wanted to avoid confrontation with the Russians in foreign policy - and consequently aimed at continuing Paasikivi's policy of low profile neutrality. The majority of the Conservatives and the Tanner faction in the SDP saw the current foreign policy as a temporary concession and supported a more Scandinavian and Western orientation in foreign policy (see Figure 1).

Because all the main groups were approximately of the same size, i.e. had an 18-25 percent electoral support, there was a possibility of a majority cycle. A coalition formed along one of the dimensions could be broken by manipulating the other dimension. This explains not only the problem involved in the 1956 presidential election, but also the more general instability typical of post-war Finnish politics.

Generally the Conservatives and the Social Democrats saw in Kekkonen a dangerous opportunist who was supposed to be willing to make deals with anybody in order to get into power. From the Agrarian point of view, the more extreme Conservatives and the Tanner faction in the SDP were irresponsible nationalists who had not understood post-war geopolitical realities. For the Communists, the Conservatives were still class enemies, but Fagerholm, the SDP candidate, was even worse - a traitor to his class. In the late forties, during the bitter fight over control of the trade unions, the SDP government had crushed the Communist-inspired strikes with an iron hand. That government had been headed by Fagerholm.

The Conservatives and the Social Democrats were not natural coalition partners either. The SDP candidate, Fagerholm, represented everything the Conservatives disliked. He stood to the left in his own party, he had no formal education, and on top of it all, he was Swedish-speaking. And for the Social Democrats, the Conservative candidate Tuomioja was a typical upper-class liberal.

Both the Swedish and the Finnish People's Party were internally divided.

Their own candidates were rather colourless, and did not have much hope of getting elected. The law allowed, moreover, the introduction of a surprise candidate in the first or the second stages of the election. This weakened the chances of Torngren and Rydman. They might otherwise have appeared as Condorcet winners, but because the agenda was not closed, and they, as persons, were not attractive to other groups, their prospects were poor.

The SKDL candidate, Kilpi, was certainly the worst choice for the electors of all the other parties. While the leading politicians were willing to look for tactical support from the Communists, a deep ideological gulf separated even the left-wing Social Democrats from the Communists. This elimination process left Tuomioja, Fagerholm and Kekkonen as the only "serious" candidates from the original set.

The preferences of different groups in the set of the three official main candidates are described in Table 1. The internal divisions inside the SFP and KP are taken into account, although they give some room for speculation.

The actual outcome of the last round reveals that in the KP at least five electors preferred Kekkonen to Fagerholm and in the SFP one or two electors had the same preferences. The position of Tuomioja in their preference orderings is more difficult to infer. However, at least two members of the KP elector group had supported Tuomioja's candidacy in the party council (Tuomioja 1986, 257). It is reasonable to suppose that they preferred him to Kekkonen.

The Course of Events

During the first stage, all groups in the electoral college voted for their own candidates. During the three hours between the first and the second ballot, all the parties engaged in intensive negotiations. In the course of the negotiations, Kilpi, Torngrert and Rydman were dropped.

The situation might be described as a "possible cycle". There was no obvious Condorcet winner, and this was certainly perceived by the actors themselves (Virolainen 1984, 272). Any kind of outcome, including a tie, seemed to be possible. As Meinander (an SFP-politician) later said, "when the electors assembled, no result could be predicted" and "all imaginable possibilities were tried" (Meinander 1978, 150). For this reason the bourgeois (the KOK, SFP and KP) introduced a surprise candidate, i.e. they manipulated the agenda. And their choice was predictable: President Paasikivi. The preferences of the groups in this new set can be inferred with great confidence (see Table 2).

Thus Paasikivi was an obvious Condorcet winner. Again, the leaders of the groups certainly perceived this. However, in the second ballot, the Communists concocted a surprise. They divided their votes between Fagerholm, was their least-preferred candidate, and Kekkonen, their favourite. Here we have an unambiguous case of strategic voting. Paasikivi received only the votes of the bourgeois bloc, and was dropped. The distribution of the votes was as follows:

Fagerholm 114 votes (the SDP 72 + SKDL 42)
Kekkonen 102 votes (the ML 88 + SKDL 14)
Paasikivi 84 votes (the KOK 57 + SFP 20 + KP 7)

Fagerholm, in other words, got 42 extra votes from the SKDL.

In the final round, Fagerholm got 149 and Kekkonen 151 votes; Kekkonen elected President with the narrowest possible margin. Because the election was secret, we cannot for certain know who voted for whom. The party strategists counted that all the 144 votes of the SKDL and ML plus five votes from the KP went to Kekkonen. One elector of the SFP (Verner Korsbåck) was a likely supporter, too. One vote remains unexplained, has caused endless speculation. Of course, it is possible that there were "defections" in both directions.

From my point of view the most interesting events occurred in the negotiations between the parties before and after the first stage. How was Paasikivi introduced? How did the Communists devise their strategy? Why did the other parties not anticipate it?

The Game

The entire game is quite complex. I shall try to divide it into several subgames
to apply a strategic analysis to all of them.

The Game Inside the Bourgeois Bloc

Prima facie, the position of the Conservative candidate Tuomioja was quite good: he could rely on the support of 57 Conservative and 20 Swedish electors in the second round. That was enough to help him into the third round. Because Kekkonen was likely to get all the votes of the Agrarians (ML) and the Communists, the final contest would be between him and Tuomioja. The problem, however, was the pivotal role of the Finnish People's Party (KP). With its seven electors, the KP could help Kekkonen to victory even in the second round. The negotiating position of the KP was thus externally strong, although weakened by the internal divisions of the party. In the negotiations with the Conservatives and the SFP, the KP

refused to back Tuomioja. Equally, it rejected the alternative proposed by the SFP, by which the negotiators of the three bourgeois groups would offer all their candidates to the Social Democrats and let them pick the candidate who was the most acceptable in their eyes. That might have produced either Torngren or Rydman as a compromise candidate (Tuomioja1986, Saukkonen 1973, 250-251; Meinander 1978, 150).

One explanation for the behaviour of the KP was that the party leadership was secretly negotiating with both the Social Democrats and with the Agrarians (ML) in order to obtain governmental posts as a direct reward for its support for their candidates, and a deal was actually made with the latter (Skog 1971, 375; Virkkunen. 1976,160-161). The party was, however, facing a difficult choice: it had very little in common with Fagerholm, the majority of KP-electors could not swallow Tuomioja whom they regarded as a "traitor" to true liberalism, but support for Kekkonen might mean "political suicide" (Skyttå 1970, 105). Thus the KP leaders persuaded the Conservatives and SFP to back Paasikivi, who was seen as an escape-route from the difficult situation. This was clearly against the deal made with the Agrarians; it is possible that the KP actually hoped that its attempt to bring in Paasikivi would fail (Virolainen 1984, 290; Saukkonen 1973, 251).

The Conservatives proposed that the bourgeois bloc should make a strategic threat to the ML. If the latter were not willing to accept Paasikivi, the bourgeois parties would vole for Fagerholm already at the second stage. The problem with this proposal was that it was not credible. In the group meetings it turned out that some electors in the Conservative and SFP groups were strongly against Fagerholm and the KP group did not give any definite answer (Suomi 1990, 491).

The introduction of Paasikivi seemed to be the only way out of the difficult situation. In the discussions with the bourgeois party negotiators, Paasikivi insisted that there should be "a substantial majority" behind him. The negotiators, however, went back to their groups and presented Paasikivi's conditional assent in an unconditional form. They were probably hoping that it would have a snowball effect so as actually to create a "substantial majority". These hopes never materialized. There is some truth in the accusation that the old man was fooled into playing their unsuccessful game (Skog 1971, 376; Saukkonen 1973, 252-253).

The Bourgeois-SDP Subgame

The major objective of both the Conservatives and the Social Democrats was to defeat Kekkonen. This was feasible only if the parties could coordinate strategies (and even then the support of the KP and SFP was needed). At the same time, both parties wanted their own candidate elected, and looked at the candidate of the other party with suspicion.

The Social Democrats had the following possible strategies:

(1) to support Tuomioja;

(2) to support some bourgeois candidate acceptable to the Conservatives
(Paasikivi was the obvious choice);

(3) to introduce a surprise SDP candidate who was acceptable to the
Conservatives (Tanner was an obvious choice);

(4) to make a compromise with other parties (the ML and the Communists);

or

(5) to stick with Fagerholm and try to force the bourgeois bloc to accept
him as a lesser evil.

Under the chairmanship of the prominent Social Democrat, Penna Tervo, the party committee of the SDP had made a ruling according to which the party could withdraw its support from Fagerholm only by a unanimous decision of all the electors, and the electoral group had confirmed the ruling. Thus, the party had intentionally tied its hands, and refused to negotiate before election day. At the last moment, the negotiators of the party, Tanner among them, tried to persuade the electors to switch their votes to Paasikivi. This proposal was rejected (Skog 1971, 374-376; Paavolainen 421-422).

The game between the KOK and the SDP somewhat resembled the Battle of the Sexes. As far as the choice was supposed to be made between Tuomioja and Fagerholm, the game could be characterized by the following reduced sub-game tree (Figure 2).

The SDP behaved like an archetypical player in the Battle of the Sexes - it tied its hands. Even threats were introduced. The Social Democratic press wrote that if the choice were to be between Tuomioja and Kekkonen, the party would back Kekkonen. It is not clear whether this expressed a real preference or whether it was a strategic threat. The Conservatives generally interpreted this threat as a piece of bluff (Tuomioja 1986, 285).

The game was changed by two factors. The Conservatives introduced Paasikivi. In response, the ML and the Communists invented the strategy of dividing their votes between Fagerholm and Kekkonen. The crucial question in the new game was whether the ML-Communist strategy was anticipated by the SDP negotiators. Assuming that it was, why didn't the party change its own strategy? In the new situation, it could switch its support to Paasikivi, who clearly was their mini-max candidate. For the Social Democrats, he was certainly a better choice than either Kekkonen or Tuomioja. And if the remaining candidates in the final stage of the election had been Paasikivi and Kekkonen, Paasikivi would certainly have won.

On the other hand, the SDP could have tried to persuade the bourgeois

parties to move behind Fagerholm by informing them about the Communists'strategy. they had agreed, the Communists would unintentionally have helped to elect Fagerholm in the second round. The original Battle of the Sexes between the KOK and the SDP was transformed into a more complex game. Assuming that the Social Democrats really did anticipate the Communists' strategy, the corresponding reduced game tree would look like this (Figure 3).

There are three possible interpretations for the behaviour of the SDP
electors:

(1) They did not anticipate the Communists' strategy. Voting for Fagerholm was intended as a tacit support for Paasikivi, which also satisfied the ideological needs of the party. Fagerholm had only 72 electors against the 84 electors supporting Paasikivi. Thus Fagerholm was likely to be eliminated, and the party could support Paasikivi in the final round without nominally rejecting Fagerholm (Skog 1971, 376). Hence there was no real risk involved.

Several things speak against this interpretation: the party negotiators'
eagerness to support Paasikivi, Tanner's comment to the Agrarian negotiatorsthat
Communists, were "able to pick up or drop anyone they

wanted" (Virolainen 1984, 275), and similar remarks of the Conservative
negotiator Saukkonen (Saukkonen 1973, 253).

(2) Perhaps because Fagerholm's chances against Kekkonen were almost even, the supporters of Fagerholm were willing to take the risk and gamble. The reason why they could not persuade Paasikivi's supporters was due to lack of time; the Communists made their final decision at the last moment.

(3) The subtlest explanation is basedon the internal struggles in the party. It was generally known that Tanner and his group did not accept Fagerholm's candidacy; they probably preferred Paasikivi to Fagerholm. It has often been noted that the necessary condition for the success of the strategy of the Communists was that the SDP electors would stick to their decision to back Fagerholm right to the end and would refuse to vote strategically. This is what happened. Now the person who orchestrated this strategy was Penna Tervo, who was on bad terms with both Tanner and Fagerholm. It is possible that Tervo actually preferred Kekkonen to Paasikivi and perhaps even to Fagerholm. Some commentators believe that Tervo did anticipate the result of the strategy adopted by the party. If this is true, Tervo played the game with an almost diabolical cunning: he led his party comrades to vote for their most preferred candidate, in order to secure that their least preferred candidate would be elected (Martin 1982, 154-156; Tuomioja 1986, 284-285).

In the only rational choice analysis made on the subject, George Tseblis
claims that

... to vote strategically, Socialist leaders would have to explain to their own party activists
and voters why they were withdrawing their quite successful candidate - a difficult task
(1990, 4).

I find Tseblis's explanation unconvincing. In all earlier presidential elections, the Social Democrats had ended up backing a bourgeois candidate. 1950, for example, they supported Paasikivi. As far as the party activists were concerned, the majority of them were against Fagerholm's candidacy. Nevertheless, Tseblis's explanation points in the right direction: the party leadership was playing two games simultaneously. One game was played against other parties, the other they played among themselves. The sub-optimal behaviour of the; party was probably an outcome, perhaps unintended, of the struggle between different factions.

The Communists' Dilemma

The idea of dividing the Communist votes in the second ballot was invented by the Agrarian negotiatiors, possibly by Kekkonen himself. But they had some difficulties in selling the idea to the Communists. One the day before the election, the party council of the ML discussed alternative strategies. The notes of Johannes Virolainen., one of the important politicians in the party, reveal how the party leaders reasoned in this complex situation:

Before the first ballot, the Agrarians negotiated with all other groups. In the negotiations it became clear that (1) the Conservatives and the Social Democrats were not willing to accept Kekkonen under any conditions, and (2) although some individual members of the KP and SFP were likely to support Kekkonen at the final stage, the parties refused to accept any joint strategy with the ML. Hence the SKDL, or the Communists, now became the most important negotiating partner for the Agrarians.

Virolainen's notes show that the possibility of strategic voting was discussed Paasikivi was brought forward as a surprise candidate. However, if the Conservatives had kept behind Tuomioja, strategic voting would have been unnecessary: the Conservatives and the SFP had enough votes to help him to the third stage. Also, as Virolainen noticed, Tuomioja was less dangerous than Paasikivi, who was a likely Condorcet winner.

In order to utilize strategic voting, or preference misrepresentation, in a plurality runoff system the actors have to know the entire preference profile of the other participants (Nurmi 1987,124). Thus, quite a lot of information is needed. The Agrarian leaders understood this. The legendary party secretary Arvo Korsimo is said to have been in personal contact with every one of the 300 electors before the final ballot. The other parties were much less effective. The Social Democrats refused to enter into discussion with the others before the election day, and Tanner even refused to be in the same room with the Communists (Skog 1971, 375-376). Of the Conservatives, Tuomioja (the son of the Conservative candidate), has testified that

. . . For example, the SKDL electors were considered to be like aliens from a different planet. (The Conservatives) neither wanted nor tried to study their thinking or aims. This led. . . to the impression that the SKDL would stand behind Paasikivi, while the fact that the SKDL divided its votes - which was mentioned as a theoretical possibility - became a bitter surprise for the bourgeois groups. (Tuomioja 1986, 248).

The Communists had two political objectives. Their aims were to ensure that the new President would continue the current foreign policy, and they wanted to break free from their parliamentary isolation. The first and the second objective ruled out both Fagerholm and Tuomioja, the second ruled out Paasikivi, who had expelled the Communists from the government in 1948. This meant that their preferences were obvious, and it also weakened their negotiating positions: no credible threats were left.

In the negotiations, the Agrarian leaders told the Communists about their strategic calculations. The negotiators agreed that Paasikivi was more dangerous to Kekkonen than Tuomioja. If the Conservatives continued to back Tuomioja, Communist strategy would be to support Kekkonen unconditionally. However, Paasikivi was a more complex case. The minimaxstrategy the party was to support him in order to prevent the

presidency of either Tuomioja or Fagerholm. Indeed, when a delegation of Finnish Communists visited Moscow before the election, the Soviet party representatives advised them to support Paasikivi. The reasons were that the Russians did not believe in Kekkonen's chances and that they feared the Social Democrats almost paranoically. On the other hand, as Hertta Kuusinen, an influential Finnish Communist, stressed, the election of Paasikivi would, because of his age, "just have postponed the decision" (Suomi 1990, 486, 488).

The Agrarian negotiators asked the Communists to divide their votes. In that strategy two risks were involved. First, while Paasikivi would almost surely defeat any other candidate, Fagerholm had some chances against Kekkonen. To help Fagerholm to the final ballot might actually make him President, but without the tactical support of the Communists, he would surely lose in the second ballot. Second, if the bourgeois bloc could anticipate the Communist strategy, it could also vote for Fagerholm and make him President already at the second stage (Virolainen 1984, 282-286, 288-291). The dilemma faced by the Communists may be depicted as shown in Fig. 4.

Both the ML and the Communist negotiators understood the basic nature of the situation, but they had certain incentives for misrepresenting it to their negotiating partners. The main task of the ML negotiators was to convince the Communists that the first risk was small.

Maalaisliitto had to show sufficient guarantees that it had secured the seven extra votes necessary for the election of Kekkonen. If the SKDL had supported Kekkonen, but he were not elected in the final ballot. . . the party would indirectly have helped to elect the feared common candidate of the Right and the Social Democrats. At the same time, the main goal in the elections, which could have been reached by supporting Paasikivi, namely to secure the continuation of the present foreign policy, would not have been reached. Therefore they unambiguously told to Maalaisliitto, that if guarantees for a sufficient support for Kekkonen were not given, the Paasikivi alternative should again be taken into consideration. (Suomi 1990, 489)

The ML negotiators claimed that they could secure Kekkonen's victory
over Fagerholm:

"The Finnish People's Party has informed us", said Sukselainen (to the Communist negotiators), they will not vote for Paasikivi or Tuomioja in the second election. At the same time, they have made it clear thai: if Kekkonen and Fagerholm are in the third ballot, five of them will vote for Kekkoinen". 'As', continued Sukselainen, according to my information, five of the Swedes will cast their vote for Kekkonen, that should be enough for victory. . ."

Kleemola (another ML negotiator): Now 13 votes are needed, for Fagerholm to beat
Paasikivi in the second ballot. (Virolainen 1984, 288-289)

Sukselainen's predictions were wrong: the Finnish People's Party in fact
supported Paasikivi in the second ballot, and only one or two of the SFP

electors supported Kekkonen in the third. It is most likely that Sukselainen intentionally overestimated Kekkonen's chances since the Communists did not have a direct information channel to the bourgeois parties. They had to rely on the information given by the Agrarians. Suomi (1990,495) claims that the Communists actually doubted the figures given by Sukselainen, but still believed that a majority could be attained.

The Communists also had an incentive to exaggerate their willingness to
consider Paasikivi as an alternative to Kekkonen so as to get concessions
from the ML:

"Paasikivi is, of course, a lesser evil from our point of view than Tuomioja or Fagerholm. I am afraid", said Hertta (Hertta Kuusinen, the Communist negotiator), "that we could not get a group decision to vote both for Kekkonen and Fagerholm in the second ballot." I think that this was tactics from Hertta's side, for she knew the situation and surely understood what the issue was. (Virolainen 1984, 289).

The final agreement between the parties was reached almost at the last
moment:

Korsimo (the Agrarian party secretary) told us very nervously that the Socialists were going to do the same trick for us as we were planning for them; they would divide their votes between Paasikivi and Fagerholm, and in the last election there would be Paasikivi against Kekkonen. He also reported that Kekkonen hopes that they would really divide their votes between Kekkonen and Fagerholm. In the end, Kuusinen promised clearly and without any ambiguity that they would do that. . . (Virolainen 1984, 291).

Again, this was most likely a means of putting pressure on the Communists.
Probably Korsimo did not have any new information about the strategy of

the Social Democrats. The last message from them recorded by Virolainen
was that their decision not to drop Fagerholm was still in force (Virolainen
1984, 290).

The game tree above (Figure 4) shows how delicate the situation was. The Social Democrats could neutralize the Communists' strategy by casting more votes for Paasikivi than the Communists were going to cast for Fagerholm. However, the Social Democrats confirmed their original decision to stay behind Fager holm. If the parties could rely on that, the right-most branch of the game tree could be ignored.

There was still a possibility that the bourgeois bloc would detect the Communists' strategy and move behind Fagerholm at the last moment. If the Communists believed this, they should, after all, have cast all their votes for Kekkonen. Then the Conservatives' best choice would have been to vote for Paasikivi. Given the sincere strategy of the SDP, all possible strategies of the SKDL and the KOK were vulnerable for some set of voters in the Farquharson sense; there were no equilibria. Virolainen has recalled in an interview how the Conservative and the SFP negotiators came to inform him that they were going to bring in Paasikivi:

"- Are you unanimously behind Paasikivi?" Virolainen asked the negotiators. The men
assured him that this was the case

- "Have you really made decisions on the issue?"

Aura and Ohman assured him thai: the groups had made binding decisions. - "Happily they didn't understand what I was asking," Virolainen said, and shook his head. - "If the other groups had grasped ouir plot and already voted for Fagerholm in the second ballot, Fagerholm would have become President." (Lehtinen 1982, 87)

According to the Conservative Saukkonen, the others actually did anticipate plot". But they could not be sure that the Communists would really act strategically. For the: Communists, Fagerholm was a much worse alternative than Paasikivi, and the Conservatives obviously hoped that the Communists would be willing to accept the old man in order to avoid Fagerholm (Tuomioja 1986,284; Saukkonen 1973,253). The Conservatives contacted the Communists amd asked whether they would be willing to support Paasikivi, but "no clear answer was given. In the (Conservative) group this created some doubts about the result" (Saukkonen 1973, 253).

Discussion

It is interesting here to compare the Communists' strategy with that adopted
by the Social Democrats in 1931. At that time, the SDP made a decision
to divide its votes in the second round, if necessary. But it also made its

decision public, and even revealed the amount of votes it was reserving for tactical use. Their antagonists among the Agrarians used that information to neutralize the strategy. In 1956, the Communists promised their support to their coalition partners only at the last moment, although their leaders had probably made the decision some time before (Suonii 1990, 495). Skyttå (1970, 107) claims that, after making the decision formally in their group, the Communists "closed themselves sound-proofed into their room. No one could go in or out before the bell rang". Postponing the formal decision and cutting themselves off from further negotiations can be seen as a rational means for keeping their opponents in uncertainty.

But the Agrarians did not close their doors, and, as Virolainen tells us, "the plan leaked". However, the Conservatives also felt that they had committed themselves by asking Paasikivi in the first place (Saukkonen 1973, 253). Thus both the Social Democrats and the bourgeois bloc had tied their hands, and the Communists were free to move. The only risk was at the third stage, and the left-most branch of the game tree (Figure 4) could be ignored, too.

Alan Gibbard has written, quite provocatively, that honest voting exists "in virtue of individual integrity, ignorance or stupidity" (Gibbard 1973, 593). All these factors might have some role in explaining why neither the bourgeois bloc nor the SDP voted strategically in the second round. "Integrity" as such was not central. Although Kekkonen's opponents were morally outraged by the Communists' stratagem, we have seen that they themselves were quite willing to make calculations in light of different strategies. Ignorance, however, seems to have been the key factor. There was an informational asymmetry between Kekkonen's supporters and his enemies, resulting from the Agrarians' ability to collect information and the Communists' ability not to reveal information. "Stupidity" also had a role, not at the individual level, but at the organizational level. Because of their heterogeneity, it was difficult for both the Social Democrats and the bourgeois bloc to agree on a common strategy.

However, the ML-Communist coalition could not be absolutely sure that the others would not vote strategically. And it was a matter of degree: while majorities were likely to follow the group decisions, individual electors could still "defect". When the fundamental agreement between the ML and SKDL on the strategy to be applied was reached, there was still the question of how many Communist votes had to be transferred to Fagerholm in order to drop Paasikivi out of the game. On the supposition that the others would vote according to their decisions, 13 votes would have been enough. In the final negotiations between the ML and the Communists, Kuusinen promised 29-30 votes for Fagerholm. At the last moment, she told Kekkonen that Fagerholm would get 30-40 Communist votes. The actual number was 42.

The optimal number can be counted in the following way. First, there was the danger that some outsiders would cast their votes for Fagerholm. If k was the number of extra votes coming from the Communists, and m the number coming from the others, the first condition for k was:

From this we get

On the other hand, some Social Democrats could vote for Paasikivi, n of
them. If they didn't, only 13 extra votes were needed. Thus,

The last condition for the optimal size of k was the requirement that
Kekkonen should not get fewer votes than Paasikivi. Thus

from which we get

For example, if the Communists estimated that over 20 Social Democrats might defect to Paasikivi, the optimal size of k would be less than 40 but more than 32. There was som«; discussion on the optimal size of k. When Kekkonen heard that he would get 30-40 extra votes, he "warned us to take care that Fagerholm would not get too many votes in the second election." (Virolainen 1984, 292).

The fact that Fagerholm got 42 extra votes while 13 would have been enough shows that the Communists estimated the risk that some Social Democrats would vote for Paasikivi as higher than the risk that some bourgeois electors would vote for Fagerholm. On the other hand, they did not expect more than 17 defections from the SDP. Given the preferences of the parties, this was reasonable. In the Social Democratic party there was a group which had initially opposed Fagerholm's candidacy and was eager to throw him overboard. Among the bourgeois groups, only the Swedish-speaking group had some sympathy for Fagerholm (Meinander 1978, 149). Even if it had cast all its votes for him (so that m = 20), it would not have been enough. In spite of Kekkonen's fears, it was reasonable to make k relatively high. Suomi (1990,495-496) states that the Communist leaders Murto and Ryomå "counted very carefully", and gave personal instructions to every elector. On the whole both the ML and SKDL played the game with great strategic sophistication.

Conclusions

My main purpose here has been to relate some parts of the theory of social choice to an interesting case taken from real life. If we try to evaluate the descriptive force of the typical social choice models, we can draw the following conclusions:

First, much as social choice approaches presume, it would appear that politicians are often willing and able to make complex strategic calculations and to collect and use the information on each other's preferences. Sometimes has been claimed that strategic voting cannot be but a marginal phenomenon, for it presupposes a lot of information about the preferences of other voters. Our example shows that in suitable conditions, sufficient information can be collected.

Second, strategic voting, agenda manipulation and the resulting political disequilibria do exist in the real world. Had the actors in our case reasoned in a slightly different way, almost anything could have happened: the election of any of the four candidates (Tuomioja, Paasikivi, Fagerholm and Kekkonen) could have been the outcome of the social choice. A further analysis of the history of Finnish presidential elections will probably lend support to Riker's conjecture, i.e. politics is generally a disequilibrium process.

Third, given the strong constitutional position of the President in Finland, the choices of individual actors had an immense practical importance. The problems studied by the social choice approach are not just curiosities. From the social choice point of view, there are several aspects in the case which complicate the modelling:

(1) In complex situations like the one discussed here, parties develop an intensive division of labour. They delegate the task of formulating a strategy to a small group of negotiators, and the other members have to rely on the information provided by them. This does not, however, guarantee that the parties always act as units. It is not always easy to identify the basic actors: sometimes they are the parties, sometimes party factions or individual members.

(2) Intra-party bargaining is sometimes as important as the more visible inter-party bargaining, and these two games can affect each other. In our case, all the major parties speculated that some of the others might split.

(3) The preferences of the parties are partially interdependent. For
example, one reason why the Conservatives could not support Kek-

konen was that the Communists did support him. It has often been
claimed that some bourgeois electors voted for Fagerholm mainly
because they disliked the Communists' stratagem. Equally, Paasikivi

became less acceptable for the Communists because his candidacy was
initiated by the Conservatives.

(4) Not only the information possessed by the actors at a particular moment, but also their ability to get additional information is asymmetrically distributed. They often have incentives, not only to lie about their own strategies and preferences, but also to misrepresent the information concerning third parties.

These conclusions do not present an argument against the use of rational models. In cases like the Finnish presidential election in 1956, there is no question of a predictive model - indeed, social choice theory tells us that the situation was essentially unpredictable. This "non-prediction" is in accordance with the subjective perceptions of the actors. The heuristic value of the basic supposition - namely that politics can be seen as a strategic game - is still considerable in understanding what happened. Finnish historians and journalists have discussed at length "the decisive vote" given for Kekkonen in the last round. The social-choice approach stresses that it is equally interesting to ask why the 299 remaining electors voted as they voted. I think that it is time for formalists and historians to take each other seriously. I also think that the case of the Finnish presidential in 1956 deserves to become a classic among those who want to continue the line of research opened up by Riker and others.

---

NOTES

1. A decision to change this system has only recently been adopted, and will take effect for the first time in connection with the presidential election scheduled for 1994.

ACKNOWLEDGMENTS

I am grateful to Dr. George Maude for linguistic corrections and substantive comments. On
the meaning of technical terms used in this article, see the Appendix.

REFERENCES

Gibbard, A. 1973. "Manipulation of Voting Schemes", Econometrica 41, 587-601.

Meinander, N. 1978. lnsyn och efterklokhet -politiska minnen 1940-1962. Borgå: Soderstrom
&Co.

Nurmi, H. 1987. Comparing Voting Systems. Dorderecht: D. Reidel.

Plott, C. 1976. "Axiomatic Social Choice Theory: An Overview and Interpretation", American
Journal of Political Science 20, 511-596.

Rasch, B. 1987. "Manipulation and Strategic Voting in the Norwegian Parliament". Public
Choice 52, 57-73.

Riker, W. 1982. Liberalism Against Populism. Chicago: W. E. Freeman.

Saukkonen, J. 1973. Vastuun sukupolvi. Helsinki: Kirjayhtymå.
Skog, E. 1971. Sosialisti ja patriootti muistelee Porvoo: WSOY.

Skyttå, K. 1970. Presidentin muotokuva 11. Helsinki: Kirjayhtymå.

Wright, S. & Riker, W. 1989. "Plurality and Runoff Systems and Numbers of Candidates",
Public Choice 60, 155-175.