Between 1954 and 1972, in two commissions appointed by the government, the four major Swedish parties bargained over constitutional reforms. In the new constitution that took effect on 1 January 1975, the Speaker of the Riksdag was given the right to appoint a candidate for Prime Minister (and thereby a new government). A voting rule which allows the proposed government to assume power unless a majority of all members of the Riksdag vote against it was also adopted.¹ In this study, public documents, archival material and a series of interviews are used to reconstruct the choice of this voting rule. I explain why a vote is required, why this vote is formulated in negative terms, and why the parties decided on an absolute majority threshold.²

This case study is important in the context of the literature on new institutionalism, coalition formation and constitutional design. The study falls within one of three major strands of new institutionalism. The sociopsychologicalstrand, by March & Olsen (1989), and the macrohistorical represented by scholars such as Evans et al. (1985) and Skocpol (1979), cast their analyses in terms of large-scale social processes.³ The scope of the present analysis is limited to the choices and behaviour of political parties in specific situations. Because of this, I favour the third, rational choice oriented strand of new institutionalism. The proponents of this strand of new institutionalism differ from "old¹" institutionalists in that

Resumé

In the early 1970s the Swedish parties designed new rules for government formation. In this study it is shown that the choice of a new voting rule stemmed from a combination of goaloriented and a process of socialization. In a cross national comparison, it is also shown that there exists an empirical correlation between the Swedish type of rule and (1) a high frequency of minority governments, (2) a high frequency of small minority governments and (3) a short government formation process. From the perspective of coalition formation theory, this is a challenging finding.

they regard as incomplete explanations which hold that institutions and rules explain outcomes. All explanations are cast in terms of the choices and behaviour of actors. Rules are important because they set limits on the range of possible behaviour and they favour certain outcomes and make certain choices more likely, but purposeful action is at the core of the analysis (Riker 1980; Shepsle 1986, 1989; Przeworski 1988; Tsebelis 1990).

This study shows that purposeful political actors expected the formation rule to have important consequences. The Swedish parties argued that there exists a relationship between formation rules and empirical outcomes. If this is true, other countries with similar formation rules should have similar empirical records. I shall show that they do. In particular, the Swedish parties expected that the formation rule selected would facilitate minority governments and a short government formation process. The other Nordic countries have similar formation rules. They also have similar empirical records.

This finding illustrates that new institutionalism can inform the literature on coalition formation. Following Riker's (1962) early example, this literature simply assumed that a winning coalition must contain more than half (an absolute majority) of the votes in the decision-making body. This assumption is problematic. For one thing, the assumed rule does not exist in any parliamentary democracy commonly included in tests of coalition theory (Bergman 1993). Secondly, Riker (1982, 56-58) himself has pointed out that some voting rules are not neutral. One of the properties of the relative majority voting rule (i.e. the alternative which gets more than half of the votes cast wins) is that it does not favour one particular alternative. In contrast, an absolute threshold, such as the one assumed by Riker (1962) or the one used in the Swedish government formation vote, does favour one particular alternative. In practice, absent and abstaining votes count in favour of one alternative. The most interesting literature on coalition formation has begun to pay attention to voting rules and other institutional arrangements (for example, Budge & Laver 1986; Laver & Schofield 1990; Strom 1990). Albeit not always explicitly, this literature represents the rational choice strand of new institutionalism. In agreement with this literature, I have argued elsewhere (Bergman 1993) that coalition theorists should study the relationship between existing formation rules and coalition formation. This article is an example of such a study.

This case study is also theoretically interesting because an important body of literature on constitutional design stresses that self-interest is a salient motive when political parties design the "rules of the game" (see, for example, Riker 1986 and von Sydow 1989). I do not doubt this thesis. However, this study shows that there is also an important element of socialization involved in the process of constitutional design. Over time the four parties came to share a strong preference for a particular form of

rules. When towards the end of the process new actors became involved,
they questioned this preference.

The article is organized as follows. In the next section the Nordic formation rules and the empirical record of government formation are presented. In the third section, how and why the Swedish voting rule was chosen is explained. How the voting rule has worked in practice is also discussed. The article then ends with a brief summary and some concluding reflections on the Swedish case and the literature on new institutionalism, coalition formation and constitutional design.

Formation Rules and the Empirical Record

In all of the Nordic countries, government formation rules are formulated in negative terms (Bergman 1993). In the two Swedish commissions on constitutional reforms, the parties and their experts referred to such rules in terms of negative parliamentarism. The major alternative was referred to in terms of positive parliamentarism. In countries with positive parliamentarism, new government must win a vote of investiture by a majority (relative or absolute) in the parliament. That is, it is required that the parliament supports the new government.

In countries with negative parliamentarism, however, such as Denmark, Finland, Iceland and Norway, there is no vote in the parliament before a new government assumes power. Instead, the ultimate power of the parliament rests with its ability to unseat an incumbent government. As mentioned above, in Sweden there is a vote in the Riksdag when a new government assumes power. Yet because a new government does not have to win this vote, Sweden too should be classified as a country with negative parliamentarism.⁴

In Table 1 the record of government formation in the five Nordic countries is compared with that of ten other parliamentary democracies in the 1945-87 period.⁵ The Nordic countries are characterized by (1) a high frequency of minority governments, (2) a more than occasional formation of very small minority governments and (3) a short government formation process.

In this sample of 15 parliamentary democracies, one-third of all governments only a minority of the votes in the parliament. More than half of all minority governments observed (61 of 116, or roughly 53 percent) were formed in the Nordic countries, whereas other countries with negative parliamentarism had a smaller proportion of minority governments. In the United Kingdom, the low frequency of minority governments is easily explained. The plurality electoral systems tend to produce a single majority party. Because of this, the empirical record of government formation is

different from that of the Nordic countries (Bergman 1993). Canada, on the other hand, has had fewer single party majorities and a higher frequency of minority governments. Finally, Portugal has only been democratic for a part of the sampled period, and a closer analysis of this deviant case is not possible in the context of this study.

As regards the Nordic countries, the proportion of minority governments is smaller in Finland and Iceland than in Denmark, Sweden and Norway. There are also some differences in the size of the minority governments. When minority governments were formed in Finland and Iceland, they were often in control of less than 40 percent of the votes in the parliament.⁶ Denmark has had a similar record. In Norway and Sweden, on the other hand, governments in control of less than 40 percent of the parliamentary vote have been more rare. Yet, small minority governments occurred more frequently in all of these countries than in most other parliamentary democracies. In fact, of all governments in control of 40 percent or less of the votes, 34 out of 39 governments occurred in the Nordic countries. If we include Portugal in this group, 36 of the 39 governments were formed in countries with negative parliamentarism.

In the Scandinavian countries (Denmark, Norway and Sweden), the

average length of the formation process was one week (7 days). Finland and Iceland, by comparison, had on the average a longer formation process (about 4 weeks). The same is true for Portugal. If a negative formation rule facilitates a short government formation process, in short, there must be other factors present in these three countries which operate in the opposite direction. Ireland, which has a positive form of parliamentarism, has also not performed according to the general expectation. However, in five of eight countries with negative parliamentarism, the government formation process has been shorter than the cross-national average (22 days). In six of seven countries with positive parliamentarism, the length of the formation process has been at or above the cross-national average.

In Table 2 the differences between countries with negative and positive parliamentarism on these three indicators are summarized. Minority governments are considerably more common in countries with negative parliamentarism. Very small minority governments form almost exclusively in these countries. As for the length of the formation process, in the eight countries with negative parliamentarism, the average length of the formation process was two weeks (14 days). In the seven countries with positive parliamentarism, the average length of the formation process was about twice as long (31 days).

In sum, this section illustrates that the Swedish parties had reason to be concerned with the consequences of the government formation rule that they chose. Different formation rules tend to be associated with different empirical outcomes. There are some deviant cases on all of the three indicators. However, the overall picture is that negative parliamentarism is associated with minority governments, small minority governments and a short government formation process. Positive parliamentarism is associated with a higher frequency of majority governments, the absence of small minority governments and a longer formation process.

The Choice of the Voting Rule

Why, then, did the Swedish parties chose a negative and absolute voting rule? Because the context in which the vote was chosen is important, it is appropriate to begin with a short review of the deliberations that preceded this choice. In an initial phase between 1954 and 1963, a government appointed commission (Forfattningsutredningen, FU) involving the Social Democratic Party, the Liberal Party, the Centre Party and the Conservative Party studied the question of large-scale constitutional reforms for nine years. After this, a second commission {Grundlagberedningen, GLB) was appointed in 1966 and presented its final report in 1972.

The First Commission

In 1963, when the first commission (FU) presented its report, a new government was (formally) appointed by the monarch and remained in power until it resigned or the Riksdag forced it to resign. The existing constitution did not specify voting rules for either the formation or the resignation process.

The four-party commission argued that it should be possible for the Riksdag to remove a government by a declaration of no confidence (misstroendeforklaring). have effect, however, it was suggested that such a declaration would have to be accepted by more than half of the members of parliament. The vote was not to be attached to any particular policy issue. The vote should simply be about whether or not the government was tolerated by the Riksdag. The distinction between tolerance of the government in general and particular policy issues was intended to ensure government stability. A party that opposed a specific policy but otherwise tolerated a government should not be "forced" to cause the downfall of the government. The commission noted that such events, in which specific policy issues had become intertwined with the issue of the survival of the government, had taken place in other countries and argued that they should be avoided. The commission also deliberately proposed a voting rule - absolute majority - in which abstaining and absent votes would (in practice) be counted in favour of the existing government. The commission argued that this voting rule would help ensure government stability and that it was a guarantee against misuse of the declaration of no confidence (SOU 1963:17, 146-148; see also SOU 1966:17, 29).

With respect to the rules for government formation, the commission wanted to keep the existing system. One important reason for this was that Jorgen Westerståhl, a professor of political science who was the commission's chief secretary as well as a member of the commission, argued in favour of negative parliamentarism. In his view, positive parliamentarism sets too high a threshold and can result in a prolonged formation process

(Westerståhl, interview, 1990). The commission concurred. It argued that the existing system worked well. Governments had been able to form without much delay, and even in complicated situations the monarch's choice of Prime Minister had not been criticized (SOU 1963:17, 134, 146).

According to the commission, the monarch helped strengthen and safeguard democracy (SOU 1963:17, see also Nyman 1981). This position, however, was strongly criticized within the Social Democratic Party (SAP-PS 26 October 1965; Westerståhl 1976, 10; von Sydow 1989, 161; Gustafsson, interview, 1992). It was also criticized by Herbert Tingsten, liberal political science professor and former editor of Sweden's largest morning newspaper, Dagens Nyheter. Tingsten (1964) criticized the monarch's involvement in the proceedings of the representative democracy. He argued that it was possible simply to let parliament elect a Prime Minister, or, should this be found to be unsuitable because of the obvious risk of tactical voting, the right to appoint the Prime Minister should be left to someone other than the monarch. Tingsten (1964, 34) suggested that the responsibility for government formation might be left to someone elected to office - the Speaker for instance. Tingsten also suggested that it would be possible firmly to establish parliamentary supremacy by requiring a vote in the Riksdag before a government could assume power. This could be done by a vote in which the Riksdag declared its support or by a vote in which the Riksdag declared that it tolerated the government. Of these alternatives, Tingsten favoured the latter (1964, 34).⁷

The Second Commission

Initially, the four political parties did not adopt any major reforms on the basis of the commission's report. There were disagreements on a series of institutional arrangements. Instead, a new commission was appointed in 1966 to work out a compromise.₈ This time a compromise was reached and the new commission proposed a number of institutional reforms (von Sydow 1989). These reforms included a revised electoral system, a national threshold of 4 percent for parties to gain representation in the Riksdag, a shortening of the tenure of the members of parliament, the provision that elections at the national and local level be held on the same day, and a one-chamber Riksdag. The reforms took effect as of the 1970 election. At this time the vote on the declaration of no confidence was also included in the constitution (Holmberg 1971). On this issue, the new commission followed the model suggested by the first commission (SOU 1967:26, 174-178) ⁹

After these partial reforms the commission continued to work on a new
constitution. Following Tingsten and other critics of the first commission's
proposal, the government had given the new commission the task of finding

a solution that either gave the Riksdag the right to elect a Prime Minister or gave the right to appoint the Prime Minister to someone other than the monarch. The instructions mentioned the Speaker as the most likely candidate. The instructions also argued that there was no need for a vote of investiture. The Riksdag's power to unseat any government it does not tolerate was seen as a sufficient guarantee against the misuse of the power to appoint a new Prime Minister (SOU 1972:15, 70). Despite this recommendation, four party commission proposed that the Riksdag should vote on the Speaker's candidate for Prime Minster. Why?

Why a Vote?

The issue of the government formation vote was intertwined with the question of the formal powers of the monarch (Hermerén, interview, 1990; Stjernquist, interview, 1990; Johansson, interview, 1990). In 1971 the four parties reached a compromise on the monarch's powers. Of particular importance here is that the compromise removed the monarch from the process of government formation. It was in this context that the four parties agreed on the new voting rule (Bergman 1992).¹⁰

With the monarch removed from government formation, someone else must be responsible for the formation process (SOU 1966:17, 20-32; SOU 1967:26, 165; SOU 1972:15, 94-96). It was seen as neither practical nor desirable to construct a system in which the Riksdag simply elected a Prime Minister. In the first place, there was an obvious risk for tactical manoeuvres that could prove publicly embarrassing. There was also the possibility that a particular party might not want actively to support a certain candidate for Prime Minister though it might be willing to tolerate the candidate. It was generally accepted among the parties that the rules for government formation ought to allow this candidate to form a government. To enable this complex and delicate selection process to work, someone had to be given the responsibility for judging the parliamentary situation and proposing a candidate.

The commission argued that the Speaker was suitable for this task because he or she holds an elected office and could be assumed to be an experienced politician.¹¹ The Speaker is also the foremost representative of the Riksdag and subject to re-election both as Member of Parliament and as Speaker (SOU 1966:17, 31; SOU 1972:15, 94, 144).

On the other hand, the commission argued, it was at least theoretically possible that the Speaker could choose a Prime Minister for partisan reasons. Once the Speaker's candidate became the Prime Minister, he would have a very strong bargaining position. Furthermore, a Prime Minister by the Speaker could dissolve the Riksdag and call a new election.¹² On these grounds, the commission argued, there ought to be a

vote in the Riksdag before a new Prime Minister assumes power. The commission also argued that is was natural that the Riksdag should have the final say in the selection of a new Prime Minister (SOU 1972:15, 94-95).

In reality, however, the voting rule was not simply a "natural" arrangement. was the Conservative Party that most strongly wanted a voting rule. Arne Gadd (interview, 1991), a younger Social Democratic member of the commission, and Hermerén (interview, 1990), the commission's expert on government formation (SOU 1970:16), do not recall any disagreements this issue. In fact, Hermerén (interview, 1991) had been advocating such a rule himself. The chief secretary, Erik Holmberg, on the other hand, believed that no vote was necessary. He considered the Riksdag's right to unseat any government to be a sufficient check on the Speaker (Holmberg, interview, 1990).

Mr. Johansson, a senior Social Democratic member of the commission, explains that the vote never became a big issue because it was readily accepted by the Social Democrats as a part of the compromise package (interview, 1990; Johansson letter). Nonetheless, Johansson (interview, 1990) believes that the Conservative Party representative, Allan Hernelius, saw the voting rule as a bargaining success for himself. Professor Stjernquist (letter 1991), the commission's constitutional expert, confirms the fact that Hernelius was active on this issue. Sterzel, a former secretary of the Riksdag's Standing Committee on the Constitution, also argues that it was the Conservative Party, supported by the Liberal Party, that most strongly favoured the new voting rule (Sterzel 1983, 81; see also Christoffersson 1985, 194).

It is plausible that the Conservatives feared that the Speaker, a position traditionally occupied by a Social Democrat, would give favours to the Social Democrats. In any case, when Hernelius was later criticized for accepting the compromise, he used the new voting rule as argument in favour of the agreement (Hernelius, letter, 1972). The new voting rule, he argued, meant that the Riksdag and not the Speaker had replaced the monarch as the one with the power to select a new government. This reluctancy to give the Speaker the right to appoint a new government explains why the parties proposed a new voting rule. But why was this voting rule formulated in negative terms?

A Negative Vote

The first commission's strong advocacy of the principle of negative parliamentarism 1963:17, 146) was readily accepted both by the second commission and the government (SOU 1967:26, 174-178; Prop. 1968:27, 204; SOU 1972:15, 94-96). In fact, the principle of negative par-

liamentarism was hardly questioned at all by the members of the second commission (Johansson, interview, 1990). The negative principle was also advocated by the chief secretary and the commission's expert on government (Holmberg, interview, 1990; Hermerén, interview, 1990). In principle, according to the chief secretary, a positive voting rule has merit. The idea that the parliament should elect the government is an intuitively appealing one. However, such a positive rule can in some circumstances make it difficult to form a government. Therefore, in internal reports he too advocated the principle of negative parliamentarism (GLB 811 Etapp 2 a 52; GLB 811 Etapp 2 a 77a).

Above all, the four parties did not want to make it more difficult to form a new government. This preference was, and still is, very strong in Sweden. Representatives of all major parties, as well as constitutional experts, stress that it is important that the country have a government and that the government formation process should be as short as possible. They argue that this is why negative parliamentarism is preferable to positive parliamentarism. is also argued that in complicated parliamentary situations, negative parliamentarism allows for flexible solutions such as minority governments. Positive parliamentarism, on the other hand, is said to force "unnatural" majority coalitions between parties with very different policy views and to prolong the government formation process (SOU 1972:15, 94; Prop. 1973:90, 178-179; Bengtsson, interview, 1990; Fiskesjo, interview, 1990; Hermerén, interview, 1990; Holmberg, interview, 1990; Johansson, interview, 1990; Stjernquist, interview, 1990). Thus, the vote was formulated negative terms because the parties preferred the consequences that they associated with negative parliamentarism over those they associated positive parliamentarism. But why did the parties choose an absolute majority threshold?

An Absolute Vote

When the vote on the declaration of no confidence was included in the Swedish constitution, the vote was made absolute to ensure government stability. Both commissions (FU and GLB) and the government argued against letting absenteeism and other temporary circumstances decide whether or not the government was tolerated by the Riksdag (SOU 1963:17, 146-148; SOU 1967:26, 174-178; Prop. 1968:27, 204).

The absolute threshold in the government formation vote was motivated on the same basis. When proposing the absolute threshold, the commission (SOU 1972:15, 95) made explicit reference to the declaration of a vote of no confidence. Both the commission (SOU 1972:15, 79, 95) and the chief secretary, Holmberg (1972, 193-194), stress that the voting rule is a copy of the one used for declarations of no confidence. Because the second

commission did not want to make it more difficult to form a new government,
chose a rule that would make it impossible for a minority in the
parliament to block the Speaker's nominee (SOU 1972:15, 145).13

In sum, the declaration of no confidence voting rule was chosen because it favours a government already in power. The government formation vote was chosen because it favours the Speaker's candidate. In both voting rules, absent and abstaining votes in practice count in favour of one of the alternatives. This design stems from a stronger preference for having a government than for having a government in control of a majority of the votes in the parliament. As long as a government is tolerated by an absolute majority of the members of the Riksdag, it should be allowed to form and to remain in power.

From a Compromise to a Constitution

Although the voting rule passed unchanged through the Riksdag, there was some opposition. The Social Democratic government raised some concerns about the possibility that having a vote on government formation could make it more difficult to form a new government. But with the voting rule being a part of a compromise package, the government ultimately accepted the commission's proposal (Prop. 1973:90, 178-179).

By 1973, the Liberal Party changed its mind on the principle of negative parliamentarism. The Liberal Party representative in the commission, Birger Lundstrom, was no longer a member of the Riksdag. The leading positions on constitutional affairs within the Liberal Party were assumed by Per Ahlmark and Bjorn Molin. Ahlmark, who was to become the next party chairman, and Molin, a well-known political scientist, had not been involved in the compromise on the formation rules and they were critical of the solution that the commission had chosen.¹⁴

In a parliamentary motion Ahlmark and Molin argued in favour of a system based on positive parliamentarism (Motion 1973:1863). The Liberal Party proposed that the support of an absolute majority be required, at least in the early rounds of voting. They also suggested that in a fourth vote any party group with more than one-twelfth of the members of parliament should be able to propose a candidate for Prime Minister. This was intended to break the Speaker's monopoly on the right to appoint the candidates.

The Liberal proposal was referred to the Standing Committee on the Constitution (KU). The committee majority argued that positive parliamentarism the kind proposed by the Liberal Party was likely to prolong the government formation process.¹⁵ The committee also argued that the Liberal Party proposal rested on a distrust of the Speaker. According to the committee, the position of the Speaker was likely to be held only by

the most trusted representatives of the Riksdag. The fear that the Speaker was to misuse his new powers was, for this reason, unfounded. Further insurance beyond the negative vote was not necessary (KU 1973:26, 32-33).

In the Riksdag debate, no other party supported the Liberal Party. The Centre Party argued that the Liberal proposal would create obstacles and prolong the government formation process (Riksdagens protokoll 1973, Nr 110, 64-66; Fiskesjo 1973, 40; Fiskesjo, interview, 1990). The Liberal Party (Ahlmark and Molin in particular) replied that the Centre Party was in favour of a compromise that put too much power in the hands of the Speaker (talmansvålde). The Social Democrats (Johansson) argued that the Liberal fear of the Speaker was completely unfounded and that this fear had no basis in Swedish political experience (Riksdagens protokoll 1973, Nr 110, 69, 91-92, 114-115). Against the votes of the Liberal Party, the Riksdag voted in favour of the negatively formulated government formation rule proposed by the commission. A second and final vote was taken after the next election. The vote was about whether to approve or reject the new constitution in its entirety. All four major parties voted for an approval (Riksdagens protokoll 1974, Nr 30).¹⁶

Since 1975, the Social Democratic leadership has repeatedly called for the abolition of the parliamentary vote on government formation, but without result (Carlsson (DN), 21 September 1991; R&D 2 November 1990; SOU 1981:15). In 1984 the Social Democratic government appointed a new commission (Folkstyrelsekommittén, FK). In its instructions the government argued that the vote should be abolished. According to the government, it is not necessary to have a vote as a check on the Speaker and such a vote can make it more difficult to form a new government (SOU 1987:6, 309). However, the commission disagreed with the government, arguing that there was no real disadvantage in having such a vote. The Speakers have managed the government formation process with integrity. Moreover, if the vote is abolished, after a general election the appointment of a new Prime Minister would probably still have to wait until the Riksdag was in session and a new Speaker was elected. Thus, the government formation process would only be marginally shorter if the vote was abolished 1987:6, 18, 236-242). The voting rule has remained in place.

The real-life outcomes of government formation processes have been consistent with what the commission expected. The overall post-World War II record of Swedish government formation was presented in Table 1. Table 3, by comparison, presents the results of the government formation votes since the new voting rule came into effect. As this table shows, the Riksdag has approved a new government with less than half of the votes cast on three of the eight occasions on which the voting rule has been in effect.

Table 4 shows the size of the governments that have formed after a vote in the Riksdag. Of the eight governments, only two have had a majority of their own. Taken together, the tenure of these two governments was about three and a half years. Thus, for more than 12 years during the period between 1975 and 1991, Sweden was governed by minority governments.

The Riksdag's Standing Committee on the Constitution (KU) has examined one of these government formations from a constitutional point of view. There has been no criticism of the way in which the Speakers have handled the government formation process. The Speaker's candidates have always been accepted on the first vote. Given that the Riksdag must be in

session before the Speaker's candidate can be approved, the committee has also concluded that the government formation process has been as short as possible (see, for example, KU 1982/83:30, 3-5; KU 1991/92:30, 5-7).

Conclusion

To sum up, the members of the second commission agreed on having a vote on the Speaker's candidate for Prime Minister. At the same time the commission did not want to make it more difficult to form a government. Because of this they ruled out a vote derived from the principle of positive parliamentarism. They wanted to stay with the principle of negative parliamentarism, for minority governments, and keep the formation process short. This is why the voting rule was negatively formulated. To ensure that it would be as easy as possible to form a government, the parties chose an absolute majority threshold. This was intended to make it impossible for a Riksdag minority to block the Speaker's candidate.

The outcomes have been the ones that the commission expected. Judging from the lack of criticism, the speakers have always made appropriate choices with regard to the candidates for Prime Minister. And as expected, minority governments have continued to form. Sometimes these have been in control of only a small share of the votes in the Riksdag, and the formation process has been short.

Compared with the formation process before 1975, the addition of the voting rule has had two important consequences. One important difference is that the parties now openly have to show whether or not they will tolerate a new government before it is formed. Because of the voting rule since 1975, doing nothing (i.e. abstaining) can more readily be interpreted by the voters as tacit support for a new government.¹⁷ The other difference is that the members of the Riksdag have become more involved in the formation process. Before the voting rule, the bargaining process was more a matter for the top leadership of the parties. With the new voting rule, the party leaders have to consult more widely to ensure that the members of their own party are willing to vote for or against a particular candidate (Isberg, letter, 1993).

The case of the design of the Swedish voting rule is also important for theoretical reasons. One reason is that this study points to a challenging and important research agenda. The finding that there exists cross-national co-variation between type of parliamentarism, minority governments and the length of the formation process, represents a challenge to coalition theorists. Either this relationship has to be explained away as spurious or a theory explaining why this relationship exists must be developed. In the

tradition of new institutionalism, a focus on the relationship between the rules and the micro-logic of the actors is necessary. The issue to be explained is why political parties more often form minority governments and spend less time forming governments in the context of negative parliamentarism than they do in the context of positive parliamentarism. How do rules interact with motives and strategies to create the observed outcome?

A second result is that constitutional design appears to involve both goaloriented and a process of socialization. In the Swedish case, the actors chose between expected consequences in a way that is best explained from a goal-oriented perspective. Two of the opposition parties, the Conservative and the Liberal Party, were reluctant to transform the largely formal power of the monarch into discretionary power for the Speaker. They wanted a check on a supposedly Social Democratic Speaker. This explains why there is a vote before a new government can assume power. The shared preference for the outcomes associated with negative parliamentarism explains why the parties chose to formulate this voting rule in negative terms.

This raises the issue of how the parties came to share a preference for negative parliamentarism. One answer is that the actors involved in constitutional design experienced a process of socialization. The actors were politicians with high positions in their own parties. They had all made their careers within the existing system. They deliberated on the issue for years. In the absence of obvious drawbacks with the existing system, they saw no reason to change it. Not having participated in debates about the pros and cons of negative parliamentarism, new actors more easily saw merits with positive parliamentarism.

Finally, on a general level, this study illustrates that from the point of view of political actors, rules have important consequences. This contrasts starkly with the lack of attention to rules and institutions that has characterized literature on coalition formation until very recently. It would seem that coalition theorists could still have more to learn from political actors on this point.

---

NOTES

1. In translation, the relevant article of the constitution (The Instrument of Government. Ch 6, Article 2) reads. "When a Prime Minister is to be appointed, the Speaker shall summon for consultation one or more representatives from each party group in the Riksdag. The Speaker shall confer with the Deputy Speakers and shall then submit a proposal to the Riksdag. The Riksdag shall proceed to vote on the proposal, no later than the fourth day thereafter, without preparation in committee. If more than half of the members of the Riksdag vote against the proposal, it is rejected. In all other circumstances it is approved" (Constitutional Documents of Sweden 1990, 47). Formally, the Prime Minister appoints the other government ministers.

---

2. A fourth important issue is the content of the vote. To ensure that it was as easy as possible to form a new government, the formation vote was simply supposed to confirm that the Speaker's proposal was appropriate and not partisan. The new vote was not supposed to involve an evaluation of the proposed government's policy programme since it was believed that this could make government formation more difficult (SOU 1967:26, 174-178; see also; SOU 1972:15, 94-96, 143-146; Stjernquist, letter, 1991). However, when the sitting government passed the commission's proposal on to the Riksdag it added that the vote was, of course, also a vote on the general policy programme of the proposed government (Prop. 1973:90, 178). This statement has led to a controversy over the content of the vote. The Left Party and the Centre Party have been the most consistent on the issue. They have, however, opposing interpretations of the content of the vote. The Left Party argues that it is about the proposed government's policy programme. The Centre Party, by comparison, argues that the vote is on whether the Speaker's candidate is an appropriate choice. The other parties have tended to interpret the vote differently in different situations. Owing to space limitations, I have chosen to focus this study on the design of the voting rule and not on the controversy over whether the vote is about the proposed government's policy programme.

3. March & Olsen (1989) argue (1) that institutions are the most important entities in social life, (2) that social behaviour is largely routine and originates from socialization, (3) that by creating social roles, institutions generate the actors understanding of themselves and (4) that politics is as much about symbols and ways of understanding the world as it is about making choices and decisions (see also Lane 1992; Olsen 1985, 1992). The macro-historical strand has other assumptions about the nature and importance of institutions. Institutions, and in particular the state, are here seen as determinants of social outcomes rather than as constitutive elements of social life.

4. This has actually been a somewhat controversial topic among Swedish constitutional experts. Nyman (1981, 58-59) argues that the Swedish system is one of negative parhamentarism since there is no requirement that a government must have the support of a majority (i.e. win a vote). Ricknell (1975, 116), on the other hånd, argues that because there is a vote before the government assumes power, the Swedish system could be seen as one based on the principle of positive parhamentarism. Between these two positions, Holmberg & Stjernquist (1980, 32, 409) and Westerståhl (1976, 10) have expressed the view that the required vote could be seen as a step towards positive parhamentarism. However, Holmberg & Stjernquist (1980, 205-206, 409) also regard the required vote as a vote designed to determine whether the government is tolerated by the Riksdag rather than as a vote intended to show whether the government is supported by the Riksdag. More recently, consensus has been growing in favour of the classification of the Swedish system as one of negative parhamentarism. Holmberg (interview, 1990) expresses the view that the Swedish government formation process should be seen as one that is based on negative parhamentarism, albeit with a negative vote before the government assumes power. Johansson (interview, 1990), also views the government formation process as an expression of the principle of negative parhamentarism. In their textbook on the constitution, Holmberg & Stjernquist 224) share this view. As evident from this study, I too share this view.

5. I am grateful to Professor Strom (1990) for allowing me to use his data set to calculate the averages for 14 of these countries. For Germany, the averages have been computed from Saalfeld (1990). Not all countries have been parliamentary democracies throughout entire period. Details on this are found in Strom (1990). A new government has been recorded at (1) every general and direct election, (2) every change of Prime Minister, (3) every change in party composition and (4) at any by-election resulting in a shift from majority to minority status, or vice versa (Strom 1990). Non-partisan governments have been excluded.

---

6. The threshold of 40 percent has been chosen because it is one standard deviation from the cross-national mean in Strom's (1990) sample of parliamentary democracies.

7. If not the first, Tingsten was certainly among the first to mention the possibility of letting the Speaker be responsible for government formation. Tingsten's views were highly influential (Stjernquist 1971; Sterzel 1983, 199; Ruin 1990, 100). According to the chief secretary of the second commission (GLB), it was not the case that Tingsten's proposals were explicitly discussed or followed. Rather, the secretary remembers reading Tingsten and believes that most of the other involved actors were familiar with Tingsten's argument (Holmberg, interview, 1991). Thus, the infiuence was indirect.

8. The commission (GLB) had been agreed upon by all four major parties (von Sydow 1989, 231). According to the Conservative representative in the commission (Hernelius, 1972; but see also Svensson 1970, 111). the only objection against the instructions came from the Conservative leader (Yngve Holmberg). He objected to the inclusion of a reference that made the issue of a monarchy versus a republic an issue for the commission to consider. The reference was included.

9. While no government has been forced to resign by a declaration of no confidence. votes have been taken on a couple of occasions. These votes have been more expressions of discontent by opposition parties than genuine efforts to remove governments power (Stjernquist 1989, 266). In February 1990 the Carlsson government resigned voluntarily after having lost an important vote in the Riksdag. However. this was not formally a vote on a declaration of no confidence.

10. The Minutes report the compromise on the powers of the monarch. They also include the agreement on the voting rule (GLB. 16-20 August 1971. author's translation): "At government formation the Speaker consults with the Deputy Speakers and with the party leaders. Thereafter he names a Prime Minister. His proposal is submitted to the Riksdag and is approved, unless more than half of the members vote againsi the proposal. If the proposal is accepted, as soon as possible thereafter the composition of the cabinet shall be reported to the Riksdag. After that, the cabinet assumes power at a meeting with the Head of State".

11. Other alternatives had been mentioned. For example, Bertil Ohlin, a leader of the Liberal Party, once suggested that the task of appointing a Prime Minister could be given to a committee of high judges (Bjorkman, minutes. 1965).

12. A rule in Chapter 3, Article 4, of the new constitution prohibits a government from calling a new election within the first three months after a general election. but this was not considered to be a strong enough insurance against the risk for misuse of the power to appoint a Prime Minister.

13. It is difficult to imagine any alternative to the absolute threshold. Once the partics had agreed on having both negative parliamentarism and a formation voting rule. what choice did they have? A negative and relative voting rule would be a novel invention, but how would it work in practice? This lack of an alternative negative voting rule was not discussed by the commission, however. The absolute threshold was chosen because it was expected to have favourable consequences.

14. As a matter of interest, one can note that after the formation of a Liberal one-party government in 1978. and only a few years after the Liberal Party had voted in favour of a system of positive parliamentarism, Molin reportedly argued that the Swedish system based on tolerance rather than support had proven itself to work well (UNT. 17 October 1978). On the other hånd, Bohman, the Conservative Party leader. whose Conservative Party had voted for the negative and absolute rule, reportedly expressed the view that it would probably have been better if Swedish governments were formed on the basis of a system of positive parliamentarism (SvD, 26 January, 1979).

15. In addition, the majority argued, the Liberal proposal to let each party appoint a candidate in the fourth vote could lead to governments with only limited parliamentary support. This is a somewhat curious argument, however, since it was because negative parliamentarism was associated with the possibility of such governments that the committee favoured a negative voting rule.

---

16. The fifth and smallest party with parliamentary representation, the Left Party-Communists, not represented in the commission. In the final vote the party voted against the new constitution (Riksdagens protokoll, 1974, Nr 30).

17. This is what happened when the Liberal Party formed a government in 1978. This government only controlled 39 out of 349 seats in the Riksdag. Because of the voting rule, the Social Democratic Party publicly had to explain why it abstained and thereby allowed the Liberal government to form. When discussing this case, two leading political scientists both concluded that under the old rules, the King would, in the end, have appointed the same government (Westerståhl, SvD, 12 October 1978; Petersson 1979, 144-146). However, the fact that the Social Democrats had publicly to abstain to allow the Liberal government to form made the government formation process much more difficult (Petersson 1979).

ACKNOWLEDGEMENTS

Research for this article has given me the opportunity to meet many interesting politicians and political scientists. I thank all of them for their valuable help. Meeting them made the research both educational and fun. Two anonymous reviewers also provided helpful advice. A special thanks to CK. Remaining errors are definitely my own responsibility.

REFERENCES

Published Sources

Bergman, T. 1992. "Multiple Goals and Constitutional Design: How the Swedish King Lost
His Formal Powers", Statsvetenskaplig tidskrift 95, 209-232.

Bergman, T. 1993. "Formation Rules and Minority Governments", European Journal of
Political Research 23, 55-66.

Budge, I. & Laver, M. 1986. "Office Seeking and Policy Pursuit in Coalition Theory",
Legislative Studies Quarterly 11, 485-506.

Carlsson, I. 1991. "Fel att rosta om statsminister", Dagens Nyheter (DN), 21 September.
Christoffersson, U. 1985. "Debatten om forfattningen", in Folkstyrelsekommittén, Makten
Från Folket. Stockholm: Liber.

Constitutional Documents of Sweden 1990. Stockholm: The Swedish Riksdag.
Evans, P. R., Rueschemeyer, D. & Skocpol, T., eds. 1985. Bringing the State Back In.
Cambridge: Cambridge University Press.

Fiskesjo, B. 1973. "Monarki eller republik", in Back, P.-E., ed., Modern demokrati: Problem
och debattfrågor, sth edition. Lund: Gleerups.

Holmberg, E. 1971. Crundlagarna i lydelse av 1 januari 1971. Stockholm: Allmånna Forlaget
(Publica).

Holmberg, E. 1972. "Grundlagberedningens forslag till ny RF och ny RO", Statsvetenskaplig
Tidskrift 75, 189-208.

Holmberg, E. & Stjernquist, N. 1980. Grundlagarna. Stockholm: P. A. Norstedt & Soners
forlag.

Holmberg, E. & Stjernquist, N. 1992. Vår forfaiming. Stockholm: Norstedts Juridik. KU 1973:26. Konstitutionsutskottets betånkande nr 26 år 1973. Stockholm: Riksdagen. KU 1982/83:30. Konstitutionsutskottets betånkande nr 30 år 1982/83. Stockholm: Riksdagen. KU 1991/92:30. Konstitutionsutskottets betånkande nr 30 år 1991/92. Stockholm: Riksdagen. Lane, J.-E. 1992. "Interests and Institutions", Statsvetenskaplig tidskrift 95, 151-161. Laver, M. & Schofield, N. 1990. Multiparty Government: The Politics of Coalition in Europe. Oxford: Oxford University Press.

March, J. G. & Olsen, J. P. 1989. Rediscovering Institutions: The Organizational Basis of
Politics. New York: The Free Press.

Motion 1973:1863. Motion Nr 1863 av herr Molin m. fl. Stockholm: Riksdagen.
Nyman, O. 1981. Parlamentariskt regeringssått. Stockholm: Bonnier Fakta.
Olsen, J. P. 1985. "Nyinstitusjonalismen og statsvitenskapen", Statsvetenskaplig tidskrift 88,

1-14.

Olsen, J. P. 1992. "Nyinstitusjonalismen: En kommentar til Bjørn Erik Rasch", Norsk
statsvitenskapelig tidsskrift 8, 148-153.

Petersson, O. 1979. Regeringsbildningen 1978. Stockholm: Rabén & Sjogren.
Prop. 1968:27. Kungl. Maj:ts proposition nr 27 år 1968. Stockholm: Riksdagen.
Prop. 1973:90. Kungl. Maj:ts proposition nr 90 år 1973. Stockholm: Riksdagen.

Przeworski, A. 1988. "Democracy as a Contingent Outcome of Conflicts", in Elster, J. &
Slagstad, R., eds., Constitutionalism and Democracy. Cambridge: Cambridge University
Press.

R&D 2 November 1990. "Socialdemokraterna vill ha långre valperioder". Från Riksdag och
Departement.

Ricknell, L. 1975. Hur vart statsskick fungerar. Stockholm: Prisma.

Riker, W. H. 1962. The Theory of Political Coalitions. New Haven: Yale University Press.
Riker, W. H. 1980. "Implications from the Disequilibrium of Majority Rules for the Study
of Institutions", American Political Science Review 74, 432-47.

Riker, W. H. 1982. Liberalism Against Populism. San Francisco: W. H. Freeman.
Riker, W. H. 1986. The Art of Political Manipulation. New Haven: Yale University Press.
Riksdagens protokoll 1973. Nr 110. 5 June. Stockholm: Riksdagen.

Riksdagens protokoll 1974. Nr 30. 27 February. Stockholm: Riksdagen.
Riksdagens Protokoll. Various issues 1976-91. Stockholm: Riksdagen.

Ruin, O. 1990. "Statsministeråmbetet: från Louis De Geer till Ingvar Carlsson". in Departementshistoriekommittén,
styra riket. Stockholm: Allmånna Forlaget.
Saalfeld, T. 1990. "Efficient but Powerless? West Germanys Bundestag After 40 Years".
Paper, ECPR Joint Sessions of Workshops, Bochum.

Shepsle, K. A. 1986. "Institutional Equilibrium and Equilibrium Institutions", in Weisberg, H. F., ed., Political Science: The Science of Politics. New York: Agathon Press. Shepsle, K. A. 1989. "Studying Institutions: Some Lessons From the Rational Choice Approach", Journal of Theoretical Politics 1, 131-147.

Skocpol, T. 1979. Slates and Social Revolutions: A Comparative Analysis of France, Russia,
and China. Cambridge: Cambridge University Press.

SOU 1963:17. Forfattningsutredningen. Del 2. Motiv. Stockholm: Statens Offentliga Utredningar.

SOU 1966:17. Arbetspromemorior i forfattningsfrågan. Stockholm: Statens Offentliga Utredningar.

SOU 1967:26. Grundlagberedningen. Partiell forfatningsreform. Stockholm: Statens
Offentliga Utredningar.

SOU 1970:16. Grundlagberedningen. Riksdagsgrupperna - Regeringsbildningen. Stockholm:
Statens Offentliga Utredningar.

SOU 1972:15. Grundlagberedningen. Ny regeringsform. Ny riksdagsordning. Stockholm:
Statens Offentliga Utredningar.

SOU 1981:15. Grundlagskommittén. Grundlagsfrågor. Stockholm: Statens Offentliga Utredningar.

SOU 1987:6. Folkstyrelsekommittén. Folkstyrelsens villkor. Stockholm: Statens Offentliga
Utredningar.

Statistisk årsbok 92 (Statistical Yearbook of Sweden) 1992. Stockholm: Statistiska Centralbyrån.

Sterzel, F. 1983. Parlamentarismen i forfattningen. Stockholm: P. A Norstedt & Soners
forlag.

Stjernquist, N. 1971. "Grundlagberedningskompromiss i statschefsfrågan". Statsvetenskaplii>
Tidskrift 74, 377-379.

Stjernquist, N. 1989. "Vad avsåg man? Hur blev det? Den svenska regeringsformens inkorningsperiod",
Festskrift tillågnad Lolo Krusius-Ahrenberg. Helsingfors: The Swedish
School of Economics and Business Administration.

Strom, K. 1990. Minority Government and Majority Rule. Cambridge: Cambridge University
Press.

SvD. 1979. "Gosta Bohman i SvD-intervju: Trepartiregering trolig vid borgerlig valseger",
Svenska Dagbladet (SvD), 26 January.

Svensson, S. 1970. Enkammarriksdag. Stockholm: Almqvist & Wiksell.

Tingsten, H. 1964. Skall kungamakten stårkas? Stockholm: Bokforlaget Aldus/Bonniers.
Tsebelis, G. 1990. Nested Games: Rational Choice in Comparative Politics. Berkeley: University
California Press.

UNT. 1978. "FP-riksdagsman efter regeringskrisen: Utred bristerna i regeringsformen",
Uppsala Nya Tidning (UNT), 17 October.

von Sydow, B. 1989. Vågen till enkammarriksdagen. Stockholm: Tidens Forlag.

12

October.

Unpublished Sources

Bengtsson, I. 1990. Interview conducted by the author on 11 December.

Bjorkman, F. 1965. Minutes from a meeting among the party leaders on 13-14 October.
Fiskesjo, B. 1990. Interview conducted by the author on 13 December.
Gadd, A. 1991. Interview conducted by the author by telephone on 23 October.
GLB, 16-20 August 1979. "Protokoll" (Minutes). Grundlagberedningen.
GLB, 811 Etapp 2 a 52. "Statschefens stallning i en parlamentarisk demokrati". Grundlagberedningen.
February 1970.

GLB, 811 Etapp 2 a 77a. "Principskiss till system for omrostning i riksdagen i samband med
regeringsbildning. Grundlagberedningen. 12 December 1970.

Gustafsson, A. 1992. Interview conducted by the author on 8 September. Hermerén, H. 1990. Interview 1 conducted by the author on 12 December. Hermerén, H. 1991. Interview 2 conducted by the author on 1 October. Hernelius, A. 1972. Letter to Per Unckel dated 26 July.

Holmberg, E. 1990. Interview 1 conducted by the author on 21 December.
Holmberg, E. 1991. Interview 2 conducted by the author by telephone on 6 October.
Isberg, M. 1993. Letter to the author dated 15 July.

Johansson, H. 1990. Interview conducted by the author on 10 December.
Johansson, H. 1991. Letter to the author dated 18 November.

SAP-PS, 26 October 1965. Sveriges Socialdemokratiska Arbetareparti, Partistyrelsens Protokoll.

Stjernquist, N. 1990. Interview conducted by the author on 5 December.
Stjernquist, N. 1991. Letter to the author dated 23 September.
Westerståhl, J. 1990. Interview conducted by the author on 20 December.