Resumé

For many years, scholars have sought to discover why governments in Western European parliamentary frequently dissolve before the time constitutionally mandated for the holding of elections. Recent research has suggested the potential importance of unpredictable; changes in the political environment for bringing about governmental dissolutions. Although not specifically described, the importance of'critical events' is suggested by both aggregate patterns; of dissolutions and anecdotal accounts of particular governments.

It is our purpose in this paper to develop further this critical events perspective and evaluate: empirically actual governmental histories in light of these developments. Two specific goals are: pursued. First, we delineate categories of critical events, thereby grounding the concept of a critical within existing knowledge of governmental stability. Second, we apply this categorical schema of events to the dissolution of Scandinavian governments during the years 1945 -1980 In so doing, we seek not only to assess the adequacy of our conceptualization of events, but also to provide a theoretically and substantively accurate account of the death of modern Scandinavian

A Critical Events Perspective

Models of Governmental Dissolution

Much of the past research on the durability and dissolution of governments may be characterized as a protracted, though not entirely cumulative, search for factors determine how long governments will survive. Taking as a starting point Lowell's (1896) assertion that single-party governments are more stable than coalition governments, Taylor & Herman (1971) and Dodd (1974, 1976) attempted to isolate those aspects of governing situations which might explain the considerable variability in the duration of governments. (See also Sanders & Herman 1977, Brass 1977, and Warwick 1979.)

Although these studies did, in fact, discover correlates of governmental longevity,the

---

* The three authors share equal responsibility for this article.

vity,theoverall result of this research was disappointing. A number of attributes, both of governmental actors and parliamentary settings, were found to be related to governmental duration. Factors such as the number of governmental and parliamentaryparties their strengths were suggested as indirect determinants of governmental duration through their impacts on the bargaining proclivities of partiesand sizes of resulting coalitions (e.g., see Dodd 1976). However, these studies proved unable to explain more than about 30 percent of the variance in governmental duration, a disheartening outcome given the more than ten years of empirical research in this area.

Recent analyses suggest that these older models have reached an upper limit of predictability, in part because they fundamentally misspecify the mechanics of governmental dissolution (Browne et al. 1984 b). These older models begin by assuming that governments possess an inherent amount of stability, which usually termed durability. Presumably, this durability depends on the willingness of governing partners to continue with their governing agreement. By and large, this willingness is assumed to be structured by conditions which are operative at the time of the government's formation. At the same time, each government has associated with it a specific tenure, that is, a number of units of time through which it has endured, which might be termed the government's duration. Older models of governmental dissolution assume that a close fit exists between governmental durability and duration. As a result, factors which structure the (inherent) durability of a government should also display a strong relationship with the government's duration, in effect determining the timing of dissolution.

More recent analyses have questioned the assumption of a close fit between governmental durability and governmental duration (see Browne et al. 1982, 1984 a). As an alternative, it is suggested that governmental dissolution is brought about by the occurrence of some critical event which destroys the willingness at least one governing partner to continue the partnership. Examples of such dissolution-precipitating events would include scandals, foreign policy crises, downturns in the economy, or decisions by some coalition partner that the moment is propitious for a change from a cooperative to a competitive strategy. The vital point about such occurrences, however, is that their timing is unrelated to the durability of a given government. The duration of a government, i.e., the timing of its dissolution, is largely determined by the occurrence of a critical event and not by its inherent durability.

For our model, we conceive governmental duration to be the result of a decisionprocess. of a government may be thought of as a series of decisionsby various governmental actors (e.g., cabinet partners) to continue supporting an existing agreement to govern. Governments must continually respondto placed upon them by the environment, the opposition, and governing partners, and their responses may take on a range of values from completesatisfaction total rejection of the demands. Dissolution is the ultimate decision to reject a demand that either cannot or will be satisfied in any effective

way. Such a decision can be either voluntary, with the government resigning, or
mandatory, with the government losing confidence and being forced from
power.

Seen in this light, the survival of governments is considered to be analogous to the survival of structures and mechanical devices which are subjected to stress and wear over time. The branch of probability theory known as reliability theory offers a variety of statistical failure models developed to describe such occurrences for example, Mann et al. 1974, Ch. 4). The general form of such models; is to assume that time-to-failure (or duration, in our terms) is a random variable with particular distribution and probability density functions. The key element for modeling the failure (or dissolution) process is to determine the functional form of the failure rate.

For the problem in question, an appropriate failure-rate function is one which describes the occurrence of shocks (critical events) which are spread randomly through time with a constant probability of occurrence in a fixed time interval. These empirical requirements are met by the mathematical properties of Poisson processes - in particular, the exponential distribution (see Mann et al. 1974, 129-125). A model of governmental dissolution, based on the expectation that dissolutions are produced by critical events which are distributed in an exponential through time, predicts an aggregate pattern of dissolutions rather than the occurrence of any particular dissolution (see Brown et al. 1984 b). Separate analyses of twelve Western countries (Brown et al. 1982) and Italy (Cioffi-Revilla have observed the aggregate dissolution patterns suggested by such a model, lending support to the idea that critical events, not inherent durability, are: the proximate causes of governmental dissolutions.

A major implication of these research findings is to suggest that further research be focused on the nature of the events which precipitate governmental dissolution. to this point, researchers have focused on systemic and parliamentary conditions at the point of cabinet fonnation to explain dissolution. It now seems more appropriate to identify what sorts of conditions specifically trigger dissolutions to theoretically relate these events to systemic, parliamentary, and intra-governmental As a first step, it is necessary to delineate important classes of critical events, classes which might be suggested by extant theoretical formulations and historical accounts. It is to this task that we now turn.

Classes of Critical Events

We have indicated that governmental dissolutions are associated with the occurrenceof events, the timing of which are not completely conditioned by the: inherent durabilities of governments. On a conceptual level, we can distinguish between critical events of sufficient severity to induce a governmental dissolution (terminal critical events) from those events which do not bring about such a dissolution(non-terminal events). In this paper we limit our concern to the descriptionof

scriptionofthe former class, terminal critical events. While it would clearly be advantageous to know the relative frequency of both terminal and non-terminal events of various sorts, it is much easier to identify and describe the events which brought about governmental dissolutions than it is to identify and describe those which did not.

We would suggest that there are several broad conceptual categories into which important attributes of terminal critical events may be grouped. The first of these categories is what might be described as the temporal nature of the event. A review of historical and popular accounts of governmental dissolutions suggests that terminal events can take on two rather distinct forms. In some instances, these events appear to be unitary in character, that is, a single environmental phenomenon to have been a sufficient condition for a change in government. An example might be the death of the prime minister, as was the case with the Benediktsson government in Iceland. For other governments, however, the terminal appears to be serial in character. Here, two or more events together appear be necessary for dissolution. This is not to be confused with a series of non-terminal events followed by a terminal event. The distinguishing characteristic of a serial terminal event is that the joint influence of the events in seriatim comprises necessary condition for governmental dissolution.

A second broad category of important attributes of terminal events is the type of activity represented by the event. An obviously important group of events consists elections, both regular and unscheduled (or extra-ordinary). The fact that many governments end because of recourse to the electorate may seem both natural obvious, but the frequency of this type of event is not unimportant. The length of time parliamentary actors expect to be in power (or in opposition) is likely to weigh heavily in their strategic calculations. The extent to which their tenures are actually bound by these electoral limits is an important empirical question.

In addition to elections, terminal events may take the form of a variety of other activities. Some may be relatively unrelated to partisan political activities, for example, scandals, international events, or the health of a key minister. In other cases, the event may be distinctly political, such as voter referenda, strikes, and inter- and intra-party disputes that alter the ability of governing parties to continue governing. We have no a priori expectations as to the relative frequency of each of these types of events; our present goal, it should be stressed, is to document the nature of terminal critical events so as to provide a baseline for further analyses of these processes.

A third category of important attributes, related to the second, is the policy arena to which the terminal events relate. Here, the most important category would be economic issues. Economic issues define one of the basic cleavages in European systems; in Scandinavia, this cleavage is the dominant one (Särlvik 1983). An interesting question to be probed is whether terminal events in Scandinaviaare from this dominant political cleavage, or whether the forging of a

government agreement basically resolves economic issues, leaving subsidiary issuedimensions
as ethnicity or the environment to generate the disputes that
disrupt the political environment and bring governments down.

The final general attribute category relates to the source of the actions which comprise the terminal event. At a gross level, we can distinguish between cabinet action and parliamentary action. Some governments end because of insoluble conflicts among cabinet members (at times, even over such things as personality conflicts). On the other hand, governments may end because of action by the parliament, in the case of a back-bench revolt that results in a no-confidence vote. A similar distinction may be made between governmental parties and opposition parties as the sources of terminal events. Finally, in some cases the actions of one individual actor may result in governmental dissolution. The prime minister is typically only single actor with the ability to bring about this result, although in some cases the constitutional head of state may also wield this degree of influence.

To reiterate, in this section we have suggested four broad dimensions which can be used to identify the significant features of terminal events. These are the temporal nature of the event, the type of activities represented by the event, the policy arena to which the event pertains, and the systemic source of the event. As our efforts are exploratory, we have suggested a variety of different ways to characterize along these four dimensions. By applying this conceptual schema to the Scandinavian setting, we hope to provide a fuller understanding of the conditions which governments fall in these nations. On a theoretical level, we see this as a first step toward integrating these phenomena into a more comprehensive of parliamentary governance, both in Scandinavia and in parliamentary generally.

The Scandinavian Setting

The countries that are of substantive interest to us in this paper are those which are commonly included in the Scandinavian group: Denmark, Finland, Iceland, Norway, and Sweden. Examination of this group of countries offers several advantages a comparative analysis. First, the nations possess a relatively high degree of similarity on a number of traits, including geography, religion, ethnicity, and a common political history. Although there are important differences among these nations, their political systems exhibit many common traits. All five nations are multi-party parliamentary democracies with at least some recent experience with coalition governments (Pesonen & Thomas 1983). All have party systems with a significant left-right cleavage that essentially defines socialist and nonsocialist blocs (Särlvik 1983). With some significant modifications, these nations generally display a five party pattern,¹ best exemplified by the Swedish system (see Pesonen & Thomas 1983, 59-73).

At the same time, there are some important differences between these systems. Chief among these is the strong constitutional head-of-state (president) in Finland, position which is likened to that of the Presidency in Fifth Republic France, and is not seen in the other Scandinavian systems (Särlvik 1983). In addition, party systems in Denmark and Iceland exhibit much less stability than in the other systems, and (at least in Denmark) there is a corresponding higher frequency of governmental formations and dissolutions (see Browne et al. 1983).

Selection of this country grouping thus corresponds to what has been termed a 'mixed systems' strategy for comparative analysis (Frendreis 1983). These five political systems are neither completely dissimilar nor completely alike. Possible causal linkages may be identified through a variety of comparisons, depending on the patterns of dissolution observed. Differences between dissolution patterns in Sweden and Norway will generate different hypotheses than differences between and Sweden or between Norway and Iceland.

Data have been collected concerning the formation, dissolution, and salient attributes every government in each of these five countries during the period 1945-1980, a total of 93 governments. The only exception to this rule is the exclusion governments which existed for some reason other than governing (e.g., caretaker or expeditionary governments).² Information on each government has been collected from a variety of sources, with final appeal to Keesing's Contemporary

Analysis of Dissolutions in Scandinavia, 1945-1980

There is considerable variability in the number of post-war governments observed these five nations. (For a listing of the governments, see the Appendix.) While Iceland, Norway, and Sweden have all witnessed 15 governments during this 35 year period, the other two nations have had greater governmental turnover, Finland displaying nearly twice as many governments. This translates into an average governmental tenure of 15 months in Finland versus 28 months in the other three nations (with an intermediate average of 21 months in Denmark). passing we note that the frequency of dissolution is related to the adequacy the stochastic 'critical events' model discussed above. In our earlier analysis et al. 1982) the best goodness-of-fit between the exponential model and actual experience was in Finland, followed by Denmark, Norway, Sweden, and Iceland (in that order).

As noted above, we have identified the critical event associated with the dissolutionof 93 governments. We have attempted to characterize these terminalevents respect to the attribute dimensions delineated earlier. The Appendixcontains description of the procedures employed in classifying terminal events. In general, the categories are not mutually exclusive; an event which emanatesfrom intra-party dispute over economic matters and results in a parliamentaryvote

mentaryvoteof no-confidence and a new election will result in several table entries.In events vary with respect to the clarity of their nature. For example, in some cases it is clear which actors brought on the dissolution, but the reasons for their actions are unclear.

Temporal Nature of Events

The first attribute dimension considered is the temporal nature of the terminal
event (Table 1).

Overall, slightly more than two-thirds of the terminal events are unitary in nature, with this class of events being dominated by constitutionally mandated elections. Half of the unitary terminal events (and over one-third of all terminal events) are scheduled elections. There is considerable variability by country in the relative frequency of unitary terminal events, ranging from a high of 87 percent in Norway a low of 57 percent in Finland. Interestingly, in the two countries with the greatest number of governments, about half of the dissolutions do not result from the occurrence of a single (unitary) event, but from a serial terminal event.

Serial event dissolutions are much more complex in their nature and structure. The number of component events varies from two to thirteen, with the frequency for each magnitude of serial event declining as the number of component events in the series rises. The complexity of a serial terminal event is illustrated by the dissolution of the Miettunen II government in Finland. A cabinet government was formed on the basis of a two-month trial period, and almost immediately the parties in government were in conflict over a tax increase to support an employmentprogram. expanded to include contention over wage and price restraints, and after nine months the government made a decision to continue governing by agreeing to accept a new economic package. Dissension continued, however, and two months later the prime minister demanded the support of the government parties for a tax proposal. When the Communists refused, Miettunenresigned, to have the president refuse to accept the resignation. In responsethe proposed a compromise permitting the Communists to vote

against the tax proposal but support the government, and the subsequent budget was passed, minus the offending proposal. However, when the next budget was proposed, the Communists again threatened to oppose sections of it, prompting a counter-threat from the Social Democrats. Miettunen again offered his government'sresignation, this time the president accepted it.

Clearly it is a subjective judgment to argue that these events represent a single series of events which jointly produced the fall of the Miettunen government. In some sense the Communist opposition to the second budget might be viewed as the beginning of a much shorter serial event. However, it does appear that this series the sufficiency to explain the dissolution without the sequence of 'on again-off again' support that was the history of the Miettunen II government. In any case, the history of this government clearly illustrates how some governmental result from a complex interaction of several forces, some of which may be present far before the actual dissolution.

Type of Activity of Events

The second attribute dimension considered is the type of activity represented by the terminal event. Table 2 describes the distribution of various types of terminal critical events in the five nations. (In this and subsequent Tables, classes of events which represent less than five percent of all terminal events are not displayed).

As noted earlier, regularly scheduled elections are the most common single cause of dissolutions,³ although Denmark and Finland are distinguished by far lower percentages for this category. The calling of unscheduled elections is the next largestclass terminal events, meaning that over half of all dissolutions are associatedwith In both Denmark and Iceland, the decision to dissolve a government is more likely than not to be associated with a call for a new election rather than an attempt to establish a new governing formula. (Eight of the fifteen

Danish governments which do not end in a regular election are terminated by an
event which includes the calling of unscheduled elections; the corresponding figurefor
is five out of nine governments.)

No other single type of event constitutes ten percent of terminal events. Few governmental dissolutions are evidently due to international events, a perhaps surprising result given the level of European interdependence seen during this period. Intraparty disputes are capable of disrupting governments, although not particularly often. (For example, the Karjalainen Ha government in Finland was brought down by the actions of the Communists in response to their own internal philosophical conflict between orthodox Communists and a more liberal party faction.)

Of some interest are the types of events not listed in Table 2. Despite their apparent drama, strikes are relatively unimportant in triggering governmenta] collapse in Scandinavia, comprising only three percent of all terminal events.⁴ Voter referenda and subnational elections comprise a similar percentage, indicating direct voter action is seldom the proximate cause of dissolution. While dramatic, scandals are the proximate causes of only two governmental dissolutions these countries during this period. Perhaps of most interest to students of coalition politics is the often discussed instance where a government falls due to a change in party strategy from being a governmental partner to an electoral contestant. only one instance in these data does a government's fall appear to be; attributable to such a strategy: Denmark's Krag II government resigned two years before scheduled elections in order to take advantage of a split in the conservative

In terms of their unpredictability, ministerial ill-health and scandals involving governmental officials most closely approximate the exponential model's re quirements for random disturbances, and together these two types of events represent one-fourth of the non-electoral terminal events. Scandals may result both current and past behavior: Norway's Borten II government resigned over the leak of a confidential state paper concerning projected EEC membership, while Finland's Sukselainen II government ended with the prime minister's resignation after he was found guilty of'administrative irregularities' in his capacity as director-general of the State Pensions Institute. 111-health (or death) of a key government official is similar to scandals in terms of lack of predictability. the health of the prime minister is involved, as in the end of the Benediktsson government in Iceland following his accidental death, although in one instance the death of a president and his replacement by the prime minister forced a government to dissolve and be reconstituted.

Policy Arena of Events

Related to the type of activity constituting a terminal event is the policy arena in
which these activities occur. Clearly, economic issues are of great importance in
this regard (Table 3).

The economy is the only general policy category containing more than five percent terminal critical events; the 37 percent of events falling in this category represents over one-half of all dissolutions not brought on by regularly scheduled elections. Perhaps the most interesting results in Table 3 are the surprisingly low percentages of economic events in Norway and Sweden. Conflict in the economic domain in these nations is apparently more likely to be seen in the context elections and government/opposition clashes than among members of the government. By inference, in these two countries issues which do not directly relate the societal economic cleavage are evidently more likely to disrupt the governing environment in such a way as to trigger a crisis and an eventual governmental This finding is even clearer for the economic subcategory of union demands, which are not seen in any of the terminal events in Norway and Sweden.

Systemic Source of Events

The final attribute dimension of interest relates to the systemic source of the terminal
(Table 4).

As we have indicated elsewhere (Browne et al. 1983), dissolution of the government parliament can result from the actions of either the government itself or the parliament. Both types of these terminal events are seen with moderate of frequency, although independent cabinet action is not seen in Iceland and independent parliamentary action is absent in Sweden. When the actions of specific parties can be pinpointed as the source of terminal events, it is about twice as likely that the actions will emanate from governmental parties than from the opposition. This clearly reflects the fact that Scandinavian governments are more likely to break apart from within than be brought down from without; opposition by one coalition partner to the actions of the government is the typical pattern in this regard.

The strength and central role of the prime minister in these parliamentary systems is underscored by the fact that one-third of the terminal events involve actions by that actor. Independent action taken by the prime minister seems to illustrate different sets of systems among these nations with respect to the dissolution In Denmark and Finland, the prime minister is an active participant 40 percent or more of the dissolutions, whereas in Norway, Sweden, and Iceland the prime minister is independently involved far less often. Overall, the fact that the prime minister is critical to the continuation of the government may be contrasted with the rather limited involvement of the head of state, who is involved only 13 percent of all terminal events. There is clearly a structural country in this regard, as over ninety percent of these cases occur in Finland. In the Finnish system, the president's role is often a stabilizing one in which he prevents hasty resignation of some governments, precludes the calling of unscheduled or expresses a preference for a particular change in the governing formula, usually toward majority or supermajority status. (Such was the case with the Karjalainen I and Ha governments.)

Comparison of the Countries

In conclusion, several points of comparison can be made with respect to these Scandinavian systems. Of the five systems, Norway and Sweden display a marked degree of similarity. Both countries are relatively stable systems in which regular elections comprise the majority of terminal events. Economic issues are unlikely to be associated with dissolution, and no single type of activity or actor appears responsible for a large number of governmental dissolutions. The only exception to this generalization relates to the role of the prime minister, who Is somewhat more active in Norway than in Sweden.

At the other extreme lie Denmark and Finland, which are far less stable and much less likely to have regularly scheduled elections as terminal events. These two nations are also distinguished by the heavy involvement of the prime minister in dissolutions and by the prevalence of economic issues in their terminal critical events. Somewhat between these two sets of countries lies Iceland, which resemblesthe

semblestheformer group in the frequency of regular elections as terminal events, but also has a higher frequency of economic events and parliamentary activity. Other Icelandic peculiarities include an unusual pattern of actions by governmentalparties induce dissolution, generally including a call for new (unscheduled)elections.

Doubtless there are certain structural and historical reasons for these divergent patterns. Swedish and Norwegian stability is partly explained by their large numbers single-party majority governments during this period (see DeSwaan 1973,254-275), while Danish governments, coalition or single party, have often not attained majority status. The constitutional innovation of a strong, independent has clearly had important consequences for the Finnish case, in effect fostering the limited stability the system displays. Great variability exists among the five nations with regard to traits which are governed by constitutional limits, like the frequency of unscheduled elections as terminal events.⁶ Where constitutional or statutory restrictions are not operative, as in the case of scandals and ministerial ill-health, the five countries display more similarity.

Conclusion

In the preceding sections, we have sought to develop a useful set of attribute dimensions describing events which result in governmental dissolutions and to apply this schema to the post-war Scandinavian setting. The conclusions that we draw from these efforts can be organized into two groups: general theoretical and research implications and observations about the five systems we have examined.

With regards to theoretical and research implications, it is clear that there exist important systemic differences, even among a set of systems which display many structural and historical similarities. Three important systemic differences appear in our analysis. First, the importance of economic issues displays surprising across systems. Although it might be suggested that economic conflicts are less severe in Sweden and Norway, the electoral histories of these nations indicate clearly that this is not the case. Rather, it would appear that in these systems institutional mechanisms have been created that channel conflict over the basic issue cleavage into arenas such as elections and the government formation process rather than into the actual governing context.

Two other important systemic differences observed are the variable propensity to call unscheduled elections and the differing systemic sources of terminal events. Although some structural reasons for these differences are already apparent, research should strive to establish more clearly the historical, cultural, (additional) structural bases for these variations across systems.

In addition, it is apparent that two types of events may disrupt the political environmentand about governmental dissolutions. One type of event is essentiallyunrelated other political variables (e.g., ministerial ill-health), while the other type is, in fact, grounded in the political context (e.g., inter-party economic

disputes). Both types of events occur with non-trivial frequency, and both need to be integrated into extant theories of the governmental process. The random qualityof unrelated type of event heightens the uncertainty of bargaining situations;this of uncertainty has not received sufficient theoretical attention. Events which are more clearly political must be related to systemic, parliamentary,and variables. Although we have offered some exploratory observations of such linkages, they must be more systematically examined.

A final research implication concerns the extent to which governments survive various sorts of critical events. By focusing on terminal critical events we have highlighted som important trends, but we have left blank the nature of non-termi nal critical events. As an example, we noted that strikes comprise only three percent terminal events. A detailed analysis of each country's weekly events may reveal that strikes occur frequently, but do not bring governments down, or alternatively, strikes in these systems are not very common, but do often result in governmental dissolutions. Each alternative is plausible and consistent with our data, but they suggest different conclusions about the impact of strikes on governments. of the relative balance of terminal versus non-terminal critical within each attribute category awaits a much more extensive and detailed collection effort.

The analysis results have also provided fairly enlightening portraits of the five Scandinavian political systems. Much of this has been discussed above, but four points bear further emphasis. First, the absence of economic issues from terminal events in Norway and Sweden is striking. Given the general relevance of the economic in all of these countries, we feel that the relative absence of this issue from this particular aspect of the political process in these two nations is noteworthy. Second, the essential absence of unscheduled elections in Norway and Sweden is also quite interesting. It is likely to be the case that the reasonable expectation (for political actors) that the electoral balance will remain unchanged for a known period of time has numerous effects in these two countries. Third, despite many similarities, the five nations are quite different in the roles played by their executive officers (prime ministers and heads-of-state) in the dissolution process. Norway, Sweden, and Denmark all have an active role for the prime minister, while Finland has two major actors, and Iceland has none. Fourth, comparing these five systems, our overall impression is one of a stable Nordic core of Sweden and Norway, with instability increasing as one moves further into the periphery from this more homogeneous center. Whether it is from ethnic diversity systemic penetration by powerful neighbours, the other three system:? display greater governmental turnover, more irregularity, and a greater frequency of basic conflict within the government.

We began this paper by noting that existing analyses have given us only limited understanding of governmental dissolutions, despite their contributions toward understanding other aspects of the governmental process. It has been our intention to provide a more complete view of governmental dissolution in Scandinavia by

focusing on the critical events which recent work has suggested are crucially importantin
process. We hope that subsequent work, both in Scandinavia and
elsewhere, finds some merit in this approach.

---

NOTES

1. Minority ethnic parties and splits within older parties have increased the number of parties beyond five in the various systems, but the party system structure still roughly corresponds to the classic five party system.

2. Governments whose tenure began before 1980 and which were still in office were also excluded, since they do not represent a relevant case (of dissolution) for our analysis.

3. Regularly scheduled elections (i.e., those constitutionally mandated) are the main exception to the rule that the various categories are not mutually exclusive. The terminal critical events constituted regularly scheduled elections do not appear within any other category of terminal events. This means that the maximum marginal percentage for all other categories is 60 percent.

4. Interestingly, strikes are not always generated by economic demands. The Federation of Icelandic Unions, for example, once called a general strike to protest American defense proposals.

5. This is exemplified by the collective resignation of the cabinet rather than the resignation of the prime minister or of one or more other specific ministers.

6. These elections constitute 17 percent of terminal critical events overall, but are scarcely seen in Finland, Sweden, and Norway. The Norwegian constitution prohibits the early dissolution of the Storting, while the Swedish Riksdag can only be dissolved by the government, and the Finnish Eduskunta can only be dissolved by the president.

REFERENCES

Brass, P. 1977: 'Party Systems and Government Stability in the Indian States', American Political
Science Review 71, 13 84-1405.

Brown, E. C, Frendreis, J. P. & Gleiber, D. W., 1982. The Survival of Cabinet Governments: The
Dissolution of Cabinets as a Stochastic Process.' Paper presented at the Annual Meeting of the
American Political Science Association, September 2-5.

Brown, E. C, Frendreis, J. P. & Gleiber, D. W., 1983. 'The Effect of Impending Elections on the
Decision to dissolve Cabinets in European Parliamentary Systems.' Paper presented at the Annual
Meeting of the Midwest Political Science Assocition, April 20-22.

Brown, E. C, Frendreis, J. P. & Gleiber, D. W., 1984a. 'An 'Events' Approach to the Problem of
Cabinet Stability', Comparative Political Studies 17, 167-197.

Brown, E. C, Frendreis, J. P. & Gleiber, D. W., 1984b. 'Testing Theories of Coalition Dissolution:
How Straightforward Empirical Approaches Can be Dead Wrong.' Unpublished Manuscript.

Brown, E. C, Frendreis, J. P. & Gleiber, D. W., 1986. 'The Process of Cabinet Dissolution: An Exponential
of Duration and Stability in Western Democracies', American Journal of Political
forthcoming.

Cioffi-Revilla, C. 1984. 'The Political Reliability of Italian Governments'. American Political
Science Review 78, 318-337.

DeSwaan, A. 1973. Coalition Theories and Cabinet Formations. Amsterdam: Elsevier Scientific
Publishing Company.

Dodd, L. C. 1974. 'Party Coalitions in Multiparty Parliaments: A Game-Theoretic Analysis', American
Science Review 67, 1093-1117.

Dodd, L. C. 1976. Coalitions in Parliamentary Government. Princeton, N. J.: Princeton University
Press.

Frendreis, J. P. 1983: 'Explanation of Variation and Detection of Covariation: The Purpose and
Logic of Comparative Analysis', Comparative Political Studies 16, 255-272.

Lowell, A. L. 1896. Goverments and Parties in Continental Europe, Vols. I and 11. Cambridge:
Harvard University Press.

Mann, N. R., Shaefer, R. E. & Singpurwalla, N. D. 1974. Methods for Statistical Analysis of Reliability
Life Data. N. Y.: Wiley.

Pesonen, P. & Thomas, A. H. 1983. 'Coalition Formation in Scandinavia.' In V. Bogdanor, ed.
Coalition Government in Western Europe. Amsterdam: Heinemann Educational Books: 59-96.

Sanders, T. & Herman, V. 1977. 'The Stability and Survival of Governments in Western Democracies'.
Politico, 12,346-377.

Särlvik, B. 1983. 'Coalition Politics and Policy Output in Scandinavia: Sweden, Denmark, and Norway.'
V. Bogdanor, ed. Coalition Government in Western Europe: 97-152.

Taylor, M. & Herman, V. 1971. 'Party Systems and Government Stability', American Political
Science Review 65, 28-37.

Warwick, P. 1979. 'The Durability of Coalition Governments in Parliamentary Democracies',
Comparative Political Studies 11,465-498.